Hopefully last target-arm queue before softfreeze; this one's largest part is the remainder of the SVE patches, but there are a selection of other minor things too. thanks -- PMM The following changes since commit 109b25045b3651f9c5d02c3766c0b3ff63e6d193: Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging (2018-06-29 12:30:29 +0100) are available in the Git repository at: git://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20180629 for you to fetch changes up to 802abf4024d23e48d45373ac3f2b580124b54b47: target/arm: Add ID_ISAR6 (2018-06-29 15:30:54 +0100) ---------------------------------------------------------------- target-arm queue: * last of the SVE patches; SVE is now enabled for aarch64 linux-user * sd: Don't trace SDRequest crc field (coverity bugfix) * target/arm: Mark PMINTENSET accesses as possibly doing IO * clean up v7VE feature bit handling * i.mx7d: minor cleanups * target/arm: support reading of CNT[VCT|FRQ]_EL0 from user-space * target/arm: Implement ARMv8.2-DotProd * virt: add addresses to dt node names (which stops dtc from complaining that they're not correctly named) * cleanups: replace error_setg(&error_fatal) by error_report() + exit() ---------------------------------------------------------------- Aaron Lindsay (3): target/arm: Add ARM_FEATURE_V7VE for v7 Virtualization Extensions target/arm: Remove redundant DIV detection for KVM target/arm: Mark PMINTENSET accesses as possibly doing IO Alex Bennée (1): target/arm: support reading of CNT[VCT|FRQ]_EL0 from user-space Eric Auger (3): device_tree: Add qemu_fdt_node_unit_path hw/arm/virt: Silence dtc /intc warnings hw/arm/virt: Silence dtc /memory warning Jean-Christophe Dubois (3): i.mx7d: Remove unused header files i.mx7d: Change SRC unimplemented device name from sdma to src i.mx7d: Change IRQ number type from hwaddr to int Peter Maydell (1): sd: Don't trace SDRequest crc field Philippe Mathieu-Daudé (4): hw/block/fdc: Replace error_setg(&error_abort) by assert() hw/arm/sysbus-fdt: Replace error_setg(&error_fatal) by error_report() + exit() device_tree: Replace error_setg(&error_fatal) by error_report() + exit() sdcard: Use the ldst API Richard Henderson (40): target/arm: Implement SVE Memory Contiguous Load Group target/arm: Implement SVE Contiguous Load, first-fault and no-fault target/arm: Implement SVE Memory Contiguous Store Group target/arm: Implement SVE load and broadcast quadword target/arm: Implement SVE integer convert to floating-point target/arm: Implement SVE floating-point arithmetic (predicated) target/arm: Implement SVE FP Multiply-Add Group target/arm: Implement SVE Floating Point Accumulating Reduction Group target/arm: Implement SVE load and broadcast element target/arm: Implement SVE store vector/predicate register target/arm: Implement SVE scatter stores target/arm: Implement SVE prefetches target/arm: Implement SVE gather loads target/arm: Implement SVE first-fault gather loads target/arm: Implement SVE scatter store vector immediate target/arm: Implement SVE floating-point compare vectors target/arm: Implement SVE floating-point arithmetic with immediate target/arm: Implement SVE Floating Point Multiply Indexed Group target/arm: Implement SVE FP Fast Reduction Group target/arm: Implement SVE Floating Point Unary Operations - Unpredicated Group target/arm: Implement SVE FP Compare with Zero Group target/arm: Implement SVE floating-point trig multiply-add coefficient target/arm: Implement SVE floating-point convert precision target/arm: Implement SVE floating-point convert to integer target/arm: Implement SVE floating-point round to integral value target/arm: Implement SVE floating-point unary operations target/arm: Implement SVE MOVPRFX target/arm: Implement SVE floating-point complex add target/arm: Implement SVE fp complex multiply add target/arm: Pass index to AdvSIMD FCMLA (indexed) target/arm: Implement SVE fp complex multiply add (indexed) target/arm: Implement SVE dot product (vectors) target/arm: Implement SVE dot product (indexed) target/arm: Enable SVE for aarch64-linux-user target/arm: Implement ARMv8.2-DotProd target/arm: Fix SVE signed division vs x86 overflow exception target/arm: Fix SVE system register access checks target/arm: Prune a57 features from max target/arm: Prune a15 features from max target/arm: Add ID_ISAR6 include/sysemu/device_tree.h | 16 + target/arm/cpu.h | 3 + target/arm/helper-sve.h | 682 +++++++++++++++ target/arm/helper.h | 44 +- device_tree.c | 78 +- hw/arm/boot.c | 41 +- hw/arm/fsl-imx7.c | 8 +- hw/arm/mcimx7d-sabre.c | 2 - hw/arm/sysbus-fdt.c | 53 +- hw/arm/virt.c | 70 +- hw/block/fdc.c | 9 +- hw/sd/bcm2835_sdhost.c | 13 +- hw/sd/core.c | 2 +- hw/sd/milkymist-memcard.c | 3 +- hw/sd/omap_mmc.c | 6 +- hw/sd/pl181.c | 11 +- hw/sd/sdhci.c | 15 +- hw/sd/ssi-sd.c | 6 +- linux-user/elfload.c | 2 + target/arm/cpu.c | 36 +- target/arm/cpu64.c | 13 +- target/arm/helper.c | 44 +- target/arm/kvm32.c | 27 +- target/arm/sve_helper.c | 1875 +++++++++++++++++++++++++++++++++++++++++- target/arm/translate-a64.c | 62 +- target/arm/translate-sve.c | 1688 ++++++++++++++++++++++++++++++++++++- target/arm/translate.c | 102 ++- target/arm/vec_helper.c | 311 ++++++- hw/sd/trace-events | 2 +- target/arm/sve.decode | 427 ++++++++++ 30 files changed, 5394 insertions(+), 257 deletions(-) From: Philippe Mathieu-Daudé <f4bug@amsat.org> Use assert() instead of error_setg(&error_abort), as suggested by the "qapi/error.h" documentation: Please don't error_setg(&error_fatal, ...), use error_report() and exit(), because that's more obvious. Likewise, don't error_setg(&error_abort, ...), use assert(). Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Acked-by: John Snow <jsnow@redhat.com> Message-id: 20180625165749.3910-2-f4bug@amsat.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/block/fdc.c | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) diff --git a/hw/block/fdc.c b/hw/block/fdc.c index XXXXXXX..XXXXXXX 100644 --- a/hw/block/fdc.c +++ b/hw/block/fdc.c @@ -XXX,XX +XXX,XX @@ static int pick_geometry(FDrive *drv) nb_sectors, FloppyDriveType_str(parse->drive)); } + assert(type_match != -1 && "misconfigured fd_format"); match = type_match; } - - /* No match of any kind found -- fd_format is misconfigured, abort. */ - if (match == -1) { - error_setg(&error_abort, "No candidate geometries present in table " - " for floppy drive type '%s'", - FloppyDriveType_str(drv->drive)); - } - parse = &(fd_formats[match]); out: -- 2.17.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> Use error_report() + exit() instead of error_setg(&error_fatal), as suggested by the "qapi/error.h" documentation: Please don't error_setg(&error_fatal, ...), use error_report() and exit(), because that's more obvious. This fixes CID 1352173: "Passing null pointer dt_name to qemu_fdt_node_path, which dereferences it." And this also fixes: hw/arm/sysbus-fdt.c:322:9: warning: Array access (from variable 'node_path') results in a null pointer dereference if (node_path[1]) { ^~~~~~~~~~~~ Fixes: Coverity CID 1352173 (Dereference after null check) Suggested-by: Eric Blake <eblake@redhat.com> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Message-id: 20180625165749.3910-3-f4bug@amsat.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/sysbus-fdt.c | 53 +++++++++++++++++++++++++-------------------- 1 file changed, 30 insertions(+), 23 deletions(-) diff --git a/hw/arm/sysbus-fdt.c b/hw/arm/sysbus-fdt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/sysbus-fdt.c +++ b/hw/arm/sysbus-fdt.c @@ -XXX,XX +XXX,XX @@ static void copy_properties_from_host(HostProperty *props, int nb_props, r = qemu_fdt_getprop(host_fdt, node_path, props[i].name, &prop_len, - props[i].optional ? &err : &error_fatal); + &err); if (r) { qemu_fdt_setprop(guest_fdt, nodename, props[i].name, r, prop_len); } else { - if (prop_len != -FDT_ERR_NOTFOUND) { - /* optional property not returned although property exists */ - error_report_err(err); - } else { + if (props[i].optional && prop_len == -FDT_ERR_NOTFOUND) { + /* optional property does not exist */ error_free(err); + } else { + error_report_err(err); + } + if (!props[i].optional) { + /* mandatory property not found: bail out */ + exit(1); } } } @@ -XXX,XX +XXX,XX @@ static void fdt_build_clock_node(void *host_fdt, void *guest_fdt, node_offset = fdt_node_offset_by_phandle(host_fdt, host_phandle); if (node_offset <= 0) { - error_setg(&error_fatal, - "not able to locate clock handle %d in host device tree", - host_phandle); + error_report("not able to locate clock handle %d in host device tree", + host_phandle); + exit(1); } node_path = g_malloc(path_len); while ((ret = fdt_get_path(host_fdt, node_offset, node_path, path_len)) @@ -XXX,XX +XXX,XX @@ static void fdt_build_clock_node(void *host_fdt, void *guest_fdt, node_path = g_realloc(node_path, path_len); } if (ret < 0) { - error_setg(&error_fatal, - "not able to retrieve node path for clock handle %d", - host_phandle); + error_report("not able to retrieve node path for clock handle %d", + host_phandle); + exit(1); } r = qemu_fdt_getprop(host_fdt, node_path, "compatible", &prop_len, &error_fatal); if (strcmp(r, "fixed-clock")) { - error_setg(&error_fatal, - "clock handle %d is not a fixed clock", host_phandle); + error_report("clock handle %d is not a fixed clock", host_phandle); + exit(1); } nodename = strrchr(node_path, '/'); @@ -XXX,XX +XXX,XX @@ static int add_amd_xgbe_fdt_node(SysBusDevice *sbdev, void *opaque) dt_name = sysfs_to_dt_name(vbasedev->name); if (!dt_name) { - error_setg(&error_fatal, "%s incorrect sysfs device name %s", - __func__, vbasedev->name); + error_report("%s incorrect sysfs device name %s", + __func__, vbasedev->name); + exit(1); } node_path = qemu_fdt_node_path(host_fdt, dt_name, vdev->compat, &error_fatal); if (!node_path || !node_path[0]) { - error_setg(&error_fatal, "%s unable to retrieve node path for %s/%s", - __func__, dt_name, vdev->compat); + error_report("%s unable to retrieve node path for %s/%s", + __func__, dt_name, vdev->compat); + exit(1); } if (node_path[1]) { - error_setg(&error_fatal, "%s more than one node matching %s/%s!", - __func__, dt_name, vdev->compat); + error_report("%s more than one node matching %s/%s!", + __func__, dt_name, vdev->compat); + exit(1); } g_free(dt_name); if (vbasedev->num_regions != 5) { - error_setg(&error_fatal, "%s Does the host dt node combine XGBE/PHY?", - __func__); + error_report("%s Does the host dt node combine XGBE/PHY?", __func__); + exit(1); } /* generate nodes for DMA_CLK and PTP_CLK */ r = qemu_fdt_getprop(host_fdt, node_path[0], "clocks", &prop_len, &error_fatal); if (prop_len != 8) { - error_setg(&error_fatal, "%s clocks property should contain 2 handles", - __func__); + error_report("%s clocks property should contain 2 handles", __func__); + exit(1); } host_clock_phandles = (uint32_t *)r; guest_clock_phandles[0] = qemu_fdt_alloc_phandle(guest_fdt); -- 2.17.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> Use error_report() + exit() instead of error_setg(&error_fatal), as suggested by the "qapi/error.h" documentation: Please don't error_setg(&error_fatal, ...), use error_report() and exit(), because that's more obvious. Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: David Gibson <david@gibson.dropbear.id.au> Message-id: 20180625165749.3910-4-f4bug@amsat.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- device_tree.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) diff --git a/device_tree.c b/device_tree.c index XXXXXXX..XXXXXXX 100644 --- a/device_tree.c +++ b/device_tree.c @@ -XXX,XX +XXX,XX @@ static void read_fstree(void *fdt, const char *dirname) const char *parent_node; if (strstr(dirname, root_dir) != dirname) { - error_setg(&error_fatal, "%s: %s must be searched within %s", - __func__, dirname, root_dir); + error_report("%s: %s must be searched within %s", + __func__, dirname, root_dir); + exit(1); } parent_node = &dirname[strlen(SYSFS_DT_BASEDIR)]; d = opendir(dirname); if (!d) { - error_setg(&error_fatal, "%s cannot open %s", __func__, dirname); - return; + error_report("%s cannot open %s", __func__, dirname); + exit(1); } while ((de = readdir(d)) != NULL) { @@ -XXX,XX +XXX,XX @@ static void read_fstree(void *fdt, const char *dirname) tmpnam = g_strdup_printf("%s/%s", dirname, de->d_name); if (lstat(tmpnam, &st) < 0) { - error_setg(&error_fatal, "%s cannot lstat %s", __func__, tmpnam); + error_report("%s cannot lstat %s", __func__, tmpnam); + exit(1); } if (S_ISREG(st.st_mode)) { @@ -XXX,XX +XXX,XX @@ static void read_fstree(void *fdt, const char *dirname) gsize len; if (!g_file_get_contents(tmpnam, &val, &len, NULL)) { - error_setg(&error_fatal, "%s not able to extract info from %s", - __func__, tmpnam); + error_report("%s not able to extract info from %s", + __func__, tmpnam); + exit(1); } if (strlen(parent_node) > 0) { @@ -XXX,XX +XXX,XX @@ void *load_device_tree_from_sysfs(void) host_fdt = create_device_tree(&host_fdt_size); read_fstree(host_fdt, SYSFS_DT_BASEDIR); if (fdt_check_header(host_fdt)) { - error_setg(&error_fatal, - "%s host device tree extracted into memory is invalid", - __func__); + error_report("%s host device tree extracted into memory is invalid", + __func__); + exit(1); } return host_fdt; } -- 2.17.1 From: Eric Auger <eric.auger@redhat.com> This helper allows to retrieve the paths of nodes whose name match node-name or node-name@unit-address patterns. Signed-off-by: Eric Auger <eric.auger@redhat.com> Message-id: 1530044492-24921-2-git-send-email-eric.auger@redhat.com Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/sysemu/device_tree.h | 16 +++++++++++ device_tree.c | 55 ++++++++++++++++++++++++++++++++++++ 2 files changed, 71 insertions(+) diff --git a/include/sysemu/device_tree.h b/include/sysemu/device_tree.h index XXXXXXX..XXXXXXX 100644 --- a/include/sysemu/device_tree.h +++ b/include/sysemu/device_tree.h @@ -XXX,XX +XXX,XX @@ void *load_device_tree_from_sysfs(void); char **qemu_fdt_node_path(void *fdt, const char *name, char *compat, Error **errp); +/** + * qemu_fdt_node_unit_path: return the paths of nodes matching a given + * node-name, ie. node-name and node-name@unit-address + * @fdt: pointer to the dt blob + * @name: node name + * @errp: handle to an error object + * + * returns a newly allocated NULL-terminated array of node paths. + * Use g_strfreev() to free it. If one or more nodes were found, the + * array contains the path of each node and the last element equals to + * NULL. If there is no error but no matching node was found, the + * returned array contains a single element equal to NULL. If an error + * was encountered when parsing the blob, the function returns NULL + */ +char **qemu_fdt_node_unit_path(void *fdt, const char *name, Error **errp); + int qemu_fdt_setprop(void *fdt, const char *node_path, const char *property, const void *val, int size); int qemu_fdt_setprop_cell(void *fdt, const char *node_path, diff --git a/device_tree.c b/device_tree.c index XXXXXXX..XXXXXXX 100644 --- a/device_tree.c +++ b/device_tree.c @@ -XXX,XX +XXX,XX @@ static int findnode_nofail(void *fdt, const char *node_path) return offset; } +char **qemu_fdt_node_unit_path(void *fdt, const char *name, Error **errp) +{ + char *prefix = g_strdup_printf("%s@", name); + unsigned int path_len = 16, n = 0; + GSList *path_list = NULL, *iter; + const char *iter_name; + int offset, len, ret; + char **path_array; + + offset = fdt_next_node(fdt, -1, NULL); + + while (offset >= 0) { + iter_name = fdt_get_name(fdt, offset, &len); + if (!iter_name) { + offset = len; + break; + } + if (!strcmp(iter_name, name) || g_str_has_prefix(iter_name, prefix)) { + char *path; + + path = g_malloc(path_len); + while ((ret = fdt_get_path(fdt, offset, path, path_len)) + == -FDT_ERR_NOSPACE) { + path_len += 16; + path = g_realloc(path, path_len); + } + path_list = g_slist_prepend(path_list, path); + n++; + } + offset = fdt_next_node(fdt, offset, NULL); + } + g_free(prefix); + + if (offset < 0 && offset != -FDT_ERR_NOTFOUND) { + error_setg(errp, "%s: abort parsing dt for %s node units: %s", + __func__, name, fdt_strerror(offset)); + for (iter = path_list; iter; iter = iter->next) { + g_free(iter->data); + } + g_slist_free(path_list); + return NULL; + } + + path_array = g_new(char *, n + 1); + path_array[n--] = NULL; + + for (iter = path_list; iter; iter = iter->next) { + path_array[n--] = iter->data; + } + + g_slist_free(path_list); + + return path_array; +} + char **qemu_fdt_node_path(void *fdt, const char *name, char *compat, Error **errp) { -- 2.17.1 From: Eric Auger <eric.auger@redhat.com> When running dtc on the guest /proc/device-tree we get the following warnings: "Warning (unit_address_vs_reg): Node <name> has a reg or ranges property, but no unit name", with name: /intc, /intc/its, /intc/v2m. Nodes should have a name in the form <name>[@<unit-address>] where unit-address is the primary address used to access the device, listed in the node's reg property. This fix seems to make dtc happy. Signed-off-by: Eric Auger <eric.auger@redhat.com> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 1530044492-24921-3-git-send-email-eric.auger@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/virt.c | 63 +++++++++++++++++++++++++++++++-------------------- 1 file changed, 39 insertions(+), 24 deletions(-) diff --git a/hw/arm/virt.c b/hw/arm/virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/virt.c +++ b/hw/arm/virt.c @@ -XXX,XX +XXX,XX @@ static void fdt_add_cpu_nodes(const VirtMachineState *vms) static void fdt_add_its_gic_node(VirtMachineState *vms) { + char *nodename; + vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt); - qemu_fdt_add_subnode(vms->fdt, "/intc/its"); - qemu_fdt_setprop_string(vms->fdt, "/intc/its", "compatible", + nodename = g_strdup_printf("/intc/its@%" PRIx64, + vms->memmap[VIRT_GIC_ITS].base); + qemu_fdt_add_subnode(vms->fdt, nodename); + qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "arm,gic-v3-its"); - qemu_fdt_setprop(vms->fdt, "/intc/its", "msi-controller", NULL, 0); - qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/its", "reg", + qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0); + qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, vms->memmap[VIRT_GIC_ITS].base, 2, vms->memmap[VIRT_GIC_ITS].size); - qemu_fdt_setprop_cell(vms->fdt, "/intc/its", "phandle", vms->msi_phandle); + qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle); + g_free(nodename); } static void fdt_add_v2m_gic_node(VirtMachineState *vms) { + char *nodename; + + nodename = g_strdup_printf("/intc/v2m@%" PRIx64, + vms->memmap[VIRT_GIC_V2M].base); vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt); - qemu_fdt_add_subnode(vms->fdt, "/intc/v2m"); - qemu_fdt_setprop_string(vms->fdt, "/intc/v2m", "compatible", + qemu_fdt_add_subnode(vms->fdt, nodename); + qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "arm,gic-v2m-frame"); - qemu_fdt_setprop(vms->fdt, "/intc/v2m", "msi-controller", NULL, 0); - qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/v2m", "reg", + qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0); + qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, vms->memmap[VIRT_GIC_V2M].base, 2, vms->memmap[VIRT_GIC_V2M].size); - qemu_fdt_setprop_cell(vms->fdt, "/intc/v2m", "phandle", vms->msi_phandle); + qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle); + g_free(nodename); } static void fdt_add_gic_node(VirtMachineState *vms) { + char *nodename; + vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt); qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle); - qemu_fdt_add_subnode(vms->fdt, "/intc"); - qemu_fdt_setprop_cell(vms->fdt, "/intc", "#interrupt-cells", 3); - qemu_fdt_setprop(vms->fdt, "/intc", "interrupt-controller", NULL, 0); - qemu_fdt_setprop_cell(vms->fdt, "/intc", "#address-cells", 0x2); - qemu_fdt_setprop_cell(vms->fdt, "/intc", "#size-cells", 0x2); - qemu_fdt_setprop(vms->fdt, "/intc", "ranges", NULL, 0); + nodename = g_strdup_printf("/intc@%" PRIx64, + vms->memmap[VIRT_GIC_DIST].base); + qemu_fdt_add_subnode(vms->fdt, nodename); + qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3); + qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0); + qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2); + qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2); + qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0); if (vms->gic_version == 3) { int nb_redist_regions = virt_gicv3_redist_region_count(vms); - qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible", + qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "arm,gic-v3"); - qemu_fdt_setprop_cell(vms->fdt, "/intc", + qemu_fdt_setprop_cell(vms->fdt, nodename, "#redistributor-regions", nb_redist_regions); if (nb_redist_regions == 1) { - qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg", + qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, vms->memmap[VIRT_GIC_DIST].base, 2, vms->memmap[VIRT_GIC_DIST].size, 2, vms->memmap[VIRT_GIC_REDIST].base, 2, vms->memmap[VIRT_GIC_REDIST].size); } else { - qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg", + qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, vms->memmap[VIRT_GIC_DIST].base, 2, vms->memmap[VIRT_GIC_DIST].size, 2, vms->memmap[VIRT_GIC_REDIST].base, @@ -XXX,XX +XXX,XX @@ static void fdt_add_gic_node(VirtMachineState *vms) } if (vms->virt) { - qemu_fdt_setprop_cells(vms->fdt, "/intc", "interrupts", + qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts", GIC_FDT_IRQ_TYPE_PPI, ARCH_GICV3_MAINT_IRQ, GIC_FDT_IRQ_FLAGS_LEVEL_HI); } } else { /* 'cortex-a15-gic' means 'GIC v2' */ - qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible", + qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "arm,cortex-a15-gic"); - qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg", + qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, vms->memmap[VIRT_GIC_DIST].base, 2, vms->memmap[VIRT_GIC_DIST].size, 2, vms->memmap[VIRT_GIC_CPU].base, 2, vms->memmap[VIRT_GIC_CPU].size); } - qemu_fdt_setprop_cell(vms->fdt, "/intc", "phandle", vms->gic_phandle); + qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle); + g_free(nodename); } static void fdt_add_pmu_nodes(const VirtMachineState *vms) -- 2.17.1 From: Eric Auger <eric.auger@redhat.com> When running dtc on the guest /proc/device-tree we get the following warning: Warning (unit_address_vs_reg): Node /memory has a reg or ranges property, but no unit name". Let's fix that by adding the unit address to the node name. We also don't create the /memory node anymore in create_fdt(). We directly create it in load_dtb. /chosen still needs to be created in create_fdt as the uart needs it. In case the user provided his own dtb, we nop all memory nodes found in root and create new one(s). Signed-off-by: Eric Auger <eric.auger@redhat.com> Message-id: 1530044492-24921-4-git-send-email-eric.auger@redhat.com Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/boot.c | 41 +++++++++++++++++++++++------------------ hw/arm/virt.c | 7 +------ 2 files changed, 24 insertions(+), 24 deletions(-) diff --git a/hw/arm/boot.c b/hw/arm/boot.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/boot.c +++ b/hw/arm/boot.c @@ -XXX,XX +XXX,XX @@ int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo, hwaddr addr_limit, AddressSpace *as) { void *fdt = NULL; - int size, rc; + int size, rc, n = 0; uint32_t acells, scells; char *nodename; unsigned int i; hwaddr mem_base, mem_len; + char **node_path; + Error *err = NULL; if (binfo->dtb_filename) { char *filename; @@ -XXX,XX +XXX,XX @@ int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo, goto fail; } + /* nop all root nodes matching /memory or /memory@unit-address */ + node_path = qemu_fdt_node_unit_path(fdt, "memory", &err); + if (err) { + error_report_err(err); + goto fail; + } + while (node_path[n]) { + if (g_str_has_prefix(node_path[n], "/memory")) { + qemu_fdt_nop_node(fdt, node_path[n]); + } + n++; + } + g_strfreev(node_path); + if (nb_numa_nodes > 0) { - /* - * Turn the /memory node created before into a NOP node, then create - * /memory@addr nodes for all numa nodes respectively. - */ - qemu_fdt_nop_node(fdt, "/memory"); mem_base = binfo->loader_start; for (i = 0; i < nb_numa_nodes; i++) { mem_len = numa_info[i].node_mem; @@ -XXX,XX +XXX,XX @@ int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo, g_free(nodename); } } else { - Error *err = NULL; + nodename = g_strdup_printf("/memory@%" PRIx64, binfo->loader_start); + qemu_fdt_add_subnode(fdt, nodename); + qemu_fdt_setprop_string(fdt, nodename, "device_type", "memory"); - rc = fdt_path_offset(fdt, "/memory"); - if (rc < 0) { - qemu_fdt_add_subnode(fdt, "/memory"); - } - - if (!qemu_fdt_getprop(fdt, "/memory", "device_type", NULL, &err)) { - qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory"); - } - - rc = qemu_fdt_setprop_sized_cells(fdt, "/memory", "reg", + rc = qemu_fdt_setprop_sized_cells(fdt, nodename, "reg", acells, binfo->loader_start, scells, binfo->ram_size); if (rc < 0) { - fprintf(stderr, "couldn't set /memory/reg\n"); + fprintf(stderr, "couldn't set %s reg\n", nodename); goto fail; } + g_free(nodename); } rc = fdt_path_offset(fdt, "/chosen"); diff --git a/hw/arm/virt.c b/hw/arm/virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/virt.c +++ b/hw/arm/virt.c @@ -XXX,XX +XXX,XX @@ static void create_fdt(VirtMachineState *vms) qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2); qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2); - /* - * /chosen and /memory nodes must exist for load_dtb - * to fill in necessary properties later - */ + /* /chosen must exist for load_dtb to fill in necessary properties later */ qemu_fdt_add_subnode(fdt, "/chosen"); - qemu_fdt_add_subnode(fdt, "/memory"); - qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory"); /* Clock node, for the benefit of the UART. The kernel device tree * binding documentation claims the PL011 node clock properties are -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-2-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 35 +++++++++ target/arm/sve_helper.c | 153 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 121 +++++++++++++++++++++++++++++ target/arm/sve.decode | 34 +++++++++ 4 files changed, 343 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_rsqrts_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_rsqrts_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_ld1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld3bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld4bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld2hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld3hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld4hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld2ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld3ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld4ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld2dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld3dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld4dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1bhu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1bsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1bdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1bhs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1bss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1bds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1hsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1hdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1hss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1hds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ld1sdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ld1sds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) return predtest_ones(d, oprsz, esz_mask); } + +/* + * Load contiguous data, protected by a governing predicate. + */ +#define DO_LD1(NAME, FN, TYPEE, TYPEM, H) \ +static void do_##NAME(CPUARMState *env, void *vd, void *vg, \ + target_ulong addr, intptr_t oprsz, \ + uintptr_t ra) \ +{ \ + intptr_t i = 0; \ + do { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m = 0; \ + if (pg & 1) { \ + m = FN(env, addr, ra); \ + } \ + *(TYPEE *)(vd + H(i)) = m; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += sizeof(TYPEM); \ + } while (i & 15); \ + } while (i < oprsz); \ +} \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + do_##NAME(env, &env->vfp.zregs[simd_data(desc)], vg, \ + addr, simd_oprsz(desc), GETPC()); \ +} + +#define DO_LD2(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m1 = 0, m2 = 0; \ + if (pg & 1) { \ + m1 = FN(env, addr, ra); \ + m2 = FN(env, addr + sizeof(TYPEM), ra); \ + } \ + *(TYPEE *)(d1 + H(i)) = m1; \ + *(TYPEE *)(d2 + H(i)) = m2; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 2 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +#define DO_LD3(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + void *d3 = &env->vfp.zregs[(rd + 2) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m1 = 0, m2 = 0, m3 = 0; \ + if (pg & 1) { \ + m1 = FN(env, addr, ra); \ + m2 = FN(env, addr + sizeof(TYPEM), ra); \ + m3 = FN(env, addr + 2 * sizeof(TYPEM), ra); \ + } \ + *(TYPEE *)(d1 + H(i)) = m1; \ + *(TYPEE *)(d2 + H(i)) = m2; \ + *(TYPEE *)(d3 + H(i)) = m3; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 3 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +#define DO_LD4(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + void *d3 = &env->vfp.zregs[(rd + 2) & 31]; \ + void *d4 = &env->vfp.zregs[(rd + 3) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m1 = 0, m2 = 0, m3 = 0, m4 = 0; \ + if (pg & 1) { \ + m1 = FN(env, addr, ra); \ + m2 = FN(env, addr + sizeof(TYPEM), ra); \ + m3 = FN(env, addr + 2 * sizeof(TYPEM), ra); \ + m4 = FN(env, addr + 3 * sizeof(TYPEM), ra); \ + } \ + *(TYPEE *)(d1 + H(i)) = m1; \ + *(TYPEE *)(d2 + H(i)) = m2; \ + *(TYPEE *)(d3 + H(i)) = m3; \ + *(TYPEE *)(d4 + H(i)) = m4; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 4 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +DO_LD1(sve_ld1bhu_r, cpu_ldub_data_ra, uint16_t, uint8_t, H1_2) +DO_LD1(sve_ld1bhs_r, cpu_ldsb_data_ra, uint16_t, int8_t, H1_2) +DO_LD1(sve_ld1bsu_r, cpu_ldub_data_ra, uint32_t, uint8_t, H1_4) +DO_LD1(sve_ld1bss_r, cpu_ldsb_data_ra, uint32_t, int8_t, H1_4) +DO_LD1(sve_ld1bdu_r, cpu_ldub_data_ra, uint64_t, uint8_t, ) +DO_LD1(sve_ld1bds_r, cpu_ldsb_data_ra, uint64_t, int8_t, ) + +DO_LD1(sve_ld1hsu_r, cpu_lduw_data_ra, uint32_t, uint16_t, H1_4) +DO_LD1(sve_ld1hss_r, cpu_ldsw_data_ra, uint32_t, int8_t, H1_4) +DO_LD1(sve_ld1hdu_r, cpu_lduw_data_ra, uint64_t, uint16_t, ) +DO_LD1(sve_ld1hds_r, cpu_ldsw_data_ra, uint64_t, int16_t, ) + +DO_LD1(sve_ld1sdu_r, cpu_ldl_data_ra, uint64_t, uint32_t, ) +DO_LD1(sve_ld1sds_r, cpu_ldl_data_ra, uint64_t, int32_t, ) + +DO_LD1(sve_ld1bb_r, cpu_ldub_data_ra, uint8_t, uint8_t, H1) +DO_LD2(sve_ld2bb_r, cpu_ldub_data_ra, uint8_t, uint8_t, H1) +DO_LD3(sve_ld3bb_r, cpu_ldub_data_ra, uint8_t, uint8_t, H1) +DO_LD4(sve_ld4bb_r, cpu_ldub_data_ra, uint8_t, uint8_t, H1) + +DO_LD1(sve_ld1hh_r, cpu_lduw_data_ra, uint16_t, uint16_t, H1_2) +DO_LD2(sve_ld2hh_r, cpu_lduw_data_ra, uint16_t, uint16_t, H1_2) +DO_LD3(sve_ld3hh_r, cpu_lduw_data_ra, uint16_t, uint16_t, H1_2) +DO_LD4(sve_ld4hh_r, cpu_lduw_data_ra, uint16_t, uint16_t, H1_2) + +DO_LD1(sve_ld1ss_r, cpu_ldl_data_ra, uint32_t, uint32_t, H1_4) +DO_LD2(sve_ld2ss_r, cpu_ldl_data_ra, uint32_t, uint32_t, H1_4) +DO_LD3(sve_ld3ss_r, cpu_ldl_data_ra, uint32_t, uint32_t, H1_4) +DO_LD4(sve_ld4ss_r, cpu_ldl_data_ra, uint32_t, uint32_t, H1_4) + +DO_LD1(sve_ld1dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) +DO_LD2(sve_ld2dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) +DO_LD3(sve_ld3dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) +DO_LD4(sve_ld4dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) + +#undef DO_LD1 +#undef DO_LD2 +#undef DO_LD3 +#undef DO_LD4 diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ typedef void gen_helper_gvec_flags_3(TCGv_i32, TCGv_ptr, TCGv_ptr, typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); +typedef void gen_helper_gvec_mem(TCGv_env, TCGv_ptr, TCGv_i64, TCGv_i32); + /* * Helpers for extracting complex instruction fields. */ @@ -XXX,XX +XXX,XX @@ static inline int expand_imm_sh8u(int x) return (uint8_t)x << (x & 0x100 ? 8 : 0); } +/* Convert a 2-bit memory size (msz) to a 4-bit data type (dtype) + * with unsigned data. C.f. SVE Memory Contiguous Load Group. + */ +static inline int msz_dtype(int msz) +{ + static const uint8_t dtype[4] = { 0, 5, 10, 15 }; + return dtype[msz]; +} + /* * Include the generated decoder. */ @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_pri(DisasContext *s, arg_rri *a, uint32_t insn) } return true; } + +/* + *** SVE Memory - Contiguous Load Group + */ + +/* The memory mode of the dtype. */ +static const TCGMemOp dtype_mop[16] = { + MO_UB, MO_UB, MO_UB, MO_UB, + MO_SL, MO_UW, MO_UW, MO_UW, + MO_SW, MO_SW, MO_UL, MO_UL, + MO_SB, MO_SB, MO_SB, MO_Q +}; + +#define dtype_msz(x) (dtype_mop[x] & MO_SIZE) + +/* The vector element size of dtype. */ +static const uint8_t dtype_esz[16] = { + 0, 1, 2, 3, + 3, 1, 2, 3, + 3, 2, 2, 3, + 3, 2, 1, 3 +}; + +static void do_mem_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr, + gen_helper_gvec_mem *fn) +{ + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr t_pg; + TCGv_i32 desc; + + /* For e.g. LD4, there are not enough arguments to pass all 4 + * registers as pointers, so encode the regno into the data field. + * For consistency, do this even for LD1. + */ + desc = tcg_const_i32(simd_desc(vsz, vsz, zt)); + t_pg = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg)); + fn(cpu_env, t_pg, addr, desc); + + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i32(desc); +} + +static void do_ld_zpa(DisasContext *s, int zt, int pg, + TCGv_i64 addr, int dtype, int nreg) +{ + static gen_helper_gvec_mem * const fns[16][4] = { + { gen_helper_sve_ld1bb_r, gen_helper_sve_ld2bb_r, + gen_helper_sve_ld3bb_r, gen_helper_sve_ld4bb_r }, + { gen_helper_sve_ld1bhu_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1bsu_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1bdu_r, NULL, NULL, NULL }, + + { gen_helper_sve_ld1sds_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1hh_r, gen_helper_sve_ld2hh_r, + gen_helper_sve_ld3hh_r, gen_helper_sve_ld4hh_r }, + { gen_helper_sve_ld1hsu_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1hdu_r, NULL, NULL, NULL }, + + { gen_helper_sve_ld1hds_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1hss_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1ss_r, gen_helper_sve_ld2ss_r, + gen_helper_sve_ld3ss_r, gen_helper_sve_ld4ss_r }, + { gen_helper_sve_ld1sdu_r, NULL, NULL, NULL }, + + { gen_helper_sve_ld1bds_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1bss_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1bhs_r, NULL, NULL, NULL }, + { gen_helper_sve_ld1dd_r, gen_helper_sve_ld2dd_r, + gen_helper_sve_ld3dd_r, gen_helper_sve_ld4dd_r }, + }; + gen_helper_gvec_mem *fn = fns[dtype][nreg]; + + /* While there are holes in the table, they are not + * accessible via the instruction encoding. + */ + assert(fn != NULL); + do_mem_zpa(s, zt, pg, addr, fn); +} + +static bool trans_LD_zprr(DisasContext *s, arg_rprr_load *a, uint32_t insn) +{ + if (a->rm == 31) { + return false; + } + if (sve_access_check(s)) { + TCGv_i64 addr = new_tmp_a64(s); + tcg_gen_muli_i64(addr, cpu_reg(s, a->rm), + (a->nreg + 1) << dtype_msz(a->dtype)); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + do_ld_zpa(s, a->rd, a->pg, addr, a->dtype, a->nreg); + } + return true; +} + +static bool trans_LD_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) +{ + if (sve_access_check(s)) { + int vsz = vec_full_reg_size(s); + int elements = vsz >> dtype_esz[a->dtype]; + TCGv_i64 addr = new_tmp_a64(s); + + tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), + (a->imm * elements * (a->nreg + 1)) + << dtype_msz(a->dtype)); + do_ld_zpa(s, a->rd, a->pg, addr, a->dtype, a->nreg); + } + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ # Unsigned 8-bit immediate, optionally shifted left by 8. %sh8_i8u 5:9 !function=expand_imm_sh8u +# Unsigned load of msz into esz=2, represented as a dtype. +%msz_dtype 23:2 !function=msz_dtype + # Either a copy of rd (at bit 0), or a different source # as propagated via the MOVPRFX instruction. %reg_movprfx 0:5 @@ -XXX,XX +XXX,XX @@ &incdec2_cnt rd rn pat esz imm d u &incdec_pred rd pg esz d u &incdec2_pred rd rn pg esz d u +&rprr_load rd pg rn rm dtype nreg +&rpri_load rd pg rn imm dtype nreg ########################################################################### # Named instruction formats. These are generally used to @@ -XXX,XX +XXX,XX @@ @incdec2_pred ........ esz:2 .... .. ..... .. pg:4 rd:5 \ &incdec2_pred rn=%reg_movprfx +# Loads; user must fill in NREG. +@rprr_load_dt ....... dtype:4 rm:5 ... pg:3 rn:5 rd:5 &rprr_load +@rpri_load_dt ....... dtype:4 . imm:s4 ... pg:3 rn:5 rd:5 &rpri_load + +@rprr_load_msz ....... .... rm:5 ... pg:3 rn:5 rd:5 \ + &rprr_load dtype=%msz_dtype +@rpri_load_msz ....... .... . imm:s4 ... pg:3 rn:5 rd:5 \ + &rpri_load dtype=%msz_dtype + ########################################################################### # Instruction patterns. Grouped according to the SVE encodingindex.xhtml. @@ -XXX,XX +XXX,XX @@ LDR_pri 10000101 10 ...... 000 ... ..... 0 .... @pd_rn_i9 # SVE load vector register LDR_zri 10000101 10 ...... 010 ... ..... ..... @rd_rn_i9 + +### SVE Memory Contiguous Load Group + +# SVE contiguous load (scalar plus scalar) +LD_zprr 1010010 .... ..... 010 ... ..... ..... @rprr_load_dt nreg=0 + +# SVE contiguous load (scalar plus immediate) +LD_zpri 1010010 .... 0.... 101 ... ..... ..... @rpri_load_dt nreg=0 + +# SVE contiguous non-temporal load (scalar plus scalar) +# LDNT1B, LDNT1H, LDNT1W, LDNT1D +# SVE load multiple structures (scalar plus scalar) +# LD2B, LD2H, LD2W, LD2D; etc. +LD_zprr 1010010 .. nreg:2 ..... 110 ... ..... ..... @rprr_load_msz + +# SVE contiguous non-temporal load (scalar plus immediate) +# LDNT1B, LDNT1H, LDNT1W, LDNT1D +# SVE load multiple structures (scalar plus immediate) +# LD2B, LD2H, LD2W, LD2D; etc. +LD_zpri 1010010 .. nreg:2 0.... 111 ... ..... ..... @rpri_load_msz -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Tested-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 40 ++++++++++ target/arm/sve_helper.c | 157 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 69 ++++++++++++++++ target/arm/sve.decode | 6 ++ 4 files changed, 272 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_ld1hds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld1sdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld1sds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldff1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bhu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bhs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1bds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldff1hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1hsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1hdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1hss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1hds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldff1ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1sdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldff1sds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldff1dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldnf1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bhu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bhs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1bds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldnf1hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1hsu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1hdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1hss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1hds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldnf1ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1sdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_ldnf1sds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_ldnf1dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_LD4(sve_ld4dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) #undef DO_LD2 #undef DO_LD3 #undef DO_LD4 + +/* + * Load contiguous data, first-fault and no-fault. + */ + +#ifdef CONFIG_USER_ONLY + +/* Fault on byte I. All bits in FFR from I are cleared. The vector + * result from I is CONSTRAINED UNPREDICTABLE; we choose the MERGE + * option, which leaves subsequent data unchanged. + */ +static void record_fault(CPUARMState *env, uintptr_t i, uintptr_t oprsz) +{ + uint64_t *ffr = env->vfp.pregs[FFR_PRED_NUM].p; + + if (i & 63) { + ffr[i / 64] &= MAKE_64BIT_MASK(0, i & 63); + i = ROUND_UP(i, 64); + } + for (; i < oprsz; i += 64) { + ffr[i / 64] = 0; + } +} + +/* Hold the mmap lock during the operation so that there is no race + * between page_check_range and the load operation. We expect the + * usual case to have no faults at all, so we check the whole range + * first and if successful defer to the normal load operation. + * + * TODO: Change mmap_lock to a rwlock so that multiple readers + * can run simultaneously. This will probably help other uses + * within QEMU as well. + */ +#define DO_LDFF1(PART, FN, TYPEE, TYPEM, H) \ +static void do_sve_ldff1##PART(CPUARMState *env, void *vd, void *vg, \ + target_ulong addr, intptr_t oprsz, \ + bool first, uintptr_t ra) \ +{ \ + intptr_t i = 0; \ + do { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m = 0; \ + if (pg & 1) { \ + if (!first && \ + unlikely(page_check_range(addr, sizeof(TYPEM), \ + PAGE_READ))) { \ + record_fault(env, i, oprsz); \ + return; \ + } \ + m = FN(env, addr, ra); \ + first = false; \ + } \ + *(TYPEE *)(vd + H(i)) = m; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += sizeof(TYPEM); \ + } while (i & 15); \ + } while (i < oprsz); \ +} \ +void HELPER(sve_ldff1##PART)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t oprsz = simd_oprsz(desc); \ + unsigned rd = simd_data(desc); \ + void *vd = &env->vfp.zregs[rd]; \ + mmap_lock(); \ + if (likely(page_check_range(addr, oprsz, PAGE_READ) == 0)) { \ + do_sve_ld1##PART(env, vd, vg, addr, oprsz, GETPC()); \ + } else { \ + do_sve_ldff1##PART(env, vd, vg, addr, oprsz, true, GETPC()); \ + } \ + mmap_unlock(); \ +} + +/* No-fault loads are like first-fault loads without the + * first faulting special case. + */ +#define DO_LDNF1(PART) \ +void HELPER(sve_ldnf1##PART)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t oprsz = simd_oprsz(desc); \ + unsigned rd = simd_data(desc); \ + void *vd = &env->vfp.zregs[rd]; \ + mmap_lock(); \ + if (likely(page_check_range(addr, oprsz, PAGE_READ) == 0)) { \ + do_sve_ld1##PART(env, vd, vg, addr, oprsz, GETPC()); \ + } else { \ + do_sve_ldff1##PART(env, vd, vg, addr, oprsz, false, GETPC()); \ + } \ + mmap_unlock(); \ +} + +#else + +/* TODO: System mode is not yet supported. + * This would probably use tlb_vaddr_to_host. + */ +#define DO_LDFF1(PART, FN, TYPEE, TYPEM, H) \ +void HELPER(sve_ldff1##PART)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + g_assert_not_reached(); \ +} + +#define DO_LDNF1(PART) \ +void HELPER(sve_ldnf1##PART)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + g_assert_not_reached(); \ +} + +#endif + +DO_LDFF1(bb_r, cpu_ldub_data_ra, uint8_t, uint8_t, H1) +DO_LDFF1(bhu_r, cpu_ldub_data_ra, uint16_t, uint8_t, H1_2) +DO_LDFF1(bhs_r, cpu_ldsb_data_ra, uint16_t, int8_t, H1_2) +DO_LDFF1(bsu_r, cpu_ldub_data_ra, uint32_t, uint8_t, H1_4) +DO_LDFF1(bss_r, cpu_ldsb_data_ra, uint32_t, int8_t, H1_4) +DO_LDFF1(bdu_r, cpu_ldub_data_ra, uint64_t, uint8_t, ) +DO_LDFF1(bds_r, cpu_ldsb_data_ra, uint64_t, int8_t, ) + +DO_LDFF1(hh_r, cpu_lduw_data_ra, uint16_t, uint16_t, H1_2) +DO_LDFF1(hsu_r, cpu_lduw_data_ra, uint32_t, uint16_t, H1_4) +DO_LDFF1(hss_r, cpu_ldsw_data_ra, uint32_t, int8_t, H1_4) +DO_LDFF1(hdu_r, cpu_lduw_data_ra, uint64_t, uint16_t, ) +DO_LDFF1(hds_r, cpu_ldsw_data_ra, uint64_t, int16_t, ) + +DO_LDFF1(ss_r, cpu_ldl_data_ra, uint32_t, uint32_t, H1_4) +DO_LDFF1(sdu_r, cpu_ldl_data_ra, uint64_t, uint32_t, ) +DO_LDFF1(sds_r, cpu_ldl_data_ra, uint64_t, int32_t, ) + +DO_LDFF1(dd_r, cpu_ldq_data_ra, uint64_t, uint64_t, ) + +#undef DO_LDFF1 + +DO_LDNF1(bb_r) +DO_LDNF1(bhu_r) +DO_LDNF1(bhs_r) +DO_LDNF1(bsu_r) +DO_LDNF1(bss_r) +DO_LDNF1(bdu_r) +DO_LDNF1(bds_r) + +DO_LDNF1(hh_r) +DO_LDNF1(hsu_r) +DO_LDNF1(hss_r) +DO_LDNF1(hdu_r) +DO_LDNF1(hds_r) + +DO_LDNF1(ss_r) +DO_LDNF1(sdu_r) +DO_LDNF1(sds_r) + +DO_LDNF1(dd_r) + +#undef DO_LDNF1 diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) } return true; } + +static bool trans_LDFF1_zprr(DisasContext *s, arg_rprr_load *a, uint32_t insn) +{ + static gen_helper_gvec_mem * const fns[16] = { + gen_helper_sve_ldff1bb_r, + gen_helper_sve_ldff1bhu_r, + gen_helper_sve_ldff1bsu_r, + gen_helper_sve_ldff1bdu_r, + + gen_helper_sve_ldff1sds_r, + gen_helper_sve_ldff1hh_r, + gen_helper_sve_ldff1hsu_r, + gen_helper_sve_ldff1hdu_r, + + gen_helper_sve_ldff1hds_r, + gen_helper_sve_ldff1hss_r, + gen_helper_sve_ldff1ss_r, + gen_helper_sve_ldff1sdu_r, + + gen_helper_sve_ldff1bds_r, + gen_helper_sve_ldff1bss_r, + gen_helper_sve_ldff1bhs_r, + gen_helper_sve_ldff1dd_r, + }; + + if (sve_access_check(s)) { + TCGv_i64 addr = new_tmp_a64(s); + tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype)); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + do_mem_zpa(s, a->rd, a->pg, addr, fns[a->dtype]); + } + return true; +} + +static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) +{ + static gen_helper_gvec_mem * const fns[16] = { + gen_helper_sve_ldnf1bb_r, + gen_helper_sve_ldnf1bhu_r, + gen_helper_sve_ldnf1bsu_r, + gen_helper_sve_ldnf1bdu_r, + + gen_helper_sve_ldnf1sds_r, + gen_helper_sve_ldnf1hh_r, + gen_helper_sve_ldnf1hsu_r, + gen_helper_sve_ldnf1hdu_r, + + gen_helper_sve_ldnf1hds_r, + gen_helper_sve_ldnf1hss_r, + gen_helper_sve_ldnf1ss_r, + gen_helper_sve_ldnf1sdu_r, + + gen_helper_sve_ldnf1bds_r, + gen_helper_sve_ldnf1bss_r, + gen_helper_sve_ldnf1bhs_r, + gen_helper_sve_ldnf1dd_r, + }; + + if (sve_access_check(s)) { + int vsz = vec_full_reg_size(s); + int elements = vsz >> dtype_esz[a->dtype]; + int off = (a->imm * elements) << dtype_msz(a->dtype); + TCGv_i64 addr = new_tmp_a64(s); + + tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), off); + do_mem_zpa(s, a->rd, a->pg, addr, fns[a->dtype]); + } + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LDR_zri 10000101 10 ...... 010 ... ..... ..... @rd_rn_i9 # SVE contiguous load (scalar plus scalar) LD_zprr 1010010 .... ..... 010 ... ..... ..... @rprr_load_dt nreg=0 +# SVE contiguous first-fault load (scalar plus scalar) +LDFF1_zprr 1010010 .... ..... 011 ... ..... ..... @rprr_load_dt nreg=0 + # SVE contiguous load (scalar plus immediate) LD_zpri 1010010 .... 0.... 101 ... ..... ..... @rpri_load_dt nreg=0 +# SVE contiguous non-fault load (scalar plus immediate) +LDNF1_zpri 1010010 .... 1.... 101 ... ..... ..... @rpri_load_dt nreg=0 + # SVE contiguous non-temporal load (scalar plus scalar) # LDNT1B, LDNT1H, LDNT1W, LDNT1D # SVE load multiple structures (scalar plus scalar) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 29 +++++ target/arm/sve_helper.c | 211 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 65 ++++++++++++ target/arm/sve.decode | 38 +++++++ 4 files changed, 343 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_ldnf1sdu_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ldnf1sds_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ldnf1dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st3bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st4bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st2hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st3hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st4hh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st2ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st3ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st4ss_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st2dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st3dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st4dd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1bh_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st1bs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st1bd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1hs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_4(sve_st1hd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_4(sve_st1sd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_LDNF1(sds_r) DO_LDNF1(dd_r) #undef DO_LDNF1 + +/* + * Store contiguous data, protected by a governing predicate. + */ +#define DO_ST1(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *vd = &env->vfp.zregs[rd]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + TYPEM m = *(TYPEE *)(vd + H(i)); \ + FN(env, addr, m, ra); \ + } \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +#define DO_ST1_D(NAME, FN, TYPEM) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc) / 8; \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + uint64_t *d = &env->vfp.zregs[rd].d[0]; \ + uint8_t *pg = vg; \ + for (i = 0; i < oprsz; i += 1) { \ + if (pg[H1(i)] & 1) { \ + FN(env, addr, d[i], ra); \ + } \ + addr += sizeof(TYPEM); \ + } \ +} + +#define DO_ST2(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + TYPEM m1 = *(TYPEE *)(d1 + H(i)); \ + TYPEM m2 = *(TYPEE *)(d2 + H(i)); \ + FN(env, addr, m1, ra); \ + FN(env, addr + sizeof(TYPEM), m2, ra); \ + } \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 2 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +#define DO_ST3(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + void *d3 = &env->vfp.zregs[(rd + 2) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + TYPEM m1 = *(TYPEE *)(d1 + H(i)); \ + TYPEM m2 = *(TYPEE *)(d2 + H(i)); \ + TYPEM m3 = *(TYPEE *)(d3 + H(i)); \ + FN(env, addr, m1, ra); \ + FN(env, addr + sizeof(TYPEM), m2, ra); \ + FN(env, addr + 2 * sizeof(TYPEM), m3, ra); \ + } \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 3 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +#define DO_ST4(NAME, FN, TYPEE, TYPEM, H) \ +void HELPER(NAME)(CPUARMState *env, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + intptr_t ra = GETPC(); \ + unsigned rd = simd_data(desc); \ + void *d1 = &env->vfp.zregs[rd]; \ + void *d2 = &env->vfp.zregs[(rd + 1) & 31]; \ + void *d3 = &env->vfp.zregs[(rd + 2) & 31]; \ + void *d4 = &env->vfp.zregs[(rd + 3) & 31]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + TYPEM m1 = *(TYPEE *)(d1 + H(i)); \ + TYPEM m2 = *(TYPEE *)(d2 + H(i)); \ + TYPEM m3 = *(TYPEE *)(d3 + H(i)); \ + TYPEM m4 = *(TYPEE *)(d4 + H(i)); \ + FN(env, addr, m1, ra); \ + FN(env, addr + sizeof(TYPEM), m2, ra); \ + FN(env, addr + 2 * sizeof(TYPEM), m3, ra); \ + FN(env, addr + 3 * sizeof(TYPEM), m4, ra); \ + } \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + addr += 4 * sizeof(TYPEM); \ + } while (i & 15); \ + } \ +} + +DO_ST1(sve_st1bh_r, cpu_stb_data_ra, uint16_t, uint8_t, H1_2) +DO_ST1(sve_st1bs_r, cpu_stb_data_ra, uint32_t, uint8_t, H1_4) +DO_ST1_D(sve_st1bd_r, cpu_stb_data_ra, uint8_t) + +DO_ST1(sve_st1hs_r, cpu_stw_data_ra, uint32_t, uint16_t, H1_4) +DO_ST1_D(sve_st1hd_r, cpu_stw_data_ra, uint16_t) + +DO_ST1_D(sve_st1sd_r, cpu_stl_data_ra, uint32_t) + +DO_ST1(sve_st1bb_r, cpu_stb_data_ra, uint8_t, uint8_t, H1) +DO_ST2(sve_st2bb_r, cpu_stb_data_ra, uint8_t, uint8_t, H1) +DO_ST3(sve_st3bb_r, cpu_stb_data_ra, uint8_t, uint8_t, H1) +DO_ST4(sve_st4bb_r, cpu_stb_data_ra, uint8_t, uint8_t, H1) + +DO_ST1(sve_st1hh_r, cpu_stw_data_ra, uint16_t, uint16_t, H1_2) +DO_ST2(sve_st2hh_r, cpu_stw_data_ra, uint16_t, uint16_t, H1_2) +DO_ST3(sve_st3hh_r, cpu_stw_data_ra, uint16_t, uint16_t, H1_2) +DO_ST4(sve_st4hh_r, cpu_stw_data_ra, uint16_t, uint16_t, H1_2) + +DO_ST1(sve_st1ss_r, cpu_stl_data_ra, uint32_t, uint32_t, H1_4) +DO_ST2(sve_st2ss_r, cpu_stl_data_ra, uint32_t, uint32_t, H1_4) +DO_ST3(sve_st3ss_r, cpu_stl_data_ra, uint32_t, uint32_t, H1_4) +DO_ST4(sve_st4ss_r, cpu_stl_data_ra, uint32_t, uint32_t, H1_4) + +DO_ST1_D(sve_st1dd_r, cpu_stq_data_ra, uint64_t) + +void HELPER(sve_st2dd_r)(CPUARMState *env, void *vg, + target_ulong addr, uint32_t desc) +{ + intptr_t i, oprsz = simd_oprsz(desc) / 8; + intptr_t ra = GETPC(); + unsigned rd = simd_data(desc); + uint64_t *d1 = &env->vfp.zregs[rd].d[0]; + uint64_t *d2 = &env->vfp.zregs[(rd + 1) & 31].d[0]; + uint8_t *pg = vg; + + for (i = 0; i < oprsz; i += 1) { + if (pg[H1(i)] & 1) { + cpu_stq_data_ra(env, addr, d1[i], ra); + cpu_stq_data_ra(env, addr + 8, d2[i], ra); + } + addr += 2 * 8; + } +} + +void HELPER(sve_st3dd_r)(CPUARMState *env, void *vg, + target_ulong addr, uint32_t desc) +{ + intptr_t i, oprsz = simd_oprsz(desc) / 8; + intptr_t ra = GETPC(); + unsigned rd = simd_data(desc); + uint64_t *d1 = &env->vfp.zregs[rd].d[0]; + uint64_t *d2 = &env->vfp.zregs[(rd + 1) & 31].d[0]; + uint64_t *d3 = &env->vfp.zregs[(rd + 2) & 31].d[0]; + uint8_t *pg = vg; + + for (i = 0; i < oprsz; i += 1) { + if (pg[H1(i)] & 1) { + cpu_stq_data_ra(env, addr, d1[i], ra); + cpu_stq_data_ra(env, addr + 8, d2[i], ra); + cpu_stq_data_ra(env, addr + 16, d3[i], ra); + } + addr += 3 * 8; + } +} + +void HELPER(sve_st4dd_r)(CPUARMState *env, void *vg, + target_ulong addr, uint32_t desc) +{ + intptr_t i, oprsz = simd_oprsz(desc) / 8; + intptr_t ra = GETPC(); + unsigned rd = simd_data(desc); + uint64_t *d1 = &env->vfp.zregs[rd].d[0]; + uint64_t *d2 = &env->vfp.zregs[(rd + 1) & 31].d[0]; + uint64_t *d3 = &env->vfp.zregs[(rd + 2) & 31].d[0]; + uint64_t *d4 = &env->vfp.zregs[(rd + 3) & 31].d[0]; + uint8_t *pg = vg; + + for (i = 0; i < oprsz; i += 1) { + if (pg[H1(i)] & 1) { + cpu_stq_data_ra(env, addr, d1[i], ra); + cpu_stq_data_ra(env, addr + 8, d2[i], ra); + cpu_stq_data_ra(env, addr + 16, d3[i], ra); + cpu_stq_data_ra(env, addr + 24, d4[i], ra); + } + addr += 4 * 8; + } +} diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) } return true; } + +static void do_st_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr, + int msz, int esz, int nreg) +{ + static gen_helper_gvec_mem * const fn_single[4][4] = { + { gen_helper_sve_st1bb_r, gen_helper_sve_st1bh_r, + gen_helper_sve_st1bs_r, gen_helper_sve_st1bd_r }, + { NULL, gen_helper_sve_st1hh_r, + gen_helper_sve_st1hs_r, gen_helper_sve_st1hd_r }, + { NULL, NULL, + gen_helper_sve_st1ss_r, gen_helper_sve_st1sd_r }, + { NULL, NULL, NULL, gen_helper_sve_st1dd_r }, + }; + static gen_helper_gvec_mem * const fn_multiple[3][4] = { + { gen_helper_sve_st2bb_r, gen_helper_sve_st2hh_r, + gen_helper_sve_st2ss_r, gen_helper_sve_st2dd_r }, + { gen_helper_sve_st3bb_r, gen_helper_sve_st3hh_r, + gen_helper_sve_st3ss_r, gen_helper_sve_st3dd_r }, + { gen_helper_sve_st4bb_r, gen_helper_sve_st4hh_r, + gen_helper_sve_st4ss_r, gen_helper_sve_st4dd_r }, + }; + gen_helper_gvec_mem *fn; + + if (nreg == 0) { + /* ST1 */ + fn = fn_single[msz][esz]; + } else { + /* ST2, ST3, ST4 -- msz == esz, enforced by encoding */ + assert(msz == esz); + fn = fn_multiple[nreg - 1][msz]; + } + assert(fn != NULL); + do_mem_zpa(s, zt, pg, addr, fn); +} + +static bool trans_ST_zprr(DisasContext *s, arg_rprr_store *a, uint32_t insn) +{ + if (a->rm == 31 || a->msz > a->esz) { + return false; + } + if (sve_access_check(s)) { + TCGv_i64 addr = new_tmp_a64(s); + tcg_gen_muli_i64(addr, cpu_reg(s, a->rm), (a->nreg + 1) << a->msz); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + do_st_zpa(s, a->rd, a->pg, addr, a->msz, a->esz, a->nreg); + } + return true; +} + +static bool trans_ST_zpri(DisasContext *s, arg_rpri_store *a, uint32_t insn) +{ + if (a->msz > a->esz) { + return false; + } + if (sve_access_check(s)) { + int vsz = vec_full_reg_size(s); + int elements = vsz >> a->esz; + TCGv_i64 addr = new_tmp_a64(s); + + tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), + (a->imm * elements * (a->nreg + 1)) << a->msz); + do_st_zpa(s, a->rd, a->pg, addr, a->msz, a->esz, a->nreg); + } + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ %imm7_22_16 22:2 16:5 %imm8_16_10 16:5 10:3 %imm9_16_10 16:s6 10:3 +%size_23 23:2 # A combination of tsz:imm3 -- extract esize. %tszimm_esz 22:2 5:5 !function=tszimm_esz @@ -XXX,XX +XXX,XX @@ &incdec2_pred rd rn pg esz d u &rprr_load rd pg rn rm dtype nreg &rpri_load rd pg rn imm dtype nreg +&rprr_store rd pg rn rm msz esz nreg +&rpri_store rd pg rn imm msz esz nreg ########################################################################### # Named instruction formats. These are generally used to @@ -XXX,XX +XXX,XX @@ @rpri_load_msz ....... .... . imm:s4 ... pg:3 rn:5 rd:5 \ &rpri_load dtype=%msz_dtype +# Stores; user must fill in ESZ, MSZ, NREG as needed. +@rprr_store ....... .. .. rm:5 ... pg:3 rn:5 rd:5 &rprr_store +@rpri_store_msz ....... msz:2 .. . imm:s4 ... pg:3 rn:5 rd:5 &rpri_store +@rprr_store_esz_n0 ....... .. esz:2 rm:5 ... pg:3 rn:5 rd:5 \ + &rprr_store nreg=0 + ########################################################################### # Instruction patterns. Grouped according to the SVE encodingindex.xhtml. @@ -XXX,XX +XXX,XX @@ LD_zprr 1010010 .. nreg:2 ..... 110 ... ..... ..... @rprr_load_msz # SVE load multiple structures (scalar plus immediate) # LD2B, LD2H, LD2W, LD2D; etc. LD_zpri 1010010 .. nreg:2 0.... 111 ... ..... ..... @rpri_load_msz + +### SVE Memory Store Group + +# SVE contiguous store (scalar plus immediate) +# ST1B, ST1H, ST1W, ST1D; require msz <= esz +ST_zpri 1110010 .. esz:2 0.... 111 ... ..... ..... \ + @rpri_store_msz nreg=0 + +# SVE contiguous store (scalar plus scalar) +# ST1B, ST1H, ST1W, ST1D; require msz <= esz +# Enumerate msz lest we conflict with STR_zri. +ST_zprr 1110010 00 .. ..... 010 ... ..... ..... \ + @rprr_store_esz_n0 msz=0 +ST_zprr 1110010 01 .. ..... 010 ... ..... ..... \ + @rprr_store_esz_n0 msz=1 +ST_zprr 1110010 10 .. ..... 010 ... ..... ..... \ + @rprr_store_esz_n0 msz=2 +ST_zprr 1110010 11 11 ..... 010 ... ..... ..... \ + @rprr_store msz=3 esz=3 nreg=0 + +# SVE contiguous non-temporal store (scalar plus immediate) (nreg == 0) +# SVE store multiple structures (scalar plus immediate) (nreg != 0) +ST_zpri 1110010 .. nreg:2 1.... 111 ... ..... ..... \ + @rpri_store_msz esz=%size_23 + +# SVE contiguous non-temporal store (scalar plus scalar) (nreg == 0) +# SVE store multiple structures (scalar plus scalar) (nreg != 0) +ST_zprr 1110010 msz:2 nreg:2 ..... 011 ... ..... ..... \ + @rprr_store esz=%size_23 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 52 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 9 +++++++ 2 files changed, 61 insertions(+) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) return true; } +static void do_ldrq(DisasContext *s, int zt, int pg, TCGv_i64 addr, int msz) +{ + static gen_helper_gvec_mem * const fns[4] = { + gen_helper_sve_ld1bb_r, gen_helper_sve_ld1hh_r, + gen_helper_sve_ld1ss_r, gen_helper_sve_ld1dd_r, + }; + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr t_pg; + TCGv_i32 desc; + + /* Load the first quadword using the normal predicated load helpers. */ + desc = tcg_const_i32(simd_desc(16, 16, zt)); + t_pg = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg)); + fns[msz](cpu_env, t_pg, addr, desc); + + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i32(desc); + + /* Replicate that first quadword. */ + if (vsz > 16) { + unsigned dofs = vec_full_reg_offset(s, zt); + tcg_gen_gvec_dup_mem(4, dofs + 16, dofs, vsz - 16, vsz - 16); + } +} + +static bool trans_LD1RQ_zprr(DisasContext *s, arg_rprr_load *a, uint32_t insn) +{ + if (a->rm == 31) { + return false; + } + if (sve_access_check(s)) { + int msz = dtype_msz(a->dtype); + TCGv_i64 addr = new_tmp_a64(s); + tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), msz); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + do_ldrq(s, a->rd, a->pg, addr, msz); + } + return true; +} + +static bool trans_LD1RQ_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) +{ + if (sve_access_check(s)) { + TCGv_i64 addr = new_tmp_a64(s); + tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), a->imm * 16); + do_ldrq(s, a->rd, a->pg, addr, dtype_msz(a->dtype)); + } + return true; +} + static void do_st_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr, int msz, int esz, int nreg) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LD_zprr 1010010 .. nreg:2 ..... 110 ... ..... ..... @rprr_load_msz # LD2B, LD2H, LD2W, LD2D; etc. LD_zpri 1010010 .. nreg:2 0.... 111 ... ..... ..... @rpri_load_msz +# SVE load and broadcast quadword (scalar plus scalar) +LD1RQ_zprr 1010010 .. 00 ..... 000 ... ..... ..... \ + @rprr_load_msz nreg=0 + +# SVE load and broadcast quadword (scalar plus immediate) +# LD1RQB, LD1RQH, LD1RQS, LD1RQD +LD1RQ_zpri 1010010 .. 00 0.... 001 ... ..... ..... \ + @rpri_load_msz nreg=0 + ### SVE Memory Store Group # SVE contiguous store (scalar plus immediate) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> [PMM: fixed typo] Message-id: 20180627043328.11531-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 30 +++++++++++++ target/arm/sve_helper.c | 38 ++++++++++++++++ target/arm/translate-sve.c | 90 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 22 ++++++++++ 4 files changed, 180 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_rsqrts_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_rsqrts_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_dh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_ss, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_sd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_ds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_scvt_dd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_ucvt_hh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_sh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_dh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_ss, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_sd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_ds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ucvt_dd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(sve_ld1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld3bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) return predtest_ones(d, oprsz, esz_mask); } +/* Fully general two-operand expander, controlled by a predicate, + * With the extra float_status parameter. + */ +#define DO_ZPZ_FP(NAME, TYPE, H, OP) \ +void HELPER(NAME)(void *vd, void *vn, void *vg, void *status, uint32_t desc) \ +{ \ + intptr_t i = simd_oprsz(desc); \ + uint64_t *g = vg; \ + do { \ + uint64_t pg = g[(i - 1) >> 6]; \ + do { \ + i -= sizeof(TYPE); \ + if (likely((pg >> (i & 63)) & 1)) { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + *(TYPE *)(vd + H(i)) = OP(nn, status); \ + } \ + } while (i & 63); \ + } while (i != 0); \ +} + +DO_ZPZ_FP(sve_scvt_hh, uint16_t, H1_2, int16_to_float16) +DO_ZPZ_FP(sve_scvt_sh, uint32_t, H1_4, int32_to_float16) +DO_ZPZ_FP(sve_scvt_ss, uint32_t, H1_4, int32_to_float32) +DO_ZPZ_FP(sve_scvt_sd, uint64_t, , int32_to_float64) +DO_ZPZ_FP(sve_scvt_dh, uint64_t, , int64_to_float16) +DO_ZPZ_FP(sve_scvt_ds, uint64_t, , int64_to_float32) +DO_ZPZ_FP(sve_scvt_dd, uint64_t, , int64_to_float64) + +DO_ZPZ_FP(sve_ucvt_hh, uint16_t, H1_2, uint16_to_float16) +DO_ZPZ_FP(sve_ucvt_sh, uint32_t, H1_4, uint32_to_float16) +DO_ZPZ_FP(sve_ucvt_ss, uint32_t, H1_4, uint32_to_float32) +DO_ZPZ_FP(sve_ucvt_sd, uint64_t, , uint32_to_float64) +DO_ZPZ_FP(sve_ucvt_dh, uint64_t, , uint64_to_float16) +DO_ZPZ_FP(sve_ucvt_ds, uint64_t, , uint64_to_float32) +DO_ZPZ_FP(sve_ucvt_dd, uint64_t, , uint64_to_float64) + +#undef DO_ZPZ_FP + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FP3(FRSQRTS, rsqrts) #undef DO_FP3 + +/* + *** SVE Floating Point Unary Operations Predicated Group + */ + +static bool do_zpz_ptr(DisasContext *s, int rd, int rn, int pg, + bool is_fp16, gen_helper_gvec_3_ptr *fn) +{ + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(is_fp16); + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + pred_full_reg_offset(s, pg), + status, vsz, vsz, 0, fn); + tcg_temp_free_ptr(status); + } + return true; +} + +static bool trans_SCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_hh); +} + +static bool trans_SCVTF_sh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_sh); +} + +static bool trans_SCVTF_dh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_dh); +} + +static bool trans_SCVTF_ss(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_scvt_ss); +} + +static bool trans_SCVTF_ds(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_scvt_ds); +} + +static bool trans_SCVTF_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_scvt_sd); +} + +static bool trans_SCVTF_dd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_scvt_dd); +} + +static bool trans_UCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_ucvt_hh); +} + +static bool trans_UCVTF_sh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_ucvt_sh); +} + +static bool trans_UCVTF_dh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_ucvt_dh); +} + +static bool trans_UCVTF_ss(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_ucvt_ss); +} + +static bool trans_UCVTF_ds(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_ucvt_ds); +} + +static bool trans_UCVTF_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_ucvt_sd); +} + +static bool trans_UCVTF_dd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_ucvt_dd); +} + /* *** SVE Memory - 32-bit Gather and Unsized Contiguous Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ @rd_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 &rpr_esz @rd_pg4_pn ........ esz:2 ... ... .. pg:4 . rn:4 rd:5 &rpr_esz +# One register operand, with governing predicate, no vector element size +@rd_pg_rn_e0 ........ .. ... ... ... pg:3 rn:5 rd:5 &rpr_esz esz=0 + # Two register operands with a 6-bit signed immediate. @rd_rn_i6 ........ ... rn:5 ..... imm:s6 rd:5 &rri @@ -XXX,XX +XXX,XX @@ FTSMUL 01100101 .. 0 ..... 000 011 ..... ..... @rd_rn_rm FRECPS 01100101 .. 0 ..... 000 110 ..... ..... @rd_rn_rm FRSQRTS 01100101 .. 0 ..... 000 111 ..... ..... @rd_rn_rm +### SVE FP Unary Operations Predicated Group + +# SVE integer convert to floating-point +SCVTF_hh 01100101 01 010 01 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_sh 01100101 01 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_dh 01100101 01 010 11 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_ss 01100101 10 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_sd 01100101 11 010 00 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_ds 01100101 11 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 +SCVTF_dd 01100101 11 010 11 0 101 ... ..... ..... @rd_pg_rn_e0 + +UCVTF_hh 01100101 01 010 01 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_sh 01100101 01 010 10 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_dh 01100101 01 010 11 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_ss 01100101 10 010 10 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_sd 01100101 11 010 00 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_ds 01100101 11 010 10 1 101 ... ..... ..... @rd_pg_rn_e0 +UCVTF_dd 01100101 11 010 11 1 101 ... ..... ..... @rd_pg_rn_e0 + ### SVE Memory - 32-bit Gather and Unsized Contiguous Group # SVE load predicate register -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 77 +++++++++++++++++++++++++++++++++ target/arm/sve_helper.c | 89 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 46 ++++++++++++++++++++ target/arm/sve.decode | 17 ++++++++ 4 files changed, 229 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_rsqrts_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_rsqrts_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadd_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadd_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadd_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fsub_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsub_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsub_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmul_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmul_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmul_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fdiv_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fdiv_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fdiv_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmin_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmin_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmin_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmax_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmax_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmax_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fminnum_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fminnum_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fminnum_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmaxnum_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxnum_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxnum_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fabd_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fabd_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fabd_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fscalbn_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fscalbn_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fscalbn_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmulx_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmulx_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmulx_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) return predtest_ones(d, oprsz, esz_mask); } +/* Fully general three-operand expander, controlled by a predicate, + * With the extra float_status parameter. + */ +#define DO_ZPZZ_FP(NAME, TYPE, H, OP) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, \ + void *status, uint32_t desc) \ +{ \ + intptr_t i = simd_oprsz(desc); \ + uint64_t *g = vg; \ + do { \ + uint64_t pg = g[(i - 1) >> 6]; \ + do { \ + i -= sizeof(TYPE); \ + if (likely((pg >> (i & 63)) & 1)) { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + TYPE mm = *(TYPE *)(vm + H(i)); \ + *(TYPE *)(vd + H(i)) = OP(nn, mm, status); \ + } \ + } while (i & 63); \ + } while (i != 0); \ +} + +DO_ZPZZ_FP(sve_fadd_h, uint16_t, H1_2, float16_add) +DO_ZPZZ_FP(sve_fadd_s, uint32_t, H1_4, float32_add) +DO_ZPZZ_FP(sve_fadd_d, uint64_t, , float64_add) + +DO_ZPZZ_FP(sve_fsub_h, uint16_t, H1_2, float16_sub) +DO_ZPZZ_FP(sve_fsub_s, uint32_t, H1_4, float32_sub) +DO_ZPZZ_FP(sve_fsub_d, uint64_t, , float64_sub) + +DO_ZPZZ_FP(sve_fmul_h, uint16_t, H1_2, float16_mul) +DO_ZPZZ_FP(sve_fmul_s, uint32_t, H1_4, float32_mul) +DO_ZPZZ_FP(sve_fmul_d, uint64_t, , float64_mul) + +DO_ZPZZ_FP(sve_fdiv_h, uint16_t, H1_2, float16_div) +DO_ZPZZ_FP(sve_fdiv_s, uint32_t, H1_4, float32_div) +DO_ZPZZ_FP(sve_fdiv_d, uint64_t, , float64_div) + +DO_ZPZZ_FP(sve_fmin_h, uint16_t, H1_2, float16_min) +DO_ZPZZ_FP(sve_fmin_s, uint32_t, H1_4, float32_min) +DO_ZPZZ_FP(sve_fmin_d, uint64_t, , float64_min) + +DO_ZPZZ_FP(sve_fmax_h, uint16_t, H1_2, float16_max) +DO_ZPZZ_FP(sve_fmax_s, uint32_t, H1_4, float32_max) +DO_ZPZZ_FP(sve_fmax_d, uint64_t, , float64_max) + +DO_ZPZZ_FP(sve_fminnum_h, uint16_t, H1_2, float16_minnum) +DO_ZPZZ_FP(sve_fminnum_s, uint32_t, H1_4, float32_minnum) +DO_ZPZZ_FP(sve_fminnum_d, uint64_t, , float64_minnum) + +DO_ZPZZ_FP(sve_fmaxnum_h, uint16_t, H1_2, float16_maxnum) +DO_ZPZZ_FP(sve_fmaxnum_s, uint32_t, H1_4, float32_maxnum) +DO_ZPZZ_FP(sve_fmaxnum_d, uint64_t, , float64_maxnum) + +static inline float16 abd_h(float16 a, float16 b, float_status *s) +{ + return float16_abs(float16_sub(a, b, s)); +} + +static inline float32 abd_s(float32 a, float32 b, float_status *s) +{ + return float32_abs(float32_sub(a, b, s)); +} + +static inline float64 abd_d(float64 a, float64 b, float_status *s) +{ + return float64_abs(float64_sub(a, b, s)); +} + +DO_ZPZZ_FP(sve_fabd_h, uint16_t, H1_2, abd_h) +DO_ZPZZ_FP(sve_fabd_s, uint32_t, H1_4, abd_s) +DO_ZPZZ_FP(sve_fabd_d, uint64_t, , abd_d) + +static inline float64 scalbn_d(float64 a, int64_t b, float_status *s) +{ + int b_int = MIN(MAX(b, INT_MIN), INT_MAX); + return float64_scalbn(a, b_int, s); +} + +DO_ZPZZ_FP(sve_fscalbn_h, int16_t, H1_2, float16_scalbn) +DO_ZPZZ_FP(sve_fscalbn_s, int32_t, H1_4, float32_scalbn) +DO_ZPZZ_FP(sve_fscalbn_d, int64_t, , scalbn_d) + +DO_ZPZZ_FP(sve_fmulx_h, uint16_t, H1_2, helper_advsimd_mulxh) +DO_ZPZZ_FP(sve_fmulx_s, uint32_t, H1_4, helper_vfp_mulxs) +DO_ZPZZ_FP(sve_fmulx_d, uint64_t, , helper_vfp_mulxd) + +#undef DO_ZPZZ_FP + /* Fully general two-operand expander, controlled by a predicate, * With the extra float_status parameter. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FP3(FRSQRTS, rsqrts) #undef DO_FP3 +/* + *** SVE Floating Point Arithmetic - Predicated Group + */ + +static bool do_zpzz_fp(DisasContext *s, arg_rprr_esz *a, + gen_helper_gvec_4_ptr *fn) +{ + if (fn == NULL) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_4_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + pred_full_reg_offset(s, a->pg), + status, vsz, vsz, 0, fn); + tcg_temp_free_ptr(status); + } + return true; +} + +#define DO_FP3(NAME, name) \ +static bool trans_##NAME(DisasContext *s, arg_rprr_esz *a, uint32_t insn) \ +{ \ + static gen_helper_gvec_4_ptr * const fns[4] = { \ + NULL, gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, gen_helper_sve_##name##_d \ + }; \ + return do_zpzz_fp(s, a, fns[a->esz]); \ +} + +DO_FP3(FADD_zpzz, fadd) +DO_FP3(FSUB_zpzz, fsub) +DO_FP3(FMUL_zpzz, fmul) +DO_FP3(FMIN_zpzz, fmin) +DO_FP3(FMAX_zpzz, fmax) +DO_FP3(FMINNM_zpzz, fminnum) +DO_FP3(FMAXNM_zpzz, fmaxnum) +DO_FP3(FABD, fabd) +DO_FP3(FSCALE, fscalbn) +DO_FP3(FDIV, fdiv) +DO_FP3(FMULX, fmulx) + +#undef DO_FP3 /* *** SVE Floating Point Unary Operations Predicated Group diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FTSMUL 01100101 .. 0 ..... 000 011 ..... ..... @rd_rn_rm FRECPS 01100101 .. 0 ..... 000 110 ..... ..... @rd_rn_rm FRSQRTS 01100101 .. 0 ..... 000 111 ..... ..... @rd_rn_rm +### SVE FP Arithmetic Predicated Group + +# SVE floating-point arithmetic (predicated) +FADD_zpzz 01100101 .. 00 0000 100 ... ..... ..... @rdn_pg_rm +FSUB_zpzz 01100101 .. 00 0001 100 ... ..... ..... @rdn_pg_rm +FMUL_zpzz 01100101 .. 00 0010 100 ... ..... ..... @rdn_pg_rm +FSUB_zpzz 01100101 .. 00 0011 100 ... ..... ..... @rdm_pg_rn # FSUBR +FMAXNM_zpzz 01100101 .. 00 0100 100 ... ..... ..... @rdn_pg_rm +FMINNM_zpzz 01100101 .. 00 0101 100 ... ..... ..... @rdn_pg_rm +FMAX_zpzz 01100101 .. 00 0110 100 ... ..... ..... @rdn_pg_rm +FMIN_zpzz 01100101 .. 00 0111 100 ... ..... ..... @rdn_pg_rm +FABD 01100101 .. 00 1000 100 ... ..... ..... @rdn_pg_rm +FSCALE 01100101 .. 00 1001 100 ... ..... ..... @rdn_pg_rm +FMULX 01100101 .. 00 1010 100 ... ..... ..... @rdn_pg_rm +FDIV 01100101 .. 00 1100 100 ... ..... ..... @rdm_pg_rn # FDIVR +FDIV 01100101 .. 00 1101 100 ... ..... ..... @rdn_pg_rm + ### SVE FP Unary Operations Predicated Group # SVE integer convert to floating-point -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 16 ++++ target/arm/sve_helper.c | 158 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 49 ++++++++++++ target/arm/sve.decode | 18 +++++ 4 files changed, 241 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_ucvt_ds, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_ucvt_dd, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_fmls_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fmls_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fmls_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_fnmla_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fnmla_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fnmla_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) + DEF_HELPER_FLAGS_4(sve_ld1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld3bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_ZPZ_FP(sve_ucvt_dd, uint64_t, , uint64_to_float64) #undef DO_ZPZ_FP +/* 4-operand predicated multiply-add. This requires 7 operands to pass + * "properly", so we need to encode some of the registers into DESC. + */ +QEMU_BUILD_BUG_ON(SIMD_DATA_SHIFT + 20 > 32); + +static void do_fmla_zpzzz_h(CPUARMState *env, void *vg, uint32_t desc, + uint16_t neg1, uint16_t neg3) +{ + intptr_t i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + i -= 2; + if (likely((pg >> (i & 63)) & 1)) { + float16 e1, e2, e3, r; + + e1 = *(uint16_t *)(vn + H1_2(i)) ^ neg1; + e2 = *(uint16_t *)(vm + H1_2(i)); + e3 = *(uint16_t *)(va + H1_2(i)) ^ neg3; + r = float16_muladd(e1, e2, e3, 0, &env->vfp.fp_status); + *(uint16_t *)(vd + H1_2(i)) = r; + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fmla_zpzzz_h)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_h(env, vg, desc, 0, 0); +} + +void HELPER(sve_fmls_zpzzz_h)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_h(env, vg, desc, 0x8000, 0); +} + +void HELPER(sve_fnmla_zpzzz_h)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_h(env, vg, desc, 0x8000, 0x8000); +} + +void HELPER(sve_fnmls_zpzzz_h)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_h(env, vg, desc, 0, 0x8000); +} + +static void do_fmla_zpzzz_s(CPUARMState *env, void *vg, uint32_t desc, + uint32_t neg1, uint32_t neg3) +{ + intptr_t i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + i -= 4; + if (likely((pg >> (i & 63)) & 1)) { + float32 e1, e2, e3, r; + + e1 = *(uint32_t *)(vn + H1_4(i)) ^ neg1; + e2 = *(uint32_t *)(vm + H1_4(i)); + e3 = *(uint32_t *)(va + H1_4(i)) ^ neg3; + r = float32_muladd(e1, e2, e3, 0, &env->vfp.fp_status); + *(uint32_t *)(vd + H1_4(i)) = r; + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fmla_zpzzz_s)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_s(env, vg, desc, 0, 0); +} + +void HELPER(sve_fmls_zpzzz_s)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_s(env, vg, desc, 0x80000000, 0); +} + +void HELPER(sve_fnmla_zpzzz_s)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_s(env, vg, desc, 0x80000000, 0x80000000); +} + +void HELPER(sve_fnmls_zpzzz_s)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_s(env, vg, desc, 0, 0x80000000); +} + +static void do_fmla_zpzzz_d(CPUARMState *env, void *vg, uint32_t desc, + uint64_t neg1, uint64_t neg3) +{ + intptr_t i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + i -= 8; + if (likely((pg >> (i & 63)) & 1)) { + float64 e1, e2, e3, r; + + e1 = *(uint64_t *)(vn + i) ^ neg1; + e2 = *(uint64_t *)(vm + i); + e3 = *(uint64_t *)(va + i) ^ neg3; + r = float64_muladd(e1, e2, e3, 0, &env->vfp.fp_status); + *(uint64_t *)(vd + i) = r; + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fmla_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_d(env, vg, desc, 0, 0); +} + +void HELPER(sve_fmls_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_d(env, vg, desc, INT64_MIN, 0); +} + +void HELPER(sve_fnmla_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_d(env, vg, desc, INT64_MIN, INT64_MIN); +} + +void HELPER(sve_fnmls_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) +{ + do_fmla_zpzzz_d(env, vg, desc, 0, INT64_MIN); +} + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FP3(FMULX, fmulx) #undef DO_FP3 +typedef void gen_helper_sve_fmla(TCGv_env, TCGv_ptr, TCGv_i32); + +static bool do_fmla(DisasContext *s, arg_rprrr_esz *a, gen_helper_sve_fmla *fn) +{ + if (fn == NULL) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = vec_full_reg_size(s); + unsigned desc; + TCGv_i32 t_desc; + TCGv_ptr pg = tcg_temp_new_ptr(); + + /* We would need 7 operands to pass these arguments "properly". + * So we encode all the register numbers into the descriptor. + */ + desc = deposit32(a->rd, 5, 5, a->rn); + desc = deposit32(desc, 10, 5, a->rm); + desc = deposit32(desc, 15, 5, a->ra); + desc = simd_desc(vsz, vsz, desc); + + t_desc = tcg_const_i32(desc); + tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg)); + fn(cpu_env, pg, t_desc); + tcg_temp_free_i32(t_desc); + tcg_temp_free_ptr(pg); + return true; +} + +#define DO_FMLA(NAME, name) \ +static bool trans_##NAME(DisasContext *s, arg_rprrr_esz *a, uint32_t insn) \ +{ \ + static gen_helper_sve_fmla * const fns[4] = { \ + NULL, gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, gen_helper_sve_##name##_d \ + }; \ + return do_fmla(s, a, fns[a->esz]); \ +} + +DO_FMLA(FMLA_zpzzz, fmla_zpzzz) +DO_FMLA(FMLS_zpzzz, fmls_zpzzz) +DO_FMLA(FNMLA_zpzzz, fnmla_zpzzz) +DO_FMLA(FNMLS_zpzzz, fnmls_zpzzz) + +#undef DO_FMLA + /* *** SVE Floating Point Unary Operations Predicated Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ &rprrr_esz ra=%reg_movprfx @rdn_pg_ra_rm ........ esz:2 . rm:5 ... pg:3 ra:5 rd:5 \ &rprrr_esz rn=%reg_movprfx +@rdn_pg_rm_ra ........ esz:2 . ra:5 ... pg:3 rm:5 rd:5 \ + &rprrr_esz rn=%reg_movprfx # One register operand, with governing predicate, vector element size @rd_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 &rpr_esz @@ -XXX,XX +XXX,XX @@ FMULX 01100101 .. 00 1010 100 ... ..... ..... @rdn_pg_rm FDIV 01100101 .. 00 1100 100 ... ..... ..... @rdm_pg_rn # FDIVR FDIV 01100101 .. 00 1101 100 ... ..... ..... @rdn_pg_rm +### SVE FP Multiply-Add Group + +# SVE floating-point multiply-accumulate writing addend +FMLA_zpzzz 01100101 .. 1 ..... 000 ... ..... ..... @rda_pg_rn_rm +FMLS_zpzzz 01100101 .. 1 ..... 001 ... ..... ..... @rda_pg_rn_rm +FNMLA_zpzzz 01100101 .. 1 ..... 010 ... ..... ..... @rda_pg_rn_rm +FNMLS_zpzzz 01100101 .. 1 ..... 011 ... ..... ..... @rda_pg_rn_rm + +# SVE floating-point multiply-accumulate writing multiplicand +# Alter the operand extraction order and reuse the helpers from above. +# FMAD, FMSB, FNMAD, FNMS +FMLA_zpzzz 01100101 .. 1 ..... 100 ... ..... ..... @rdn_pg_rm_ra +FMLS_zpzzz 01100101 .. 1 ..... 101 ... ..... ..... @rdn_pg_rm_ra +FNMLA_zpzzz 01100101 .. 1 ..... 110 ... ..... ..... @rdn_pg_rm_ra +FNMLS_zpzzz 01100101 .. 1 ..... 111 ... ..... ..... @rdn_pg_rm_ra + ### SVE FP Unary Operations Predicated Group # SVE integer convert to floating-point -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 7 +++++ target/arm/sve_helper.c | 56 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 45 ++++++++++++++++++++++++++++++ target/arm/sve.decode | 5 ++++ 4 files changed, 113 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_rsqrts_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_rsqrts_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fadda_h, TCG_CALL_NO_RWG, + i64, i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fadda_s, TCG_CALL_NO_RWG, + i64, i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fadda_d, TCG_CALL_NO_RWG, + i64, i64, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_6(sve_fadd_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_6(sve_fadd_s, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) return predtest_ones(d, oprsz, esz_mask); } +uint64_t HELPER(sve_fadda_h)(uint64_t nn, void *vm, void *vg, + void *status, uint32_t desc) +{ + intptr_t i = 0, opr_sz = simd_oprsz(desc); + float16 result = nn; + + do { + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); + do { + if (pg & 1) { + float16 mm = *(float16 *)(vm + H1_2(i)); + result = float16_add(result, mm, status); + } + i += sizeof(float16), pg >>= sizeof(float16); + } while (i & 15); + } while (i < opr_sz); + + return result; +} + +uint64_t HELPER(sve_fadda_s)(uint64_t nn, void *vm, void *vg, + void *status, uint32_t desc) +{ + intptr_t i = 0, opr_sz = simd_oprsz(desc); + float32 result = nn; + + do { + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); + do { + if (pg & 1) { + float32 mm = *(float32 *)(vm + H1_2(i)); + result = float32_add(result, mm, status); + } + i += sizeof(float32), pg >>= sizeof(float32); + } while (i & 15); + } while (i < opr_sz); + + return result; +} + +uint64_t HELPER(sve_fadda_d)(uint64_t nn, void *vm, void *vg, + void *status, uint32_t desc) +{ + intptr_t i = 0, opr_sz = simd_oprsz(desc) / 8; + uint64_t *m = vm; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i++) { + if (pg[H1(i)] & 1) { + nn = float64_add(nn, m[i], status); + } + } + + return nn; +} + /* Fully general three-operand expander, controlled by a predicate, * With the extra float_status parameter. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_ZZI(UMIN, umin) #undef DO_ZZI +/* + *** SVE Floating Point Accumulating Reduction Group + */ + +static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a, uint32_t insn) +{ + typedef void fadda_fn(TCGv_i64, TCGv_i64, TCGv_ptr, + TCGv_ptr, TCGv_ptr, TCGv_i32); + static fadda_fn * const fns[3] = { + gen_helper_sve_fadda_h, + gen_helper_sve_fadda_s, + gen_helper_sve_fadda_d, + }; + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr t_rm, t_pg, t_fpst; + TCGv_i64 t_val; + TCGv_i32 t_desc; + + if (a->esz == 0) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + t_val = load_esz(cpu_env, vec_reg_offset(s, a->rn, 0, a->esz), a->esz); + t_rm = tcg_temp_new_ptr(); + t_pg = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(t_rm, cpu_env, vec_full_reg_offset(s, a->rm)); + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->pg)); + t_fpst = get_fpstatus_ptr(a->esz == MO_16); + t_desc = tcg_const_i32(simd_desc(vsz, vsz, 0)); + + fns[a->esz - 1](t_val, t_val, t_rm, t_pg, t_fpst, t_desc); + + tcg_temp_free_i32(t_desc); + tcg_temp_free_ptr(t_fpst); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_ptr(t_rm); + + write_fp_dreg(s, a->rd, t_val); + tcg_temp_free_i64(t_val); + return true; +} + /* *** SVE Floating Point Arithmetic - Unpredicated Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u # SVE integer multiply immediate (unpredicated) MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s +### SVE FP Accumulating Reduction Group + +# SVE floating-point serial reduction (predicated) +FADDA 01100101 .. 011 000 001 ... ..... ..... @rdn_pg_rm + ### SVE Floating Point Arithmetic - Unpredicated Group # SVE floating-point arithmetic (unpredicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-10-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 5 +++ target/arm/sve_helper.c | 41 +++++++++++++++++++++++++ target/arm/translate-sve.c | 62 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 5 +++ 4 files changed, 113 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(sve_clr_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_clr_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_clr_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_movz_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_movz_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_movz_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_movz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(sve_asr_zpzi_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_asr_zpzi_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_asr_zpzi_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_clr_d)(void *vd, void *vg, uint32_t desc) } } +/* Copy Zn into Zd, and store zero into inactive elements. */ +void HELPER(sve_movz_b)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + for (i = 0; i < opr_sz; i += 1) { + d[i] = n[i] & expand_pred_b(pg[H1(i)]); + } +} + +void HELPER(sve_movz_h)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + for (i = 0; i < opr_sz; i += 1) { + d[i] = n[i] & expand_pred_h(pg[H1(i)]); + } +} + +void HELPER(sve_movz_s)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + for (i = 0; i < opr_sz; i += 1) { + d[i] = n[i] & expand_pred_s(pg[H1(i)]); + } +} + +void HELPER(sve_movz_d)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + for (i = 0; i < opr_sz; i += 1) { + d[i] = n[1] & -(uint64_t)(pg[H1(i)] & 1); + } +} + /* Three-operand expander, immediate operand, controlled by a predicate. */ #define DO_ZPZI(NAME, TYPE, H, OP) \ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_clr_zp(DisasContext *s, int rd, int pg, int esz) return true; } +/* Copy Zn into Zd, storing zeros into inactive elements. */ +static void do_movz_zpz(DisasContext *s, int rd, int rn, int pg, int esz) +{ + static gen_helper_gvec_3 * const fns[4] = { + gen_helper_sve_movz_b, gen_helper_sve_movz_h, + gen_helper_sve_movz_s, gen_helper_sve_movz_d, + }; + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + pred_full_reg_offset(s, pg), + vsz, vsz, 0, fns[esz]); +} + static bool do_zpzi_ool(DisasContext *s, arg_rpri_esz *a, gen_helper_gvec_3 *fn) { @@ -XXX,XX +XXX,XX @@ static bool trans_LD1RQ_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) return true; } +/* Load and broadcast element. */ +static bool trans_LD1R_zpri(DisasContext *s, arg_rpri_load *a, uint32_t insn) +{ + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = vec_full_reg_size(s); + unsigned psz = pred_full_reg_size(s); + unsigned esz = dtype_esz[a->dtype]; + TCGLabel *over = gen_new_label(); + TCGv_i64 temp; + + /* If the guarding predicate has no bits set, no load occurs. */ + if (psz <= 8) { + /* Reduce the pred_esz_masks value simply to reduce the + * size of the code generated here. + */ + uint64_t psz_mask = MAKE_64BIT_MASK(0, psz * 8); + temp = tcg_temp_new_i64(); + tcg_gen_ld_i64(temp, cpu_env, pred_full_reg_offset(s, a->pg)); + tcg_gen_andi_i64(temp, temp, pred_esz_masks[esz] & psz_mask); + tcg_gen_brcondi_i64(TCG_COND_EQ, temp, 0, over); + tcg_temp_free_i64(temp); + } else { + TCGv_i32 t32 = tcg_temp_new_i32(); + find_last_active(s, t32, esz, a->pg); + tcg_gen_brcondi_i32(TCG_COND_LT, t32, 0, over); + tcg_temp_free_i32(t32); + } + + /* Load the data. */ + temp = tcg_temp_new_i64(); + tcg_gen_addi_i64(temp, cpu_reg_sp(s, a->rn), a->imm << esz); + tcg_gen_qemu_ld_i64(temp, temp, get_mem_index(s), + s->be_data | dtype_mop[a->dtype]); + + /* Broadcast to *all* elements. */ + tcg_gen_gvec_dup_i64(esz, vec_full_reg_offset(s, a->rd), + vsz, vsz, temp); + tcg_temp_free_i64(temp); + + /* Zero the inactive elements. */ + gen_set_label(over); + do_movz_zpz(s, a->rd, a->rd, a->pg, esz); + return true; +} + static void do_st_zpa(DisasContext *s, int zt, int pg, TCGv_i64 addr, int msz, int esz, int nreg) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ %imm8_16_10 16:5 10:3 %imm9_16_10 16:s6 10:3 %size_23 23:2 +%dtype_23_13 23:2 13:2 # A combination of tsz:imm3 -- extract esize. %tszimm_esz 22:2 5:5 !function=tszimm_esz @@ -XXX,XX +XXX,XX @@ LDR_pri 10000101 10 ...... 000 ... ..... 0 .... @pd_rn_i9 # SVE load vector register LDR_zri 10000101 10 ...... 010 ... ..... ..... @rd_rn_i9 +# SVE load and broadcast element +LD1R_zpri 1000010 .. 1 imm:6 1.. pg:3 rn:5 rd:5 \ + &rpri_load dtype=%dtype_23_13 nreg=0 + ### SVE Memory Contiguous Load Group # SVE contiguous load (scalar plus scalar) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-11-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 103 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 6 +++ 2 files changed, 109 insertions(+) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, uint32_t len, tcg_temp_free_i64(t0); } +/* Similarly for stores. */ +static void do_str(DisasContext *s, uint32_t vofs, uint32_t len, + int rn, int imm) +{ + uint32_t len_align = QEMU_ALIGN_DOWN(len, 8); + uint32_t len_remain = len % 8; + uint32_t nparts = len / 8 + ctpop8(len_remain); + int midx = get_mem_index(s); + TCGv_i64 addr, t0; + + addr = tcg_temp_new_i64(); + t0 = tcg_temp_new_i64(); + + /* Note that unpredicated load/store of vector/predicate registers + * are defined as a stream of bytes, which equates to little-endian + * operations on larger quantities. There is no nice way to force + * a little-endian store for aarch64_be-linux-user out of line. + * + * Attempt to keep code expansion to a minimum by limiting the + * amount of unrolling done. + */ + if (nparts <= 4) { + int i; + + for (i = 0; i < len_align; i += 8) { + tcg_gen_ld_i64(t0, cpu_env, vofs + i); + tcg_gen_addi_i64(addr, cpu_reg_sp(s, rn), imm + i); + tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEQ); + } + } else { + TCGLabel *loop = gen_new_label(); + TCGv_ptr t2, i = tcg_const_local_ptr(0); + + gen_set_label(loop); + + t2 = tcg_temp_new_ptr(); + tcg_gen_add_ptr(t2, cpu_env, i); + tcg_gen_ld_i64(t0, t2, vofs); + + /* Minimize the number of local temps that must be re-read from + * the stack each iteration. Instead, re-compute values other + * than the loop counter. + */ + tcg_gen_addi_ptr(t2, i, imm); + tcg_gen_extu_ptr_i64(addr, t2); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, rn)); + tcg_temp_free_ptr(t2); + + tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEQ); + + tcg_gen_addi_ptr(i, i, 8); + + tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop); + tcg_temp_free_ptr(i); + } + + /* Predicate register stores can be any multiple of 2. */ + if (len_remain) { + tcg_gen_ld_i64(t0, cpu_env, vofs + len_align); + tcg_gen_addi_i64(addr, cpu_reg_sp(s, rn), imm + len_align); + + switch (len_remain) { + case 2: + case 4: + case 8: + tcg_gen_qemu_st_i64(t0, addr, midx, MO_LE | ctz32(len_remain)); + break; + + case 6: + tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEUL); + tcg_gen_addi_i64(addr, addr, 4); + tcg_gen_shri_i64(t0, t0, 32); + tcg_gen_qemu_st_i64(t0, addr, midx, MO_LEUW); + break; + + default: + g_assert_not_reached(); + } + } + tcg_temp_free_i64(addr); + tcg_temp_free_i64(t0); +} + static bool trans_LDR_zri(DisasContext *s, arg_rri *a, uint32_t insn) { if (sve_access_check(s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_pri(DisasContext *s, arg_rri *a, uint32_t insn) return true; } +static bool trans_STR_zri(DisasContext *s, arg_rri *a, uint32_t insn) +{ + if (sve_access_check(s)) { + int size = vec_full_reg_size(s); + int off = vec_full_reg_offset(s, a->rd); + do_str(s, off, size, a->rn, a->imm * size); + } + return true; +} + +static bool trans_STR_pri(DisasContext *s, arg_rri *a, uint32_t insn) +{ + if (sve_access_check(s)) { + int size = pred_full_reg_size(s); + int off = pred_full_reg_offset(s, a->rd); + do_str(s, off, size, a->rn, a->imm * size); + } + return true; +} + /* *** SVE Memory - Contiguous Load Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LD1RQ_zpri 1010010 .. 00 0.... 001 ... ..... ..... \ ### SVE Memory Store Group +# SVE store predicate register +STR_pri 1110010 11 0. ..... 000 ... ..... 0 .... @pd_rn_i9 + +# SVE store vector register +STR_zri 1110010 11 0. ..... 010 ... ..... ..... @rd_rn_i9 + # SVE contiguous store (scalar plus immediate) # ST1B, ST1H, ST1W, ST1D; require msz <= esz ST_zpri 1110010 .. esz:2 0.... 111 ... ..... ..... \ -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-12-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 41 +++++++++++++++++++++ target/arm/sve_helper.c | 61 +++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 75 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 39 ++++++++++++++++++++ 4 files changed, 216 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_st1hs_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_st1hd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_st1sd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_stbs_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_sths_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stss_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_stbs_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_sths_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stss_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_stbd_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_sthd_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stsd_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stdd_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_stbd_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_sthd_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stsd_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stdd_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_stbd_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_sthd_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stsd_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_stdd_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_st4dd_r)(CPUARMState *env, void *vg, addr += 4 * 8; } } + +/* Stores with a vector index. */ + +#define DO_ST1_ZPZ_S(NAME, TYPEI, FN) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + unsigned scale = simd_data(desc); \ + uintptr_t ra = GETPC(); \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (likely(pg & 1)) { \ + target_ulong off = *(TYPEI *)(vm + H1_4(i)); \ + uint32_t d = *(uint32_t *)(vd + H1_4(i)); \ + FN(env, base + (off << scale), d, ra); \ + } \ + i += sizeof(uint32_t), pg >>= sizeof(uint32_t); \ + } while (i & 15); \ + } \ +} + +#define DO_ST1_ZPZ_D(NAME, TYPEI, FN) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc) / 8; \ + unsigned scale = simd_data(desc); \ + uintptr_t ra = GETPC(); \ + uint64_t *d = vd, *m = vm; uint8_t *pg = vg; \ + for (i = 0; i < oprsz; i++) { \ + if (likely(pg[H1(i)] & 1)) { \ + target_ulong off = (target_ulong)(TYPEI)m[i] << scale; \ + FN(env, base + off, d[i], ra); \ + } \ + } \ +} + +DO_ST1_ZPZ_S(sve_stbs_zsu, uint32_t, cpu_stb_data_ra) +DO_ST1_ZPZ_S(sve_sths_zsu, uint32_t, cpu_stw_data_ra) +DO_ST1_ZPZ_S(sve_stss_zsu, uint32_t, cpu_stl_data_ra) + +DO_ST1_ZPZ_S(sve_stbs_zss, int32_t, cpu_stb_data_ra) +DO_ST1_ZPZ_S(sve_sths_zss, int32_t, cpu_stw_data_ra) +DO_ST1_ZPZ_S(sve_stss_zss, int32_t, cpu_stl_data_ra) + +DO_ST1_ZPZ_D(sve_stbd_zsu, uint32_t, cpu_stb_data_ra) +DO_ST1_ZPZ_D(sve_sthd_zsu, uint32_t, cpu_stw_data_ra) +DO_ST1_ZPZ_D(sve_stsd_zsu, uint32_t, cpu_stl_data_ra) +DO_ST1_ZPZ_D(sve_stdd_zsu, uint32_t, cpu_stq_data_ra) + +DO_ST1_ZPZ_D(sve_stbd_zss, int32_t, cpu_stb_data_ra) +DO_ST1_ZPZ_D(sve_sthd_zss, int32_t, cpu_stw_data_ra) +DO_ST1_ZPZ_D(sve_stsd_zss, int32_t, cpu_stl_data_ra) +DO_ST1_ZPZ_D(sve_stdd_zss, int32_t, cpu_stq_data_ra) + +DO_ST1_ZPZ_D(sve_stbd_zd, uint64_t, cpu_stb_data_ra) +DO_ST1_ZPZ_D(sve_sthd_zd, uint64_t, cpu_stw_data_ra) +DO_ST1_ZPZ_D(sve_stsd_zd, uint64_t, cpu_stl_data_ra) +DO_ST1_ZPZ_D(sve_stdd_zd, uint64_t, cpu_stq_data_ra) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void gen_helper_gvec_mem(TCGv_env, TCGv_ptr, TCGv_i64, TCGv_i32); +typedef void gen_helper_gvec_mem_scatter(TCGv_env, TCGv_ptr, TCGv_ptr, + TCGv_ptr, TCGv_i64, TCGv_i32); /* * Helpers for extracting complex instruction fields. @@ -XXX,XX +XXX,XX @@ static bool trans_ST_zpri(DisasContext *s, arg_rpri_store *a, uint32_t insn) } return true; } + +/* + *** SVE gather loads / scatter stores + */ + +static void do_mem_zpz(DisasContext *s, int zt, int pg, int zm, int scale, + TCGv_i64 scalar, gen_helper_gvec_mem_scatter *fn) +{ + unsigned vsz = vec_full_reg_size(s); + TCGv_i32 desc = tcg_const_i32(simd_desc(vsz, vsz, scale)); + TCGv_ptr t_zm = tcg_temp_new_ptr(); + TCGv_ptr t_pg = tcg_temp_new_ptr(); + TCGv_ptr t_zt = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg)); + tcg_gen_addi_ptr(t_zm, cpu_env, vec_full_reg_offset(s, zm)); + tcg_gen_addi_ptr(t_zt, cpu_env, vec_full_reg_offset(s, zt)); + fn(cpu_env, t_zt, t_pg, t_zm, scalar, desc); + + tcg_temp_free_ptr(t_zt); + tcg_temp_free_ptr(t_zm); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i32(desc); +} + +static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) +{ + /* Indexed by [xs][msz]. */ + static gen_helper_gvec_mem_scatter * const fn32[2][3] = { + { gen_helper_sve_stbs_zsu, + gen_helper_sve_sths_zsu, + gen_helper_sve_stss_zsu, }, + { gen_helper_sve_stbs_zss, + gen_helper_sve_sths_zss, + gen_helper_sve_stss_zss, }, + }; + /* Note that we overload xs=2 to indicate 64-bit offset. */ + static gen_helper_gvec_mem_scatter * const fn64[3][4] = { + { gen_helper_sve_stbd_zsu, + gen_helper_sve_sthd_zsu, + gen_helper_sve_stsd_zsu, + gen_helper_sve_stdd_zsu, }, + { gen_helper_sve_stbd_zss, + gen_helper_sve_sthd_zss, + gen_helper_sve_stsd_zss, + gen_helper_sve_stdd_zss, }, + { gen_helper_sve_stbd_zd, + gen_helper_sve_sthd_zd, + gen_helper_sve_stsd_zd, + gen_helper_sve_stdd_zd, }, + }; + gen_helper_gvec_mem_scatter *fn; + + if (a->esz < a->msz || (a->msz == 0 && a->scale)) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + switch (a->esz) { + case MO_32: + fn = fn32[a->xs][a->msz]; + break; + case MO_64: + fn = fn64[a->xs][a->msz]; + break; + default: + g_assert_not_reached(); + } + do_mem_zpz(s, a->rd, a->pg, a->rm, a->scale * a->msz, + cpu_reg_sp(s, a->rn), fn); + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ &rpri_load rd pg rn imm dtype nreg &rprr_store rd pg rn rm msz esz nreg &rpri_store rd pg rn imm msz esz nreg +&rprr_scatter_store rd pg rn rm esz msz xs scale ########################################################################### # Named instruction formats. These are generally used to @@ -XXX,XX +XXX,XX @@ @rpri_store_msz ....... msz:2 .. . imm:s4 ... pg:3 rn:5 rd:5 &rpri_store @rprr_store_esz_n0 ....... .. esz:2 rm:5 ... pg:3 rn:5 rd:5 \ &rprr_store nreg=0 +@rprr_scatter_store ....... msz:2 .. rm:5 ... pg:3 rn:5 rd:5 \ + &rprr_scatter_store ########################################################################### # Instruction patterns. Grouped according to the SVE encodingindex.xhtml. @@ -XXX,XX +XXX,XX @@ ST_zpri 1110010 .. nreg:2 1.... 111 ... ..... ..... \ # SVE store multiple structures (scalar plus scalar) (nreg != 0) ST_zprr 1110010 msz:2 nreg:2 ..... 011 ... ..... ..... \ @rprr_store esz=%size_23 + +# SVE 32-bit scatter store (scalar plus 32-bit scaled offsets) +# Require msz > 0 && msz <= esz. +ST1_zprz 1110010 .. 11 ..... 100 ... ..... ..... \ + @rprr_scatter_store xs=0 esz=2 scale=1 +ST1_zprz 1110010 .. 11 ..... 110 ... ..... ..... \ + @rprr_scatter_store xs=1 esz=2 scale=1 + +# SVE 32-bit scatter store (scalar plus 32-bit unscaled offsets) +# Require msz <= esz. +ST1_zprz 1110010 .. 10 ..... 100 ... ..... ..... \ + @rprr_scatter_store xs=0 esz=2 scale=0 +ST1_zprz 1110010 .. 10 ..... 110 ... ..... ..... \ + @rprr_scatter_store xs=1 esz=2 scale=0 + +# SVE 64-bit scatter store (scalar plus 64-bit scaled offset) +# Require msz > 0 +ST1_zprz 1110010 .. 01 ..... 101 ... ..... ..... \ + @rprr_scatter_store xs=2 esz=3 scale=1 + +# SVE 64-bit scatter store (scalar plus 64-bit unscaled offset) +ST1_zprz 1110010 .. 00 ..... 101 ... ..... ..... \ + @rprr_scatter_store xs=2 esz=3 scale=0 + +# SVE 64-bit scatter store (scalar plus unpacked 32-bit scaled offset) +# Require msz > 0 +ST1_zprz 1110010 .. 01 ..... 100 ... ..... ..... \ + @rprr_scatter_store xs=0 esz=3 scale=1 +ST1_zprz 1110010 .. 01 ..... 110 ... ..... ..... \ + @rprr_scatter_store xs=1 esz=3 scale=1 + +# SVE 64-bit scatter store (scalar plus unpacked 32-bit unscaled offset) +ST1_zprz 1110010 .. 00 ..... 100 ... ..... ..... \ + @rprr_scatter_store xs=0 esz=3 scale=0 +ST1_zprz 1110010 .. 00 ..... 110 ... ..... ..... \ + @rprr_scatter_store xs=1 esz=3 scale=0 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-13-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 21 +++++++++++++++++++++ target/arm/sve.decode | 23 +++++++++++++++++++++++ 2 files changed, 44 insertions(+) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) cpu_reg_sp(s, a->rn), fn); return true; } + +/* + * Prefetches + */ + +static bool trans_PRF(DisasContext *s, arg_PRF *a, uint32_t insn) +{ + /* Prefetch is a nop within QEMU. */ + sve_access_check(s); + return true; +} + +static bool trans_PRF_rr(DisasContext *s, arg_PRF_rr *a, uint32_t insn) +{ + if (a->rm == 31) { + return false; + } + /* Prefetch is a nop within QEMU. */ + sve_access_check(s); + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LD1RQ_zprr 1010010 .. 00 ..... 000 ... ..... ..... \ LD1RQ_zpri 1010010 .. 00 0.... 001 ... ..... ..... \ @rpri_load_msz nreg=0 +# SVE 32-bit gather prefetch (scalar plus 32-bit scaled offsets) +PRF 1000010 00 -1 ----- 0-- --- ----- 0 ---- + +# SVE 32-bit gather prefetch (vector plus immediate) +PRF 1000010 -- 00 ----- 111 --- ----- 0 ---- + +# SVE contiguous prefetch (scalar plus immediate) +PRF 1000010 11 1- ----- 0-- --- ----- 0 ---- + +# SVE contiguous prefetch (scalar plus scalar) +PRF_rr 1000010 -- 00 rm:5 110 --- ----- 0 ---- + +### SVE Memory 64-bit Gather Group + +# SVE 64-bit gather prefetch (scalar plus 64-bit scaled offsets) +PRF 1100010 00 11 ----- 1-- --- ----- 0 ---- + +# SVE 64-bit gather prefetch (scalar plus unpacked 32-bit scaled offsets) +PRF 1100010 00 -1 ----- 0-- --- ----- 0 ---- + +# SVE 64-bit gather prefetch (vector plus immediate) +PRF 1100010 -- 00 ----- 111 --- ----- 0 ---- + ### SVE Memory Store Group # SVE store predicate register -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-14-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 67 +++++++++++++++++++++++++ target/arm/sve_helper.c | 77 ++++++++++++++++++++++++++++ target/arm/translate-sve.c | 100 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 57 +++++++++++++++++++++ 4 files changed, 301 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_st1hd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_st1sd_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbsu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhsu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldssu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbss_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhss_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldbsu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhsu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldssu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbss_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhss_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldbdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldddu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldbdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldddu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldbdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldddu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldbds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldhds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldsds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + DEF_HELPER_FLAGS_6(sve_stbs_zsu, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_6(sve_sths_zsu, TCG_CALL_NO_WG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_st4dd_r)(CPUARMState *env, void *vg, } } +/* Loads with a vector index. */ + +#define DO_LD1_ZPZ_S(NAME, TYPEI, TYPEM, FN) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + unsigned scale = simd_data(desc); \ + uintptr_t ra = GETPC(); \ + for (i = 0; i < oprsz; i++) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m = 0; \ + if (pg & 1) { \ + target_ulong off = *(TYPEI *)(vm + H1_4(i)); \ + m = FN(env, base + (off << scale), ra); \ + } \ + *(uint32_t *)(vd + H1_4(i)) = m; \ + i += 4, pg >>= 4; \ + } while (i & 15); \ + } \ +} + +#define DO_LD1_ZPZ_D(NAME, TYPEI, TYPEM, FN) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc) / 8; \ + unsigned scale = simd_data(desc); \ + uintptr_t ra = GETPC(); \ + uint64_t *d = vd, *m = vm; uint8_t *pg = vg; \ + for (i = 0; i < oprsz; i++) { \ + TYPEM mm = 0; \ + if (pg[H1(i)] & 1) { \ + target_ulong off = (TYPEI)m[i]; \ + mm = FN(env, base + (off << scale), ra); \ + } \ + d[i] = mm; \ + } \ +} + +DO_LD1_ZPZ_S(sve_ldbsu_zsu, uint32_t, uint8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_S(sve_ldhsu_zsu, uint32_t, uint16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_S(sve_ldssu_zsu, uint32_t, uint32_t, cpu_ldl_data_ra) +DO_LD1_ZPZ_S(sve_ldbss_zsu, uint32_t, int8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_S(sve_ldhss_zsu, uint32_t, int16_t, cpu_lduw_data_ra) + +DO_LD1_ZPZ_S(sve_ldbsu_zss, int32_t, uint8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_S(sve_ldhsu_zss, int32_t, uint16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_S(sve_ldssu_zss, int32_t, uint32_t, cpu_ldl_data_ra) +DO_LD1_ZPZ_S(sve_ldbss_zss, int32_t, int8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_S(sve_ldhss_zss, int32_t, int16_t, cpu_lduw_data_ra) + +DO_LD1_ZPZ_D(sve_ldbdu_zsu, uint32_t, uint8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhdu_zsu, uint32_t, uint16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsdu_zsu, uint32_t, uint32_t, cpu_ldl_data_ra) +DO_LD1_ZPZ_D(sve_ldddu_zsu, uint32_t, uint64_t, cpu_ldq_data_ra) +DO_LD1_ZPZ_D(sve_ldbds_zsu, uint32_t, int8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhds_zsu, uint32_t, int16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsds_zsu, uint32_t, int32_t, cpu_ldl_data_ra) + +DO_LD1_ZPZ_D(sve_ldbdu_zss, int32_t, uint8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhdu_zss, int32_t, uint16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsdu_zss, int32_t, uint32_t, cpu_ldl_data_ra) +DO_LD1_ZPZ_D(sve_ldddu_zss, int32_t, uint64_t, cpu_ldq_data_ra) +DO_LD1_ZPZ_D(sve_ldbds_zss, int32_t, int8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhds_zss, int32_t, int16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsds_zss, int32_t, int32_t, cpu_ldl_data_ra) + +DO_LD1_ZPZ_D(sve_ldbdu_zd, uint64_t, uint8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhdu_zd, uint64_t, uint16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsdu_zd, uint64_t, uint32_t, cpu_ldl_data_ra) +DO_LD1_ZPZ_D(sve_ldddu_zd, uint64_t, uint64_t, cpu_ldq_data_ra) +DO_LD1_ZPZ_D(sve_ldbds_zd, uint64_t, int8_t, cpu_ldub_data_ra) +DO_LD1_ZPZ_D(sve_ldhds_zd, uint64_t, int16_t, cpu_lduw_data_ra) +DO_LD1_ZPZ_D(sve_ldsds_zd, uint64_t, int32_t, cpu_ldl_data_ra) + /* Stores with a vector index. */ #define DO_ST1_ZPZ_S(NAME, TYPEI, FN) \ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static void do_mem_zpz(DisasContext *s, int zt, int pg, int zm, int scale, tcg_temp_free_i32(desc); } +/* Indexed by [ff][xs][u][msz]. */ +static gen_helper_gvec_mem_scatter * const gather_load_fn32[2][2][2][3] = { + { { { gen_helper_sve_ldbss_zsu, + gen_helper_sve_ldhss_zsu, + NULL, }, + { gen_helper_sve_ldbsu_zsu, + gen_helper_sve_ldhsu_zsu, + gen_helper_sve_ldssu_zsu, } }, + { { gen_helper_sve_ldbss_zss, + gen_helper_sve_ldhss_zss, + NULL, }, + { gen_helper_sve_ldbsu_zss, + gen_helper_sve_ldhsu_zss, + gen_helper_sve_ldssu_zss, } } }, + /* TODO fill in first-fault handlers */ +}; + +/* Note that we overload xs=2 to indicate 64-bit offset. */ +static gen_helper_gvec_mem_scatter * const gather_load_fn64[2][3][2][4] = { + { { { gen_helper_sve_ldbds_zsu, + gen_helper_sve_ldhds_zsu, + gen_helper_sve_ldsds_zsu, + NULL, }, + { gen_helper_sve_ldbdu_zsu, + gen_helper_sve_ldhdu_zsu, + gen_helper_sve_ldsdu_zsu, + gen_helper_sve_ldddu_zsu, } }, + { { gen_helper_sve_ldbds_zss, + gen_helper_sve_ldhds_zss, + gen_helper_sve_ldsds_zss, + NULL, }, + { gen_helper_sve_ldbdu_zss, + gen_helper_sve_ldhdu_zss, + gen_helper_sve_ldsdu_zss, + gen_helper_sve_ldddu_zss, } }, + { { gen_helper_sve_ldbds_zd, + gen_helper_sve_ldhds_zd, + gen_helper_sve_ldsds_zd, + NULL, }, + { gen_helper_sve_ldbdu_zd, + gen_helper_sve_ldhdu_zd, + gen_helper_sve_ldsdu_zd, + gen_helper_sve_ldddu_zd, } } }, + /* TODO fill in first-fault handlers */ +}; + +static bool trans_LD1_zprz(DisasContext *s, arg_LD1_zprz *a, uint32_t insn) +{ + gen_helper_gvec_mem_scatter *fn = NULL; + + if (!sve_access_check(s)) { + return true; + } + + switch (a->esz) { + case MO_32: + fn = gather_load_fn32[a->ff][a->xs][a->u][a->msz]; + break; + case MO_64: + fn = gather_load_fn64[a->ff][a->xs][a->u][a->msz]; + break; + } + assert(fn != NULL); + + do_mem_zpz(s, a->rd, a->pg, a->rm, a->scale * a->msz, + cpu_reg_sp(s, a->rn), fn); + return true; +} + +static bool trans_LD1_zpiz(DisasContext *s, arg_LD1_zpiz *a, uint32_t insn) +{ + gen_helper_gvec_mem_scatter *fn = NULL; + TCGv_i64 imm; + + if (a->esz < a->msz || (a->esz == a->msz && !a->u)) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + switch (a->esz) { + case MO_32: + fn = gather_load_fn32[a->ff][0][a->u][a->msz]; + break; + case MO_64: + fn = gather_load_fn64[a->ff][2][a->u][a->msz]; + break; + } + assert(fn != NULL); + + /* Treat LD1_zpiz (zn[x] + imm) the same way as LD1_zprz (rn + zm[x]) + * by loading the immediate into the scalar parameter. + */ + imm = tcg_const_i64(a->imm << a->msz); + do_mem_zpz(s, a->rd, a->pg, a->rn, 0, imm, fn); + tcg_temp_free_i64(imm); + return true; +} + static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) { /* Indexed by [xs][msz]. */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ &rpri_load rd pg rn imm dtype nreg &rprr_store rd pg rn rm msz esz nreg &rpri_store rd pg rn imm msz esz nreg +&rprr_gather_load rd pg rn rm esz msz u ff xs scale +&rpri_gather_load rd pg rn imm esz msz u ff &rprr_scatter_store rd pg rn rm esz msz xs scale ########################################################################### @@ -XXX,XX +XXX,XX @@ @rpri_load_msz ....... .... . imm:s4 ... pg:3 rn:5 rd:5 \ &rpri_load dtype=%msz_dtype +# Gather Loads. +@rprr_g_load_u ....... .. . . rm:5 . u:1 ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load xs=2 +@rprr_g_load_xs_u ....... .. xs:1 . rm:5 . u:1 ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load +@rprr_g_load_xs_u_sc ....... .. xs:1 scale:1 rm:5 . u:1 ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load +@rprr_g_load_xs_sc ....... .. xs:1 scale:1 rm:5 . . ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load +@rprr_g_load_u_sc ....... .. . scale:1 rm:5 . u:1 ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load xs=2 +@rprr_g_load_sc ....... .. . scale:1 rm:5 . . ff:1 pg:3 rn:5 rd:5 \ + &rprr_gather_load xs=2 +@rpri_g_load ....... msz:2 .. imm:5 . u:1 ff:1 pg:3 rn:5 rd:5 \ + &rpri_gather_load + # Stores; user must fill in ESZ, MSZ, NREG as needed. @rprr_store ....... .. .. rm:5 ... pg:3 rn:5 rd:5 &rprr_store @rpri_store_msz ....... msz:2 .. . imm:s4 ... pg:3 rn:5 rd:5 &rpri_store @@ -XXX,XX +XXX,XX @@ LDR_zri 10000101 10 ...... 010 ... ..... ..... @rd_rn_i9 LD1R_zpri 1000010 .. 1 imm:6 1.. pg:3 rn:5 rd:5 \ &rpri_load dtype=%dtype_23_13 nreg=0 +# SVE 32-bit gather load (scalar plus 32-bit unscaled offsets) +# SVE 32-bit gather load (scalar plus 32-bit scaled offsets) +LD1_zprz 1000010 00 .0 ..... 0.. ... ..... ..... \ + @rprr_g_load_xs_u esz=2 msz=0 scale=0 +LD1_zprz 1000010 01 .. ..... 0.. ... ..... ..... \ + @rprr_g_load_xs_u_sc esz=2 msz=1 +LD1_zprz 1000010 10 .. ..... 01. ... ..... ..... \ + @rprr_g_load_xs_sc esz=2 msz=2 u=1 + +# SVE 32-bit gather load (vector plus immediate) +LD1_zpiz 1000010 .. 01 ..... 1.. ... ..... ..... \ + @rpri_g_load esz=2 + ### SVE Memory Contiguous Load Group # SVE contiguous load (scalar plus scalar) @@ -XXX,XX +XXX,XX @@ PRF_rr 1000010 -- 00 rm:5 110 --- ----- 0 ---- ### SVE Memory 64-bit Gather Group +# SVE 64-bit gather load (scalar plus 32-bit unpacked unscaled offsets) +# SVE 64-bit gather load (scalar plus 32-bit unpacked scaled offsets) +LD1_zprz 1100010 00 .0 ..... 0.. ... ..... ..... \ + @rprr_g_load_xs_u esz=3 msz=0 scale=0 +LD1_zprz 1100010 01 .. ..... 0.. ... ..... ..... \ + @rprr_g_load_xs_u_sc esz=3 msz=1 +LD1_zprz 1100010 10 .. ..... 0.. ... ..... ..... \ + @rprr_g_load_xs_u_sc esz=3 msz=2 +LD1_zprz 1100010 11 .. ..... 01. ... ..... ..... \ + @rprr_g_load_xs_sc esz=3 msz=3 u=1 + +# SVE 64-bit gather load (scalar plus 64-bit unscaled offsets) +# SVE 64-bit gather load (scalar plus 64-bit scaled offsets) +LD1_zprz 1100010 00 10 ..... 1.. ... ..... ..... \ + @rprr_g_load_u esz=3 msz=0 scale=0 +LD1_zprz 1100010 01 1. ..... 1.. ... ..... ..... \ + @rprr_g_load_u_sc esz=3 msz=1 +LD1_zprz 1100010 10 1. ..... 1.. ... ..... ..... \ + @rprr_g_load_u_sc esz=3 msz=2 +LD1_zprz 1100010 11 1. ..... 11. ... ..... ..... \ + @rprr_g_load_sc esz=3 msz=3 u=1 + +# SVE 64-bit gather load (vector plus immediate) +LD1_zpiz 1100010 .. 01 ..... 1.. ... ..... ..... \ + @rpri_g_load esz=3 + # SVE 64-bit gather prefetch (scalar plus 64-bit scaled offsets) PRF 1100010 00 11 ----- 1-- --- ----- 0 ---- -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-15-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 67 +++++++++++++++++++++++++++++ target/arm/sve_helper.c | 88 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 40 ++++++++++++++++- 3 files changed, 193 insertions(+), 2 deletions(-) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(sve_ldhds_zd, TCG_CALL_NO_WG, DEF_HELPER_FLAGS_6(sve_ldsds_zd, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbsu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhsu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffssu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbss_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhss_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldffbsu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhsu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffssu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbss_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhss_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldffbdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsdu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffddu_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsds_zsu, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldffbdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsdu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffddu_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsds_zss, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_6(sve_ldffbdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsdu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffddu_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffbds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffhds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_6(sve_ldffsds_zd, TCG_CALL_NO_WG, + void, env, ptr, ptr, ptr, tl, i32) + DEF_HELPER_FLAGS_6(sve_stbs_zsu, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_6(sve_sths_zsu, TCG_CALL_NO_WG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_LD1_ZPZ_D(sve_ldbds_zd, uint64_t, int8_t, cpu_ldub_data_ra) DO_LD1_ZPZ_D(sve_ldhds_zd, uint64_t, int16_t, cpu_lduw_data_ra) DO_LD1_ZPZ_D(sve_ldsds_zd, uint64_t, int32_t, cpu_ldl_data_ra) +/* First fault loads with a vector index. */ + +#ifdef CONFIG_USER_ONLY + +#define DO_LDFF1_ZPZ(NAME, TYPEE, TYPEI, TYPEM, FN, H) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + unsigned scale = simd_data(desc); \ + uintptr_t ra = GETPC(); \ + bool first = true; \ + mmap_lock(); \ + for (i = 0; i < oprsz; i++) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPEM m = 0; \ + if (pg & 1) { \ + target_ulong off = *(TYPEI *)(vm + H(i)); \ + target_ulong addr = base + (off << scale); \ + if (!first && \ + page_check_range(addr, sizeof(TYPEM), PAGE_READ)) { \ + record_fault(env, i, oprsz); \ + goto exit; \ + } \ + m = FN(env, addr, ra); \ + first = false; \ + } \ + *(TYPEE *)(vd + H(i)) = m; \ + i += sizeof(TYPEE), pg >>= sizeof(TYPEE); \ + } while (i & 15); \ + } \ + exit: \ + mmap_unlock(); \ +} + +#else + +#define DO_LDFF1_ZPZ(NAME, TYPEE, TYPEI, TYPEM, FN, H) \ +void HELPER(NAME)(CPUARMState *env, void *vd, void *vg, void *vm, \ + target_ulong base, uint32_t desc) \ +{ \ + g_assert_not_reached(); \ +} + +#endif + +#define DO_LDFF1_ZPZ_S(NAME, TYPEI, TYPEM, FN) \ + DO_LDFF1_ZPZ(NAME, uint32_t, TYPEI, TYPEM, FN, H1_4) +#define DO_LDFF1_ZPZ_D(NAME, TYPEI, TYPEM, FN) \ + DO_LDFF1_ZPZ(NAME, uint64_t, TYPEI, TYPEM, FN, ) + +DO_LDFF1_ZPZ_S(sve_ldffbsu_zsu, uint32_t, uint8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffhsu_zsu, uint32_t, uint16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffssu_zsu, uint32_t, uint32_t, cpu_ldl_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffbss_zsu, uint32_t, int8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffhss_zsu, uint32_t, int16_t, cpu_lduw_data_ra) + +DO_LDFF1_ZPZ_S(sve_ldffbsu_zss, int32_t, uint8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffhsu_zss, int32_t, uint16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffssu_zss, int32_t, uint32_t, cpu_ldl_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffbss_zss, int32_t, int8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_S(sve_ldffhss_zss, int32_t, int16_t, cpu_lduw_data_ra) + +DO_LDFF1_ZPZ_D(sve_ldffbdu_zsu, uint32_t, uint8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhdu_zsu, uint32_t, uint16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsdu_zsu, uint32_t, uint32_t, cpu_ldl_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffddu_zsu, uint32_t, uint64_t, cpu_ldq_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffbds_zsu, uint32_t, int8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhds_zsu, uint32_t, int16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsds_zsu, uint32_t, int32_t, cpu_ldl_data_ra) + +DO_LDFF1_ZPZ_D(sve_ldffbdu_zss, int32_t, uint8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhdu_zss, int32_t, uint16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsdu_zss, int32_t, uint32_t, cpu_ldl_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffddu_zss, int32_t, uint64_t, cpu_ldq_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffbds_zss, int32_t, int8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhds_zss, int32_t, int16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsds_zss, int32_t, int32_t, cpu_ldl_data_ra) + +DO_LDFF1_ZPZ_D(sve_ldffbdu_zd, uint64_t, uint8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhdu_zd, uint64_t, uint16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsdu_zd, uint64_t, uint32_t, cpu_ldl_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffddu_zd, uint64_t, uint64_t, cpu_ldq_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffbds_zd, uint64_t, int8_t, cpu_ldub_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffhds_zd, uint64_t, int16_t, cpu_lduw_data_ra) +DO_LDFF1_ZPZ_D(sve_ldffsds_zd, uint64_t, int32_t, cpu_ldl_data_ra) + /* Stores with a vector index. */ #define DO_ST1_ZPZ_S(NAME, TYPEI, FN) \ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_mem_scatter * const gather_load_fn32[2][2][2][3] = { { gen_helper_sve_ldbsu_zss, gen_helper_sve_ldhsu_zss, gen_helper_sve_ldssu_zss, } } }, - /* TODO fill in first-fault handlers */ + + { { { gen_helper_sve_ldffbss_zsu, + gen_helper_sve_ldffhss_zsu, + NULL, }, + { gen_helper_sve_ldffbsu_zsu, + gen_helper_sve_ldffhsu_zsu, + gen_helper_sve_ldffssu_zsu, } }, + { { gen_helper_sve_ldffbss_zss, + gen_helper_sve_ldffhss_zss, + NULL, }, + { gen_helper_sve_ldffbsu_zss, + gen_helper_sve_ldffhsu_zss, + gen_helper_sve_ldffssu_zss, } } } }; /* Note that we overload xs=2 to indicate 64-bit offset. */ @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_mem_scatter * const gather_load_fn64[2][3][2][4] = { gen_helper_sve_ldhdu_zd, gen_helper_sve_ldsdu_zd, gen_helper_sve_ldddu_zd, } } }, - /* TODO fill in first-fault handlers */ + + { { { gen_helper_sve_ldffbds_zsu, + gen_helper_sve_ldffhds_zsu, + gen_helper_sve_ldffsds_zsu, + NULL, }, + { gen_helper_sve_ldffbdu_zsu, + gen_helper_sve_ldffhdu_zsu, + gen_helper_sve_ldffsdu_zsu, + gen_helper_sve_ldffddu_zsu, } }, + { { gen_helper_sve_ldffbds_zss, + gen_helper_sve_ldffhds_zss, + gen_helper_sve_ldffsds_zss, + NULL, }, + { gen_helper_sve_ldffbdu_zss, + gen_helper_sve_ldffhdu_zss, + gen_helper_sve_ldffsdu_zss, + gen_helper_sve_ldffddu_zss, } }, + { { gen_helper_sve_ldffbds_zd, + gen_helper_sve_ldffhds_zd, + gen_helper_sve_ldffsds_zd, + NULL, }, + { gen_helper_sve_ldffbdu_zd, + gen_helper_sve_ldffhdu_zd, + gen_helper_sve_ldffsdu_zd, + gen_helper_sve_ldffddu_zd, } } } }; static bool trans_LD1_zprz(DisasContext *s, arg_LD1_zprz *a, uint32_t insn) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-16-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 85 ++++++++++++++++++++++++++------------ target/arm/sve.decode | 11 +++++ 2 files changed, 70 insertions(+), 26 deletions(-) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD1_zpiz(DisasContext *s, arg_LD1_zpiz *a, uint32_t insn) return true; } +/* Indexed by [xs][msz]. */ +static gen_helper_gvec_mem_scatter * const scatter_store_fn32[2][3] = { + { gen_helper_sve_stbs_zsu, + gen_helper_sve_sths_zsu, + gen_helper_sve_stss_zsu, }, + { gen_helper_sve_stbs_zss, + gen_helper_sve_sths_zss, + gen_helper_sve_stss_zss, }, +}; + +/* Note that we overload xs=2 to indicate 64-bit offset. */ +static gen_helper_gvec_mem_scatter * const scatter_store_fn64[3][4] = { + { gen_helper_sve_stbd_zsu, + gen_helper_sve_sthd_zsu, + gen_helper_sve_stsd_zsu, + gen_helper_sve_stdd_zsu, }, + { gen_helper_sve_stbd_zss, + gen_helper_sve_sthd_zss, + gen_helper_sve_stsd_zss, + gen_helper_sve_stdd_zss, }, + { gen_helper_sve_stbd_zd, + gen_helper_sve_sthd_zd, + gen_helper_sve_stsd_zd, + gen_helper_sve_stdd_zd, }, +}; + static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) { - /* Indexed by [xs][msz]. */ - static gen_helper_gvec_mem_scatter * const fn32[2][3] = { - { gen_helper_sve_stbs_zsu, - gen_helper_sve_sths_zsu, - gen_helper_sve_stss_zsu, }, - { gen_helper_sve_stbs_zss, - gen_helper_sve_sths_zss, - gen_helper_sve_stss_zss, }, - }; - /* Note that we overload xs=2 to indicate 64-bit offset. */ - static gen_helper_gvec_mem_scatter * const fn64[3][4] = { - { gen_helper_sve_stbd_zsu, - gen_helper_sve_sthd_zsu, - gen_helper_sve_stsd_zsu, - gen_helper_sve_stdd_zsu, }, - { gen_helper_sve_stbd_zss, - gen_helper_sve_sthd_zss, - gen_helper_sve_stsd_zss, - gen_helper_sve_stdd_zss, }, - { gen_helper_sve_stbd_zd, - gen_helper_sve_sthd_zd, - gen_helper_sve_stsd_zd, - gen_helper_sve_stdd_zd, }, - }; gen_helper_gvec_mem_scatter *fn; if (a->esz < a->msz || (a->msz == 0 && a->scale)) { @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) } switch (a->esz) { case MO_32: - fn = fn32[a->xs][a->msz]; + fn = scatter_store_fn32[a->xs][a->msz]; break; case MO_64: - fn = fn64[a->xs][a->msz]; + fn = scatter_store_fn64[a->xs][a->msz]; break; default: g_assert_not_reached(); @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a, uint32_t insn) return true; } +static bool trans_ST1_zpiz(DisasContext *s, arg_ST1_zpiz *a, uint32_t insn) +{ + gen_helper_gvec_mem_scatter *fn = NULL; + TCGv_i64 imm; + + if (a->esz < a->msz) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + switch (a->esz) { + case MO_32: + fn = scatter_store_fn32[0][a->msz]; + break; + case MO_64: + fn = scatter_store_fn64[2][a->msz]; + break; + } + assert(fn != NULL); + + /* Treat ST1_zpiz (zn[x] + imm) the same way as ST1_zprz (rn + zm[x]) + * by loading the immediate into the scalar parameter. + */ + imm = tcg_const_i64(a->imm << a->msz); + do_mem_zpz(s, a->rd, a->pg, a->rn, 0, imm, fn); + tcg_temp_free_i64(imm); + return true; +} + /* * Prefetches */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ &rprr_gather_load rd pg rn rm esz msz u ff xs scale &rpri_gather_load rd pg rn imm esz msz u ff &rprr_scatter_store rd pg rn rm esz msz xs scale +&rpri_scatter_store rd pg rn imm esz msz ########################################################################### # Named instruction formats. These are generally used to @@ -XXX,XX +XXX,XX @@ &rprr_store nreg=0 @rprr_scatter_store ....... msz:2 .. rm:5 ... pg:3 rn:5 rd:5 \ &rprr_scatter_store +@rpri_scatter_store ....... msz:2 .. imm:5 ... pg:3 rn:5 rd:5 \ + &rpri_scatter_store ########################################################################### # Instruction patterns. Grouped according to the SVE encodingindex.xhtml. @@ -XXX,XX +XXX,XX @@ ST1_zprz 1110010 .. 01 ..... 101 ... ..... ..... \ ST1_zprz 1110010 .. 00 ..... 101 ... ..... ..... \ @rprr_scatter_store xs=2 esz=3 scale=0 +# SVE 64-bit scatter store (vector plus immediate) +ST1_zpiz 1110010 .. 10 ..... 101 ... ..... ..... \ + @rpri_scatter_store esz=3 + +# SVE 32-bit scatter store (vector plus immediate) +ST1_zpiz 1110010 .. 11 ..... 101 ... ..... ..... \ + @rpri_scatter_store esz=2 + # SVE 64-bit scatter store (scalar plus unpacked 32-bit scaled offset) # Require msz > 0 ST1_zprz 1110010 .. 01 ..... 100 ... ..... ..... \ -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-17-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 49 ++++++++++++++++++++++++++++++ target/arm/sve_helper.c | 62 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 40 ++++++++++++++++++++++++ target/arm/sve.decode | 11 +++++++ 4 files changed, 162 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_ucvt_ds, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_ucvt_dd, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmge_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmge_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmge_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fcmgt_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmgt_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmgt_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fcmeq_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmeq_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmeq_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fcmne_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmne_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmne_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fcmuo_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmuo_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcmuo_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_facge_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_facge_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_facge_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_facgt_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_facgt_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_facgt_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_fnmls_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) do_fmla_zpzzz_d(env, vg, desc, 0, INT64_MIN); } +/* Two operand floating-point comparison controlled by a predicate. + * Unlike the integer version, we are not allowed to optimistically + * compare operands, since the comparison may have side effects wrt + * the FPSR. + */ +#define DO_FPCMP_PPZZ(NAME, TYPE, H, OP) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, \ + void *status, uint32_t desc) \ +{ \ + intptr_t i = simd_oprsz(desc), j = (i - 1) >> 6; \ + uint64_t *d = vd, *g = vg; \ + do { \ + uint64_t out = 0, pg = g[j]; \ + do { \ + i -= sizeof(TYPE), out <<= sizeof(TYPE); \ + if (likely((pg >> (i & 63)) & 1)) { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + TYPE mm = *(TYPE *)(vm + H(i)); \ + out |= OP(TYPE, nn, mm, status); \ + } \ + } while (i & 63); \ + d[j--] = out; \ + } while (i > 0); \ +} + +#define DO_FPCMP_PPZZ_H(NAME, OP) \ + DO_FPCMP_PPZZ(NAME##_h, float16, H1_2, OP) +#define DO_FPCMP_PPZZ_S(NAME, OP) \ + DO_FPCMP_PPZZ(NAME##_s, float32, H1_4, OP) +#define DO_FPCMP_PPZZ_D(NAME, OP) \ + DO_FPCMP_PPZZ(NAME##_d, float64, , OP) + +#define DO_FPCMP_PPZZ_ALL(NAME, OP) \ + DO_FPCMP_PPZZ_H(NAME, OP) \ + DO_FPCMP_PPZZ_S(NAME, OP) \ + DO_FPCMP_PPZZ_D(NAME, OP) + +#define DO_FCMGE(TYPE, X, Y, ST) TYPE##_compare(Y, X, ST) <= 0 +#define DO_FCMGT(TYPE, X, Y, ST) TYPE##_compare(Y, X, ST) < 0 +#define DO_FCMEQ(TYPE, X, Y, ST) TYPE##_compare_quiet(X, Y, ST) == 0 +#define DO_FCMNE(TYPE, X, Y, ST) TYPE##_compare_quiet(X, Y, ST) != 0 +#define DO_FCMUO(TYPE, X, Y, ST) \ + TYPE##_compare_quiet(X, Y, ST) == float_relation_unordered +#define DO_FACGE(TYPE, X, Y, ST) \ + TYPE##_compare(TYPE##_abs(Y), TYPE##_abs(X), ST) <= 0 +#define DO_FACGT(TYPE, X, Y, ST) \ + TYPE##_compare(TYPE##_abs(Y), TYPE##_abs(X), ST) < 0 + +DO_FPCMP_PPZZ_ALL(sve_fcmge, DO_FCMGE) +DO_FPCMP_PPZZ_ALL(sve_fcmgt, DO_FCMGT) +DO_FPCMP_PPZZ_ALL(sve_fcmeq, DO_FCMEQ) +DO_FPCMP_PPZZ_ALL(sve_fcmne, DO_FCMNE) +DO_FPCMP_PPZZ_ALL(sve_fcmuo, DO_FCMUO) +DO_FPCMP_PPZZ_ALL(sve_facge, DO_FACGE) +DO_FPCMP_PPZZ_ALL(sve_facgt, DO_FACGT) + +#undef DO_FPCMP_PPZZ_ALL +#undef DO_FPCMP_PPZZ_D +#undef DO_FPCMP_PPZZ_S +#undef DO_FPCMP_PPZZ_H +#undef DO_FPCMP_PPZZ + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FP3(FMULX, fmulx) #undef DO_FP3 +static bool do_fp_cmp(DisasContext *s, arg_rprr_esz *a, + gen_helper_gvec_4_ptr *fn) +{ + if (fn == NULL) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_4_ptr(pred_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + pred_full_reg_offset(s, a->pg), + status, vsz, vsz, 0, fn); + tcg_temp_free_ptr(status); + } + return true; +} + +#define DO_FPCMP(NAME, name) \ +static bool trans_##NAME##_ppzz(DisasContext *s, arg_rprr_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_gvec_4_ptr * const fns[4] = { \ + NULL, gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, gen_helper_sve_##name##_d \ + }; \ + return do_fp_cmp(s, a, fns[a->esz]); \ +} + +DO_FPCMP(FCMGE, fcmge) +DO_FPCMP(FCMGT, fcmgt) +DO_FPCMP(FCMEQ, fcmeq) +DO_FPCMP(FCMNE, fcmne) +DO_FPCMP(FCMUO, fcmuo) +DO_FPCMP(FACGE, facge) +DO_FPCMP(FACGT, facgt) + +#undef DO_FPCMP + typedef void gen_helper_sve_fmla(TCGv_env, TCGv_ptr, TCGv_i32); static bool do_fmla(DisasContext *s, arg_rprrr_esz *a, gen_helper_sve_fmla *fn) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ UXTH 00000100 .. 010 011 101 ... ..... ..... @rd_pg_rn SXTW 00000100 .. 010 100 101 ... ..... ..... @rd_pg_rn UXTW 00000100 .. 010 101 101 ... ..... ..... @rd_pg_rn +### SVE Floating Point Compare - Vectors Group + +# SVE floating-point compare vectors +FCMGE_ppzz 01100101 .. 0 ..... 010 ... ..... 0 .... @pd_pg_rn_rm +FCMGT_ppzz 01100101 .. 0 ..... 010 ... ..... 1 .... @pd_pg_rn_rm +FCMEQ_ppzz 01100101 .. 0 ..... 011 ... ..... 0 .... @pd_pg_rn_rm +FCMNE_ppzz 01100101 .. 0 ..... 011 ... ..... 1 .... @pd_pg_rn_rm +FCMUO_ppzz 01100101 .. 0 ..... 110 ... ..... 0 .... @pd_pg_rn_rm +FACGE_ppzz 01100101 .. 0 ..... 110 ... ..... 1 .... @pd_pg_rn_rm +FACGT_ppzz 01100101 .. 0 ..... 111 ... ..... 1 .... @pd_pg_rn_rm + ### SVE Integer Multiply-Add Group # SVE integer multiply-add writing addend (predicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-18-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 56 ++++++++++++++++++++++++++++ target/arm/sve_helper.c | 69 +++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 75 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 14 +++++++ 4 files changed, 214 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(sve_fmulx_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_6(sve_fmulx_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadds_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadds_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fadds_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fsubs_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsubs_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsubs_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmuls_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmuls_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmuls_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fsubrs_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsubrs_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fsubrs_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmaxnms_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxnms_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxnms_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fminnms_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fminnms_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fminnms_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmaxs_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxs_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmaxs_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + +DEF_HELPER_FLAGS_6(sve_fmins_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmins_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fmins_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, i64, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_ZPZZ_FP(sve_fmulx_d, uint64_t, , helper_vfp_mulxd) #undef DO_ZPZZ_FP +/* Three-operand expander, with one scalar operand, controlled by + * a predicate, with the extra float_status parameter. + */ +#define DO_ZPZS_FP(NAME, TYPE, H, OP) \ +void HELPER(NAME)(void *vd, void *vn, void *vg, uint64_t scalar, \ + void *status, uint32_t desc) \ +{ \ + intptr_t i = simd_oprsz(desc); \ + uint64_t *g = vg; \ + TYPE mm = scalar; \ + do { \ + uint64_t pg = g[(i - 1) >> 6]; \ + do { \ + i -= sizeof(TYPE); \ + if (likely((pg >> (i & 63)) & 1)) { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + *(TYPE *)(vd + H(i)) = OP(nn, mm, status); \ + } \ + } while (i & 63); \ + } while (i != 0); \ +} + +DO_ZPZS_FP(sve_fadds_h, float16, H1_2, float16_add) +DO_ZPZS_FP(sve_fadds_s, float32, H1_4, float32_add) +DO_ZPZS_FP(sve_fadds_d, float64, , float64_add) + +DO_ZPZS_FP(sve_fsubs_h, float16, H1_2, float16_sub) +DO_ZPZS_FP(sve_fsubs_s, float32, H1_4, float32_sub) +DO_ZPZS_FP(sve_fsubs_d, float64, , float64_sub) + +DO_ZPZS_FP(sve_fmuls_h, float16, H1_2, float16_mul) +DO_ZPZS_FP(sve_fmuls_s, float32, H1_4, float32_mul) +DO_ZPZS_FP(sve_fmuls_d, float64, , float64_mul) + +static inline float16 subr_h(float16 a, float16 b, float_status *s) +{ + return float16_sub(b, a, s); +} + +static inline float32 subr_s(float32 a, float32 b, float_status *s) +{ + return float32_sub(b, a, s); +} + +static inline float64 subr_d(float64 a, float64 b, float_status *s) +{ + return float64_sub(b, a, s); +} + +DO_ZPZS_FP(sve_fsubrs_h, float16, H1_2, subr_h) +DO_ZPZS_FP(sve_fsubrs_s, float32, H1_4, subr_s) +DO_ZPZS_FP(sve_fsubrs_d, float64, , subr_d) + +DO_ZPZS_FP(sve_fmaxnms_h, float16, H1_2, float16_maxnum) +DO_ZPZS_FP(sve_fmaxnms_s, float32, H1_4, float32_maxnum) +DO_ZPZS_FP(sve_fmaxnms_d, float64, , float64_maxnum) + +DO_ZPZS_FP(sve_fminnms_h, float16, H1_2, float16_minnum) +DO_ZPZS_FP(sve_fminnms_s, float32, H1_4, float32_minnum) +DO_ZPZS_FP(sve_fminnms_d, float64, , float64_minnum) + +DO_ZPZS_FP(sve_fmaxs_h, float16, H1_2, float16_max) +DO_ZPZS_FP(sve_fmaxs_s, float32, H1_4, float32_max) +DO_ZPZS_FP(sve_fmaxs_d, float64, , float64_max) + +DO_ZPZS_FP(sve_fmins_h, float16, H1_2, float16_min) +DO_ZPZS_FP(sve_fmins_s, float32, H1_4, float32_min) +DO_ZPZS_FP(sve_fmins_d, float64, , float64_min) + /* Fully general two-operand expander, controlled by a predicate, * With the extra float_status parameter. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ #include "exec/log.h" #include "trace-tcg.h" #include "translate-a64.h" +#include "fpu/softfloat.h" typedef void GVecGen2sFn(unsigned, uint32_t, uint32_t, @@ -XXX,XX +XXX,XX @@ DO_FP3(FMULX, fmulx) #undef DO_FP3 +typedef void gen_helper_sve_fp2scalar(TCGv_ptr, TCGv_ptr, TCGv_ptr, + TCGv_i64, TCGv_ptr, TCGv_i32); + +static void do_fp_scalar(DisasContext *s, int zd, int zn, int pg, bool is_fp16, + TCGv_i64 scalar, gen_helper_sve_fp2scalar *fn) +{ + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr t_zd, t_zn, t_pg, status; + TCGv_i32 desc; + + t_zd = tcg_temp_new_ptr(); + t_zn = tcg_temp_new_ptr(); + t_pg = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, zd)); + tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, zn)); + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg)); + + status = get_fpstatus_ptr(is_fp16); + desc = tcg_const_i32(simd_desc(vsz, vsz, 0)); + fn(t_zd, t_zn, t_pg, scalar, status, desc); + + tcg_temp_free_i32(desc); + tcg_temp_free_ptr(status); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_ptr(t_zn); + tcg_temp_free_ptr(t_zd); +} + +static void do_fp_imm(DisasContext *s, arg_rpri_esz *a, uint64_t imm, + gen_helper_sve_fp2scalar *fn) +{ + TCGv_i64 temp = tcg_const_i64(imm); + do_fp_scalar(s, a->rd, a->rn, a->pg, a->esz == MO_16, temp, fn); + tcg_temp_free_i64(temp); +} + +#define DO_FP_IMM(NAME, name, const0, const1) \ +static bool trans_##NAME##_zpzi(DisasContext *s, arg_rpri_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_sve_fp2scalar * const fns[3] = { \ + gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, \ + gen_helper_sve_##name##_d \ + }; \ + static uint64_t const val[3][2] = { \ + { float16_##const0, float16_##const1 }, \ + { float32_##const0, float32_##const1 }, \ + { float64_##const0, float64_##const1 }, \ + }; \ + if (a->esz == 0) { \ + return false; \ + } \ + if (sve_access_check(s)) { \ + do_fp_imm(s, a, val[a->esz - 1][a->imm], fns[a->esz - 1]); \ + } \ + return true; \ +} + +#define float16_two make_float16(0x4000) +#define float32_two make_float32(0x40000000) +#define float64_two make_float64(0x4000000000000000ULL) + +DO_FP_IMM(FADD, fadds, half, one) +DO_FP_IMM(FSUB, fsubs, half, one) +DO_FP_IMM(FMUL, fmuls, half, two) +DO_FP_IMM(FSUBR, fsubrs, half, one) +DO_FP_IMM(FMAXNM, fmaxnms, zero, one) +DO_FP_IMM(FMINNM, fminnms, zero, one) +DO_FP_IMM(FMAX, fmaxs, zero, one) +DO_FP_IMM(FMIN, fmins, zero, one) + +#undef DO_FP_IMM + static bool do_fp_cmp(DisasContext *s, arg_rprr_esz *a, gen_helper_gvec_4_ptr *fn) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ @rdn_pg4 ........ esz:2 .. pg:4 ... ........ rd:5 \ &rpri_esz rn=%reg_movprfx +# Two register operand, one one-bit floating-point operand. +@rdn_i1 ........ esz:2 ......... pg:3 .... imm:1 rd:5 \ + &rpri_esz rn=%reg_movprfx + # Two register operand, one encoded bitmask. @rdn_dbm ........ .. .... dbm:13 rd:5 \ &rr_dbm rn=%reg_movprfx @@ -XXX,XX +XXX,XX @@ FMULX 01100101 .. 00 1010 100 ... ..... ..... @rdn_pg_rm FDIV 01100101 .. 00 1100 100 ... ..... ..... @rdm_pg_rn # FDIVR FDIV 01100101 .. 00 1101 100 ... ..... ..... @rdn_pg_rm +# SVE floating-point arithmetic with immediate (predicated) +FADD_zpzi 01100101 .. 011 000 100 ... 0000 . ..... @rdn_i1 +FSUB_zpzi 01100101 .. 011 001 100 ... 0000 . ..... @rdn_i1 +FMUL_zpzi 01100101 .. 011 010 100 ... 0000 . ..... @rdn_i1 +FSUBR_zpzi 01100101 .. 011 011 100 ... 0000 . ..... @rdn_i1 +FMAXNM_zpzi 01100101 .. 011 100 100 ... 0000 . ..... @rdn_i1 +FMINNM_zpzi 01100101 .. 011 101 100 ... 0000 . ..... @rdn_i1 +FMAX_zpzi 01100101 .. 011 110 100 ... 0000 . ..... @rdn_i1 +FMIN_zpzi 01100101 .. 011 111 100 ... 0000 . ..... @rdn_i1 + ### SVE FP Multiply-Add Group # SVE floating-point multiply-accumulate writing addend -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-19-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 14 +++++++++++ target/arm/translate-sve.c | 50 ++++++++++++++++++++++++++++++++++++++ target/arm/vec_helper.c | 48 ++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 19 +++++++++++++++ 4 files changed, 131 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_ftsmul_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_ftsmul_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fmul_idx_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fmul_idx_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fmul_idx_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_6(gvec_fmla_idx_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(gvec_fmla_idx_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(gvec_fmla_idx_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + #ifdef TARGET_AARCH64 #include "helper-a64.h" #include "helper-sve.h" diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_ZZI(UMIN, umin) #undef DO_ZZI +/* + *** SVE Floating Point Multiply-Add Indexed Group + */ + +static bool trans_FMLA_zzxz(DisasContext *s, arg_FMLA_zzxz *a, uint32_t insn) +{ + static gen_helper_gvec_4_ptr * const fns[3] = { + gen_helper_gvec_fmla_idx_h, + gen_helper_gvec_fmla_idx_s, + gen_helper_gvec_fmla_idx_d, + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_4_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + vec_full_reg_offset(s, a->ra), + status, vsz, vsz, (a->index << 1) | a->sub, + fns[a->esz - 1]); + tcg_temp_free_ptr(status); + } + return true; +} + +/* + *** SVE Floating Point Multiply Indexed Group + */ + +static bool trans_FMUL_zzx(DisasContext *s, arg_FMUL_zzx *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[3] = { + gen_helper_gvec_fmul_idx_h, + gen_helper_gvec_fmul_idx_s, + gen_helper_gvec_fmul_idx_d, + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + status, vsz, vsz, a->index, fns[a->esz - 1]); + tcg_temp_free_ptr(status); + } + return true; +} + /* *** SVE Floating Point Accumulating Reduction Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ DO_3OP(gvec_rsqrts_d, helper_rsqrtsf_f64, float64) #endif #undef DO_3OP + +/* For the indexed ops, SVE applies the index per 128-bit vector segment. + * For AdvSIMD, there is of course only one such vector segment. + */ + +#define DO_MUL_IDX(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, void *stat, uint32_t desc) \ +{ \ + intptr_t i, j, oprsz = simd_oprsz(desc), segment = 16 / sizeof(TYPE); \ + intptr_t idx = simd_data(desc); \ + TYPE *d = vd, *n = vn, *m = vm; \ + for (i = 0; i < oprsz / sizeof(TYPE); i += segment) { \ + TYPE mm = m[H(i + idx)]; \ + for (j = 0; j < segment; j++) { \ + d[i + j] = TYPE##_mul(n[i + j], mm, stat); \ + } \ + } \ +} + +DO_MUL_IDX(gvec_fmul_idx_h, float16, H2) +DO_MUL_IDX(gvec_fmul_idx_s, float32, H4) +DO_MUL_IDX(gvec_fmul_idx_d, float64, ) + +#undef DO_MUL_IDX + +#define DO_FMLA_IDX(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, void *va, \ + void *stat, uint32_t desc) \ +{ \ + intptr_t i, j, oprsz = simd_oprsz(desc), segment = 16 / sizeof(TYPE); \ + TYPE op1_neg = extract32(desc, SIMD_DATA_SHIFT, 1); \ + intptr_t idx = desc >> (SIMD_DATA_SHIFT + 1); \ + TYPE *d = vd, *n = vn, *m = vm, *a = va; \ + op1_neg <<= (8 * sizeof(TYPE) - 1); \ + for (i = 0; i < oprsz / sizeof(TYPE); i += segment) { \ + TYPE mm = m[H(i + idx)]; \ + for (j = 0; j < segment; j++) { \ + d[i + j] = TYPE##_muladd(n[i + j] ^ op1_neg, \ + mm, a[i + j], 0, stat); \ + } \ + } \ +} + +DO_FMLA_IDX(gvec_fmla_idx_h, float16, H2) +DO_FMLA_IDX(gvec_fmla_idx_s, float32, H4) +DO_FMLA_IDX(gvec_fmla_idx_d, float64, ) + +#undef DO_FMLA_IDX diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ %imm9_16_10 16:s6 10:3 %size_23 23:2 %dtype_23_13 23:2 13:2 +%index3_22_19 22:1 19:2 # A combination of tsz:imm3 -- extract esize. %tszimm_esz 22:2 5:5 !function=tszimm_esz @@ -XXX,XX +XXX,XX @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u # SVE integer multiply immediate (unpredicated) MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s +### SVE FP Multiply-Add Indexed Group + +# SVE floating-point multiply-add (indexed) +FMLA_zzxz 01100100 0.1 .. rm:3 00000 sub:1 rn:5 rd:5 \ + ra=%reg_movprfx index=%index3_22_19 esz=1 +FMLA_zzxz 01100100 101 index:2 rm:3 00000 sub:1 rn:5 rd:5 \ + ra=%reg_movprfx esz=2 +FMLA_zzxz 01100100 111 index:1 rm:4 00000 sub:1 rn:5 rd:5 \ + ra=%reg_movprfx esz=3 + +### SVE FP Multiply Indexed Group + +# SVE floating-point multiply (indexed) +FMUL_zzx 01100100 0.1 .. rm:3 001000 rn:5 rd:5 \ + index=%index3_22_19 esz=1 +FMUL_zzx 01100100 101 index:2 rm:3 001000 rn:5 rd:5 esz=2 +FMUL_zzx 01100100 111 index:1 rm:4 001000 rn:5 rd:5 esz=3 + ### SVE FP Accumulating Reduction Group # SVE floating-point serial reduction (predicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-20-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 35 ++++++++++++++++++++++ target/arm/sve_helper.c | 61 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 57 +++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 8 +++++ 4 files changed, 161 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_rsqrts_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_rsqrts_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_faddv_h, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_faddv_s, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_faddv_d, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_fmaxnmv_h, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fmaxnmv_s, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fmaxnmv_d, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_fminnmv_h, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fminnmv_s, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fminnmv_d, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_fmaxv_h, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fmaxv_s, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fmaxv_d, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_fminv_h, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fminv_s, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_fminv_d, TCG_CALL_NO_RWG, + i64, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_fadda_h, TCG_CALL_NO_RWG, i64, i64, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_fadda_s, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) return predtest_ones(d, oprsz, esz_mask); } +/* Recursive reduction on a function; + * C.f. the ARM ARM function ReducePredicated. + * + * While it would be possible to write this without the DATA temporary, + * it is much simpler to process the predicate register this way. + * The recursion is bounded to depth 7 (128 fp16 elements), so there's + * little to gain with a more complex non-recursive form. + */ +#define DO_REDUCE(NAME, TYPE, H, FUNC, IDENT) \ +static TYPE NAME##_reduce(TYPE *data, float_status *status, uintptr_t n) \ +{ \ + if (n == 1) { \ + return *data; \ + } else { \ + uintptr_t half = n / 2; \ + TYPE lo = NAME##_reduce(data, status, half); \ + TYPE hi = NAME##_reduce(data + half, status, half); \ + return TYPE##_##FUNC(lo, hi, status); \ + } \ +} \ +uint64_t HELPER(NAME)(void *vn, void *vg, void *vs, uint32_t desc) \ +{ \ + uintptr_t i, oprsz = simd_oprsz(desc), maxsz = simd_maxsz(desc); \ + TYPE data[sizeof(ARMVectorReg) / sizeof(TYPE)]; \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + *(TYPE *)((void *)data + i) = (pg & 1 ? nn : IDENT); \ + i += sizeof(TYPE), pg >>= sizeof(TYPE); \ + } while (i & 15); \ + } \ + for (; i < maxsz; i += sizeof(TYPE)) { \ + *(TYPE *)((void *)data + i) = IDENT; \ + } \ + return NAME##_reduce(data, vs, maxsz / sizeof(TYPE)); \ +} + +DO_REDUCE(sve_faddv_h, float16, H1_2, add, float16_zero) +DO_REDUCE(sve_faddv_s, float32, H1_4, add, float32_zero) +DO_REDUCE(sve_faddv_d, float64, , add, float64_zero) + +/* Identity is floatN_default_nan, without the function call. */ +DO_REDUCE(sve_fminnmv_h, float16, H1_2, minnum, 0x7E00) +DO_REDUCE(sve_fminnmv_s, float32, H1_4, minnum, 0x7FC00000) +DO_REDUCE(sve_fminnmv_d, float64, , minnum, 0x7FF8000000000000ULL) + +DO_REDUCE(sve_fmaxnmv_h, float16, H1_2, maxnum, 0x7E00) +DO_REDUCE(sve_fmaxnmv_s, float32, H1_4, maxnum, 0x7FC00000) +DO_REDUCE(sve_fmaxnmv_d, float64, , maxnum, 0x7FF8000000000000ULL) + +DO_REDUCE(sve_fminv_h, float16, H1_2, min, float16_infinity) +DO_REDUCE(sve_fminv_s, float32, H1_4, min, float32_infinity) +DO_REDUCE(sve_fminv_d, float64, , min, float64_infinity) + +DO_REDUCE(sve_fmaxv_h, float16, H1_2, max, float16_chs(float16_infinity)) +DO_REDUCE(sve_fmaxv_s, float32, H1_4, max, float32_chs(float32_infinity)) +DO_REDUCE(sve_fmaxv_d, float64, , max, float64_chs(float64_infinity)) + +#undef DO_REDUCE + uint64_t HELPER(sve_fadda_h)(uint64_t nn, void *vm, void *vg, void *status, uint32_t desc) { diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FMUL_zzx(DisasContext *s, arg_FMUL_zzx *a, uint32_t insn) return true; } +/* + *** SVE Floating Point Fast Reduction Group + */ + +typedef void gen_helper_fp_reduce(TCGv_i64, TCGv_ptr, TCGv_ptr, + TCGv_ptr, TCGv_i32); + +static void do_reduce(DisasContext *s, arg_rpr_esz *a, + gen_helper_fp_reduce *fn) +{ + unsigned vsz = vec_full_reg_size(s); + unsigned p2vsz = pow2ceil(vsz); + TCGv_i32 t_desc = tcg_const_i32(simd_desc(vsz, p2vsz, 0)); + TCGv_ptr t_zn, t_pg, status; + TCGv_i64 temp; + + temp = tcg_temp_new_i64(); + t_zn = tcg_temp_new_ptr(); + t_pg = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, a->pg)); + status = get_fpstatus_ptr(a->esz == MO_16); + + fn(temp, t_zn, t_pg, status, t_desc); + tcg_temp_free_ptr(t_zn); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_ptr(status); + tcg_temp_free_i32(t_desc); + + write_fp_dreg(s, a->rd, temp); + tcg_temp_free_i64(temp); +} + +#define DO_VPZ(NAME, name) \ +static bool trans_##NAME(DisasContext *s, arg_rpr_esz *a, uint32_t insn) \ +{ \ + static gen_helper_fp_reduce * const fns[3] = { \ + gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, \ + gen_helper_sve_##name##_d, \ + }; \ + if (a->esz == 0) { \ + return false; \ + } \ + if (sve_access_check(s)) { \ + do_reduce(s, a, fns[a->esz - 1]); \ + } \ + return true; \ +} + +DO_VPZ(FADDV, faddv) +DO_VPZ(FMINNMV, fminnmv) +DO_VPZ(FMAXNMV, fmaxnmv) +DO_VPZ(FMINV, fminv) +DO_VPZ(FMAXV, fmaxv) + /* *** SVE Floating Point Accumulating Reduction Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FMUL_zzx 01100100 0.1 .. rm:3 001000 rn:5 rd:5 \ FMUL_zzx 01100100 101 index:2 rm:3 001000 rn:5 rd:5 esz=2 FMUL_zzx 01100100 111 index:1 rm:4 001000 rn:5 rd:5 esz=3 +### SVE FP Fast Reduction Group + +FADDV 01100101 .. 000 000 001 ... ..... ..... @rd_pg_rn +FMAXNMV 01100101 .. 000 100 001 ... ..... ..... @rd_pg_rn +FMINNMV 01100101 .. 000 101 001 ... ..... ..... @rd_pg_rn +FMAXV 01100101 .. 000 110 001 ... ..... ..... @rd_pg_rn +FMINV 01100101 .. 000 111 001 ... ..... ..... @rd_pg_rn + ### SVE FP Accumulating Reduction Group # SVE floating-point serial reduction (predicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-21-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 8 +++++++ target/arm/translate-sve.c | 47 ++++++++++++++++++++++++++++++++++++++ target/arm/vec_helper.c | 20 ++++++++++++++++ target/arm/sve.decode | 5 ++++ 4 files changed, 80 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_fcmlas_idx, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_fcmlad, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_frecpe_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_frecpe_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_frecpe_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(gvec_frsqrte_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_frsqrte_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_frsqrte_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(gvec_fadd_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_fadd_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_fadd_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_VPZ(FMAXNMV, fmaxnmv) DO_VPZ(FMINV, fminv) DO_VPZ(FMAXV, fmaxv) +/* + *** SVE Floating Point Unary Operations - Unpredicated Group + */ + +static void do_zz_fp(DisasContext *s, arg_rr_esz *a, gen_helper_gvec_2_ptr *fn) +{ + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + + tcg_gen_gvec_2_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + status, vsz, vsz, 0, fn); + tcg_temp_free_ptr(status); +} + +static bool trans_FRECPE(DisasContext *s, arg_rr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_2_ptr * const fns[3] = { + gen_helper_gvec_frecpe_h, + gen_helper_gvec_frecpe_s, + gen_helper_gvec_frecpe_d, + }; + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + do_zz_fp(s, a, fns[a->esz - 1]); + } + return true; +} + +static bool trans_FRSQRTE(DisasContext *s, arg_rr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_2_ptr * const fns[3] = { + gen_helper_gvec_frsqrte_h, + gen_helper_gvec_frsqrte_s, + gen_helper_gvec_frsqrte_d, + }; + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + do_zz_fp(s, a, fns[a->esz - 1]); + } + return true; +} + /* *** SVE Floating Point Accumulating Reduction Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_fcmlad)(void *vd, void *vn, void *vm, clear_tail(d, opr_sz, simd_maxsz(desc)); } +#define DO_2OP(NAME, FUNC, TYPE) \ +void HELPER(NAME)(void *vd, void *vn, void *stat, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + TYPE *d = vd, *n = vn; \ + for (i = 0; i < oprsz / sizeof(TYPE); i++) { \ + d[i] = FUNC(n[i], stat); \ + } \ +} + +DO_2OP(gvec_frecpe_h, helper_recpe_f16, float16) +DO_2OP(gvec_frecpe_s, helper_recpe_f32, float32) +DO_2OP(gvec_frecpe_d, helper_recpe_f64, float64) + +DO_2OP(gvec_frsqrte_h, helper_rsqrte_f16, float16) +DO_2OP(gvec_frsqrte_s, helper_rsqrte_f32, float32) +DO_2OP(gvec_frsqrte_d, helper_rsqrte_f64, float64) + +#undef DO_2OP + /* Floating-point trigonometric starting value. * See the ARM ARM pseudocode function FPTrigSMul. */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FMINNMV 01100101 .. 000 101 001 ... ..... ..... @rd_pg_rn FMAXV 01100101 .. 000 110 001 ... ..... ..... @rd_pg_rn FMINV 01100101 .. 000 111 001 ... ..... ..... @rd_pg_rn +## SVE Floating Point Unary Operations - Unpredicated Group + +FRECPE 01100101 .. 001 110 001100 ..... ..... @rd_rn +FRSQRTE 01100101 .. 001 111 001100 ..... ..... @rd_rn + ### SVE FP Accumulating Reduction Group # SVE floating-point serial reduction (predicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-22-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 42 +++++++++++++++++++++++++++++++++++++ target/arm/sve_helper.c | 43 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 43 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 10 +++++++++ 4 files changed, 138 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_fadda_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_fadda_d, TCG_CALL_NO_RWG, i64, i64, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmge0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmge0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmge0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcmgt0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmgt0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmgt0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcmlt0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmlt0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmlt0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcmle0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmle0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmle0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcmeq0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmeq0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmeq0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcmne0_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmne0_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcmne0_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_6(sve_fadd_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_6(sve_fadd_s, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, \ #define DO_FCMGE(TYPE, X, Y, ST) TYPE##_compare(Y, X, ST) <= 0 #define DO_FCMGT(TYPE, X, Y, ST) TYPE##_compare(Y, X, ST) < 0 +#define DO_FCMLE(TYPE, X, Y, ST) TYPE##_compare(X, Y, ST) <= 0 +#define DO_FCMLT(TYPE, X, Y, ST) TYPE##_compare(X, Y, ST) < 0 #define DO_FCMEQ(TYPE, X, Y, ST) TYPE##_compare_quiet(X, Y, ST) == 0 #define DO_FCMNE(TYPE, X, Y, ST) TYPE##_compare_quiet(X, Y, ST) != 0 #define DO_FCMUO(TYPE, X, Y, ST) \ @@ -XXX,XX +XXX,XX @@ DO_FPCMP_PPZZ_ALL(sve_facgt, DO_FACGT) #undef DO_FPCMP_PPZZ_H #undef DO_FPCMP_PPZZ +/* One operand floating-point comparison against zero, controlled + * by a predicate. + */ +#define DO_FPCMP_PPZ0(NAME, TYPE, H, OP) \ +void HELPER(NAME)(void *vd, void *vn, void *vg, \ + void *status, uint32_t desc) \ +{ \ + intptr_t i = simd_oprsz(desc), j = (i - 1) >> 6; \ + uint64_t *d = vd, *g = vg; \ + do { \ + uint64_t out = 0, pg = g[j]; \ + do { \ + i -= sizeof(TYPE), out <<= sizeof(TYPE); \ + if ((pg >> (i & 63)) & 1) { \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + out |= OP(TYPE, nn, 0, status); \ + } \ + } while (i & 63); \ + d[j--] = out; \ + } while (i > 0); \ +} + +#define DO_FPCMP_PPZ0_H(NAME, OP) \ + DO_FPCMP_PPZ0(NAME##_h, float16, H1_2, OP) +#define DO_FPCMP_PPZ0_S(NAME, OP) \ + DO_FPCMP_PPZ0(NAME##_s, float32, H1_4, OP) +#define DO_FPCMP_PPZ0_D(NAME, OP) \ + DO_FPCMP_PPZ0(NAME##_d, float64, , OP) + +#define DO_FPCMP_PPZ0_ALL(NAME, OP) \ + DO_FPCMP_PPZ0_H(NAME, OP) \ + DO_FPCMP_PPZ0_S(NAME, OP) \ + DO_FPCMP_PPZ0_D(NAME, OP) + +DO_FPCMP_PPZ0_ALL(sve_fcmge0, DO_FCMGE) +DO_FPCMP_PPZ0_ALL(sve_fcmgt0, DO_FCMGT) +DO_FPCMP_PPZ0_ALL(sve_fcmle0, DO_FCMLE) +DO_FPCMP_PPZ0_ALL(sve_fcmlt0, DO_FCMLT) +DO_FPCMP_PPZ0_ALL(sve_fcmeq0, DO_FCMEQ) +DO_FPCMP_PPZ0_ALL(sve_fcmne0, DO_FCMNE) + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FRSQRTE(DisasContext *s, arg_rr_esz *a, uint32_t insn) return true; } +/* + *** SVE Floating Point Compare with Zero Group + */ + +static void do_ppz_fp(DisasContext *s, arg_rpr_esz *a, + gen_helper_gvec_3_ptr *fn) +{ + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + + tcg_gen_gvec_3_ptr(pred_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + pred_full_reg_offset(s, a->pg), + status, vsz, vsz, 0, fn); + tcg_temp_free_ptr(status); +} + +#define DO_PPZ(NAME, name) \ +static bool trans_##NAME(DisasContext *s, arg_rpr_esz *a, uint32_t insn) \ +{ \ + static gen_helper_gvec_3_ptr * const fns[3] = { \ + gen_helper_sve_##name##_h, \ + gen_helper_sve_##name##_s, \ + gen_helper_sve_##name##_d, \ + }; \ + if (a->esz == 0) { \ + return false; \ + } \ + if (sve_access_check(s)) { \ + do_ppz_fp(s, a, fns[a->esz - 1]); \ + } \ + return true; \ +} + +DO_PPZ(FCMGE_ppz0, fcmge0) +DO_PPZ(FCMGT_ppz0, fcmgt0) +DO_PPZ(FCMLE_ppz0, fcmle0) +DO_PPZ(FCMLT_ppz0, fcmlt0) +DO_PPZ(FCMEQ_ppz0, fcmeq0) +DO_PPZ(FCMNE_ppz0, fcmne0) + +#undef DO_PPZ + /* *** SVE Floating Point Accumulating Reduction Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ # One register operand, with governing predicate, vector element size @rd_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 &rpr_esz @rd_pg4_pn ........ esz:2 ... ... .. pg:4 . rn:4 rd:5 &rpr_esz +@pd_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 . rd:4 &rpr_esz # One register operand, with governing predicate, no vector element size @rd_pg_rn_e0 ........ .. ... ... ... pg:3 rn:5 rd:5 &rpr_esz esz=0 @@ -XXX,XX +XXX,XX @@ FMINV 01100101 .. 000 111 001 ... ..... ..... @rd_pg_rn FRECPE 01100101 .. 001 110 001100 ..... ..... @rd_rn FRSQRTE 01100101 .. 001 111 001100 ..... ..... @rd_rn +### SVE FP Compare with Zero Group + +FCMGE_ppz0 01100101 .. 0100 00 001 ... ..... 0 .... @pd_pg_rn +FCMGT_ppz0 01100101 .. 0100 00 001 ... ..... 1 .... @pd_pg_rn +FCMLT_ppz0 01100101 .. 0100 01 001 ... ..... 0 .... @pd_pg_rn +FCMLE_ppz0 01100101 .. 0100 01 001 ... ..... 1 .... @pd_pg_rn +FCMEQ_ppz0 01100101 .. 0100 10 001 ... ..... 0 .... @pd_pg_rn +FCMNE_ppz0 01100101 .. 0100 11 001 ... ..... 0 .... @pd_pg_rn + ### SVE FP Accumulating Reduction Group # SVE floating-point serial reduction (predicated) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-23-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 4 +++ target/arm/sve_helper.c | 70 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 27 +++++++++++++++ target/arm/sve.decode | 3 ++ 4 files changed, 104 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ftmad_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ftmad_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_ftmad_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(sve_ld1bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld2bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) DEF_HELPER_FLAGS_4(sve_ld3bb_r, TCG_CALL_NO_WG, void, env, ptr, tl, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_FPCMP_PPZ0_ALL(sve_fcmlt0, DO_FCMLT) DO_FPCMP_PPZ0_ALL(sve_fcmeq0, DO_FCMEQ) DO_FPCMP_PPZ0_ALL(sve_fcmne0, DO_FCMNE) +/* FP Trig Multiply-Add. */ + +void HELPER(sve_ftmad_h)(void *vd, void *vn, void *vm, void *vs, uint32_t desc) +{ + static const float16 coeff[16] = { + 0x3c00, 0xb155, 0x2030, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x3c00, 0xb800, 0x293a, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + }; + intptr_t i, opr_sz = simd_oprsz(desc) / sizeof(float16); + intptr_t x = simd_data(desc); + float16 *d = vd, *n = vn, *m = vm; + for (i = 0; i < opr_sz; i++) { + float16 mm = m[i]; + intptr_t xx = x; + if (float16_is_neg(mm)) { + mm = float16_abs(mm); + xx += 8; + } + d[i] = float16_muladd(n[i], mm, coeff[xx], 0, vs); + } +} + +void HELPER(sve_ftmad_s)(void *vd, void *vn, void *vm, void *vs, uint32_t desc) +{ + static const float32 coeff[16] = { + 0x3f800000, 0xbe2aaaab, 0x3c088886, 0xb95008b9, + 0x36369d6d, 0x00000000, 0x00000000, 0x00000000, + 0x3f800000, 0xbf000000, 0x3d2aaaa6, 0xbab60705, + 0x37cd37cc, 0x00000000, 0x00000000, 0x00000000, + }; + intptr_t i, opr_sz = simd_oprsz(desc) / sizeof(float32); + intptr_t x = simd_data(desc); + float32 *d = vd, *n = vn, *m = vm; + for (i = 0; i < opr_sz; i++) { + float32 mm = m[i]; + intptr_t xx = x; + if (float32_is_neg(mm)) { + mm = float32_abs(mm); + xx += 8; + } + d[i] = float32_muladd(n[i], mm, coeff[xx], 0, vs); + } +} + +void HELPER(sve_ftmad_d)(void *vd, void *vn, void *vm, void *vs, uint32_t desc) +{ + static const float64 coeff[16] = { + 0x3ff0000000000000ull, 0xbfc5555555555543ull, + 0x3f8111111110f30cull, 0xbf2a01a019b92fc6ull, + 0x3ec71de351f3d22bull, 0xbe5ae5e2b60f7b91ull, + 0x3de5d8408868552full, 0x0000000000000000ull, + 0x3ff0000000000000ull, 0xbfe0000000000000ull, + 0x3fa5555555555536ull, 0xbf56c16c16c13a0bull, + 0x3efa01a019b1e8d8ull, 0xbe927e4f7282f468ull, + 0x3e21ee96d2641b13ull, 0xbda8f76380fbb401ull, + }; + intptr_t i, opr_sz = simd_oprsz(desc) / sizeof(float64); + intptr_t x = simd_data(desc); + float64 *d = vd, *n = vn, *m = vm; + for (i = 0; i < opr_sz; i++) { + float64 mm = m[i]; + intptr_t xx = x; + if (float64_is_neg(mm)) { + mm = float64_abs(mm); + xx += 8; + } + d[i] = float64_muladd(n[i], mm, coeff[xx], 0, vs); + } +} + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_PPZ(FCMNE_ppz0, fcmne0) #undef DO_PPZ +/* + *** SVE floating-point trig multiply-add coefficient + */ + +static bool trans_FTMAD(DisasContext *s, arg_FTMAD *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[3] = { + gen_helper_sve_ftmad_h, + gen_helper_sve_ftmad_s, + gen_helper_sve_ftmad_d, + }; + + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + status, vsz, vsz, a->imm, fns[a->esz - 1]); + tcg_temp_free_ptr(status); + } + return true; +} + /* *** SVE Floating Point Accumulating Reduction Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FMINNM_zpzi 01100101 .. 011 101 100 ... 0000 . ..... @rdn_i1 FMAX_zpzi 01100101 .. 011 110 100 ... 0000 . ..... @rdn_i1 FMIN_zpzi 01100101 .. 011 111 100 ... 0000 . ..... @rdn_i1 +# SVE floating-point trig multiply-add coefficient +FTMAD 01100101 esz:2 010 imm:3 100000 rm:5 rd:5 rn=%reg_movprfx + ### SVE FP Multiply-Add Group # SVE floating-point multiply-accumulate writing addend -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-24-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 13 +++++++++ target/arm/sve_helper.c | 55 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 30 +++++++++++++++++++++ target/arm/sve.decode | 8 ++++++ 4 files changed, 106 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(sve_fmins_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_6(sve_fmins_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i64, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_sh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_dh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_hs, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_ds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_hd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvt_sd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(NAME)(void *vd, void *vn, void *vg, void *status, uint32_t desc) \ } while (i != 0); \ } +/* SVE fp16 conversions always use IEEE mode. Like AdvSIMD, they ignore + * FZ16. When converting from fp16, this affects flushing input denormals; + * when converting to fp16, this affects flushing output denormals. + */ +static inline float32 sve_f16_to_f32(float16 f, float_status *fpst) +{ + flag save = get_flush_inputs_to_zero(fpst); + float32 ret; + + set_flush_inputs_to_zero(false, fpst); + ret = float16_to_float32(f, true, fpst); + set_flush_inputs_to_zero(save, fpst); + return ret; +} + +static inline float64 sve_f16_to_f64(float16 f, float_status *fpst) +{ + flag save = get_flush_inputs_to_zero(fpst); + float64 ret; + + set_flush_inputs_to_zero(false, fpst); + ret = float16_to_float64(f, true, fpst); + set_flush_inputs_to_zero(save, fpst); + return ret; +} + +static inline float16 sve_f32_to_f16(float32 f, float_status *fpst) +{ + flag save = get_flush_to_zero(fpst); + float16 ret; + + set_flush_to_zero(false, fpst); + ret = float32_to_float16(f, true, fpst); + set_flush_to_zero(save, fpst); + return ret; +} + +static inline float16 sve_f64_to_f16(float64 f, float_status *fpst) +{ + flag save = get_flush_to_zero(fpst); + float16 ret; + + set_flush_to_zero(false, fpst); + ret = float64_to_float16(f, true, fpst); + set_flush_to_zero(save, fpst); + return ret; +} + +DO_ZPZ_FP(sve_fcvt_sh, uint32_t, H1_4, sve_f32_to_f16) +DO_ZPZ_FP(sve_fcvt_hs, uint32_t, H1_4, sve_f16_to_f32) +DO_ZPZ_FP(sve_fcvt_dh, uint64_t, , sve_f64_to_f16) +DO_ZPZ_FP(sve_fcvt_hd, uint64_t, , sve_f16_to_f64) +DO_ZPZ_FP(sve_fcvt_ds, uint64_t, , float64_to_float32) +DO_ZPZ_FP(sve_fcvt_sd, uint64_t, , float32_to_float64) + DO_ZPZ_FP(sve_scvt_hh, uint16_t, H1_2, int16_to_float16) DO_ZPZ_FP(sve_scvt_sh, uint32_t, H1_4, int32_to_float16) DO_ZPZ_FP(sve_scvt_ss, uint32_t, H1_4, int32_to_float32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_zpz_ptr(DisasContext *s, int rd, int rn, int pg, return true; } +static bool trans_FCVT_sh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvt_sh); +} + +static bool trans_FCVT_hs(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvt_hs); +} + +static bool trans_FCVT_dh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvt_dh); +} + +static bool trans_FCVT_hd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvt_hd); +} + +static bool trans_FCVT_ds(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvt_ds); +} + +static bool trans_FCVT_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvt_sd); +} + static bool trans_SCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) { return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_hh); diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FNMLS_zpzzz 01100101 .. 1 ..... 111 ... ..... ..... @rdn_pg_rm_ra ### SVE FP Unary Operations Predicated Group +# SVE floating-point convert precision +FCVT_sh 01100101 10 0010 00 101 ... ..... ..... @rd_pg_rn_e0 +FCVT_hs 01100101 10 0010 01 101 ... ..... ..... @rd_pg_rn_e0 +FCVT_dh 01100101 11 0010 00 101 ... ..... ..... @rd_pg_rn_e0 +FCVT_hd 01100101 11 0010 01 101 ... ..... ..... @rd_pg_rn_e0 +FCVT_ds 01100101 11 0010 10 101 ... ..... ..... @rd_pg_rn_e0 +FCVT_sd 01100101 11 0010 11 101 ... ..... ..... @rd_pg_rn_e0 + # SVE integer convert to floating-point SCVTF_hh 01100101 01 010 01 0 101 ... ..... ..... @rd_pg_rn_e0 SCVTF_sh 01100101 01 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-25-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 30 +++++++++++++ target/arm/helper.h | 12 +++--- target/arm/helper.c | 2 +- target/arm/sve_helper.c | 88 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 70 ++++++++++++++++++++++++++++++ target/arm/sve.decode | 16 +++++++ 6 files changed, 211 insertions(+), 7 deletions(-) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_fcvt_hd, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_fcvt_sd, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_hh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_hs, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_ss, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_ds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_hd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_sd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzs_dd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fcvtzu_hh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_hs, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_ss, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_ds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_hd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_sd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fcvtzu_dd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_2(vfp_touid, i32, f64, ptr) DEF_HELPER_2(vfp_touizh, i32, f16, ptr) DEF_HELPER_2(vfp_touizs, i32, f32, ptr) DEF_HELPER_2(vfp_touizd, i32, f64, ptr) -DEF_HELPER_2(vfp_tosih, i32, f16, ptr) -DEF_HELPER_2(vfp_tosis, i32, f32, ptr) -DEF_HELPER_2(vfp_tosid, i32, f64, ptr) -DEF_HELPER_2(vfp_tosizh, i32, f16, ptr) -DEF_HELPER_2(vfp_tosizs, i32, f32, ptr) -DEF_HELPER_2(vfp_tosizd, i32, f64, ptr) +DEF_HELPER_2(vfp_tosih, s32, f16, ptr) +DEF_HELPER_2(vfp_tosis, s32, f32, ptr) +DEF_HELPER_2(vfp_tosid, s32, f64, ptr) +DEF_HELPER_2(vfp_tosizh, s32, f16, ptr) +DEF_HELPER_2(vfp_tosizs, s32, f32, ptr) +DEF_HELPER_2(vfp_tosizd, s32, f64, ptr) DEF_HELPER_3(vfp_toshs_round_to_zero, i32, f32, i32, ptr) DEF_HELPER_3(vfp_tosls_round_to_zero, i32, f32, i32, ptr) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ ftype HELPER(name)(uint32_t x, void *fpstp) \ } #define CONV_FTOI(name, ftype, fsz, sign, round) \ -uint32_t HELPER(name)(ftype x, void *fpstp) \ +sign##int32_t HELPER(name)(ftype x, void *fpstp) \ { \ float_status *fpst = fpstp; \ if (float##fsz##_is_any_nan(x)) { \ diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ static inline float16 sve_f64_to_f16(float64 f, float_status *fpst) return ret; } +static inline int16_t vfp_float16_to_int16_rtz(float16 f, float_status *s) +{ + if (float16_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float16_to_int16_round_to_zero(f, s); +} + +static inline int64_t vfp_float16_to_int64_rtz(float16 f, float_status *s) +{ + if (float16_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float16_to_int64_round_to_zero(f, s); +} + +static inline int64_t vfp_float32_to_int64_rtz(float32 f, float_status *s) +{ + if (float32_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float32_to_int64_round_to_zero(f, s); +} + +static inline int64_t vfp_float64_to_int64_rtz(float64 f, float_status *s) +{ + if (float64_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float64_to_int64_round_to_zero(f, s); +} + +static inline uint16_t vfp_float16_to_uint16_rtz(float16 f, float_status *s) +{ + if (float16_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float16_to_uint16_round_to_zero(f, s); +} + +static inline uint64_t vfp_float16_to_uint64_rtz(float16 f, float_status *s) +{ + if (float16_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float16_to_uint64_round_to_zero(f, s); +} + +static inline uint64_t vfp_float32_to_uint64_rtz(float32 f, float_status *s) +{ + if (float32_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float32_to_uint64_round_to_zero(f, s); +} + +static inline uint64_t vfp_float64_to_uint64_rtz(float64 f, float_status *s) +{ + if (float64_is_any_nan(f)) { + float_raise(float_flag_invalid, s); + return 0; + } + return float64_to_uint64_round_to_zero(f, s); +} + DO_ZPZ_FP(sve_fcvt_sh, uint32_t, H1_4, sve_f32_to_f16) DO_ZPZ_FP(sve_fcvt_hs, uint32_t, H1_4, sve_f16_to_f32) DO_ZPZ_FP(sve_fcvt_dh, uint64_t, , sve_f64_to_f16) @@ -XXX,XX +XXX,XX @@ DO_ZPZ_FP(sve_fcvt_hd, uint64_t, , sve_f16_to_f64) DO_ZPZ_FP(sve_fcvt_ds, uint64_t, , float64_to_float32) DO_ZPZ_FP(sve_fcvt_sd, uint64_t, , float32_to_float64) +DO_ZPZ_FP(sve_fcvtzs_hh, uint16_t, H1_2, vfp_float16_to_int16_rtz) +DO_ZPZ_FP(sve_fcvtzs_hs, uint32_t, H1_4, helper_vfp_tosizh) +DO_ZPZ_FP(sve_fcvtzs_ss, uint32_t, H1_4, helper_vfp_tosizs) +DO_ZPZ_FP(sve_fcvtzs_hd, uint64_t, , vfp_float16_to_int64_rtz) +DO_ZPZ_FP(sve_fcvtzs_sd, uint64_t, , vfp_float32_to_int64_rtz) +DO_ZPZ_FP(sve_fcvtzs_ds, uint64_t, , helper_vfp_tosizd) +DO_ZPZ_FP(sve_fcvtzs_dd, uint64_t, , vfp_float64_to_int64_rtz) + +DO_ZPZ_FP(sve_fcvtzu_hh, uint16_t, H1_2, vfp_float16_to_uint16_rtz) +DO_ZPZ_FP(sve_fcvtzu_hs, uint32_t, H1_4, helper_vfp_touizh) +DO_ZPZ_FP(sve_fcvtzu_ss, uint32_t, H1_4, helper_vfp_touizs) +DO_ZPZ_FP(sve_fcvtzu_hd, uint64_t, , vfp_float16_to_uint64_rtz) +DO_ZPZ_FP(sve_fcvtzu_sd, uint64_t, , vfp_float32_to_uint64_rtz) +DO_ZPZ_FP(sve_fcvtzu_ds, uint64_t, , helper_vfp_touizd) +DO_ZPZ_FP(sve_fcvtzu_dd, uint64_t, , vfp_float64_to_uint64_rtz) + DO_ZPZ_FP(sve_scvt_hh, uint16_t, H1_2, int16_to_float16) DO_ZPZ_FP(sve_scvt_sh, uint32_t, H1_4, int32_to_float16) DO_ZPZ_FP(sve_scvt_ss, uint32_t, H1_4, int32_to_float32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FCVT_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvt_sd); } +static bool trans_FCVTZS_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzs_hh); +} + +static bool trans_FCVTZU_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzu_hh); +} + +static bool trans_FCVTZS_hs(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzs_hs); +} + +static bool trans_FCVTZU_hs(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzu_hs); +} + +static bool trans_FCVTZS_hd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzs_hd); +} + +static bool trans_FCVTZU_hd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_fcvtzu_hd); +} + +static bool trans_FCVTZS_ss(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzs_ss); +} + +static bool trans_FCVTZU_ss(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzu_ss); +} + +static bool trans_FCVTZS_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzs_sd); +} + +static bool trans_FCVTZU_sd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzu_sd); +} + +static bool trans_FCVTZS_ds(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzs_ds); +} + +static bool trans_FCVTZU_ds(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzu_ds); +} + +static bool trans_FCVTZS_dd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzs_dd); +} + +static bool trans_FCVTZU_dd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzu_dd); +} + static bool trans_SCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) { return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_hh); diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FCVT_hd 01100101 11 0010 01 101 ... ..... ..... @rd_pg_rn_e0 FCVT_ds 01100101 11 0010 10 101 ... ..... ..... @rd_pg_rn_e0 FCVT_sd 01100101 11 0010 11 101 ... ..... ..... @rd_pg_rn_e0 +# SVE floating-point convert to integer +FCVTZS_hh 01100101 01 011 01 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_hh 01100101 01 011 01 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_hs 01100101 01 011 10 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_hs 01100101 01 011 10 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_hd 01100101 01 011 11 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_hd 01100101 01 011 11 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_ss 01100101 10 011 10 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_ss 01100101 10 011 10 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_ds 01100101 11 011 00 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_ds 01100101 11 011 00 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_sd 01100101 11 011 10 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_sd 01100101 11 011 10 1 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZS_dd 01100101 11 011 11 0 101 ... ..... ..... @rd_pg_rn_e0 +FCVTZU_dd 01100101 11 011 11 1 101 ... ..... ..... @rd_pg_rn_e0 + # SVE integer convert to floating-point SCVTF_hh 01100101 01 010 01 0 101 ... ..... ..... @rd_pg_rn_e0 SCVTF_sh 01100101 01 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-26-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 14 +++++++ target/arm/sve_helper.c | 8 ++++ target/arm/translate-sve.c | 77 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 9 +++++ 4 files changed, 108 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_fcvtzu_sd, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_fcvtzu_dd, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frint_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frint_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frint_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_frintx_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frintx_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frintx_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_ZPZ_FP(sve_fcvtzu_sd, uint64_t, , vfp_float32_to_uint64_rtz) DO_ZPZ_FP(sve_fcvtzu_ds, uint64_t, , helper_vfp_touizd) DO_ZPZ_FP(sve_fcvtzu_dd, uint64_t, , vfp_float64_to_uint64_rtz) +DO_ZPZ_FP(sve_frint_h, uint16_t, H1_2, helper_advsimd_rinth) +DO_ZPZ_FP(sve_frint_s, uint32_t, H1_4, helper_rints) +DO_ZPZ_FP(sve_frint_d, uint64_t, , helper_rintd) + +DO_ZPZ_FP(sve_frintx_h, uint16_t, H1_2, float16_round_to_int) +DO_ZPZ_FP(sve_frintx_s, uint32_t, H1_4, float32_round_to_int) +DO_ZPZ_FP(sve_frintx_d, uint64_t, , float64_round_to_int) + DO_ZPZ_FP(sve_scvt_hh, uint16_t, H1_2, int16_to_float16) DO_ZPZ_FP(sve_scvt_sh, uint32_t, H1_4, int32_to_float16) DO_ZPZ_FP(sve_scvt_ss, uint32_t, H1_4, int32_to_float32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FCVTZU_dd(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_zpz_ptr(s, a->rd, a->rn, a->pg, false, gen_helper_sve_fcvtzu_dd); } +static gen_helper_gvec_3_ptr * const frint_fns[3] = { + gen_helper_sve_frint_h, + gen_helper_sve_frint_s, + gen_helper_sve_frint_d +}; + +static bool trans_FRINTI(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + if (a->esz == 0) { + return false; + } + return do_zpz_ptr(s, a->rd, a->rn, a->pg, a->esz == MO_16, + frint_fns[a->esz - 1]); +} + +static bool trans_FRINTX(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[3] = { + gen_helper_sve_frintx_h, + gen_helper_sve_frintx_s, + gen_helper_sve_frintx_d + }; + if (a->esz == 0) { + return false; + } + return do_zpz_ptr(s, a->rd, a->rn, a->pg, a->esz == MO_16, fns[a->esz - 1]); +} + +static bool do_frint_mode(DisasContext *s, arg_rpr_esz *a, int mode) +{ + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_i32 tmode = tcg_const_i32(mode); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + + gen_helper_set_rmode(tmode, tmode, status); + + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + pred_full_reg_offset(s, a->pg), + status, vsz, vsz, 0, frint_fns[a->esz - 1]); + + gen_helper_set_rmode(tmode, tmode, status); + tcg_temp_free_i32(tmode); + tcg_temp_free_ptr(status); + } + return true; +} + +static bool trans_FRINTN(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_frint_mode(s, a, float_round_nearest_even); +} + +static bool trans_FRINTP(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_frint_mode(s, a, float_round_up); +} + +static bool trans_FRINTM(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_frint_mode(s, a, float_round_down); +} + +static bool trans_FRINTZ(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_frint_mode(s, a, float_round_to_zero); +} + +static bool trans_FRINTA(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_frint_mode(s, a, float_round_ties_away); +} + static bool trans_SCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) { return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_hh); diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FCVTZU_sd 01100101 11 011 10 1 101 ... ..... ..... @rd_pg_rn_e0 FCVTZS_dd 01100101 11 011 11 0 101 ... ..... ..... @rd_pg_rn_e0 FCVTZU_dd 01100101 11 011 11 1 101 ... ..... ..... @rd_pg_rn_e0 +# SVE floating-point round to integral value +FRINTN 01100101 .. 000 000 101 ... ..... ..... @rd_pg_rn +FRINTP 01100101 .. 000 001 101 ... ..... ..... @rd_pg_rn +FRINTM 01100101 .. 000 010 101 ... ..... ..... @rd_pg_rn +FRINTZ 01100101 .. 000 011 101 ... ..... ..... @rd_pg_rn +FRINTA 01100101 .. 000 100 101 ... ..... ..... @rd_pg_rn +FRINTX 01100101 .. 000 110 101 ... ..... ..... @rd_pg_rn +FRINTI 01100101 .. 000 111 101 ... ..... ..... @rd_pg_rn + # SVE integer convert to floating-point SCVTF_hh 01100101 01 010 01 0 101 ... ..... ..... @rd_pg_rn_e0 SCVTF_sh 01100101 01 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-27-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 14 ++++++++++++++ target/arm/sve_helper.c | 8 ++++++++ target/arm/translate-sve.c | 26 ++++++++++++++++++++++++++ target/arm/sve.decode | 4 ++++ 4 files changed, 52 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_frintx_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_frintx_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frecpx_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frecpx_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_frecpx_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_fsqrt_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fsqrt_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_fsqrt_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_scvt_hh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_scvt_sh, TCG_CALL_NO_RWG, diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_ZPZ_FP(sve_frintx_h, uint16_t, H1_2, float16_round_to_int) DO_ZPZ_FP(sve_frintx_s, uint32_t, H1_4, float32_round_to_int) DO_ZPZ_FP(sve_frintx_d, uint64_t, , float64_round_to_int) +DO_ZPZ_FP(sve_frecpx_h, uint16_t, H1_2, helper_frecpx_f16) +DO_ZPZ_FP(sve_frecpx_s, uint32_t, H1_4, helper_frecpx_f32) +DO_ZPZ_FP(sve_frecpx_d, uint64_t, , helper_frecpx_f64) + +DO_ZPZ_FP(sve_fsqrt_h, uint16_t, H1_2, float16_sqrt) +DO_ZPZ_FP(sve_fsqrt_s, uint32_t, H1_4, float32_sqrt) +DO_ZPZ_FP(sve_fsqrt_d, uint64_t, , float64_sqrt) + DO_ZPZ_FP(sve_scvt_hh, uint16_t, H1_2, int16_to_float16) DO_ZPZ_FP(sve_scvt_sh, uint32_t, H1_4, int32_to_float16) DO_ZPZ_FP(sve_scvt_ss, uint32_t, H1_4, int32_to_float32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FRINTA(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_frint_mode(s, a, float_round_ties_away); } +static bool trans_FRECPX(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[3] = { + gen_helper_sve_frecpx_h, + gen_helper_sve_frecpx_s, + gen_helper_sve_frecpx_d + }; + if (a->esz == 0) { + return false; + } + return do_zpz_ptr(s, a->rd, a->rn, a->pg, a->esz == MO_16, fns[a->esz - 1]); +} + +static bool trans_FSQRT(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[3] = { + gen_helper_sve_fsqrt_h, + gen_helper_sve_fsqrt_s, + gen_helper_sve_fsqrt_d + }; + if (a->esz == 0) { + return false; + } + return do_zpz_ptr(s, a->rd, a->rn, a->pg, a->esz == MO_16, fns[a->esz - 1]); +} + static bool trans_SCVTF_hh(DisasContext *s, arg_rpr_esz *a, uint32_t insn) { return do_zpz_ptr(s, a->rd, a->rn, a->pg, true, gen_helper_sve_scvt_hh); diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FRINTA 01100101 .. 000 100 101 ... ..... ..... @rd_pg_rn FRINTX 01100101 .. 000 110 101 ... ..... ..... @rd_pg_rn FRINTI 01100101 .. 000 111 101 ... ..... ..... @rd_pg_rn +# SVE floating-point unary operations +FRECPX 01100101 .. 001 100 101 ... ..... ..... @rd_pg_rn +FSQRT 01100101 .. 001 101 101 ... ..... ..... @rd_pg_rn + # SVE integer convert to floating-point SCVTF_hh 01100101 01 010 01 0 101 ... ..... ..... @rd_pg_rn_e0 SCVTF_sh 01100101 01 010 10 0 101 ... ..... ..... @rd_pg_rn_e0 -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-28-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 60 +++++++++++++++++++++++++++++++++++++- target/arm/sve.decode | 7 +++++ 2 files changed, 66 insertions(+), 1 deletion(-) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_zpzz_ool(DisasContext *s, arg_rprr_esz *a, gen_helper_gvec_4 *fn) return true; } +/* Select active elememnts from Zn and inactive elements from Zm, + * storing the result in Zd. + */ +static void do_sel_z(DisasContext *s, int rd, int rn, int rm, int pg, int esz) +{ + static gen_helper_gvec_4 * const fns[4] = { + gen_helper_sve_sel_zpzz_b, gen_helper_sve_sel_zpzz_h, + gen_helper_sve_sel_zpzz_s, gen_helper_sve_sel_zpzz_d + }; + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_4_ool(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + vec_full_reg_offset(s, rm), + pred_full_reg_offset(s, pg), + vsz, vsz, 0, fns[esz]); +} + #define DO_ZPZZ(NAME, name) \ static bool trans_##NAME##_zpzz(DisasContext *s, arg_rprr_esz *a, \ uint32_t insn) \ @@ -XXX,XX +XXX,XX @@ static bool trans_UDIV_zpzz(DisasContext *s, arg_rprr_esz *a, uint32_t insn) return do_zpzz_ool(s, a, fns[a->esz]); } -DO_ZPZZ(SEL, sel) +static bool trans_SEL_zpzz(DisasContext *s, arg_rprr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_sel_z(s, a->rd, a->rn, a->rm, a->pg, a->esz); + } + return true; +} #undef DO_ZPZZ @@ -XXX,XX +XXX,XX @@ static bool trans_PRF_rr(DisasContext *s, arg_PRF_rr *a, uint32_t insn) sve_access_check(s); return true; } + +/* + * Move Prefix + * + * TODO: The implementation so far could handle predicated merging movprfx. + * The helper functions as written take an extra source register to + * use in the operation, but the result is only written when predication + * succeeds. For unpredicated movprfx, we need to rearrange the helpers + * to allow the final write back to the destination to be unconditional. + * For predicated zeroing movprfx, we need to rearrange the helpers to + * allow the final write back to zero inactives. + * + * In the meantime, just emit the moves. + */ + +static bool trans_MOVPRFX(DisasContext *s, arg_MOVPRFX *a, uint32_t insn) +{ + return do_mov_z(s, a->rd, a->rn); +} + +static bool trans_MOVPRFX_m(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_sel_z(s, a->rd, a->rn, a->rd, a->pg, a->esz); + } + return true; +} + +static bool trans_MOVPRFX_z(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_movz_zpz(s, a->rd, a->rn, a->pg, a->esz); + } + return true; +} diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ ORV 00000100 .. 011 000 001 ... ..... ..... @rd_pg_rn EORV 00000100 .. 011 001 001 ... ..... ..... @rd_pg_rn ANDV 00000100 .. 011 010 001 ... ..... ..... @rd_pg_rn +# SVE constructive prefix (predicated) +MOVPRFX_z 00000100 .. 010 000 001 ... ..... ..... @rd_pg_rn +MOVPRFX_m 00000100 .. 010 001 001 ... ..... ..... @rd_pg_rn + # SVE integer add reduction (predicated) # Note that saddv requires size != 3. UADDV 00000100 .. 000 001 001 ... ..... ..... @rd_pg_rn @@ -XXX,XX +XXX,XX @@ ADR_p64 00000100 11 1 ..... 1010 .. ..... ..... @rd_rn_msz_rm ### SVE Integer Misc - Unpredicated Group +# SVE constructive prefix (unpredicated) +MOVPRFX 00000100 00 1 00000 101111 rn:5 rd:5 + # SVE floating-point exponential accelerator # Note esz != 0 FEXPA 00000100 .. 1 00000 101110 ..... ..... @rd_rn -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-29-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 7 +++ target/arm/sve_helper.c | 100 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 24 +++++++++ target/arm/sve.decode | 4 ++ 4 files changed, 135 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(sve_facgt_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_6(sve_facgt_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcadd_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcadd_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sve_fcadd_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fmla_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_ftmad_d)(void *vd, void *vn, void *vm, void *vs, uint32_t desc) } } +/* + * FP Complex Add + */ + +void HELPER(sve_fcadd_h)(void *vd, void *vn, void *vm, void *vg, + void *vs, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + uint64_t *g = vg; + float16 neg_imag = float16_set_sign(0, simd_data(desc)); + float16 neg_real = float16_chs(neg_imag); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float16 e0, e1, e2, e3; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float16); + i -= 2 * sizeof(float16); + + e0 = *(float16 *)(vn + H1_2(i)); + e1 = *(float16 *)(vm + H1_2(j)) ^ neg_real; + e2 = *(float16 *)(vn + H1_2(j)); + e3 = *(float16 *)(vm + H1_2(i)) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + *(float16 *)(vd + H1_2(i)) = float16_add(e0, e1, vs); + } + if (likely((pg >> (j & 63)) & 1)) { + *(float16 *)(vd + H1_2(j)) = float16_add(e2, e3, vs); + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fcadd_s)(void *vd, void *vn, void *vm, void *vg, + void *vs, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + uint64_t *g = vg; + float32 neg_imag = float32_set_sign(0, simd_data(desc)); + float32 neg_real = float32_chs(neg_imag); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float32 e0, e1, e2, e3; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float32); + i -= 2 * sizeof(float32); + + e0 = *(float32 *)(vn + H1_2(i)); + e1 = *(float32 *)(vm + H1_2(j)) ^ neg_real; + e2 = *(float32 *)(vn + H1_2(j)); + e3 = *(float32 *)(vm + H1_2(i)) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + *(float32 *)(vd + H1_2(i)) = float32_add(e0, e1, vs); + } + if (likely((pg >> (j & 63)) & 1)) { + *(float32 *)(vd + H1_2(j)) = float32_add(e2, e3, vs); + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fcadd_d)(void *vd, void *vn, void *vm, void *vg, + void *vs, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + uint64_t *g = vg; + float64 neg_imag = float64_set_sign(0, simd_data(desc)); + float64 neg_real = float64_chs(neg_imag); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float64 e0, e1, e2, e3; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float64); + i -= 2 * sizeof(float64); + + e0 = *(float64 *)(vn + H1_2(i)); + e1 = *(float64 *)(vm + H1_2(j)) ^ neg_real; + e2 = *(float64 *)(vn + H1_2(j)); + e3 = *(float64 *)(vm + H1_2(i)) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + *(float64 *)(vd + H1_2(i)) = float64_add(e0, e1, vs); + } + if (likely((pg >> (j & 63)) & 1)) { + *(float64 *)(vd + H1_2(j)) = float64_add(e2, e3, vs); + } + } while (i & 63); + } while (i != 0); +} + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FPCMP(FACGT, facgt) #undef DO_FPCMP +static bool trans_FCADD(DisasContext *s, arg_FCADD *a, uint32_t insn) +{ + static gen_helper_gvec_4_ptr * const fns[3] = { + gen_helper_sve_fcadd_h, + gen_helper_sve_fcadd_s, + gen_helper_sve_fcadd_d + }; + + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_4_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + pred_full_reg_offset(s, a->pg), + status, vsz, vsz, a->rot, fns[a->esz - 1]); + tcg_temp_free_ptr(status); + } + return true; +} + typedef void gen_helper_sve_fmla(TCGv_env, TCGv_ptr, TCGv_i32); static bool do_fmla(DisasContext *s, arg_rprrr_esz *a, gen_helper_sve_fmla *fn) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u # SVE integer multiply immediate (unpredicated) MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s +# SVE floating-point complex add (predicated) +FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \ + rn=%reg_movprfx + ### SVE FP Multiply-Add Indexed Group # SVE floating-point multiply-add (indexed) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-30-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 4 + target/arm/sve_helper.c | 162 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 37 +++++++++ target/arm/sve.decode | 4 + 4 files changed, 207 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(sve_fnmls_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fcmla_zpzzz_h, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fcmla_zpzzz_s, TCG_CALL_NO_RWG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(sve_fcmla_zpzzz_d, TCG_CALL_NO_RWG, void, env, ptr, i32) + DEF_HELPER_FLAGS_5(sve_ftmad_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_ftmad_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_ftmad_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_fcadd_d)(void *vd, void *vn, void *vm, void *vg, } while (i != 0); } +/* + * FP Complex Multiply + */ + +QEMU_BUILD_BUG_ON(SIMD_DATA_SHIFT + 22 > 32); + +void HELPER(sve_fcmla_zpzzz_h)(CPUARMState *env, void *vg, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + unsigned rot = extract32(desc, SIMD_DATA_SHIFT + 20, 2); + bool flip = rot & 1; + float16 neg_imag, neg_real; + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + neg_imag = float16_set_sign(0, (rot & 2) != 0); + neg_real = float16_set_sign(0, rot == 1 || rot == 2); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float16 e1, e2, e3, e4, nr, ni, mr, mi, d; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float16); + i -= 2 * sizeof(float16); + + nr = *(float16 *)(vn + H1_2(i)); + ni = *(float16 *)(vn + H1_2(j)); + mr = *(float16 *)(vm + H1_2(i)); + mi = *(float16 *)(vm + H1_2(j)); + + e2 = (flip ? ni : nr); + e1 = (flip ? mi : mr) ^ neg_real; + e4 = e2; + e3 = (flip ? mr : mi) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + d = *(float16 *)(va + H1_2(i)); + d = float16_muladd(e2, e1, d, 0, &env->vfp.fp_status_f16); + *(float16 *)(vd + H1_2(i)) = d; + } + if (likely((pg >> (j & 63)) & 1)) { + d = *(float16 *)(va + H1_2(j)); + d = float16_muladd(e4, e3, d, 0, &env->vfp.fp_status_f16); + *(float16 *)(vd + H1_2(j)) = d; + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fcmla_zpzzz_s)(CPUARMState *env, void *vg, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + unsigned rot = extract32(desc, SIMD_DATA_SHIFT + 20, 2); + bool flip = rot & 1; + float32 neg_imag, neg_real; + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + neg_imag = float32_set_sign(0, (rot & 2) != 0); + neg_real = float32_set_sign(0, rot == 1 || rot == 2); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float32 e1, e2, e3, e4, nr, ni, mr, mi, d; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float32); + i -= 2 * sizeof(float32); + + nr = *(float32 *)(vn + H1_2(i)); + ni = *(float32 *)(vn + H1_2(j)); + mr = *(float32 *)(vm + H1_2(i)); + mi = *(float32 *)(vm + H1_2(j)); + + e2 = (flip ? ni : nr); + e1 = (flip ? mi : mr) ^ neg_real; + e4 = e2; + e3 = (flip ? mr : mi) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + d = *(float32 *)(va + H1_2(i)); + d = float32_muladd(e2, e1, d, 0, &env->vfp.fp_status); + *(float32 *)(vd + H1_2(i)) = d; + } + if (likely((pg >> (j & 63)) & 1)) { + d = *(float32 *)(va + H1_2(j)); + d = float32_muladd(e4, e3, d, 0, &env->vfp.fp_status); + *(float32 *)(vd + H1_2(j)) = d; + } + } while (i & 63); + } while (i != 0); +} + +void HELPER(sve_fcmla_zpzzz_d)(CPUARMState *env, void *vg, uint32_t desc) +{ + intptr_t j, i = simd_oprsz(desc); + unsigned rd = extract32(desc, SIMD_DATA_SHIFT, 5); + unsigned rn = extract32(desc, SIMD_DATA_SHIFT + 5, 5); + unsigned rm = extract32(desc, SIMD_DATA_SHIFT + 10, 5); + unsigned ra = extract32(desc, SIMD_DATA_SHIFT + 15, 5); + unsigned rot = extract32(desc, SIMD_DATA_SHIFT + 20, 2); + bool flip = rot & 1; + float64 neg_imag, neg_real; + void *vd = &env->vfp.zregs[rd]; + void *vn = &env->vfp.zregs[rn]; + void *vm = &env->vfp.zregs[rm]; + void *va = &env->vfp.zregs[ra]; + uint64_t *g = vg; + + neg_imag = float64_set_sign(0, (rot & 2) != 0); + neg_real = float64_set_sign(0, rot == 1 || rot == 2); + + do { + uint64_t pg = g[(i - 1) >> 6]; + do { + float64 e1, e2, e3, e4, nr, ni, mr, mi, d; + + /* I holds the real index; J holds the imag index. */ + j = i - sizeof(float64); + i -= 2 * sizeof(float64); + + nr = *(float64 *)(vn + H1_2(i)); + ni = *(float64 *)(vn + H1_2(j)); + mr = *(float64 *)(vm + H1_2(i)); + mi = *(float64 *)(vm + H1_2(j)); + + e2 = (flip ? ni : nr); + e1 = (flip ? mi : mr) ^ neg_real; + e4 = e2; + e3 = (flip ? mr : mi) ^ neg_imag; + + if (likely((pg >> (i & 63)) & 1)) { + d = *(float64 *)(va + H1_2(i)); + d = float64_muladd(e2, e1, d, 0, &env->vfp.fp_status); + *(float64 *)(vd + H1_2(i)) = d; + } + if (likely((pg >> (j & 63)) & 1)) { + d = *(float64 *)(va + H1_2(j)); + d = float64_muladd(e4, e3, d, 0, &env->vfp.fp_status); + *(float64 *)(vd + H1_2(j)) = d; + } + } while (i & 63); + } while (i != 0); +} + /* * Load contiguous data, protected by a governing predicate. */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_FMLA(FNMLS_zpzzz, fnmls_zpzzz) #undef DO_FMLA +static bool trans_FCMLA_zpzzz(DisasContext *s, + arg_FCMLA_zpzzz *a, uint32_t insn) +{ + static gen_helper_sve_fmla * const fns[3] = { + gen_helper_sve_fcmla_zpzzz_h, + gen_helper_sve_fcmla_zpzzz_s, + gen_helper_sve_fcmla_zpzzz_d, + }; + + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + unsigned desc; + TCGv_i32 t_desc; + TCGv_ptr pg = tcg_temp_new_ptr(); + + /* We would need 7 operands to pass these arguments "properly". + * So we encode all the register numbers into the descriptor. + */ + desc = deposit32(a->rd, 5, 5, a->rn); + desc = deposit32(desc, 10, 5, a->rm); + desc = deposit32(desc, 15, 5, a->ra); + desc = deposit32(desc, 20, 2, a->rot); + desc = sextract32(desc, 0, 22); + desc = simd_desc(vsz, vsz, desc); + + t_desc = tcg_const_i32(desc); + tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg)); + fns[a->esz - 1](cpu_env, pg, t_desc); + tcg_temp_free_i32(t_desc); + tcg_temp_free_ptr(pg); + } + return true; +} + /* *** SVE Floating Point Unary Operations Predicated Group */ diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \ rn=%reg_movprfx +# SVE floating-point complex multiply-add (predicated) +FCMLA_zpzzz 01100100 esz:2 0 rm:5 0 rot:2 pg:3 rn:5 rd:5 \ + ra=%reg_movprfx + ### SVE FP Multiply-Add Indexed Group # SVE floating-point multiply-add (indexed) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> For aa64 advsimd, we had been passing the pre-indexed vector. However, sve applies the index to each 128-bit segment, so we need to pass in the index separately. For aa32 advsimd, the fp32 operation always has index 0, but we failed to interpret the fp16 index correctly. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-31-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 21 ++++++++++++--------- target/arm/translate.c | 32 +++++++++++++++++++++++--------- target/arm/vec_helper.c | 10 ++++++---- 3 files changed, 41 insertions(+), 22 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void disas_simd_indexed(DisasContext *s, uint32_t insn) case 0x13: /* FCMLA #90 */ case 0x15: /* FCMLA #180 */ case 0x17: /* FCMLA #270 */ - tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd), - vec_full_reg_offset(s, rn), - vec_reg_offset(s, rm, index, size), fpst, - is_q ? 16 : 8, vec_full_reg_size(s), - extract32(insn, 13, 2), /* rot */ - size == MO_64 - ? gen_helper_gvec_fcmlas_idx - : gen_helper_gvec_fcmlah_idx); - tcg_temp_free_ptr(fpst); + { + int rot = extract32(insn, 13, 2); + int data = (index << 2) | rot; + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + vec_full_reg_offset(s, rm), fpst, + is_q ? 16 : 8, vec_full_reg_size(s), data, + size == MO_64 + ? gen_helper_gvec_fcmlas_idx + : gen_helper_gvec_fcmlah_idx); + tcg_temp_free_ptr(fpst); + } return; } diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) { - int rd, rn, rm, rot, size, opr_sz; + gen_helper_gvec_3_ptr *fn_gvec_ptr; + int rd, rn, rm, opr_sz, data; TCGv_ptr fpst; bool q; q = extract32(insn, 6, 1); VFP_DREG_D(rd, insn); VFP_DREG_N(rn, insn); - VFP_DREG_M(rm, insn); if ((rd | rn) & q) { return 1; } if ((insn & 0xff000f10) == 0xfe000800) { /* VCMLA (indexed) -- 1111 1110 S.RR .... .... 1000 ...0 .... */ - rot = extract32(insn, 20, 2); - size = extract32(insn, 23, 1); - if (!arm_dc_feature(s, ARM_FEATURE_V8_FCMA) - || (!size && !arm_dc_feature(s, ARM_FEATURE_V8_FP16))) { + int rot = extract32(insn, 20, 2); + int size = extract32(insn, 23, 1); + int index; + + if (!arm_dc_feature(s, ARM_FEATURE_V8_FCMA)) { return 1; } + if (size == 0) { + if (!arm_dc_feature(s, ARM_FEATURE_V8_FP16)) { + return 1; + } + /* For fp16, rm is just Vm, and index is M. */ + rm = extract32(insn, 0, 4); + index = extract32(insn, 5, 1); + } else { + /* For fp32, rm is the usual M:Vm, and index is 0. */ + VFP_DREG_M(rm, insn); + index = 0; + } + data = (index << 2) | rot; + fn_gvec_ptr = (size ? gen_helper_gvec_fcmlas_idx + : gen_helper_gvec_fcmlah_idx); } else { return 1; } @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), vfp_reg_offset(1, rn), vfp_reg_offset(1, rm), fpst, - opr_sz, opr_sz, rot, - size ? gen_helper_gvec_fcmlas_idx - : gen_helper_gvec_fcmlah_idx); + opr_sz, opr_sz, data, fn_gvec_ptr); tcg_temp_free_ptr(fpst); return 0; } diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_fcmlah_idx)(void *vd, void *vn, void *vm, float_status *fpst = vfpst; intptr_t flip = extract32(desc, SIMD_DATA_SHIFT, 1); uint32_t neg_imag = extract32(desc, SIMD_DATA_SHIFT + 1, 1); + intptr_t index = extract32(desc, SIMD_DATA_SHIFT + 2, 2); uint32_t neg_real = flip ^ neg_imag; uintptr_t i; - float16 e1 = m[H2(flip)]; - float16 e3 = m[H2(1 - flip)]; + float16 e1 = m[H2(2 * index + flip)]; + float16 e3 = m[H2(2 * index + 1 - flip)]; /* Shift boolean to the sign bit so we can xor to negate. */ neg_real <<= 15; @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_fcmlas_idx)(void *vd, void *vn, void *vm, float_status *fpst = vfpst; intptr_t flip = extract32(desc, SIMD_DATA_SHIFT, 1); uint32_t neg_imag = extract32(desc, SIMD_DATA_SHIFT + 1, 1); + intptr_t index = extract32(desc, SIMD_DATA_SHIFT + 2, 2); uint32_t neg_real = flip ^ neg_imag; uintptr_t i; - float32 e1 = m[H4(flip)]; - float32 e3 = m[H4(1 - flip)]; + float32 e1 = m[H4(2 * index + flip)]; + float32 e3 = m[H4(2 * index + 1 - flip)]; /* Shift boolean to the sign bit so we can xor to negate. */ neg_real <<= 31; -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Enhance the existing helpers to support SVE, which takes the index from each 128-bit segment. The change has no effect for AdvSIMD, since there is only one such segment. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-32-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 23 ++++++++++++++++++ target/arm/vec_helper.c | 50 +++++++++++++++++++++++--------------- target/arm/sve.decode | 6 +++++ 3 files changed, 59 insertions(+), 20 deletions(-) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_FCMLA_zpzzz(DisasContext *s, return true; } +static bool trans_FCMLA_zzxz(DisasContext *s, arg_FCMLA_zzxz *a, uint32_t insn) +{ + static gen_helper_gvec_3_ptr * const fns[2] = { + gen_helper_gvec_fcmlah_idx, + gen_helper_gvec_fcmlas_idx, + }; + + tcg_debug_assert(a->esz == 1 || a->esz == 2); + tcg_debug_assert(a->rd == a->ra); + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_ptr status = get_fpstatus_ptr(a->esz == MO_16); + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + status, vsz, vsz, + a->index * 4 + a->rot, + fns[a->esz - 1]); + tcg_temp_free_ptr(status); + } + return true; +} + /* *** SVE Floating Point Unary Operations Predicated Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_fcmlah_idx)(void *vd, void *vn, void *vm, uint32_t neg_imag = extract32(desc, SIMD_DATA_SHIFT + 1, 1); intptr_t index = extract32(desc, SIMD_DATA_SHIFT + 2, 2); uint32_t neg_real = flip ^ neg_imag; - uintptr_t i; - float16 e1 = m[H2(2 * index + flip)]; - float16 e3 = m[H2(2 * index + 1 - flip)]; + intptr_t elements = opr_sz / sizeof(float16); + intptr_t eltspersegment = 16 / sizeof(float16); + intptr_t i, j; /* Shift boolean to the sign bit so we can xor to negate. */ neg_real <<= 15; neg_imag <<= 15; - e1 ^= neg_real; - e3 ^= neg_imag; - for (i = 0; i < opr_sz / 2; i += 2) { - float16 e2 = n[H2(i + flip)]; - float16 e4 = e2; + for (i = 0; i < elements; i += eltspersegment) { + float16 mr = m[H2(i + 2 * index + 0)]; + float16 mi = m[H2(i + 2 * index + 1)]; + float16 e1 = neg_real ^ (flip ? mi : mr); + float16 e3 = neg_imag ^ (flip ? mr : mi); - d[H2(i)] = float16_muladd(e2, e1, d[H2(i)], 0, fpst); - d[H2(i + 1)] = float16_muladd(e4, e3, d[H2(i + 1)], 0, fpst); + for (j = i; j < i + eltspersegment; j += 2) { + float16 e2 = n[H2(j + flip)]; + float16 e4 = e2; + + d[H2(j)] = float16_muladd(e2, e1, d[H2(j)], 0, fpst); + d[H2(j + 1)] = float16_muladd(e4, e3, d[H2(j + 1)], 0, fpst); + } } clear_tail(d, opr_sz, simd_maxsz(desc)); } @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_fcmlas_idx)(void *vd, void *vn, void *vm, uint32_t neg_imag = extract32(desc, SIMD_DATA_SHIFT + 1, 1); intptr_t index = extract32(desc, SIMD_DATA_SHIFT + 2, 2); uint32_t neg_real = flip ^ neg_imag; - uintptr_t i; - float32 e1 = m[H4(2 * index + flip)]; - float32 e3 = m[H4(2 * index + 1 - flip)]; + intptr_t elements = opr_sz / sizeof(float32); + intptr_t eltspersegment = 16 / sizeof(float32); + intptr_t i, j; /* Shift boolean to the sign bit so we can xor to negate. */ neg_real <<= 31; neg_imag <<= 31; - e1 ^= neg_real; - e3 ^= neg_imag; - for (i = 0; i < opr_sz / 4; i += 2) { - float32 e2 = n[H4(i + flip)]; - float32 e4 = e2; + for (i = 0; i < elements; i += eltspersegment) { + float32 mr = m[H4(i + 2 * index + 0)]; + float32 mi = m[H4(i + 2 * index + 1)]; + float32 e1 = neg_real ^ (flip ? mi : mr); + float32 e3 = neg_imag ^ (flip ? mr : mi); - d[H4(i)] = float32_muladd(e2, e1, d[H4(i)], 0, fpst); - d[H4(i + 1)] = float32_muladd(e4, e3, d[H4(i + 1)], 0, fpst); + for (j = i; j < i + eltspersegment; j += 2) { + float32 e2 = n[H4(j + flip)]; + float32 e4 = e2; + + d[H4(j)] = float32_muladd(e2, e1, d[H4(j)], 0, fpst); + d[H4(j + 1)] = float32_muladd(e4, e3, d[H4(j + 1)], 0, fpst); + } } clear_tail(d, opr_sz, simd_maxsz(desc)); } diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \ FCMLA_zpzzz 01100100 esz:2 0 rm:5 0 rot:2 pg:3 rn:5 rd:5 \ ra=%reg_movprfx +# SVE floating-point complex multiply-add (indexed) +FCMLA_zzxz 01100100 10 1 index:2 rm:3 0001 rot:2 rn:5 rd:5 \ + ra=%reg_movprfx esz=1 +FCMLA_zzxz 01100100 11 1 index:1 rm:4 0001 rot:2 rn:5 rd:5 \ + ra=%reg_movprfx esz=2 + ### SVE FP Multiply-Add Indexed Group # SVE floating-point multiply-add (indexed) -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-33-richard.henderson@linaro.org [PMM: moved 'ra=%reg_movprfx' here from following patch] Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 5 +++ target/arm/translate-sve.c | 17 ++++++++++ target/arm/vec_helper.c | 67 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 3 ++ 4 files changed, 92 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(gvec_qrdmlah_s32, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_qrdmlsh_s32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_fcadds, TCG_CALL_NO_RWG, diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_ZZI(UMIN, umin) #undef DO_ZZI +static bool trans_DOT_zzz(DisasContext *s, arg_DOT_zzz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[2][2] = { + { gen_helper_gvec_sdot_b, gen_helper_gvec_sdot_h }, + { gen_helper_gvec_udot_b, gen_helper_gvec_udot_h } + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + vsz, vsz, 0, fns[a->u][a->sz]); + } + return true; +} + /* *** SVE Floating Point Multiply-Add Indexed Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_qrdmlsh_s32)(void *vd, void *vn, void *vm, clear_tail(d, opr_sz, simd_maxsz(desc)); } +/* Integer 8 and 16-bit dot-product. + * + * Note that for the loops herein, host endianness does not matter + * with respect to the ordering of data within the 64-bit lanes. + * All elements are treated equally, no matter where they are. + */ + +void HELPER(gvec_sdot_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc); + uint32_t *d = vd; + int8_t *n = vn, *m = vm; + + for (i = 0; i < opr_sz / 4; ++i) { + d[i] += n[i * 4 + 0] * m[i * 4 + 0] + + n[i * 4 + 1] * m[i * 4 + 1] + + n[i * 4 + 2] * m[i * 4 + 2] + + n[i * 4 + 3] * m[i * 4 + 3]; + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc); + uint32_t *d = vd; + uint8_t *n = vn, *m = vm; + + for (i = 0; i < opr_sz / 4; ++i) { + d[i] += n[i * 4 + 0] * m[i * 4 + 0] + + n[i * 4 + 1] * m[i * 4 + 1] + + n[i * 4 + 2] * m[i * 4 + 2] + + n[i * 4 + 3] * m[i * 4 + 3]; + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_sdot_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc); + uint64_t *d = vd; + int16_t *n = vn, *m = vm; + + for (i = 0; i < opr_sz / 8; ++i) { + d[i] += (int64_t)n[i * 4 + 0] * m[i * 4 + 0] + + (int64_t)n[i * 4 + 1] * m[i * 4 + 1] + + (int64_t)n[i * 4 + 2] * m[i * 4 + 2] + + (int64_t)n[i * 4 + 3] * m[i * 4 + 3]; + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc); + uint64_t *d = vd; + uint16_t *n = vn, *m = vm; + + for (i = 0; i < opr_sz / 8; ++i) { + d[i] += (uint64_t)n[i * 4 + 0] * m[i * 4 + 0] + + (uint64_t)n[i * 4 + 1] * m[i * 4 + 1] + + (uint64_t)n[i * 4 + 2] * m[i * 4 + 2] + + (uint64_t)n[i * 4 + 3] * m[i * 4 + 3]; + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + void HELPER(gvec_fcaddh)(void *vd, void *vn, void *vm, void *vfpst, uint32_t desc) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u # SVE integer multiply immediate (unpredicated) MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s +# SVE integer dot product (unpredicated) +DOT_zzz 01000100 1 sz:1 0 rm:5 00000 u:1 rn:5 rd:5 ra=%reg_movprfx + # SVE floating-point complex add (predicated) FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \ rn=%reg_movprfx -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180627043328.11531-34-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 5 ++ target/arm/translate-sve.c | 18 ++++++ target/arm/vec_helper.c | 124 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 6 ++ 4 files changed, 153 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(gvec_udot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sdot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_udot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_idx_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_idx_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_idx_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_idx_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_fcadds, TCG_CALL_NO_RWG, diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_DOT_zzz(DisasContext *s, arg_DOT_zzz *a, uint32_t insn) return true; } +static bool trans_DOT_zzx(DisasContext *s, arg_DOT_zzx *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[2][2] = { + { gen_helper_gvec_sdot_idx_b, gen_helper_gvec_sdot_idx_h }, + { gen_helper_gvec_udot_idx_b, gen_helper_gvec_udot_idx_h } + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + vsz, vsz, a->index, fns[a->u][a->sz]); + } + return true; +} + + /* *** SVE Floating Point Multiply-Add Indexed Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_udot_h)(void *vd, void *vn, void *vm, uint32_t desc) clear_tail(d, opr_sz, simd_maxsz(desc)); } +void HELPER(gvec_sdot_idx_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4; + intptr_t index = simd_data(desc); + uint32_t *d = vd; + int8_t *n = vn; + int8_t *m_indexed = (int8_t *)vm + index * 4; + + /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd. + * Otherwise opr_sz is a multiple of 16. + */ + segend = MIN(4, opr_sz_4); + i = 0; + do { + int8_t m0 = m_indexed[i * 4 + 0]; + int8_t m1 = m_indexed[i * 4 + 1]; + int8_t m2 = m_indexed[i * 4 + 2]; + int8_t m3 = m_indexed[i * 4 + 3]; + + do { + d[i] += n[i * 4 + 0] * m0 + + n[i * 4 + 1] * m1 + + n[i * 4 + 2] * m2 + + n[i * 4 + 3] * m3; + } while (++i < segend); + segend = i + 4; + } while (i < opr_sz_4); + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_idx_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4; + intptr_t index = simd_data(desc); + uint32_t *d = vd; + uint8_t *n = vn; + uint8_t *m_indexed = (uint8_t *)vm + index * 4; + + /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd. + * Otherwise opr_sz is a multiple of 16. + */ + segend = MIN(4, opr_sz_4); + i = 0; + do { + uint8_t m0 = m_indexed[i * 4 + 0]; + uint8_t m1 = m_indexed[i * 4 + 1]; + uint8_t m2 = m_indexed[i * 4 + 2]; + uint8_t m3 = m_indexed[i * 4 + 3]; + + do { + d[i] += n[i * 4 + 0] * m0 + + n[i * 4 + 1] * m1 + + n[i * 4 + 2] * m2 + + n[i * 4 + 3] * m3; + } while (++i < segend); + segend = i + 4; + } while (i < opr_sz_4); + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_sdot_idx_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc), opr_sz_8 = opr_sz / 8; + intptr_t index = simd_data(desc); + uint64_t *d = vd; + int16_t *n = vn; + int16_t *m_indexed = (int16_t *)vm + index * 4; + + /* This is supported by SVE only, so opr_sz is always a multiple of 16. + * Process the entire segment all at once, writing back the results + * only after we've consumed all of the inputs. + */ + for (i = 0; i < opr_sz_8 ; i += 2) { + uint64_t d0, d1; + + d0 = n[i * 4 + 0] * (int64_t)m_indexed[i * 4 + 0]; + d0 += n[i * 4 + 1] * (int64_t)m_indexed[i * 4 + 1]; + d0 += n[i * 4 + 2] * (int64_t)m_indexed[i * 4 + 2]; + d0 += n[i * 4 + 3] * (int64_t)m_indexed[i * 4 + 3]; + d1 = n[i * 4 + 4] * (int64_t)m_indexed[i * 4 + 0]; + d1 += n[i * 4 + 5] * (int64_t)m_indexed[i * 4 + 1]; + d1 += n[i * 4 + 6] * (int64_t)m_indexed[i * 4 + 2]; + d1 += n[i * 4 + 7] * (int64_t)m_indexed[i * 4 + 3]; + + d[i + 0] += d0; + d[i + 1] += d1; + } + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_idx_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc), opr_sz_8 = opr_sz / 8; + intptr_t index = simd_data(desc); + uint64_t *d = vd; + uint16_t *n = vn; + uint16_t *m_indexed = (uint16_t *)vm + index * 4; + + /* This is supported by SVE only, so opr_sz is always a multiple of 16. + * Process the entire segment all at once, writing back the results + * only after we've consumed all of the inputs. + */ + for (i = 0; i < opr_sz_8 ; i += 2) { + uint64_t d0, d1; + + d0 = n[i * 4 + 0] * (uint64_t)m_indexed[i * 4 + 0]; + d0 += n[i * 4 + 1] * (uint64_t)m_indexed[i * 4 + 1]; + d0 += n[i * 4 + 2] * (uint64_t)m_indexed[i * 4 + 2]; + d0 += n[i * 4 + 3] * (uint64_t)m_indexed[i * 4 + 3]; + d1 = n[i * 4 + 4] * (uint64_t)m_indexed[i * 4 + 0]; + d1 += n[i * 4 + 5] * (uint64_t)m_indexed[i * 4 + 1]; + d1 += n[i * 4 + 6] * (uint64_t)m_indexed[i * 4 + 2]; + d1 += n[i * 4 + 7] * (uint64_t)m_indexed[i * 4 + 3]; + + d[i + 0] += d0; + d[i + 1] += d1; + } + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + void HELPER(gvec_fcaddh)(void *vd, void *vn, void *vm, void *vfpst, uint32_t desc) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s # SVE integer dot product (unpredicated) DOT_zzz 01000100 1 sz:1 0 rm:5 00000 u:1 rn:5 rd:5 ra=%reg_movprfx +# SVE integer dot product (indexed) +DOT_zzx 01000100 101 index:2 rm:3 00000 u:1 rn:5 rd:5 \ + sz=0 ra=%reg_movprfx +DOT_zzx 01000100 111 index:1 rm:4 00000 u:1 rn:5 rd:5 \ + sz=1 ra=%reg_movprfx + # SVE floating-point complex add (predicated) FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \ rn=%reg_movprfx -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Enable ARM_FEATURE_SVE for the generic "max" cpu. Tested-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-35-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/elfload.c | 1 + target/arm/cpu.c | 7 +++++++ target/arm/cpu64.c | 1 + 3 files changed, 9 insertions(+) diff --git a/linux-user/elfload.c b/linux-user/elfload.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/elfload.c +++ b/linux-user/elfload.c @@ -XXX,XX +XXX,XX @@ static uint32_t get_elf_hwcap(void) GET_FEATURE(ARM_FEATURE_V8_ATOMICS, ARM_HWCAP_A64_ATOMICS); GET_FEATURE(ARM_FEATURE_V8_RDM, ARM_HWCAP_A64_ASIMDRDM); GET_FEATURE(ARM_FEATURE_V8_FCMA, ARM_HWCAP_A64_FCMA); + GET_FEATURE(ARM_FEATURE_SVE, ARM_HWCAP_A64_SVE); #undef GET_FEATURE return hwcaps; diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_reset(CPUState *s) env->cp15.sctlr_el[1] |= SCTLR_UCT | SCTLR_UCI | SCTLR_DZE; /* and to the FP/Neon instructions */ env->cp15.cpacr_el1 = deposit64(env->cp15.cpacr_el1, 20, 2, 3); + /* and to the SVE instructions */ + env->cp15.cpacr_el1 = deposit64(env->cp15.cpacr_el1, 16, 2, 3); + env->cp15.cptr_el[3] |= CPTR_EZ; + /* with maximum vector length */ + env->vfp.zcr_el[1] = ARM_MAX_VQ - 1; + env->vfp.zcr_el[2] = ARM_MAX_VQ - 1; + env->vfp.zcr_el[3] = ARM_MAX_VQ - 1; #else /* Reset into the highest available EL */ if (arm_feature(env, ARM_FEATURE_EL3)) { diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) set_feature(&cpu->env, ARM_FEATURE_V8_RDM); set_feature(&cpu->env, ARM_FEATURE_V8_FP16); set_feature(&cpu->env, ARM_FEATURE_V8_FCMA); + set_feature(&cpu->env, ARM_FEATURE_SVE); /* For usermode -cpu max we can use a larger and more efficient DCZ * blocksize since we don't have to follow what the hardware does. */ -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> We've already added the helpers with an SVE patch, all that remains is to wire up the aa64 and aa32 translators. Enable the feature within -cpu max for CONFIG_USER_ONLY. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180627043328.11531-36-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 1 + linux-user/elfload.c | 1 + target/arm/cpu.c | 1 + target/arm/cpu64.c | 1 + target/arm/translate-a64.c | 36 +++++++++++++++++++ target/arm/translate.c | 74 +++++++++++++++++++++++++++----------- 6 files changed, 93 insertions(+), 21 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ enum arm_features { ARM_FEATURE_V8_SM4, /* implements SM4 part of v8 Crypto Extensions */ ARM_FEATURE_V8_ATOMICS, /* ARMv8.1-Atomics feature */ ARM_FEATURE_V8_RDM, /* implements v8.1 simd round multiply */ + ARM_FEATURE_V8_DOTPROD, /* implements v8.2 simd dot product */ ARM_FEATURE_V8_FP16, /* implements v8.2 half-precision float */ ARM_FEATURE_V8_FCMA, /* has complex number part of v8.3 extensions. */ ARM_FEATURE_M_MAIN, /* M profile Main Extension */ diff --git a/linux-user/elfload.c b/linux-user/elfload.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/elfload.c +++ b/linux-user/elfload.c @@ -XXX,XX +XXX,XX @@ static uint32_t get_elf_hwcap(void) ARM_HWCAP_A64_FPHP | ARM_HWCAP_A64_ASIMDHP); GET_FEATURE(ARM_FEATURE_V8_ATOMICS, ARM_HWCAP_A64_ATOMICS); GET_FEATURE(ARM_FEATURE_V8_RDM, ARM_HWCAP_A64_ASIMDRDM); + GET_FEATURE(ARM_FEATURE_V8_DOTPROD, ARM_HWCAP_A64_ASIMDDP); GET_FEATURE(ARM_FEATURE_V8_FCMA, ARM_HWCAP_A64_FCMA); GET_FEATURE(ARM_FEATURE_SVE, ARM_HWCAP_A64_SVE); #undef GET_FEATURE diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_max_initfn(Object *obj) set_feature(&cpu->env, ARM_FEATURE_V8_PMULL); set_feature(&cpu->env, ARM_FEATURE_CRC); set_feature(&cpu->env, ARM_FEATURE_V8_RDM); + set_feature(&cpu->env, ARM_FEATURE_V8_DOTPROD); set_feature(&cpu->env, ARM_FEATURE_V8_FCMA); #endif } diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) set_feature(&cpu->env, ARM_FEATURE_CRC); set_feature(&cpu->env, ARM_FEATURE_V8_ATOMICS); set_feature(&cpu->env, ARM_FEATURE_V8_RDM); + set_feature(&cpu->env, ARM_FEATURE_V8_DOTPROD); set_feature(&cpu->env, ARM_FEATURE_V8_FP16); set_feature(&cpu->env, ARM_FEATURE_V8_FCMA); set_feature(&cpu->env, ARM_FEATURE_SVE); diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void gen_gvec_op3(DisasContext *s, bool is_q, int rd, vec_full_reg_size(s), gvec_op); } +/* Expand a 3-operand operation using an out-of-line helper. */ +static void gen_gvec_op3_ool(DisasContext *s, bool is_q, int rd, + int rn, int rm, int data, gen_helper_gvec_3 *fn) +{ + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + vec_full_reg_offset(s, rm), + is_q ? 16 : 8, vec_full_reg_size(s), data, fn); +} + /* Expand a 3-operand + env pointer operation using * an out-of-line helper. */ @@ -XXX,XX +XXX,XX @@ static void disas_simd_three_reg_same_extra(DisasContext *s, uint32_t insn) } feature = ARM_FEATURE_V8_RDM; break; + case 0x02: /* SDOT (vector) */ + case 0x12: /* UDOT (vector) */ + if (size != MO_32) { + unallocated_encoding(s); + return; + } + feature = ARM_FEATURE_V8_DOTPROD; + break; case 0x8: /* FCMLA, #0 */ case 0x9: /* FCMLA, #90 */ case 0xa: /* FCMLA, #180 */ @@ -XXX,XX +XXX,XX @@ static void disas_simd_three_reg_same_extra(DisasContext *s, uint32_t insn) } return; + case 0x2: /* SDOT / UDOT */ + gen_gvec_op3_ool(s, is_q, rd, rn, rm, 0, + u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b); + return; + case 0x8: /* FCMLA, #0 */ case 0x9: /* FCMLA, #90 */ case 0xa: /* FCMLA, #180 */ @@ -XXX,XX +XXX,XX @@ static void disas_simd_indexed(DisasContext *s, uint32_t insn) return; } break; + case 0x0e: /* SDOT */ + case 0x1e: /* UDOT */ + if (size != MO_32 || !arm_dc_feature(s, ARM_FEATURE_V8_DOTPROD)) { + unallocated_encoding(s); + return; + } + break; case 0x11: /* FCMLA #0 */ case 0x13: /* FCMLA #90 */ case 0x15: /* FCMLA #180 */ @@ -XXX,XX +XXX,XX @@ static void disas_simd_indexed(DisasContext *s, uint32_t insn) } switch (16 * u + opcode) { + case 0x0e: /* SDOT */ + case 0x1e: /* UDOT */ + gen_gvec_op3_ool(s, is_q, rd, rn, rm, index, + u ? gen_helper_gvec_udot_idx_b + : gen_helper_gvec_sdot_idx_b); + return; case 0x11: /* FCMLA #0 */ case 0x13: /* FCMLA #90 */ case 0x15: /* FCMLA #180 */ diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) */ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) { - gen_helper_gvec_3_ptr *fn_gvec_ptr; - int rd, rn, rm, rot, size, opr_sz; - TCGv_ptr fpst; + gen_helper_gvec_3 *fn_gvec = NULL; + gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; + int rd, rn, rm, opr_sz; + int data = 0; bool q; q = extract32(insn, 6, 1); @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) if ((insn & 0xfe200f10) == 0xfc200800) { /* VCMLA -- 1111 110R R.1S .... .... 1000 ...0 .... */ - size = extract32(insn, 20, 1); - rot = extract32(insn, 23, 2); + int size = extract32(insn, 20, 1); + data = extract32(insn, 23, 2); /* rot */ if (!arm_dc_feature(s, ARM_FEATURE_V8_FCMA) || (!size && !arm_dc_feature(s, ARM_FEATURE_V8_FP16))) { return 1; @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) fn_gvec_ptr = size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; } else if ((insn & 0xfea00f10) == 0xfc800800) { /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ - size = extract32(insn, 20, 1); - rot = extract32(insn, 24, 1); + int size = extract32(insn, 20, 1); + data = extract32(insn, 24, 1); /* rot */ if (!arm_dc_feature(s, ARM_FEATURE_V8_FCMA) || (!size && !arm_dc_feature(s, ARM_FEATURE_V8_FP16))) { return 1; } fn_gvec_ptr = size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; + } else if ((insn & 0xfeb00f00) == 0xfc200d00) { + /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ + bool u = extract32(insn, 4, 1); + if (!arm_dc_feature(s, ARM_FEATURE_V8_DOTPROD)) { + return 1; + } + fn_gvec = u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; } else { return 1; } @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) } opr_sz = (1 + q) * 8; - fpst = get_fpstatus_ptr(1); - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), - vfp_reg_offset(1, rn), - vfp_reg_offset(1, rm), fpst, - opr_sz, opr_sz, rot, fn_gvec_ptr); - tcg_temp_free_ptr(fpst); + if (fn_gvec_ptr) { + TCGv_ptr fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), + vfp_reg_offset(1, rn), + vfp_reg_offset(1, rm), fpst, + opr_sz, opr_sz, data, fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + } else { + tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), + vfp_reg_offset(1, rn), + vfp_reg_offset(1, rm), + opr_sz, opr_sz, data, fn_gvec); + } return 0; } @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) { - gen_helper_gvec_3_ptr *fn_gvec_ptr; + gen_helper_gvec_3 *fn_gvec = NULL; + gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; int rd, rn, rm, opr_sz, data; - TCGv_ptr fpst; bool q; q = extract32(insn, 6, 1); @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) data = (index << 2) | rot; fn_gvec_ptr = (size ? gen_helper_gvec_fcmlas_idx : gen_helper_gvec_fcmlah_idx); + } else if ((insn & 0xffb00f00) == 0xfe200d00) { + /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ + int u = extract32(insn, 4, 1); + if (!arm_dc_feature(s, ARM_FEATURE_V8_DOTPROD)) { + return 1; + } + fn_gvec = u ? gen_helper_gvec_udot_idx_b : gen_helper_gvec_sdot_idx_b; + /* rm is just Vm, and index is M. */ + data = extract32(insn, 5, 1); /* index */ + rm = extract32(insn, 0, 4); } else { return 1; } @@ -XXX,XX +XXX,XX @@ static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) } opr_sz = (1 + q) * 8; - fpst = get_fpstatus_ptr(1); - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), - vfp_reg_offset(1, rn), - vfp_reg_offset(1, rm), fpst, - opr_sz, opr_sz, data, fn_gvec_ptr); - tcg_temp_free_ptr(fpst); + if (fn_gvec_ptr) { + TCGv_ptr fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), + vfp_reg_offset(1, rn), + vfp_reg_offset(1, rm), fpst, + opr_sz, opr_sz, data, fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + } else { + tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), + vfp_reg_offset(1, rn), + vfp_reg_offset(1, rm), + opr_sz, opr_sz, data, fn_gvec); + } return 0; } -- 2.17.1 From: Alex Bennée <alex.bennee@linaro.org> Since kernel commit a86bd139f2 (arm64: arch_timer: Enable CNTVCT_EL0 trap..), released in kernel version v4.12, user-space has been able to read these system registers. As we can't use QEMUTimer's in linux-user mode we just directly call cpu_get_clock(). Signed-off-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180625160009.17437-2-alex.bennee@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 27 ++++++++++++++++++++++++--- 1 file changed, 24 insertions(+), 3 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo generic_timer_cp_reginfo[] = { }; #else -/* In user-mode none of the generic timer registers are accessible, - * and their implementation depends on QEMU_CLOCK_VIRTUAL and qdev gpio outputs, - * so instead just don't register any of them. + +/* In user-mode most of the generic timer registers are inaccessible + * however modern kernels (4.12+) allow access to cntvct_el0 */ + +static uint64_t gt_virt_cnt_read(CPUARMState *env, const ARMCPRegInfo *ri) +{ + /* Currently we have no support for QEMUTimer in linux-user so we + * can't call gt_get_countervalue(env), instead we directly + * call the lower level functions. + */ + return cpu_get_clock() / GTIMER_SCALE; +} + static const ARMCPRegInfo generic_timer_cp_reginfo[] = { + { .name = "CNTFRQ_EL0", .state = ARM_CP_STATE_AA64, + .opc0 = 3, .opc1 = 3, .crn = 14, .crm = 0, .opc2 = 0, + .type = ARM_CP_CONST, .access = PL0_R /* no PL1_RW in linux-user */, + .fieldoffset = offsetof(CPUARMState, cp15.c14_cntfrq), + .resetvalue = NANOSECONDS_PER_SECOND / GTIMER_SCALE, + }, + { .name = "CNTVCT_EL0", .state = ARM_CP_STATE_AA64, + .opc0 = 3, .opc1 = 3, .crn = 14, .crm = 0, .opc2 = 2, + .access = PL0_R, .type = ARM_CP_NO_RAW | ARM_CP_IO, + .readfn = gt_virt_cnt_read, + }, REGINFO_SENTINEL }; -- 2.17.1 From: Jean-Christophe Dubois <jcd@tribudubois.net> Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/mcimx7d-sabre.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/hw/arm/mcimx7d-sabre.c b/hw/arm/mcimx7d-sabre.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/mcimx7d-sabre.c +++ b/hw/arm/mcimx7d-sabre.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/fsl-imx7.h" #include "hw/boards.h" #include "sysemu/sysemu.h" -#include "sysemu/device_tree.h" #include "qemu/error-report.h" #include "sysemu/qtest.h" -#include "net/net.h" typedef struct { FslIMX7State soc; -- 2.17.1 From: Jean-Christophe Dubois <jcd@tribudubois.net> Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/fsl-imx7.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/fsl-imx7.c b/hw/arm/fsl-imx7.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/fsl-imx7.c +++ b/hw/arm/fsl-imx7.c @@ -XXX,XX +XXX,XX @@ static void fsl_imx7_realize(DeviceState *dev, Error **errp) /* * SRC */ - create_unimplemented_device("sdma", FSL_IMX7_SRC_ADDR, FSL_IMX7_SRC_SIZE); + create_unimplemented_device("src", FSL_IMX7_SRC_ADDR, FSL_IMX7_SRC_SIZE); /* * Watchdog -- 2.17.1 From: Jean-Christophe Dubois <jcd@tribudubois.net> The qdev_get_gpio_in() function accept an int as second parameter. Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/fsl-imx7.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/hw/arm/fsl-imx7.c b/hw/arm/fsl-imx7.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/fsl-imx7.c +++ b/hw/arm/fsl-imx7.c @@ -XXX,XX +XXX,XX @@ static void fsl_imx7_realize(DeviceState *dev, Error **errp) FSL_IMX7_ECSPI4_ADDR, }; - static const hwaddr FSL_IMX7_SPIn_IRQ[FSL_IMX7_NUM_ECSPIS] = { + static const int FSL_IMX7_SPIn_IRQ[FSL_IMX7_NUM_ECSPIS] = { FSL_IMX7_ECSPI1_IRQ, FSL_IMX7_ECSPI2_IRQ, FSL_IMX7_ECSPI3_IRQ, @@ -XXX,XX +XXX,XX @@ static void fsl_imx7_realize(DeviceState *dev, Error **errp) FSL_IMX7_I2C4_ADDR, }; - static const hwaddr FSL_IMX7_I2Cn_IRQ[FSL_IMX7_NUM_I2CS] = { + static const int FSL_IMX7_I2Cn_IRQ[FSL_IMX7_NUM_I2CS] = { FSL_IMX7_I2C1_IRQ, FSL_IMX7_I2C2_IRQ, FSL_IMX7_I2C3_IRQ, @@ -XXX,XX +XXX,XX @@ static void fsl_imx7_realize(DeviceState *dev, Error **errp) FSL_IMX7_USB3_ADDR, }; - static const hwaddr FSL_IMX7_USBn_IRQ[FSL_IMX7_NUM_USBS] = { + static const int FSL_IMX7_USBn_IRQ[FSL_IMX7_NUM_USBS] = { FSL_IMX7_USB1_IRQ, FSL_IMX7_USB2_IRQ, FSL_IMX7_USB3_IRQ, -- 2.17.1 From: Aaron Lindsay <alindsay@codeaurora.org> Signed-off-by: Aaron Lindsay <alindsay@codeaurora.org> Message-id: 1529699547-17044-5-git-send-email-alindsay@codeaurora.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 1 + target/arm/cpu.c | 21 ++++++++++++++------- target/arm/kvm32.c | 8 ++++---- 3 files changed, 19 insertions(+), 11 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ enum arm_features { ARM_FEATURE_OMAPCP, /* OMAP specific CP15 ops handling. */ ARM_FEATURE_THUMB2EE, ARM_FEATURE_V7MP, /* v7 Multiprocessing Extensions */ + ARM_FEATURE_V7VE, /* v7 Virtualization Extensions (non-EL2 parts) */ ARM_FEATURE_V4T, ARM_FEATURE_V5, ARM_FEATURE_STRONGARM, diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_realizefn(DeviceState *dev, Error **errp) /* Some features automatically imply others: */ if (arm_feature(env, ARM_FEATURE_V8)) { - set_feature(env, ARM_FEATURE_V7); + set_feature(env, ARM_FEATURE_V7VE); + } + if (arm_feature(env, ARM_FEATURE_V7VE)) { + /* v7 Virtualization Extensions. In real hardware this implies + * EL2 and also the presence of the Security Extensions. + * For QEMU, for backwards-compatibility we implement some + * CPUs or CPU configs which have no actual EL2 or EL3 but do + * include the various other features that V7VE implies. + * Presence of EL2 itself is ARM_FEATURE_EL2, and of the + * Security Extensions is ARM_FEATURE_EL3. + */ set_feature(env, ARM_FEATURE_ARM_DIV); set_feature(env, ARM_FEATURE_LPAE); + set_feature(env, ARM_FEATURE_V7); } if (arm_feature(env, ARM_FEATURE_V7)) { set_feature(env, ARM_FEATURE_VAPA); @@ -XXX,XX +XXX,XX @@ static void cortex_a7_initfn(Object *obj) ARMCPU *cpu = ARM_CPU(obj); cpu->dtb_compatible = "arm,cortex-a7"; - set_feature(&cpu->env, ARM_FEATURE_V7); + set_feature(&cpu->env, ARM_FEATURE_V7VE); set_feature(&cpu->env, ARM_FEATURE_VFP4); set_feature(&cpu->env, ARM_FEATURE_NEON); set_feature(&cpu->env, ARM_FEATURE_THUMB2EE); - set_feature(&cpu->env, ARM_FEATURE_ARM_DIV); set_feature(&cpu->env, ARM_FEATURE_GENERIC_TIMER); set_feature(&cpu->env, ARM_FEATURE_DUMMY_C15_REGS); set_feature(&cpu->env, ARM_FEATURE_CBAR_RO); - set_feature(&cpu->env, ARM_FEATURE_LPAE); set_feature(&cpu->env, ARM_FEATURE_EL3); cpu->kvm_target = QEMU_KVM_ARM_TARGET_CORTEX_A7; cpu->midr = 0x410fc075; @@ -XXX,XX +XXX,XX @@ static void cortex_a15_initfn(Object *obj) ARMCPU *cpu = ARM_CPU(obj); cpu->dtb_compatible = "arm,cortex-a15"; - set_feature(&cpu->env, ARM_FEATURE_V7); + set_feature(&cpu->env, ARM_FEATURE_V7VE); set_feature(&cpu->env, ARM_FEATURE_VFP4); set_feature(&cpu->env, ARM_FEATURE_NEON); set_feature(&cpu->env, ARM_FEATURE_THUMB2EE); - set_feature(&cpu->env, ARM_FEATURE_ARM_DIV); set_feature(&cpu->env, ARM_FEATURE_GENERIC_TIMER); set_feature(&cpu->env, ARM_FEATURE_DUMMY_C15_REGS); set_feature(&cpu->env, ARM_FEATURE_CBAR_RO); - set_feature(&cpu->env, ARM_FEATURE_LPAE); set_feature(&cpu->env, ARM_FEATURE_EL3); cpu->kvm_target = QEMU_KVM_ARM_TARGET_CORTEX_A15; cpu->midr = 0x412fc0f1; diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm32.c +++ b/target/arm/kvm32.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) /* Now we've retrieved all the register information we can * set the feature bits based on the ID register fields. * We can assume any KVM supporting CPU is at least a v7 - * with VFPv3, LPAE and the generic timers; this in turn implies - * most of the other feature bits, but a few must be tested. + * with VFPv3, virtualization extensions, and the generic + * timers; this in turn implies most of the other feature + * bits, but a few must be tested. */ - set_feature(&features, ARM_FEATURE_V7); + set_feature(&features, ARM_FEATURE_V7VE); set_feature(&features, ARM_FEATURE_VFP3); - set_feature(&features, ARM_FEATURE_LPAE); set_feature(&features, ARM_FEATURE_GENERIC_TIMER); switch (extract32(id_isar0, 24, 4)) { -- 2.17.1 From: Aaron Lindsay <alindsay@codeaurora.org> KVM implies V7VE, which implies ARM_DIV and THUMB_DIV. The conditional detection here is therefore unnecessary. Because V7VE is already unconditionally specified for all KVM hosts, ARM_DIV and THUMB_DIV are already indirectly specified and do not need to be included here at all. Signed-off-by: Aaron Lindsay <alindsay@codeaurora.org> Message-id: 1529699547-17044-6-git-send-email-alindsay@codeaurora.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/kvm32.c | 19 +------------------ 1 file changed, 1 insertion(+), 18 deletions(-) diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm32.c +++ b/target/arm/kvm32.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) * and then query that CPU for the relevant ID registers. */ int i, ret, fdarray[3]; - uint32_t midr, id_pfr0, id_isar0, mvfr1; + uint32_t midr, id_pfr0, mvfr1; uint64_t features = 0; /* Old kernels may not know about the PREFERRED_TARGET ioctl: however * we know these will only support creating one kind of guest CPU, @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) | ENCODE_CP_REG(15, 0, 0, 0, 1, 0, 0), .addr = (uintptr_t)&id_pfr0, }, - { - .id = KVM_REG_ARM | KVM_REG_SIZE_U32 - | ENCODE_CP_REG(15, 0, 0, 0, 2, 0, 0), - .addr = (uintptr_t)&id_isar0, - }, { .id = KVM_REG_ARM | KVM_REG_SIZE_U32 | KVM_REG_ARM_VFP | KVM_REG_ARM_VFP_MVFR1, @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) set_feature(&features, ARM_FEATURE_VFP3); set_feature(&features, ARM_FEATURE_GENERIC_TIMER); - switch (extract32(id_isar0, 24, 4)) { - case 1: - set_feature(&features, ARM_FEATURE_THUMB_DIV); - break; - case 2: - set_feature(&features, ARM_FEATURE_ARM_DIV); - set_feature(&features, ARM_FEATURE_THUMB_DIV); - break; - default: - break; - } - if (extract32(id_pfr0, 12, 4) == 1) { set_feature(&features, ARM_FEATURE_THUMB2EE); } -- 2.17.1 From: Aaron Lindsay <alindsay@codeaurora.org> This makes it match its AArch64 equivalent, PMINTENSET_EL1 Signed-off-by: Aaron Lindsay <alindsay@codeaurora.org> Message-id: 1529699547-17044-13-git-send-email-alindsay@codeaurora.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v7_cp_reginfo[] = { .writefn = pmuserenr_write, .raw_writefn = raw_write }, { .name = "PMINTENSET", .cp = 15, .crn = 9, .crm = 14, .opc1 = 0, .opc2 = 1, .access = PL1_RW, .accessfn = access_tpm, - .type = ARM_CP_ALIAS, + .type = ARM_CP_ALIAS | ARM_CP_IO, .fieldoffset = offsetoflow32(CPUARMState, cp15.c9_pminten), .resetvalue = 0, .writefn = pmintenset_write, .raw_writefn = raw_write }, -- 2.17.1 We don't actually implement SD command CRC checking, because for almost all of our SD controllers the CRC generation is done in hardware, and so modelling CRC generation and checking would be a bit pointless. (The exception is that milkymist-memcard makes the guest software compute the CRC.) As a result almost all of our SD controller models don't bother to set the SDRequest crc field, and the SD card model doesn't check it. So the tracing of it in sdbus_do_command() provokes Coverity warnings about use of uninitialized data. Drop the CRC field from the trace; we can always add it back if and when we do anything useful with the CRC. Fixes Coverity issues 1386072, 1386074, 1386076, 1390571. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180626180324.5537-1-peter.maydell@linaro.org --- hw/sd/core.c | 2 +- hw/sd/trace-events | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/hw/sd/core.c b/hw/sd/core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/core.c +++ b/hw/sd/core.c @@ -XXX,XX +XXX,XX @@ int sdbus_do_command(SDBus *sdbus, SDRequest *req, uint8_t *response) { SDState *card = get_card(sdbus); - trace_sdbus_command(sdbus_name(sdbus), req->cmd, req->arg, req->crc); + trace_sdbus_command(sdbus_name(sdbus), req->cmd, req->arg); if (card) { SDCardClass *sc = SD_CARD_GET_CLASS(card); diff --git a/hw/sd/trace-events b/hw/sd/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/trace-events +++ b/hw/sd/trace-events @@ -XXX,XX +XXX,XX @@ bcm2835_sdhost_edm_change(const char *why, uint32_t edm) "(%s) EDM now 0x%x" bcm2835_sdhost_update_irq(uint32_t irq) "IRQ bits 0x%x\n" # hw/sd/core.c -sdbus_command(const char *bus_name, uint8_t cmd, uint32_t arg, uint8_t crc) "@%s CMD%02d arg 0x%08x crc 0x%02x" +sdbus_command(const char *bus_name, uint8_t cmd, uint32_t arg) "@%s CMD%02d arg 0x%08x" sdbus_read(const char *bus_name, uint8_t value) "@%s value 0x%02x" sdbus_write(const char *bus_name, uint8_t value) "@%s value 0x%02x" sdbus_set_voltage(const char *bus_name, uint16_t millivolts) "@%s %u (mV)" -- 2.17.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> The load/store API will ease further code movement. Per the Physical Layer Simplified Spec. "3.6 Bus Protocol": "In the CMD line the Most Significant Bit (MSB) is transmitted first, the Least Significant Bit (LSB) is the last." Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/sd/bcm2835_sdhost.c | 13 +++++-------- hw/sd/milkymist-memcard.c | 3 +-- hw/sd/omap_mmc.c | 6 ++---- hw/sd/pl181.c | 11 ++++------- hw/sd/sdhci.c | 15 +++++---------- hw/sd/ssi-sd.c | 6 ++---- 6 files changed, 19 insertions(+), 35 deletions(-) diff --git a/hw/sd/bcm2835_sdhost.c b/hw/sd/bcm2835_sdhost.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/bcm2835_sdhost.c +++ b/hw/sd/bcm2835_sdhost.c @@ -XXX,XX +XXX,XX @@ static void bcm2835_sdhost_send_command(BCM2835SDHostState *s) goto error; } if (!(s->cmd & SDCMD_NO_RESPONSE)) { -#define RWORD(n) (((uint32_t)rsp[n] << 24) | (rsp[n + 1] << 16) \ - | (rsp[n + 2] << 8) | rsp[n + 3]) if (rlen == 0 || (rlen == 4 && (s->cmd & SDCMD_LONG_RESPONSE))) { goto error; } @@ -XXX,XX +XXX,XX @@ static void bcm2835_sdhost_send_command(BCM2835SDHostState *s) goto error; } if (rlen == 4) { - s->rsp[0] = RWORD(0); + s->rsp[0] = ldl_be_p(&rsp[0]); s->rsp[1] = s->rsp[2] = s->rsp[3] = 0; } else { - s->rsp[0] = RWORD(12); - s->rsp[1] = RWORD(8); - s->rsp[2] = RWORD(4); - s->rsp[3] = RWORD(0); + s->rsp[0] = ldl_be_p(&rsp[12]); + s->rsp[1] = ldl_be_p(&rsp[8]); + s->rsp[2] = ldl_be_p(&rsp[4]); + s->rsp[3] = ldl_be_p(&rsp[0]); } -#undef RWORD } /* We never really delay commands, so if this was a 'busywait' command * then we've completed it now and can raise the interrupt. diff --git a/hw/sd/milkymist-memcard.c b/hw/sd/milkymist-memcard.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/milkymist-memcard.c +++ b/hw/sd/milkymist-memcard.c @@ -XXX,XX +XXX,XX @@ static void memcard_sd_command(MilkymistMemcardState *s) SDRequest req; req.cmd = s->command[0] & 0x3f; - req.arg = (s->command[1] << 24) | (s->command[2] << 16) - | (s->command[3] << 8) | s->command[4]; + req.arg = ldl_be_p(s->command + 1); req.crc = s->command[5]; s->response[0] = req.cmd; diff --git a/hw/sd/omap_mmc.c b/hw/sd/omap_mmc.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/omap_mmc.c +++ b/hw/sd/omap_mmc.c @@ -XXX,XX +XXX,XX @@ static void omap_mmc_command(struct omap_mmc_s *host, int cmd, int dir, CID_CSD_OVERWRITE; if (host->sdio & (1 << 13)) mask |= AKE_SEQ_ERROR; - rspstatus = (response[0] << 24) | (response[1] << 16) | - (response[2] << 8) | (response[3] << 0); + rspstatus = ldl_be_p(response); break; case sd_r2: @@ -XXX,XX +XXX,XX @@ static void omap_mmc_command(struct omap_mmc_s *host, int cmd, int dir, } rsplen = 4; - rspstatus = (response[0] << 24) | (response[1] << 16) | - (response[2] << 8) | (response[3] << 0); + rspstatus = ldl_be_p(response); if (rspstatus & 0x80000000) host->status &= 0xe000; else diff --git a/hw/sd/pl181.c b/hw/sd/pl181.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/pl181.c +++ b/hw/sd/pl181.c @@ -XXX,XX +XXX,XX @@ static void pl181_send_command(PL181State *s) if (rlen < 0) goto error; if (s->cmd & PL181_CMD_RESPONSE) { -#define RWORD(n) (((uint32_t)response[n] << 24) | (response[n + 1] << 16) \ - | (response[n + 2] << 8) | response[n + 3]) if (rlen == 0 || (rlen == 4 && (s->cmd & PL181_CMD_LONGRESP))) goto error; if (rlen != 4 && rlen != 16) goto error; - s->response[0] = RWORD(0); + s->response[0] = ldl_be_p(&response[0]); if (rlen == 4) { s->response[1] = s->response[2] = s->response[3] = 0; } else { - s->response[1] = RWORD(4); - s->response[2] = RWORD(8); - s->response[3] = RWORD(12) & ~1; + s->response[1] = ldl_be_p(&response[4]); + s->response[2] = ldl_be_p(&response[8]); + s->response[3] = ldl_be_p(&response[12]) & ~1; } DPRINTF("Response received\n"); s->status |= PL181_STATUS_CMDRESPEND; -#undef RWORD } else { DPRINTF("Command sent\n"); s->status |= PL181_STATUS_CMDSENT; diff --git a/hw/sd/sdhci.c b/hw/sd/sdhci.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/sdhci.c +++ b/hw/sd/sdhci.c @@ -XXX,XX +XXX,XX @@ static void sdhci_send_command(SDHCIState *s) if (s->cmdreg & SDHC_CMD_RESPONSE) { if (rlen == 4) { - s->rspreg[0] = (response[0] << 24) | (response[1] << 16) | - (response[2] << 8) | response[3]; + s->rspreg[0] = ldl_be_p(response); s->rspreg[1] = s->rspreg[2] = s->rspreg[3] = 0; trace_sdhci_response4(s->rspreg[0]); } else if (rlen == 16) { - s->rspreg[0] = (response[11] << 24) | (response[12] << 16) | - (response[13] << 8) | response[14]; - s->rspreg[1] = (response[7] << 24) | (response[8] << 16) | - (response[9] << 8) | response[10]; - s->rspreg[2] = (response[3] << 24) | (response[4] << 16) | - (response[5] << 8) | response[6]; + s->rspreg[0] = ldl_be_p(&response[11]); + s->rspreg[1] = ldl_be_p(&response[7]); + s->rspreg[2] = ldl_be_p(&response[3]); s->rspreg[3] = (response[0] << 16) | (response[1] << 8) | response[2]; trace_sdhci_response16(s->rspreg[3], s->rspreg[2], @@ -XXX,XX +XXX,XX @@ static void sdhci_end_transfer(SDHCIState *s) trace_sdhci_end_transfer(request.cmd, request.arg); sdbus_do_command(&s->sdbus, &request, response); /* Auto CMD12 response goes to the upper Response register */ - s->rspreg[3] = (response[0] << 24) | (response[1] << 16) | - (response[2] << 8) | response[3]; + s->rspreg[3] = ldl_be_p(response); } s->prnsts &= ~(SDHC_DOING_READ | SDHC_DOING_WRITE | diff --git a/hw/sd/ssi-sd.c b/hw/sd/ssi-sd.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sd/ssi-sd.c +++ b/hw/sd/ssi-sd.c @@ -XXX,XX +XXX,XX @@ static uint32_t ssi_sd_transfer(SSISlave *dev, uint32_t val) uint8_t longresp[16]; /* FIXME: Check CRC. */ request.cmd = s->cmd; - request.arg = (s->cmdarg[0] << 24) | (s->cmdarg[1] << 16) - | (s->cmdarg[2] << 8) | s->cmdarg[3]; + request.arg = ldl_be_p(s->cmdarg); DPRINTF("CMD%d arg 0x%08x\n", s->cmd, request.arg); s->arglen = sdbus_do_command(&s->sdbus, &request, longresp); if (s->arglen <= 0) { @@ -XXX,XX +XXX,XX @@ static uint32_t ssi_sd_transfer(SSISlave *dev, uint32_t val) /* CMD13 returns a 2-byte statuse work. Other commands only return the first byte. */ s->arglen = (s->cmd == 13) ? 2 : 1; - cardstatus = (longresp[0] << 24) | (longresp[1] << 16) - | (longresp[2] << 8) | longresp[3]; + cardstatus = ldl_be_p(longresp); status = 0; if (((cardstatus >> 9) & 0xf) < 4) status |= SSI_SDR_IDLE; -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> We already check for the same condition within the normal integer sdiv and sdiv64 helpers. Use a slightly different formation that does not require deducing the expression type. Fixes: f97cfd596ed Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20180629001538.11415-2-richard.henderson@linaro.org [PMM: reworded a comment] Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sve_helper.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-) diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) \ #define DO_MIN(N, M) ((N) >= (M) ? (M) : (N)) #define DO_ABD(N, M) ((N) >= (M) ? (N) - (M) : (M) - (N)) #define DO_MUL(N, M) (N * M) -#define DO_DIV(N, M) (M ? N / M : 0) + + +/* + * We must avoid the C undefined behaviour cases: division by + * zero and signed division of INT_MIN by -1. Both of these + * have architecturally defined required results for Arm. + * We special case all signed divisions by -1 to avoid having + * to deduce the minimum integer for the type involved. + */ +#define DO_SDIV(N, M) (unlikely(M == 0) ? 0 : unlikely(M == -1) ? -N : N / M) +#define DO_UDIV(N, M) (unlikely(M == 0) ? 0 : N / M) DO_ZPZZ(sve_and_zpzz_b, uint8_t, H1, DO_AND) DO_ZPZZ(sve_and_zpzz_h, uint16_t, H1_2, DO_AND) @@ -XXX,XX +XXX,XX @@ DO_ZPZZ(sve_umulh_zpzz_h, uint16_t, H1_2, do_mulh_h) DO_ZPZZ(sve_umulh_zpzz_s, uint32_t, H1_4, do_mulh_s) DO_ZPZZ_D(sve_umulh_zpzz_d, uint64_t, do_umulh_d) -DO_ZPZZ(sve_sdiv_zpzz_s, int32_t, H1_4, DO_DIV) -DO_ZPZZ_D(sve_sdiv_zpzz_d, int64_t, DO_DIV) +DO_ZPZZ(sve_sdiv_zpzz_s, int32_t, H1_4, DO_SDIV) +DO_ZPZZ_D(sve_sdiv_zpzz_d, int64_t, DO_SDIV) -DO_ZPZZ(sve_udiv_zpzz_s, uint32_t, H1_4, DO_DIV) -DO_ZPZZ_D(sve_udiv_zpzz_d, uint64_t, DO_DIV) +DO_ZPZZ(sve_udiv_zpzz_s, uint32_t, H1_4, DO_UDIV) +DO_ZPZZ_D(sve_udiv_zpzz_d, uint64_t, DO_UDIV) /* Note that all bits of the shift are significant and not modulo the element size. */ -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Leave ARM_CP_SVE, removing ARM_CP_FPU; the sve_access_check produced by the flag already includes fp_access_check. If we also check ARM_CP_FPU the double fp_access_check asserts. Reported-by: Laurent Desnogues <laurent.desnogues@gmail.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Laurent Desnogues <laurent.desnogues@gmail.com> Message-id: 20180629001538.11415-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 8 ++++---- target/arm/translate-a64.c | 5 ++--- 2 files changed, 6 insertions(+), 7 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static void zcr_write(CPUARMState *env, const ARMCPRegInfo *ri, static const ARMCPRegInfo zcr_el1_reginfo = { .name = "ZCR_EL1", .state = ARM_CP_STATE_AA64, .opc0 = 3, .opc1 = 0, .crn = 1, .crm = 2, .opc2 = 0, - .access = PL1_RW, .type = ARM_CP_SVE | ARM_CP_FPU, + .access = PL1_RW, .type = ARM_CP_SVE, .fieldoffset = offsetof(CPUARMState, vfp.zcr_el[1]), .writefn = zcr_write, .raw_writefn = raw_write }; @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo zcr_el1_reginfo = { static const ARMCPRegInfo zcr_el2_reginfo = { .name = "ZCR_EL2", .state = ARM_CP_STATE_AA64, .opc0 = 3, .opc1 = 4, .crn = 1, .crm = 2, .opc2 = 0, - .access = PL2_RW, .type = ARM_CP_SVE | ARM_CP_FPU, + .access = PL2_RW, .type = ARM_CP_SVE, .fieldoffset = offsetof(CPUARMState, vfp.zcr_el[2]), .writefn = zcr_write, .raw_writefn = raw_write }; @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo zcr_el2_reginfo = { static const ARMCPRegInfo zcr_no_el2_reginfo = { .name = "ZCR_EL2", .state = ARM_CP_STATE_AA64, .opc0 = 3, .opc1 = 4, .crn = 1, .crm = 2, .opc2 = 0, - .access = PL2_RW, .type = ARM_CP_SVE | ARM_CP_FPU, + .access = PL2_RW, .type = ARM_CP_SVE, .readfn = arm_cp_read_zero, .writefn = arm_cp_write_ignore }; static const ARMCPRegInfo zcr_el3_reginfo = { .name = "ZCR_EL3", .state = ARM_CP_STATE_AA64, .opc0 = 3, .opc1 = 6, .crn = 1, .crm = 2, .opc2 = 0, - .access = PL3_RW, .type = ARM_CP_SVE | ARM_CP_FPU, + .access = PL3_RW, .type = ARM_CP_SVE, .fieldoffset = offsetof(CPUARMState, vfp.zcr_el[3]), .writefn = zcr_write, .raw_writefn = raw_write }; diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread, default: break; } - if ((ri->type & ARM_CP_SVE) && !sve_access_check(s)) { - return; - } if ((ri->type & ARM_CP_FPU) && !fp_access_check(s)) { return; + } else if ((ri->type & ARM_CP_SVE) && !sve_access_check(s)) { + return; } if ((tb_cflags(s->base.tb) & CF_USE_ICOUNT) && (ri->type & ARM_CP_IO)) { -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> There is no need to re-set these 9 features already implied by the call to aarch64_a57_initfn. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20180629001538.11415-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu64.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) * whereas the architecture requires them to be present in both if * present in either. */ - set_feature(&cpu->env, ARM_FEATURE_V8); - set_feature(&cpu->env, ARM_FEATURE_VFP4); - set_feature(&cpu->env, ARM_FEATURE_NEON); - set_feature(&cpu->env, ARM_FEATURE_AARCH64); - set_feature(&cpu->env, ARM_FEATURE_V8_AES); - set_feature(&cpu->env, ARM_FEATURE_V8_SHA1); - set_feature(&cpu->env, ARM_FEATURE_V8_SHA256); set_feature(&cpu->env, ARM_FEATURE_V8_SHA512); set_feature(&cpu->env, ARM_FEATURE_V8_SHA3); set_feature(&cpu->env, ARM_FEATURE_V8_SM3); set_feature(&cpu->env, ARM_FEATURE_V8_SM4); - set_feature(&cpu->env, ARM_FEATURE_V8_PMULL); - set_feature(&cpu->env, ARM_FEATURE_CRC); set_feature(&cpu->env, ARM_FEATURE_V8_ATOMICS); set_feature(&cpu->env, ARM_FEATURE_V8_RDM); set_feature(&cpu->env, ARM_FEATURE_V8_DOTPROD); -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> There is no need to re-set these 3 features already implied by the call to aarch64_a15_initfn. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20180629001538.11415-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_max_initfn(Object *obj) * since we don't correctly set the ID registers to advertise them, */ set_feature(&cpu->env, ARM_FEATURE_V8); - set_feature(&cpu->env, ARM_FEATURE_VFP4); - set_feature(&cpu->env, ARM_FEATURE_NEON); - set_feature(&cpu->env, ARM_FEATURE_THUMB2EE); set_feature(&cpu->env, ARM_FEATURE_V8_AES); set_feature(&cpu->env, ARM_FEATURE_V8_SHA1); set_feature(&cpu->env, ARM_FEATURE_V8_SHA256); -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> This register was added to aa32 state by ARMv8.2. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20180629001538.11415-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 1 + target/arm/cpu.c | 4 ++++ target/arm/cpu64.c | 2 ++ target/arm/helper.c | 5 ++--- 4 files changed, 9 insertions(+), 3 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ struct ARMCPU { uint32_t id_isar3; uint32_t id_isar4; uint32_t id_isar5; + uint32_t id_isar6; uint64_t id_aa64pfr0; uint64_t id_aa64pfr1; uint64_t id_aa64dfr0; diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void cortex_m3_initfn(Object *obj) cpu->id_isar3 = 0x01111110; cpu->id_isar4 = 0x01310102; cpu->id_isar5 = 0x00000000; + cpu->id_isar6 = 0x00000000; } static void cortex_m4_initfn(Object *obj) @@ -XXX,XX +XXX,XX @@ static void cortex_m4_initfn(Object *obj) cpu->id_isar3 = 0x01111110; cpu->id_isar4 = 0x01310102; cpu->id_isar5 = 0x00000000; + cpu->id_isar6 = 0x00000000; } static void cortex_m33_initfn(Object *obj) @@ -XXX,XX +XXX,XX @@ static void cortex_m33_initfn(Object *obj) cpu->id_isar3 = 0x01111131; cpu->id_isar4 = 0x01310132; cpu->id_isar5 = 0x00000000; + cpu->id_isar6 = 0x00000000; cpu->clidr = 0x00000000; cpu->ctr = 0x8000c000; } @@ -XXX,XX +XXX,XX @@ static void cortex_r5_initfn(Object *obj) cpu->id_isar3 = 0x01112131; cpu->id_isar4 = 0x0010142; cpu->id_isar5 = 0x0; + cpu->id_isar6 = 0x0; cpu->mp_is_up = true; cpu->pmsav7_dregion = 16; define_arm_cp_regs(cpu, cortexr5_cp_reginfo); diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_a57_initfn(Object *obj) cpu->id_isar3 = 0x01112131; cpu->id_isar4 = 0x00011142; cpu->id_isar5 = 0x00011121; + cpu->id_isar6 = 0; cpu->id_aa64pfr0 = 0x00002222; cpu->id_aa64dfr0 = 0x10305106; cpu->pmceid0 = 0x00000000; @@ -XXX,XX +XXX,XX @@ static void aarch64_a53_initfn(Object *obj) cpu->id_isar3 = 0x01112131; cpu->id_isar4 = 0x00011142; cpu->id_isar5 = 0x00011121; + cpu->id_isar6 = 0; cpu->id_aa64pfr0 = 0x00002222; cpu->id_aa64dfr0 = 0x10305106; cpu->id_aa64isar0 = 0x00011120; diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ void register_cp_regs_for_features(ARMCPU *cpu) .opc0 = 3, .opc1 = 0, .crn = 0, .crm = 2, .opc2 = 6, .access = PL1_R, .type = ARM_CP_CONST, .resetvalue = cpu->id_mmfr4 }, - /* 7 is as yet unallocated and must RAZ */ - { .name = "ID_ISAR7_RESERVED", .state = ARM_CP_STATE_BOTH, + { .name = "ID_ISAR6", .state = ARM_CP_STATE_BOTH, .opc0 = 3, .opc1 = 0, .crn = 0, .crm = 2, .opc2 = 7, .access = PL1_R, .type = ARM_CP_CONST, - .resetvalue = 0 }, + .resetvalue = cpu->id_isar6 }, REGINFO_SENTINEL }; define_arm_cp_regs(cpu, v6_idregs); -- 2.17.1 target-arm queue: two bug fixes, plus the KVM/SVE patchset, which is a new feature but one which was in my pre-softfreeze pullreq (it just had to be dropped due to an unexpected test failure.) thanks -- PMM The following changes since commit b7c9a7f353c0e260519bf735ff0d4aa01e72784b: Merge remote-tracking branch 'remotes/jnsnow/tags/ide-pull-request' into staging (2019-10-31 15:57:30 +0000) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20191101-1 for you to fetch changes up to d9ae7624b659362cb2bb2b04fee53bf50829ca56: target/arm: Allow reading flags from FPSCR for M-profile (2019-11-01 08:49:10 +0000) ---------------------------------------------------------------- target-arm queue: * Support SVE in KVM guests * Don't UNDEF on M-profile 'vmrs apsr_nzcv, fpscr' * Update hflags after boot.c modifies CPU state ---------------------------------------------------------------- Andrew Jones (9): target/arm/monitor: Introduce qmp_query_cpu_model_expansion tests: arm: Introduce cpu feature tests target/arm: Allow SVE to be disabled via a CPU property target/arm/cpu64: max cpu: Introduce sve<N> properties target/arm/kvm64: Add kvm_arch_get/put_sve target/arm/kvm64: max cpu: Enable SVE when available target/arm/kvm: scratch vcpu: Preserve input kvm_vcpu_init features target/arm/cpu64: max cpu: Support sve properties with KVM target/arm/kvm: host cpu: Add support for sve<N> properties Christophe Lyon (1): target/arm: Allow reading flags from FPSCR for M-profile Edgar E. Iglesias (1): hw/arm/boot: Rebuild hflags when modifying CPUState at boot tests/Makefile.include | 5 +- qapi/machine-target.json | 6 +- include/qemu/bitops.h | 1 + target/arm/cpu.h | 21 ++ target/arm/kvm_arm.h | 39 +++ hw/arm/boot.c | 1 + target/arm/cpu.c | 25 +- target/arm/cpu64.c | 364 +++++++++++++++++++++++++-- target/arm/helper.c | 10 +- target/arm/kvm.c | 25 +- target/arm/kvm32.c | 6 +- target/arm/kvm64.c | 325 +++++++++++++++++++++--- target/arm/monitor.c | 158 ++++++++++++ target/arm/translate-vfp.inc.c | 5 +- tests/arm-cpu-features.c | 551 +++++++++++++++++++++++++++++++++++++++++ docs/arm-cpu-features.rst | 317 ++++++++++++++++++++++++ 16 files changed, 1795 insertions(+), 64 deletions(-) create mode 100644 tests/arm-cpu-features.c create mode 100644 docs/arm-cpu-features.rst From: Andrew Jones <drjones@redhat.com> Add support for the query-cpu-model-expansion QMP command to Arm. We do this selectively, only exposing CPU properties which represent optional CPU features which the user may want to enable/disable. Additionally we restrict the list of queryable cpu models to 'max', 'host', or the current type when KVM is in use. And, finally, we only implement expansion type 'full', as Arm does not yet have a "base" CPU type. More details and example queries are described in a new document (docs/arm-cpu-features.rst). Note, certainly more features may be added to the list of advertised features, e.g. 'vfp' and 'neon'. The only requirement is that we can detect invalid configurations and emit failures at QMP query time. For 'vfp' and 'neon' this will require some refactoring to share a validation function between the QMP query and the CPU realize functions. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Reviewed-by: Beata Michalska <beata.michalska@linaro.org> Message-id: 20191031142734.8590-2-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- qapi/machine-target.json | 6 +- target/arm/monitor.c | 146 ++++++++++++++++++++++++++++++++++++++ docs/arm-cpu-features.rst | 137 +++++++++++++++++++++++++++++++++++ 3 files changed, 286 insertions(+), 3 deletions(-) create mode 100644 docs/arm-cpu-features.rst diff --git a/qapi/machine-target.json b/qapi/machine-target.json index XXXXXXX..XXXXXXX 100644 --- a/qapi/machine-target.json +++ b/qapi/machine-target.json @@ -XXX,XX +XXX,XX @@ ## { 'struct': 'CpuModelExpansionInfo', 'data': { 'model': 'CpuModelInfo' }, - 'if': 'defined(TARGET_S390X) || defined(TARGET_I386)' } + 'if': 'defined(TARGET_S390X) || defined(TARGET_I386) || defined(TARGET_ARM)' } ## # @query-cpu-model-expansion: @@ -XXX,XX +XXX,XX @@ # query-cpu-model-expansion while using these is not advised. # # Some architectures may not support all expansion types. s390x supports -# "full" and "static". +# "full" and "static". Arm only supports "full". # # Returns: a CpuModelExpansionInfo. Returns an error if expanding CPU models is # not supported, if the model cannot be expanded, if the model contains @@ -XXX,XX +XXX,XX @@ 'data': { 'type': 'CpuModelExpansionType', 'model': 'CpuModelInfo' }, 'returns': 'CpuModelExpansionInfo', - 'if': 'defined(TARGET_S390X) || defined(TARGET_I386)' } + 'if': 'defined(TARGET_S390X) || defined(TARGET_I386) || defined(TARGET_ARM)' } ## # @CpuDefinitionInfo: diff --git a/target/arm/monitor.c b/target/arm/monitor.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/monitor.c +++ b/target/arm/monitor.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "hw/boards.h" #include "kvm_arm.h" +#include "qapi/error.h" +#include "qapi/visitor.h" +#include "qapi/qobject-input-visitor.h" +#include "qapi/qapi-commands-machine-target.h" #include "qapi/qapi-commands-misc-target.h" +#include "qapi/qmp/qerror.h" +#include "qapi/qmp/qdict.h" +#include "qom/qom-qobject.h" static GICCapability *gic_cap_new(int version) { @@ -XXX,XX +XXX,XX @@ GICCapabilityList *qmp_query_gic_capabilities(Error **errp) return head; } + +/* + * These are cpu model features we want to advertise. The order here + * matters as this is the order in which qmp_query_cpu_model_expansion + * will attempt to set them. If there are dependencies between features, + * then the order that considers those dependencies must be used. + */ +static const char *cpu_model_advertised_features[] = { + "aarch64", "pmu", + NULL +}; + +CpuModelExpansionInfo *qmp_query_cpu_model_expansion(CpuModelExpansionType type, + CpuModelInfo *model, + Error **errp) +{ + CpuModelExpansionInfo *expansion_info; + const QDict *qdict_in = NULL; + QDict *qdict_out; + ObjectClass *oc; + Object *obj; + const char *name; + int i; + + if (type != CPU_MODEL_EXPANSION_TYPE_FULL) { + error_setg(errp, "The requested expansion type is not supported"); + return NULL; + } + + if (!kvm_enabled() && !strcmp(model->name, "host")) { + error_setg(errp, "The CPU type '%s' requires KVM", model->name); + return NULL; + } + + oc = cpu_class_by_name(TYPE_ARM_CPU, model->name); + if (!oc) { + error_setg(errp, "The CPU type '%s' is not a recognized ARM CPU type", + model->name); + return NULL; + } + + if (kvm_enabled()) { + const char *cpu_type = current_machine->cpu_type; + int len = strlen(cpu_type) - strlen(ARM_CPU_TYPE_SUFFIX); + bool supported = false; + + if (!strcmp(model->name, "host") || !strcmp(model->name, "max")) { + /* These are kvmarm's recommended cpu types */ + supported = true; + } else if (strlen(model->name) == len && + !strncmp(model->name, cpu_type, len)) { + /* KVM is enabled and we're using this type, so it works. */ + supported = true; + } + if (!supported) { + error_setg(errp, "We cannot guarantee the CPU type '%s' works " + "with KVM on this host", model->name); + return NULL; + } + } + + if (model->props) { + qdict_in = qobject_to(QDict, model->props); + if (!qdict_in) { + error_setg(errp, QERR_INVALID_PARAMETER_TYPE, "props", "dict"); + return NULL; + } + } + + obj = object_new(object_class_get_name(oc)); + + if (qdict_in) { + Visitor *visitor; + Error *err = NULL; + + visitor = qobject_input_visitor_new(model->props); + visit_start_struct(visitor, NULL, NULL, 0, &err); + if (err) { + visit_free(visitor); + object_unref(obj); + error_propagate(errp, err); + return NULL; + } + + i = 0; + while ((name = cpu_model_advertised_features[i++]) != NULL) { + if (qdict_get(qdict_in, name)) { + object_property_set(obj, visitor, name, &err); + if (err) { + break; + } + } + } + + if (!err) { + visit_check_struct(visitor, &err); + } + visit_end_struct(visitor, NULL); + visit_free(visitor); + if (err) { + object_unref(obj); + error_propagate(errp, err); + return NULL; + } + } + + expansion_info = g_new0(CpuModelExpansionInfo, 1); + expansion_info->model = g_malloc0(sizeof(*expansion_info->model)); + expansion_info->model->name = g_strdup(model->name); + + qdict_out = qdict_new(); + + i = 0; + while ((name = cpu_model_advertised_features[i++]) != NULL) { + ObjectProperty *prop = object_property_find(obj, name, NULL); + if (prop) { + Error *err = NULL; + QObject *value; + + assert(prop->get); + value = object_property_get_qobject(obj, name, &err); + assert(!err); + + qdict_put_obj(qdict_out, name, value); + } + } + + if (!qdict_size(qdict_out)) { + qobject_unref(qdict_out); + } else { + expansion_info->model->props = QOBJECT(qdict_out); + expansion_info->model->has_props = true; + } + + object_unref(obj); + + return expansion_info; +} diff --git a/docs/arm-cpu-features.rst b/docs/arm-cpu-features.rst new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/docs/arm-cpu-features.rst @@ -XXX,XX +XXX,XX @@ +================ +ARM CPU Features +================ + +Examples of probing and using ARM CPU features + +Introduction +============ + +CPU features are optional features that a CPU of supporting type may +choose to implement or not. In QEMU, optional CPU features have +corresponding boolean CPU proprieties that, when enabled, indicate +that the feature is implemented, and, conversely, when disabled, +indicate that it is not implemented. An example of an ARM CPU feature +is the Performance Monitoring Unit (PMU). CPU types such as the +Cortex-A15 and the Cortex-A57, which respectively implement ARM +architecture reference manuals ARMv7-A and ARMv8-A, may both optionally +implement PMUs. For example, if a user wants to use a Cortex-A15 without +a PMU, then the `-cpu` parameter should contain `pmu=off` on the QEMU +command line, i.e. `-cpu cortex-a15,pmu=off`. + +As not all CPU types support all optional CPU features, then whether or +not a CPU property exists depends on the CPU type. For example, CPUs +that implement the ARMv8-A architecture reference manual may optionally +support the AArch32 CPU feature, which may be enabled by disabling the +`aarch64` CPU property. A CPU type such as the Cortex-A15, which does +not implement ARMv8-A, will not have the `aarch64` CPU property. + +QEMU's support may be limited for some CPU features, only partially +supporting the feature or only supporting the feature under certain +configurations. For example, the `aarch64` CPU feature, which, when +disabled, enables the optional AArch32 CPU feature, is only supported +when using the KVM accelerator and when running on a host CPU type that +supports the feature. + +CPU Feature Probing +=================== + +Determining which CPU features are available and functional for a given +CPU type is possible with the `query-cpu-model-expansion` QMP command. +Below are some examples where `scripts/qmp/qmp-shell` (see the top comment +block in the script for usage) is used to issue the QMP commands. + +(1) Determine which CPU features are available for the `max` CPU type + (Note, we started QEMU with qemu-system-aarch64, so `max` is + implementing the ARMv8-A reference manual in this case):: + + (QEMU) query-cpu-model-expansion type=full model={"name":"max"} + { "return": { + "model": { "name": "max", "props": { + "pmu": true, "aarch64": true + }}}} + +We see that the `max` CPU type has the `pmu` and `aarch64` CPU features. +We also see that the CPU features are enabled, as they are all `true`. + +(2) Let's try to disable the PMU:: + + (QEMU) query-cpu-model-expansion type=full model={"name":"max","props":{"pmu":false}} + { "return": { + "model": { "name": "max", "props": { + "pmu": false, "aarch64": true + }}}} + +We see it worked, as `pmu` is now `false`. + +(3) Let's try to disable `aarch64`, which enables the AArch32 CPU feature:: + + (QEMU) query-cpu-model-expansion type=full model={"name":"max","props":{"aarch64":false}} + {"error": { + "class": "GenericError", "desc": + "'aarch64' feature cannot be disabled unless KVM is enabled and 32-bit EL1 is supported" + }} + +It looks like this feature is limited to a configuration we do not +currently have. + +(4) Let's try probing CPU features for the Cortex-A15 CPU type:: + + (QEMU) query-cpu-model-expansion type=full model={"name":"cortex-a15"} + {"return": {"model": {"name": "cortex-a15", "props": {"pmu": true}}}} + +Only the `pmu` CPU feature is available. + +A note about CPU feature dependencies +------------------------------------- + +It's possible for features to have dependencies on other features. I.e. +it may be possible to change one feature at a time without error, but +when attempting to change all features at once an error could occur +depending on the order they are processed. It's also possible changing +all at once doesn't generate an error, because a feature's dependencies +are satisfied with other features, but the same feature cannot be changed +independently without error. For these reasons callers should always +attempt to make their desired changes all at once in order to ensure the +collection is valid. + +A note about CPU models and KVM +------------------------------- + +Named CPU models generally do not work with KVM. There are a few cases +that do work, e.g. using the named CPU model `cortex-a57` with KVM on a +seattle host, but mostly if KVM is enabled the `host` CPU type must be +used. This means the guest is provided all the same CPU features as the +host CPU type has. And, for this reason, the `host` CPU type should +enable all CPU features that the host has by default. Indeed it's even +a bit strange to allow disabling CPU features that the host has when using +the `host` CPU type, but in the absence of CPU models it's the best we can +do if we want to launch guests without all the host's CPU features enabled. + +Enabling KVM also affects the `query-cpu-model-expansion` QMP command. The +affect is not only limited to specific features, as pointed out in example +(3) of "CPU Feature Probing", but also to which CPU types may be expanded. +When KVM is enabled, only the `max`, `host`, and current CPU type may be +expanded. This restriction is necessary as it's not possible to know all +CPU types that may work with KVM, but it does impose a small risk of users +experiencing unexpected errors. For example on a seattle, as mentioned +above, the `cortex-a57` CPU type is also valid when KVM is enabled. +Therefore a user could use the `host` CPU type for the current type, but +then attempt to query `cortex-a57`, however that query will fail with our +restrictions. This shouldn't be an issue though as management layers and +users have been preferring the `host` CPU type for use with KVM for quite +some time. Additionally, if the KVM-enabled QEMU instance running on a +seattle host is using the `cortex-a57` CPU type, then querying `cortex-a57` +will work. + +Using CPU Features +================== + +After determining which CPU features are available and supported for a +given CPU type, then they may be selectively enabled or disabled on the +QEMU command line with that CPU type:: + + $ qemu-system-aarch64 -M virt -cpu max,pmu=off + +The example above disables the PMU for the `max` CPU type. + -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Now that Arm CPUs have advertised features lets add tests to ensure we maintain their expected availability with and without KVM. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Message-id: 20191031142734.8590-3-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- tests/Makefile.include | 5 +- tests/arm-cpu-features.c | 253 +++++++++++++++++++++++++++++++++++++++ 2 files changed, 257 insertions(+), 1 deletion(-) create mode 100644 tests/arm-cpu-features.c diff --git a/tests/Makefile.include b/tests/Makefile.include index XXXXXXX..XXXXXXX 100644 --- a/tests/Makefile.include +++ b/tests/Makefile.include @@ -XXX,XX +XXX,XX @@ check-qtest-sparc64-$(CONFIG_ISA_TESTDEV) = tests/endianness-test$(EXESUF) check-qtest-sparc64-y += tests/prom-env-test$(EXESUF) check-qtest-sparc64-y += tests/boot-serial-test$(EXESUF) +check-qtest-arm-y += tests/arm-cpu-features$(EXESUF) check-qtest-arm-y += tests/microbit-test$(EXESUF) check-qtest-arm-y += tests/m25p80-test$(EXESUF) check-qtest-arm-y += tests/test-arm-mptimer$(EXESUF) @@ -XXX,XX +XXX,XX @@ check-qtest-arm-y += tests/boot-serial-test$(EXESUF) check-qtest-arm-y += tests/hexloader-test$(EXESUF) check-qtest-arm-$(CONFIG_PFLASH_CFI02) += tests/pflash-cfi02-test$(EXESUF) -check-qtest-aarch64-y = tests/numa-test$(EXESUF) +check-qtest-aarch64-y += tests/arm-cpu-features$(EXESUF) +check-qtest-aarch64-y += tests/numa-test$(EXESUF) check-qtest-aarch64-y += tests/boot-serial-test$(EXESUF) check-qtest-aarch64-y += tests/migration-test$(EXESUF) # TODO: once aarch64 TCG is fixed on ARM 32 bit host, make test unconditional @@ -XXX,XX +XXX,XX @@ tests/test-qapi-util$(EXESUF): tests/test-qapi-util.o $(test-util-obj-y) tests/numa-test$(EXESUF): tests/numa-test.o tests/vmgenid-test$(EXESUF): tests/vmgenid-test.o tests/boot-sector.o tests/acpi-utils.o tests/cdrom-test$(EXESUF): tests/cdrom-test.o tests/boot-sector.o $(libqos-obj-y) +tests/arm-cpu-features$(EXESUF): tests/arm-cpu-features.o tests/migration/stress$(EXESUF): tests/migration/stress.o $(call quiet-command, $(LINKPROG) -static -O3 $(PTHREAD_LIB) -o $@ $< ,"LINK","$(TARGET_DIR)$@") diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ +/* + * Arm CPU feature test cases + * + * Copyright (c) 2019 Red Hat Inc. + * Authors: + * Andrew Jones <drjones@redhat.com> + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ +#include "qemu/osdep.h" +#include "libqtest.h" +#include "qapi/qmp/qdict.h" +#include "qapi/qmp/qjson.h" + +#define MACHINE "-machine virt,gic-version=max,accel=tcg " +#define MACHINE_KVM "-machine virt,gic-version=max,accel=kvm:tcg " +#define QUERY_HEAD "{ 'execute': 'query-cpu-model-expansion', " \ + " 'arguments': { 'type': 'full', " +#define QUERY_TAIL "}}" + +static bool kvm_enabled(QTestState *qts) +{ + QDict *resp, *qdict; + bool enabled; + + resp = qtest_qmp(qts, "{ 'execute': 'query-kvm' }"); + g_assert(qdict_haskey(resp, "return")); + qdict = qdict_get_qdict(resp, "return"); + g_assert(qdict_haskey(qdict, "enabled")); + enabled = qdict_get_bool(qdict, "enabled"); + qobject_unref(resp); + + return enabled; +} + +static QDict *do_query_no_props(QTestState *qts, const char *cpu_type) +{ + return qtest_qmp(qts, QUERY_HEAD "'model': { 'name': %s }" + QUERY_TAIL, cpu_type); +} + +static QDict *do_query(QTestState *qts, const char *cpu_type, + const char *fmt, ...) +{ + QDict *resp; + + if (fmt) { + QDict *args; + va_list ap; + + va_start(ap, fmt); + args = qdict_from_vjsonf_nofail(fmt, ap); + va_end(ap); + + resp = qtest_qmp(qts, QUERY_HEAD "'model': { 'name': %s, " + "'props': %p }" + QUERY_TAIL, cpu_type, args); + } else { + resp = do_query_no_props(qts, cpu_type); + } + + return resp; +} + +static const char *resp_get_error(QDict *resp) +{ + QDict *qdict; + + g_assert(resp); + + qdict = qdict_get_qdict(resp, "error"); + if (qdict) { + return qdict_get_str(qdict, "desc"); + } + + return NULL; +} + +#define assert_error(qts, cpu_type, expected_error, fmt, ...) \ +({ \ + QDict *_resp; \ + const char *_error; \ + \ + _resp = do_query(qts, cpu_type, fmt, ##__VA_ARGS__); \ + g_assert(_resp); \ + _error = resp_get_error(_resp); \ + g_assert(_error); \ + g_assert(g_str_equal(_error, expected_error)); \ + qobject_unref(_resp); \ +}) + +static bool resp_has_props(QDict *resp) +{ + QDict *qdict; + + g_assert(resp); + + if (!qdict_haskey(resp, "return")) { + return false; + } + qdict = qdict_get_qdict(resp, "return"); + + if (!qdict_haskey(qdict, "model")) { + return false; + } + qdict = qdict_get_qdict(qdict, "model"); + + return qdict_haskey(qdict, "props"); +} + +static QDict *resp_get_props(QDict *resp) +{ + QDict *qdict; + + g_assert(resp); + g_assert(resp_has_props(resp)); + + qdict = qdict_get_qdict(resp, "return"); + qdict = qdict_get_qdict(qdict, "model"); + qdict = qdict_get_qdict(qdict, "props"); + + return qdict; +} + +#define assert_has_feature(qts, cpu_type, feature) \ +({ \ + QDict *_resp = do_query_no_props(qts, cpu_type); \ + g_assert(_resp); \ + g_assert(resp_has_props(_resp)); \ + g_assert(qdict_get(resp_get_props(_resp), feature)); \ + qobject_unref(_resp); \ +}) + +#define assert_has_not_feature(qts, cpu_type, feature) \ +({ \ + QDict *_resp = do_query_no_props(qts, cpu_type); \ + g_assert(_resp); \ + g_assert(!resp_has_props(_resp) || \ + !qdict_get(resp_get_props(_resp), feature)); \ + qobject_unref(_resp); \ +}) + +static void assert_type_full(QTestState *qts) +{ + const char *error; + QDict *resp; + + resp = qtest_qmp(qts, "{ 'execute': 'query-cpu-model-expansion', " + "'arguments': { 'type': 'static', " + "'model': { 'name': 'foo' }}}"); + g_assert(resp); + error = resp_get_error(resp); + g_assert(error); + g_assert(g_str_equal(error, + "The requested expansion type is not supported")); + qobject_unref(resp); +} + +static void assert_bad_props(QTestState *qts, const char *cpu_type) +{ + const char *error; + QDict *resp; + + resp = qtest_qmp(qts, "{ 'execute': 'query-cpu-model-expansion', " + "'arguments': { 'type': 'full', " + "'model': { 'name': %s, " + "'props': false }}}", + cpu_type); + g_assert(resp); + error = resp_get_error(resp); + g_assert(error); + g_assert(g_str_equal(error, + "Invalid parameter type for 'props', expected: dict")); + qobject_unref(resp); +} + +static void test_query_cpu_model_expansion(const void *data) +{ + QTestState *qts; + + qts = qtest_init(MACHINE "-cpu max"); + + /* Test common query-cpu-model-expansion input validation */ + assert_type_full(qts); + assert_bad_props(qts, "max"); + assert_error(qts, "foo", "The CPU type 'foo' is not a recognized " + "ARM CPU type", NULL); + assert_error(qts, "max", "Parameter 'not-a-prop' is unexpected", + "{ 'not-a-prop': false }"); + assert_error(qts, "host", "The CPU type 'host' requires KVM", NULL); + + /* Test expected feature presence/absence for some cpu types */ + assert_has_feature(qts, "max", "pmu"); + assert_has_feature(qts, "cortex-a15", "pmu"); + assert_has_not_feature(qts, "cortex-a15", "aarch64"); + + if (g_str_equal(qtest_get_arch(), "aarch64")) { + assert_has_feature(qts, "max", "aarch64"); + assert_has_feature(qts, "cortex-a57", "pmu"); + assert_has_feature(qts, "cortex-a57", "aarch64"); + + /* Test that features that depend on KVM generate errors without. */ + assert_error(qts, "max", + "'aarch64' feature cannot be disabled " + "unless KVM is enabled and 32-bit EL1 " + "is supported", + "{ 'aarch64': false }"); + } + + qtest_quit(qts); +} + +static void test_query_cpu_model_expansion_kvm(const void *data) +{ + QTestState *qts; + + qts = qtest_init(MACHINE_KVM "-cpu max"); + + /* + * These tests target the 'host' CPU type, so KVM must be enabled. + */ + if (!kvm_enabled(qts)) { + qtest_quit(qts); + return; + } + + if (g_str_equal(qtest_get_arch(), "aarch64")) { + assert_has_feature(qts, "host", "aarch64"); + assert_has_feature(qts, "host", "pmu"); + + assert_error(qts, "cortex-a15", + "We cannot guarantee the CPU type 'cortex-a15' works " + "with KVM on this host", NULL); + } else { + assert_has_not_feature(qts, "host", "aarch64"); + assert_has_not_feature(qts, "host", "pmu"); + } + + qtest_quit(qts); +} + +int main(int argc, char **argv) +{ + g_test_init(&argc, &argv, NULL); + + qtest_add_data_func("/arm/query-cpu-model-expansion", + NULL, test_query_cpu_model_expansion); + qtest_add_data_func("/arm/kvm/query-cpu-model-expansion", + NULL, test_query_cpu_model_expansion_kvm); + + return g_test_run(); +} -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Since 97a28b0eeac14 ("target/arm: Allow VFP and Neon to be disabled via a CPU property") we can disable the 'max' cpu model's VFP and neon features, but there's no way to disable SVE. Add the 'sve=on|off' property to give it that flexibility. We also rename cpu_max_get/set_sve_vq to cpu_max_get/set_sve_max_vq in order for them to follow the typical *_get/set_<property-name> pattern. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Reviewed-by: Beata Michalska <beata.michalska@linaro.org> Message-id: 20191031142734.8590-4-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 3 ++- target/arm/cpu64.c | 52 ++++++++++++++++++++++++++++++++++------ target/arm/monitor.c | 2 +- tests/arm-cpu-features.c | 1 + 4 files changed, 49 insertions(+), 9 deletions(-) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_reset(CPUState *s) env->cp15.cpacr_el1 = deposit64(env->cp15.cpacr_el1, 16, 2, 3); env->cp15.cptr_el[3] |= CPTR_EZ; /* with maximum vector length */ - env->vfp.zcr_el[1] = cpu->sve_max_vq - 1; + env->vfp.zcr_el[1] = cpu_isar_feature(aa64_sve, cpu) ? + cpu->sve_max_vq - 1 : 0; env->vfp.zcr_el[2] = env->vfp.zcr_el[1]; env->vfp.zcr_el[3] = env->vfp.zcr_el[1]; /* diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_a72_initfn(Object *obj) define_arm_cp_regs(cpu, cortex_a72_a57_a53_cp_reginfo); } -static void cpu_max_get_sve_vq(Object *obj, Visitor *v, const char *name, - void *opaque, Error **errp) +static void cpu_max_get_sve_max_vq(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) { ARMCPU *cpu = ARM_CPU(obj); - visit_type_uint32(v, name, &cpu->sve_max_vq, errp); + uint32_t value; + + /* All vector lengths are disabled when SVE is off. */ + if (!cpu_isar_feature(aa64_sve, cpu)) { + value = 0; + } else { + value = cpu->sve_max_vq; + } + visit_type_uint32(v, name, &value, errp); } -static void cpu_max_set_sve_vq(Object *obj, Visitor *v, const char *name, - void *opaque, Error **errp) +static void cpu_max_set_sve_max_vq(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) { ARMCPU *cpu = ARM_CPU(obj); Error *err = NULL; @@ -XXX,XX +XXX,XX @@ static void cpu_max_set_sve_vq(Object *obj, Visitor *v, const char *name, error_propagate(errp, err); } +static void cpu_arm_get_sve(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) +{ + ARMCPU *cpu = ARM_CPU(obj); + bool value = cpu_isar_feature(aa64_sve, cpu); + + visit_type_bool(v, name, &value, errp); +} + +static void cpu_arm_set_sve(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) +{ + ARMCPU *cpu = ARM_CPU(obj); + Error *err = NULL; + bool value; + uint64_t t; + + visit_type_bool(v, name, &value, &err); + if (err) { + error_propagate(errp, err); + return; + } + + t = cpu->isar.id_aa64pfr0; + t = FIELD_DP64(t, ID_AA64PFR0, SVE, value); + cpu->isar.id_aa64pfr0 = t; +} + /* -cpu max: if KVM is enabled, like -cpu host (best possible with this host); * otherwise, a CPU with as many features enabled as our emulation supports. * The version of '-cpu max' for qemu-system-arm is defined in cpu.c; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) #endif cpu->sve_max_vq = ARM_MAX_VQ; - object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_vq, - cpu_max_set_sve_vq, NULL, NULL, &error_fatal); + object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, + cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); + object_property_add(obj, "sve", "bool", cpu_arm_get_sve, + cpu_arm_set_sve, NULL, NULL, &error_fatal); } } diff --git a/target/arm/monitor.c b/target/arm/monitor.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/monitor.c +++ b/target/arm/monitor.c @@ -XXX,XX +XXX,XX @@ GICCapabilityList *qmp_query_gic_capabilities(Error **errp) * then the order that considers those dependencies must be used. */ static const char *cpu_model_advertised_features[] = { - "aarch64", "pmu", + "aarch64", "pmu", "sve", NULL }; diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c index XXXXXXX..XXXXXXX 100644 --- a/tests/arm-cpu-features.c +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion(const void *data) if (g_str_equal(qtest_get_arch(), "aarch64")) { assert_has_feature(qts, "max", "aarch64"); + assert_has_feature(qts, "max", "sve"); assert_has_feature(qts, "cortex-a57", "pmu"); assert_has_feature(qts, "cortex-a57", "aarch64"); -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Introduce cpu properties to give fine control over SVE vector lengths. We introduce a property for each valid length up to the current maximum supported, which is 2048-bits. The properties are named, e.g. sve128, sve256, sve384, sve512, ..., where the number is the number of bits. See the updates to docs/arm-cpu-features.rst for a description of the semantics and for example uses. Note, as sve-max-vq is still present and we'd like to be able to support qmp_query_cpu_model_expansion with guests launched with e.g. -cpu max,sve-max-vq=8 on their command lines, then we do allow sve-max-vq and sve<N> properties to be provided at the same time, but this is not recommended, and is why sve-max-vq is not mentioned in the document. If sve-max-vq is provided then it enables all lengths smaller than and including the max and disables all lengths larger. It also has the side-effect that no larger lengths may be enabled and that the max itself cannot be disabled. Smaller non-power-of-two lengths may, however, be disabled, e.g. -cpu max,sve-max-vq=4,sve384=off provides a guest the vector lengths 128, 256, and 512 bits. This patch has been co-authored with Richard Henderson, who reworked the target/arm/cpu64.c changes in order to push all the validation and auto-enabling/disabling steps into the finalizer, resulting in a nice LOC reduction. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Reviewed-by: Beata Michalska <beata.michalska@linaro.org> Message-id: 20191031142734.8590-5-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/qemu/bitops.h | 1 + target/arm/cpu.h | 19 ++++ target/arm/cpu.c | 19 ++++ target/arm/cpu64.c | 192 ++++++++++++++++++++++++++++++++++++- target/arm/helper.c | 10 +- target/arm/monitor.c | 12 +++ tests/arm-cpu-features.c | 194 ++++++++++++++++++++++++++++++++++++++ docs/arm-cpu-features.rst | 168 +++++++++++++++++++++++++++++++-- 8 files changed, 606 insertions(+), 9 deletions(-) diff --git a/include/qemu/bitops.h b/include/qemu/bitops.h index XXXXXXX..XXXXXXX 100644 --- a/include/qemu/bitops.h +++ b/include/qemu/bitops.h @@ -XXX,XX +XXX,XX @@ #define BITS_PER_LONG (sizeof (unsigned long) * BITS_PER_BYTE) #define BIT(nr) (1UL << (nr)) +#define BIT_ULL(nr) (1ULL << (nr)) #define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_LONG)) #define BIT_WORD(nr) ((nr) / BITS_PER_LONG) #define BITS_TO_LONGS(nr) DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long)) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ typedef struct { #ifdef TARGET_AARCH64 # define ARM_MAX_VQ 16 +void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp); +uint32_t arm_cpu_vq_map_next_smaller(ARMCPU *cpu, uint32_t vq); #else # define ARM_MAX_VQ 1 +static inline void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) { } +static inline uint32_t arm_cpu_vq_map_next_smaller(ARMCPU *cpu, uint32_t vq) +{ return 0; } #endif typedef struct ARMVectorReg { @@ -XXX,XX +XXX,XX @@ struct ARMCPU { /* Used to set the maximum vector length the cpu will support. */ uint32_t sve_max_vq; + + /* + * In sve_vq_map each set bit is a supported vector length of + * (bit-number + 1) * 16 bytes, i.e. each bit number + 1 is the vector + * length in quadwords. + * + * While processing properties during initialization, corresponding + * sve_vq_init bits are set for bits in sve_vq_map that have been + * set by properties. + */ + DECLARE_BITMAP(sve_vq_map, ARM_MAX_VQ); + DECLARE_BITMAP(sve_vq_init, ARM_MAX_VQ); }; void arm_cpu_post_init(Object *obj); @@ -XXX,XX +XXX,XX @@ static inline int arm_feature(CPUARMState *env, int feature) return (env->features & (1ULL << feature)) != 0; } +void arm_cpu_finalize_features(ARMCPU *cpu, Error **errp); + #if !defined(CONFIG_USER_ONLY) /* Return true if exception levels below EL3 are in secure state, * or would be following an exception return to that level. diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_finalizefn(Object *obj) #endif } +void arm_cpu_finalize_features(ARMCPU *cpu, Error **errp) +{ + Error *local_err = NULL; + + if (arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) { + arm_cpu_sve_finalize(cpu, &local_err); + if (local_err != NULL) { + error_propagate(errp, local_err); + return; + } + } +} + static void arm_cpu_realizefn(DeviceState *dev, Error **errp) { CPUState *cs = CPU(dev); @@ -XXX,XX +XXX,XX @@ static void arm_cpu_realizefn(DeviceState *dev, Error **errp) return; } + arm_cpu_finalize_features(cpu, &local_err); + if (local_err != NULL) { + error_propagate(errp, local_err); + return; + } + if (arm_feature(env, ARM_FEATURE_AARCH64) && cpu->has_vfp != cpu->has_neon) { /* diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_a72_initfn(Object *obj) define_arm_cp_regs(cpu, cortex_a72_a57_a53_cp_reginfo); } +void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) +{ + /* + * If any vector lengths are explicitly enabled with sve<N> properties, + * then all other lengths are implicitly disabled. If sve-max-vq is + * specified then it is the same as explicitly enabling all lengths + * up to and including the specified maximum, which means all larger + * lengths will be implicitly disabled. If no sve<N> properties + * are enabled and sve-max-vq is not specified, then all lengths not + * explicitly disabled will be enabled. Additionally, all power-of-two + * vector lengths less than the maximum enabled length will be + * automatically enabled and all vector lengths larger than the largest + * disabled power-of-two vector length will be automatically disabled. + * Errors are generated if the user provided input that interferes with + * any of the above. Finally, if SVE is not disabled, then at least one + * vector length must be enabled. + */ + DECLARE_BITMAP(tmp, ARM_MAX_VQ); + uint32_t vq, max_vq = 0; + + /* + * Process explicit sve<N> properties. + * From the properties, sve_vq_map<N> implies sve_vq_init<N>. + * Check first for any sve<N> enabled. + */ + if (!bitmap_empty(cpu->sve_vq_map, ARM_MAX_VQ)) { + max_vq = find_last_bit(cpu->sve_vq_map, ARM_MAX_VQ) + 1; + + if (cpu->sve_max_vq && max_vq > cpu->sve_max_vq) { + error_setg(errp, "cannot enable sve%d", max_vq * 128); + error_append_hint(errp, "sve%d is larger than the maximum vector " + "length, sve-max-vq=%d (%d bits)\n", + max_vq * 128, cpu->sve_max_vq, + cpu->sve_max_vq * 128); + return; + } + + /* Propagate enabled bits down through required powers-of-two. */ + for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { + if (!test_bit(vq - 1, cpu->sve_vq_init)) { + set_bit(vq - 1, cpu->sve_vq_map); + } + } + } else if (cpu->sve_max_vq == 0) { + /* + * No explicit bits enabled, and no implicit bits from sve-max-vq. + */ + if (!cpu_isar_feature(aa64_sve, cpu)) { + /* SVE is disabled and so are all vector lengths. Good. */ + return; + } + + /* Disabling a power-of-two disables all larger lengths. */ + if (test_bit(0, cpu->sve_vq_init)) { + error_setg(errp, "cannot disable sve128"); + error_append_hint(errp, "Disabling sve128 results in all vector " + "lengths being disabled.\n"); + error_append_hint(errp, "With SVE enabled, at least one vector " + "length must be enabled.\n"); + return; + } + for (vq = 2; vq <= ARM_MAX_VQ; vq <<= 1) { + if (test_bit(vq - 1, cpu->sve_vq_init)) { + break; + } + } + max_vq = vq <= ARM_MAX_VQ ? vq - 1 : ARM_MAX_VQ; + + bitmap_complement(cpu->sve_vq_map, cpu->sve_vq_init, max_vq); + max_vq = find_last_bit(cpu->sve_vq_map, max_vq) + 1; + } + + /* + * Process the sve-max-vq property. + * Note that we know from the above that no bit above + * sve-max-vq is currently set. + */ + if (cpu->sve_max_vq != 0) { + max_vq = cpu->sve_max_vq; + + if (!test_bit(max_vq - 1, cpu->sve_vq_map) && + test_bit(max_vq - 1, cpu->sve_vq_init)) { + error_setg(errp, "cannot disable sve%d", max_vq * 128); + error_append_hint(errp, "The maximum vector length must be " + "enabled, sve-max-vq=%d (%d bits)\n", + max_vq, max_vq * 128); + return; + } + + /* Set all bits not explicitly set within sve-max-vq. */ + bitmap_complement(tmp, cpu->sve_vq_init, max_vq); + bitmap_or(cpu->sve_vq_map, cpu->sve_vq_map, tmp, max_vq); + } + + /* + * We should know what max-vq is now. Also, as we're done + * manipulating sve-vq-map, we ensure any bits above max-vq + * are clear, just in case anybody looks. + */ + assert(max_vq != 0); + bitmap_clear(cpu->sve_vq_map, max_vq, ARM_MAX_VQ - max_vq); + + /* Ensure all required powers-of-two are enabled. */ + for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { + if (!test_bit(vq - 1, cpu->sve_vq_map)) { + error_setg(errp, "cannot disable sve%d", vq * 128); + error_append_hint(errp, "sve%d is required as it " + "is a power-of-two length smaller than " + "the maximum, sve%d\n", + vq * 128, max_vq * 128); + return; + } + } + + /* + * Now that we validated all our vector lengths, the only question + * left to answer is if we even want SVE at all. + */ + if (!cpu_isar_feature(aa64_sve, cpu)) { + error_setg(errp, "cannot enable sve%d", max_vq * 128); + error_append_hint(errp, "SVE must be enabled to enable vector " + "lengths.\n"); + error_append_hint(errp, "Add sve=on to the CPU property list.\n"); + return; + } + + /* From now on sve_max_vq is the actual maximum supported length. */ + cpu->sve_max_vq = max_vq; +} + +uint32_t arm_cpu_vq_map_next_smaller(ARMCPU *cpu, uint32_t vq) +{ + uint32_t bitnum; + + /* + * We allow vq == ARM_MAX_VQ + 1 to be input because the caller may want + * to find the maximum vq enabled, which may be ARM_MAX_VQ, but this + * function always returns the next smaller than the input. + */ + assert(vq && vq <= ARM_MAX_VQ + 1); + + bitnum = find_last_bit(cpu->sve_vq_map, vq - 1); + return bitnum == vq - 1 ? 0 : bitnum + 1; +} + static void cpu_max_get_sve_max_vq(Object *obj, Visitor *v, const char *name, void *opaque, Error **errp) { @@ -XXX,XX +XXX,XX @@ static void cpu_max_set_sve_max_vq(Object *obj, Visitor *v, const char *name, error_propagate(errp, err); } +static void cpu_arm_get_sve_vq(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) +{ + ARMCPU *cpu = ARM_CPU(obj); + uint32_t vq = atoi(&name[3]) / 128; + bool value; + + /* All vector lengths are disabled when SVE is off. */ + if (!cpu_isar_feature(aa64_sve, cpu)) { + value = false; + } else { + value = test_bit(vq - 1, cpu->sve_vq_map); + } + visit_type_bool(v, name, &value, errp); +} + +static void cpu_arm_set_sve_vq(Object *obj, Visitor *v, const char *name, + void *opaque, Error **errp) +{ + ARMCPU *cpu = ARM_CPU(obj); + uint32_t vq = atoi(&name[3]) / 128; + Error *err = NULL; + bool value; + + visit_type_bool(v, name, &value, &err); + if (err) { + error_propagate(errp, err); + return; + } + + if (value) { + set_bit(vq - 1, cpu->sve_vq_map); + } else { + clear_bit(vq - 1, cpu->sve_vq_map); + } + set_bit(vq - 1, cpu->sve_vq_init); +} + static void cpu_arm_get_sve(Object *obj, Visitor *v, const char *name, void *opaque, Error **errp) { @@ -XXX,XX +XXX,XX @@ static void cpu_arm_set_sve(Object *obj, Visitor *v, const char *name, static void aarch64_max_initfn(Object *obj) { ARMCPU *cpu = ARM_CPU(obj); + uint32_t vq; if (kvm_enabled()) { kvm_arm_set_cpu_features_from_host(cpu); @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) cpu->dcz_blocksize = 7; /* 512 bytes */ #endif - cpu->sve_max_vq = ARM_MAX_VQ; object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); object_property_add(obj, "sve", "bool", cpu_arm_get_sve, cpu_arm_set_sve, NULL, NULL, &error_fatal); + + for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { + char name[8]; + sprintf(name, "sve%d", vq * 128); + object_property_add(obj, name, "bool", cpu_arm_get_sve_vq, + cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); + } } } diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ int sve_exception_el(CPUARMState *env, int el) return 0; } +static uint32_t sve_zcr_get_valid_len(ARMCPU *cpu, uint32_t start_len) +{ + uint32_t start_vq = (start_len & 0xf) + 1; + + return arm_cpu_vq_map_next_smaller(cpu, start_vq + 1) - 1; +} + /* * Given that SVE is enabled, return the vector length for EL. */ @@ -XXX,XX +XXX,XX @@ uint32_t sve_zcr_len_for_el(CPUARMState *env, int el) if (arm_feature(env, ARM_FEATURE_EL3)) { zcr_len = MIN(zcr_len, 0xf & (uint32_t)env->vfp.zcr_el[3]); } - return zcr_len; + + return sve_zcr_get_valid_len(cpu, zcr_len); } static void zcr_write(CPUARMState *env, const ARMCPRegInfo *ri, diff --git a/target/arm/monitor.c b/target/arm/monitor.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/monitor.c +++ b/target/arm/monitor.c @@ -XXX,XX +XXX,XX @@ GICCapabilityList *qmp_query_gic_capabilities(Error **errp) return head; } +QEMU_BUILD_BUG_ON(ARM_MAX_VQ > 16); + /* * These are cpu model features we want to advertise. The order here * matters as this is the order in which qmp_query_cpu_model_expansion @@ -XXX,XX +XXX,XX @@ GICCapabilityList *qmp_query_gic_capabilities(Error **errp) */ static const char *cpu_model_advertised_features[] = { "aarch64", "pmu", "sve", + "sve128", "sve256", "sve384", "sve512", + "sve640", "sve768", "sve896", "sve1024", "sve1152", "sve1280", + "sve1408", "sve1536", "sve1664", "sve1792", "sve1920", "sve2048", NULL }; @@ -XXX,XX +XXX,XX @@ CpuModelExpansionInfo *qmp_query_cpu_model_expansion(CpuModelExpansionType type, if (!err) { visit_check_struct(visitor, &err); } + if (!err) { + arm_cpu_finalize_features(ARM_CPU(obj), &err); + } visit_end_struct(visitor, NULL); visit_free(visitor); if (err) { @@ -XXX,XX +XXX,XX @@ CpuModelExpansionInfo *qmp_query_cpu_model_expansion(CpuModelExpansionType type, error_propagate(errp, err); return NULL; } + } else { + Error *err = NULL; + arm_cpu_finalize_features(ARM_CPU(obj), &err); + assert(err == NULL); } expansion_info = g_new0(CpuModelExpansionInfo, 1); diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c index XXXXXXX..XXXXXXX 100644 --- a/tests/arm-cpu-features.c +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ * See the COPYING file in the top-level directory. */ #include "qemu/osdep.h" +#include "qemu/bitops.h" #include "libqtest.h" #include "qapi/qmp/qdict.h" #include "qapi/qmp/qjson.h" +/* + * We expect the SVE max-vq to be 16. Also it must be <= 64 + * for our test code, otherwise 'vls' can't just be a uint64_t. + */ +#define SVE_MAX_VQ 16 + #define MACHINE "-machine virt,gic-version=max,accel=tcg " #define MACHINE_KVM "-machine virt,gic-version=max,accel=kvm:tcg " #define QUERY_HEAD "{ 'execute': 'query-cpu-model-expansion', " \ @@ -XXX,XX +XXX,XX @@ static void assert_bad_props(QTestState *qts, const char *cpu_type) qobject_unref(resp); } +static uint64_t resp_get_sve_vls(QDict *resp) +{ + QDict *props; + const QDictEntry *e; + uint64_t vls = 0; + int n = 0; + + g_assert(resp); + g_assert(resp_has_props(resp)); + + props = resp_get_props(resp); + + for (e = qdict_first(props); e; e = qdict_next(props, e)) { + if (strlen(e->key) > 3 && !strncmp(e->key, "sve", 3) && + g_ascii_isdigit(e->key[3])) { + char *endptr; + int bits; + + bits = g_ascii_strtoll(&e->key[3], &endptr, 10); + if (!bits || *endptr != '\0') { + continue; + } + + if (qdict_get_bool(props, e->key)) { + vls |= BIT_ULL((bits / 128) - 1); + } + ++n; + } + } + + g_assert(n == SVE_MAX_VQ); + + return vls; +} + +#define assert_sve_vls(qts, cpu_type, expected_vls, fmt, ...) \ +({ \ + QDict *_resp = do_query(qts, cpu_type, fmt, ##__VA_ARGS__); \ + g_assert(_resp); \ + g_assert(resp_has_props(_resp)); \ + g_assert(resp_get_sve_vls(_resp) == expected_vls); \ + qobject_unref(_resp); \ +}) + +static void sve_tests_default(QTestState *qts, const char *cpu_type) +{ + /* + * With no sve-max-vq or sve<N> properties on the command line + * the default is to have all vector lengths enabled. This also + * tests that 'sve' is 'on' by default. + */ + assert_sve_vls(qts, cpu_type, BIT_ULL(SVE_MAX_VQ) - 1, NULL); + + /* With SVE off, all vector lengths should also be off. */ + assert_sve_vls(qts, cpu_type, 0, "{ 'sve': false }"); + + /* With SVE on, we must have at least one vector length enabled. */ + assert_error(qts, cpu_type, "cannot disable sve128", "{ 'sve128': false }"); + + /* Basic enable/disable tests. */ + assert_sve_vls(qts, cpu_type, 0x7, "{ 'sve384': true }"); + assert_sve_vls(qts, cpu_type, ((BIT_ULL(SVE_MAX_VQ) - 1) & ~BIT_ULL(2)), + "{ 'sve384': false }"); + + /* + * --------------------------------------------------------------------- + * power-of-two(vq) all-power- can can + * of-two(< vq) enable disable + * --------------------------------------------------------------------- + * vq < max_vq no MUST* yes yes + * vq < max_vq yes MUST* yes no + * --------------------------------------------------------------------- + * vq == max_vq n/a MUST* yes** yes** + * --------------------------------------------------------------------- + * vq > max_vq n/a no no yes + * vq > max_vq n/a yes yes yes + * --------------------------------------------------------------------- + * + * [*] "MUST" means this requirement must already be satisfied, + * otherwise 'max_vq' couldn't itself be enabled. + * + * [**] Not testable with the QMP interface, only with the command line. + */ + + /* max_vq := 8 */ + assert_sve_vls(qts, cpu_type, 0x8b, "{ 'sve1024': true }"); + + /* max_vq := 8, vq < max_vq, !power-of-two(vq) */ + assert_sve_vls(qts, cpu_type, 0x8f, + "{ 'sve1024': true, 'sve384': true }"); + assert_sve_vls(qts, cpu_type, 0x8b, + "{ 'sve1024': true, 'sve384': false }"); + + /* max_vq := 8, vq < max_vq, power-of-two(vq) */ + assert_sve_vls(qts, cpu_type, 0x8b, + "{ 'sve1024': true, 'sve256': true }"); + assert_error(qts, cpu_type, "cannot disable sve256", + "{ 'sve1024': true, 'sve256': false }"); + + /* max_vq := 3, vq > max_vq, !all-power-of-two(< vq) */ + assert_error(qts, cpu_type, "cannot disable sve512", + "{ 'sve384': true, 'sve512': false, 'sve640': true }"); + + /* + * We can disable power-of-two vector lengths when all larger lengths + * are also disabled. We only need to disable the power-of-two length, + * as all non-enabled larger lengths will then be auto-disabled. + */ + assert_sve_vls(qts, cpu_type, 0x7, "{ 'sve512': false }"); + + /* max_vq := 3, vq > max_vq, all-power-of-two(< vq) */ + assert_sve_vls(qts, cpu_type, 0x1f, + "{ 'sve384': true, 'sve512': true, 'sve640': true }"); + assert_sve_vls(qts, cpu_type, 0xf, + "{ 'sve384': true, 'sve512': true, 'sve640': false }"); +} + +static void sve_tests_sve_max_vq_8(const void *data) +{ + QTestState *qts; + + qts = qtest_init(MACHINE "-cpu max,sve-max-vq=8"); + + assert_sve_vls(qts, "max", BIT_ULL(8) - 1, NULL); + + /* + * Disabling the max-vq set by sve-max-vq is not allowed, but + * of course enabling it is OK. + */ + assert_error(qts, "max", "cannot disable sve1024", "{ 'sve1024': false }"); + assert_sve_vls(qts, "max", 0xff, "{ 'sve1024': true }"); + + /* + * Enabling anything larger than max-vq set by sve-max-vq is not + * allowed, but of course disabling everything larger is OK. + */ + assert_error(qts, "max", "cannot enable sve1152", "{ 'sve1152': true }"); + assert_sve_vls(qts, "max", 0xff, "{ 'sve1152': false }"); + + /* + * We can enable/disable non power-of-two lengths smaller than the + * max-vq set by sve-max-vq, but, while we can enable power-of-two + * lengths, we can't disable them. + */ + assert_sve_vls(qts, "max", 0xff, "{ 'sve384': true }"); + assert_sve_vls(qts, "max", 0xfb, "{ 'sve384': false }"); + assert_sve_vls(qts, "max", 0xff, "{ 'sve256': true }"); + assert_error(qts, "max", "cannot disable sve256", "{ 'sve256': false }"); + + qtest_quit(qts); +} + +static void sve_tests_sve_off(const void *data) +{ + QTestState *qts; + + qts = qtest_init(MACHINE "-cpu max,sve=off"); + + /* SVE is off, so the map should be empty. */ + assert_sve_vls(qts, "max", 0, NULL); + + /* The map stays empty even if we turn lengths off. */ + assert_sve_vls(qts, "max", 0, "{ 'sve128': false }"); + + /* It's an error to enable lengths when SVE is off. */ + assert_error(qts, "max", "cannot enable sve128", "{ 'sve128': true }"); + + /* With SVE re-enabled we should get all vector lengths enabled. */ + assert_sve_vls(qts, "max", BIT_ULL(SVE_MAX_VQ) - 1, "{ 'sve': true }"); + + /* Or enable SVE with just specific vector lengths. */ + assert_sve_vls(qts, "max", 0x3, + "{ 'sve': true, 'sve128': true, 'sve256': true }"); + + qtest_quit(qts); +} + static void test_query_cpu_model_expansion(const void *data) { QTestState *qts; @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion(const void *data) if (g_str_equal(qtest_get_arch(), "aarch64")) { assert_has_feature(qts, "max", "aarch64"); assert_has_feature(qts, "max", "sve"); + assert_has_feature(qts, "max", "sve128"); assert_has_feature(qts, "cortex-a57", "pmu"); assert_has_feature(qts, "cortex-a57", "aarch64"); + sve_tests_default(qts, "max"); + /* Test that features that depend on KVM generate errors without. */ assert_error(qts, "max", "'aarch64' feature cannot be disabled " @@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv) qtest_add_data_func("/arm/kvm/query-cpu-model-expansion", NULL, test_query_cpu_model_expansion_kvm); + if (g_str_equal(qtest_get_arch(), "aarch64")) { + qtest_add_data_func("/arm/max/query-cpu-model-expansion/sve-max-vq-8", + NULL, sve_tests_sve_max_vq_8); + qtest_add_data_func("/arm/max/query-cpu-model-expansion/sve-off", + NULL, sve_tests_sve_off); + } + return g_test_run(); } diff --git a/docs/arm-cpu-features.rst b/docs/arm-cpu-features.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/arm-cpu-features.rst +++ b/docs/arm-cpu-features.rst @@ -XXX,XX +XXX,XX @@ block in the script for usage) is used to issue the QMP commands. (QEMU) query-cpu-model-expansion type=full model={"name":"max"} { "return": { "model": { "name": "max", "props": { - "pmu": true, "aarch64": true + "sve1664": true, "pmu": true, "sve1792": true, "sve1920": true, + "sve128": true, "aarch64": true, "sve1024": true, "sve": true, + "sve640": true, "sve768": true, "sve1408": true, "sve256": true, + "sve1152": true, "sve512": true, "sve384": true, "sve1536": true, + "sve896": true, "sve1280": true, "sve2048": true }}}} -We see that the `max` CPU type has the `pmu` and `aarch64` CPU features. -We also see that the CPU features are enabled, as they are all `true`. +We see that the `max` CPU type has the `pmu`, `aarch64`, `sve`, and many +`sve<N>` CPU features. We also see that all the CPU features are +enabled, as they are all `true`. (The `sve<N>` CPU features are all +optional SVE vector lengths (see "SVE CPU Properties"). While with TCG +all SVE vector lengths can be supported, when KVM is in use it's more +likely that only a few lengths will be supported, if SVE is supported at +all.) (2) Let's try to disable the PMU:: (QEMU) query-cpu-model-expansion type=full model={"name":"max","props":{"pmu":false}} { "return": { "model": { "name": "max", "props": { - "pmu": false, "aarch64": true + "sve1664": true, "pmu": false, "sve1792": true, "sve1920": true, + "sve128": true, "aarch64": true, "sve1024": true, "sve": true, + "sve640": true, "sve768": true, "sve1408": true, "sve256": true, + "sve1152": true, "sve512": true, "sve384": true, "sve1536": true, + "sve896": true, "sve1280": true, "sve2048": true }}}} We see it worked, as `pmu` is now `false`. @@ -XXX,XX +XXX,XX @@ We see it worked, as `pmu` is now `false`. It looks like this feature is limited to a configuration we do not currently have. -(4) Let's try probing CPU features for the Cortex-A15 CPU type:: +(4) Let's disable `sve` and see what happens to all the optional SVE + vector lengths:: + + (QEMU) query-cpu-model-expansion type=full model={"name":"max","props":{"sve":false}} + { "return": { + "model": { "name": "max", "props": { + "sve1664": false, "pmu": true, "sve1792": false, "sve1920": false, + "sve128": false, "aarch64": true, "sve1024": false, "sve": false, + "sve640": false, "sve768": false, "sve1408": false, "sve256": false, + "sve1152": false, "sve512": false, "sve384": false, "sve1536": false, + "sve896": false, "sve1280": false, "sve2048": false + }}}} + +As expected they are now all `false`. + +(5) Let's try probing CPU features for the Cortex-A15 CPU type:: (QEMU) query-cpu-model-expansion type=full model={"name":"cortex-a15"} {"return": {"model": {"name": "cortex-a15", "props": {"pmu": true}}}} @@ -XXX,XX +XXX,XX @@ After determining which CPU features are available and supported for a given CPU type, then they may be selectively enabled or disabled on the QEMU command line with that CPU type:: - $ qemu-system-aarch64 -M virt -cpu max,pmu=off + $ qemu-system-aarch64 -M virt -cpu max,pmu=off,sve=on,sve128=on,sve256=on -The example above disables the PMU for the `max` CPU type. +The example above disables the PMU and enables the first two SVE vector +lengths for the `max` CPU type. Note, the `sve=on` isn't actually +necessary, because, as we observed above with our probe of the `max` CPU +type, `sve` is already on by default. Also, based on our probe of +defaults, it would seem we need to disable many SVE vector lengths, rather +than only enabling the two we want. This isn't the case, because, as +disabling many SVE vector lengths would be quite verbose, the `sve<N>` CPU +properties have special semantics (see "SVE CPU Property Parsing +Semantics"). + +SVE CPU Properties +================== + +There are two types of SVE CPU properties: `sve` and `sve<N>`. The first +is used to enable or disable the entire SVE feature, just as the `pmu` +CPU property completely enables or disables the PMU. The second type +is used to enable or disable specific vector lengths, where `N` is the +number of bits of the length. The `sve<N>` CPU properties have special +dependencies and constraints, see "SVE CPU Property Dependencies and +Constraints" below. Additionally, as we want all supported vector lengths +to be enabled by default, then, in order to avoid overly verbose command +lines (command lines full of `sve<N>=off`, for all `N` not wanted), we +provide the parsing semantics listed in "SVE CPU Property Parsing +Semantics". + +SVE CPU Property Dependencies and Constraints +--------------------------------------------- + + 1) At least one vector length must be enabled when `sve` is enabled. + + 2) If a vector length `N` is enabled, then all power-of-two vector + lengths smaller than `N` must also be enabled. E.g. if `sve512` + is enabled, then the 128-bit and 256-bit vector lengths must also + be enabled. + +SVE CPU Property Parsing Semantics +---------------------------------- + + 1) If SVE is disabled (`sve=off`), then which SVE vector lengths + are enabled or disabled is irrelevant to the guest, as the entire + SVE feature is disabled and that disables all vector lengths for + the guest. However QEMU will still track any `sve<N>` CPU + properties provided by the user. If later an `sve=on` is provided, + then the guest will get only the enabled lengths. If no `sve=on` + is provided and there are explicitly enabled vector lengths, then + an error is generated. + + 2) If SVE is enabled (`sve=on`), but no `sve<N>` CPU properties are + provided, then all supported vector lengths are enabled, including + the non-power-of-two lengths. + + 3) If SVE is enabled, then an error is generated when attempting to + disable the last enabled vector length (see constraint (1) of "SVE + CPU Property Dependencies and Constraints"). + + 4) If one or more vector lengths have been explicitly enabled and at + at least one of the dependency lengths of the maximum enabled length + has been explicitly disabled, then an error is generated (see + constraint (2) of "SVE CPU Property Dependencies and Constraints"). + + 5) If one or more `sve<N>` CPU properties are set `off`, but no `sve<N>`, + CPU properties are set `on`, then the specified vector lengths are + disabled but the default for any unspecified lengths remains enabled. + Disabling a power-of-two vector length also disables all vector + lengths larger than the power-of-two length (see constraint (2) of + "SVE CPU Property Dependencies and Constraints"). + + 6) If one or more `sve<N>` CPU properties are set to `on`, then they + are enabled and all unspecified lengths default to disabled, except + for the required lengths per constraint (2) of "SVE CPU Property + Dependencies and Constraints", which will even be auto-enabled if + they were not explicitly enabled. + + 7) If SVE was disabled (`sve=off`), allowing all vector lengths to be + explicitly disabled (i.e. avoiding the error specified in (3) of + "SVE CPU Property Parsing Semantics"), then if later an `sve=on` is + provided an error will be generated. To avoid this error, one must + enable at least one vector length prior to enabling SVE. + +SVE CPU Property Examples +------------------------- + + 1) Disable SVE:: + + $ qemu-system-aarch64 -M virt -cpu max,sve=off + + 2) Implicitly enable all vector lengths for the `max` CPU type:: + + $ qemu-system-aarch64 -M virt -cpu max + + 3) Only enable the 128-bit vector length:: + + $ qemu-system-aarch64 -M virt -cpu max,sve128=on + + 4) Disable the 512-bit vector length and all larger vector lengths, + since 512 is a power-of-two. This results in all the smaller, + uninitialized lengths (128, 256, and 384) defaulting to enabled:: + + $ qemu-system-aarch64 -M virt -cpu max,sve512=off + + 5) Enable the 128-bit, 256-bit, and 512-bit vector lengths:: + + $ qemu-system-aarch64 -M virt -cpu max,sve128=on,sve256=on,sve512=on + + 6) The same as (5), but since the 128-bit and 256-bit vector + lengths are required for the 512-bit vector length to be enabled, + then allow them to be auto-enabled:: + + $ qemu-system-aarch64 -M virt -cpu max,sve512=on + + 7) Do the same as (6), but by first disabling SVE and then re-enabling it:: + + $ qemu-system-aarch64 -M virt -cpu max,sve=off,sve512=on,sve=on + + 8) Force errors regarding the last vector length:: + + $ qemu-system-aarch64 -M virt -cpu max,sve128=off + $ qemu-system-aarch64 -M virt -cpu max,sve=off,sve128=off,sve=on + +SVE CPU Property Recommendations +-------------------------------- + +The examples in "SVE CPU Property Examples" exhibit many ways to select +vector lengths which developers may find useful in order to avoid overly +verbose command lines. However, the recommended way to select vector +lengths is to explicitly enable each desired length. Therefore only +example's (1), (3), and (5) exhibit recommended uses of the properties. -- 2.20.1 From: Andrew Jones <drjones@redhat.com> These are the SVE equivalents to kvm_arch_get/put_fpsimd. Note, the swabbing is different than it is for fpsmid because the vector format is a little-endian stream of words. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Message-id: 20191031142734.8590-6-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/kvm64.c | 185 ++++++++++++++++++++++++++++++++++++++------- 1 file changed, 156 insertions(+), 29 deletions(-) diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ int kvm_arch_destroy_vcpu(CPUState *cs) bool kvm_arm_reg_syncs_via_cpreg_list(uint64_t regidx) { /* Return true if the regidx is a register we should synchronize - * via the cpreg_tuples array (ie is not a core reg we sync by - * hand in kvm_arch_get/put_registers()) + * via the cpreg_tuples array (ie is not a core or sve reg that + * we sync by hand in kvm_arch_get/put_registers()) */ switch (regidx & KVM_REG_ARM_COPROC_MASK) { case KVM_REG_ARM_CORE: + case KVM_REG_ARM64_SVE: return false; default: return true; @@ -XXX,XX +XXX,XX @@ int kvm_arm_cpreg_level(uint64_t regidx) static int kvm_arch_put_fpsimd(CPUState *cs) { - ARMCPU *cpu = ARM_CPU(cs); - CPUARMState *env = &cpu->env; + CPUARMState *env = &ARM_CPU(cs)->env; struct kvm_one_reg reg; - uint32_t fpr; int i, ret; for (i = 0; i < 32; i++) { @@ -XXX,XX +XXX,XX @@ static int kvm_arch_put_fpsimd(CPUState *cs) } } - reg.addr = (uintptr_t)(&fpr); - fpr = vfp_get_fpsr(env); - reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpsr); - ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); - if (ret) { - return ret; + return 0; +} + +/* + * SVE registers are encoded in KVM's memory in an endianness-invariant format. + * The byte at offset i from the start of the in-memory representation contains + * the bits [(7 + 8 * i) : (8 * i)] of the register value. As this means the + * lowest offsets are stored in the lowest memory addresses, then that nearly + * matches QEMU's representation, which is to use an array of host-endian + * uint64_t's, where the lower offsets are at the lower indices. To complete + * the translation we just need to byte swap the uint64_t's on big-endian hosts. + */ +static uint64_t *sve_bswap64(uint64_t *dst, uint64_t *src, int nr) +{ +#ifdef HOST_WORDS_BIGENDIAN + int i; + + for (i = 0; i < nr; ++i) { + dst[i] = bswap64(src[i]); } - reg.addr = (uintptr_t)(&fpr); - fpr = vfp_get_fpcr(env); - reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpcr); + return dst; +#else + return src; +#endif +} + +/* + * KVM SVE registers come in slices where ZREGs have a slice size of 2048 bits + * and PREGS and the FFR have a slice size of 256 bits. However we simply hard + * code the slice index to zero for now as it's unlikely we'll need more than + * one slice for quite some time. + */ +static int kvm_arch_put_sve(CPUState *cs) +{ + ARMCPU *cpu = ARM_CPU(cs); + CPUARMState *env = &cpu->env; + uint64_t tmp[ARM_MAX_VQ * 2]; + uint64_t *r; + struct kvm_one_reg reg; + int n, ret; + + for (n = 0; n < KVM_ARM64_SVE_NUM_ZREGS; ++n) { + r = sve_bswap64(tmp, &env->vfp.zregs[n].d[0], cpu->sve_max_vq * 2); + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_ZREG(n, 0); + ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); + if (ret) { + return ret; + } + } + + for (n = 0; n < KVM_ARM64_SVE_NUM_PREGS; ++n) { + r = sve_bswap64(tmp, r = &env->vfp.pregs[n].p[0], + DIV_ROUND_UP(cpu->sve_max_vq * 2, 8)); + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_PREG(n, 0); + ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); + if (ret) { + return ret; + } + } + + r = sve_bswap64(tmp, &env->vfp.pregs[FFR_PRED_NUM].p[0], + DIV_ROUND_UP(cpu->sve_max_vq * 2, 8)); + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_FFR(0); ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); if (ret) { return ret; @@ -XXX,XX +XXX,XX @@ int kvm_arch_put_registers(CPUState *cs, int level) { struct kvm_one_reg reg; uint64_t val; + uint32_t fpr; int i, ret; unsigned int el; @@ -XXX,XX +XXX,XX @@ int kvm_arch_put_registers(CPUState *cs, int level) } } - ret = kvm_arch_put_fpsimd(cs); + if (cpu_isar_feature(aa64_sve, cpu)) { + ret = kvm_arch_put_sve(cs); + } else { + ret = kvm_arch_put_fpsimd(cs); + } + if (ret) { + return ret; + } + + reg.addr = (uintptr_t)(&fpr); + fpr = vfp_get_fpsr(env); + reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpsr); + ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); + if (ret) { + return ret; + } + + reg.addr = (uintptr_t)(&fpr); + fpr = vfp_get_fpcr(env); + reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpcr); + ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); if (ret) { return ret; } @@ -XXX,XX +XXX,XX @@ int kvm_arch_put_registers(CPUState *cs, int level) static int kvm_arch_get_fpsimd(CPUState *cs) { - ARMCPU *cpu = ARM_CPU(cs); - CPUARMState *env = &cpu->env; + CPUARMState *env = &ARM_CPU(cs)->env; struct kvm_one_reg reg; - uint32_t fpr; int i, ret; for (i = 0; i < 32; i++) { @@ -XXX,XX +XXX,XX @@ static int kvm_arch_get_fpsimd(CPUState *cs) } } - reg.addr = (uintptr_t)(&fpr); - reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpsr); - ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); - if (ret) { - return ret; - } - vfp_set_fpsr(env, fpr); + return 0; +} - reg.addr = (uintptr_t)(&fpr); - reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpcr); +/* + * KVM SVE registers come in slices where ZREGs have a slice size of 2048 bits + * and PREGS and the FFR have a slice size of 256 bits. However we simply hard + * code the slice index to zero for now as it's unlikely we'll need more than + * one slice for quite some time. + */ +static int kvm_arch_get_sve(CPUState *cs) +{ + ARMCPU *cpu = ARM_CPU(cs); + CPUARMState *env = &cpu->env; + struct kvm_one_reg reg; + uint64_t *r; + int n, ret; + + for (n = 0; n < KVM_ARM64_SVE_NUM_ZREGS; ++n) { + r = &env->vfp.zregs[n].d[0]; + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_ZREG(n, 0); + ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); + if (ret) { + return ret; + } + sve_bswap64(r, r, cpu->sve_max_vq * 2); + } + + for (n = 0; n < KVM_ARM64_SVE_NUM_PREGS; ++n) { + r = &env->vfp.pregs[n].p[0]; + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_PREG(n, 0); + ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); + if (ret) { + return ret; + } + sve_bswap64(r, r, DIV_ROUND_UP(cpu->sve_max_vq * 2, 8)); + } + + r = &env->vfp.pregs[FFR_PRED_NUM].p[0]; + reg.addr = (uintptr_t)r; + reg.id = KVM_REG_ARM64_SVE_FFR(0); ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); if (ret) { return ret; } - vfp_set_fpcr(env, fpr); + sve_bswap64(r, r, DIV_ROUND_UP(cpu->sve_max_vq * 2, 8)); return 0; } @@ -XXX,XX +XXX,XX @@ int kvm_arch_get_registers(CPUState *cs) struct kvm_one_reg reg; uint64_t val; unsigned int el; + uint32_t fpr; int i, ret; ARMCPU *cpu = ARM_CPU(cs); @@ -XXX,XX +XXX,XX @@ int kvm_arch_get_registers(CPUState *cs) env->spsr = env->banked_spsr[i]; } - ret = kvm_arch_get_fpsimd(cs); + if (cpu_isar_feature(aa64_sve, cpu)) { + ret = kvm_arch_get_sve(cs); + } else { + ret = kvm_arch_get_fpsimd(cs); + } if (ret) { return ret; } + reg.addr = (uintptr_t)(&fpr); + reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpsr); + ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); + if (ret) { + return ret; + } + vfp_set_fpsr(env, fpr); + + reg.addr = (uintptr_t)(&fpr); + reg.id = AARCH64_SIMD_CTRL_REG(fp_regs.fpcr); + ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg); + if (ret) { + return ret; + } + vfp_set_fpcr(env, fpr); + ret = kvm_get_vcpu_events(cpu); if (ret) { return ret; -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Enable SVE in the KVM guest when the 'max' cpu type is configured and KVM supports it. KVM SVE requires use of the new finalize vcpu ioctl, so we add that now too. For starters SVE can only be turned on or off, getting all vector lengths the host CPU supports when on. We'll add the other SVE CPU properties in later patches. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Reviewed-by: Beata Michalska <beata.michalska@linaro.org> Message-id: 20191031142734.8590-7-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/kvm_arm.h | 27 +++++++++++++++++++++++++++ target/arm/cpu64.c | 17 ++++++++++++++--- target/arm/kvm.c | 5 +++++ target/arm/kvm64.c | 20 +++++++++++++++++++- tests/arm-cpu-features.c | 4 ++++ 5 files changed, 69 insertions(+), 4 deletions(-) diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm_arm.h +++ b/target/arm/kvm_arm.h @@ -XXX,XX +XXX,XX @@ */ int kvm_arm_vcpu_init(CPUState *cs); +/** + * kvm_arm_vcpu_finalize + * @cs: CPUState + * @feature: int + * + * Finalizes the configuration of the specified VCPU feature by + * invoking the KVM_ARM_VCPU_FINALIZE ioctl. Features requiring + * this are documented in the "KVM_ARM_VCPU_FINALIZE" section of + * KVM's API documentation. + * + * Returns: 0 if success else < 0 error code + */ +int kvm_arm_vcpu_finalize(CPUState *cs, int feature); + /** * kvm_arm_register_device: * @mr: memory region for this device @@ -XXX,XX +XXX,XX @@ bool kvm_arm_aarch32_supported(CPUState *cs); */ bool kvm_arm_pmu_supported(CPUState *cs); +/** + * bool kvm_arm_sve_supported: + * @cs: CPUState + * + * Returns true if the KVM VCPU can enable SVE and false otherwise. + */ +bool kvm_arm_sve_supported(CPUState *cs); + /** * kvm_arm_get_max_vm_ipa_size - Returns the number of bits in the * IPA address space supported by KVM @@ -XXX,XX +XXX,XX @@ static inline bool kvm_arm_pmu_supported(CPUState *cs) return false; } +static inline bool kvm_arm_sve_supported(CPUState *cs) +{ + return false; +} + static inline int kvm_arm_get_max_vm_ipa_size(MachineState *ms) { return -ENOENT; diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void cpu_arm_set_sve(Object *obj, Visitor *v, const char *name, return; } + if (value && kvm_enabled() && !kvm_arm_sve_supported(CPU(cpu))) { + error_setg(errp, "'sve' feature not supported by KVM on this host"); + return; + } + t = cpu->isar.id_aa64pfr0; t = FIELD_DP64(t, ID_AA64PFR0, SVE, value); cpu->isar.id_aa64pfr0 = t; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) { ARMCPU *cpu = ARM_CPU(obj); uint32_t vq; + uint64_t t; if (kvm_enabled()) { kvm_arm_set_cpu_features_from_host(cpu); + if (kvm_arm_sve_supported(CPU(cpu))) { + t = cpu->isar.id_aa64pfr0; + t = FIELD_DP64(t, ID_AA64PFR0, SVE, 1); + cpu->isar.id_aa64pfr0 = t; + } } else { - uint64_t t; uint32_t u; aarch64_a57_initfn(obj); @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); - object_property_add(obj, "sve", "bool", cpu_arm_get_sve, - cpu_arm_set_sve, NULL, NULL, &error_fatal); for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { char name[8]; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); } } + + object_property_add(obj, "sve", "bool", cpu_arm_get_sve, + cpu_arm_set_sve, NULL, NULL, &error_fatal); } struct ARMCPUInfo { diff --git a/target/arm/kvm.c b/target/arm/kvm.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm.c +++ b/target/arm/kvm.c @@ -XXX,XX +XXX,XX @@ int kvm_arm_vcpu_init(CPUState *cs) return kvm_vcpu_ioctl(cs, KVM_ARM_VCPU_INIT, &init); } +int kvm_arm_vcpu_finalize(CPUState *cs, int feature) +{ + return kvm_vcpu_ioctl(cs, KVM_ARM_VCPU_FINALIZE, &feature); +} + void kvm_arm_init_serror_injection(CPUState *cs) { cap_has_inject_serror_esr = kvm_check_extension(cs->kvm_state, diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_aarch32_supported(CPUState *cpu) return kvm_check_extension(s, KVM_CAP_ARM_EL1_32BIT); } +bool kvm_arm_sve_supported(CPUState *cpu) +{ + KVMState *s = KVM_STATE(current_machine->accelerator); + + return kvm_check_extension(s, KVM_CAP_ARM_SVE); +} + #define ARM_CPU_ID_MPIDR 3, 0, 0, 0, 5 int kvm_arch_init_vcpu(CPUState *cs) @@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs) cpu->kvm_init_features[0] |= 1 << KVM_ARM_VCPU_EL1_32BIT; } if (!kvm_check_extension(cs->kvm_state, KVM_CAP_ARM_PMU_V3)) { - cpu->has_pmu = false; + cpu->has_pmu = false; } if (cpu->has_pmu) { cpu->kvm_init_features[0] |= 1 << KVM_ARM_VCPU_PMU_V3; } else { unset_feature(&env->features, ARM_FEATURE_PMU); } + if (cpu_isar_feature(aa64_sve, cpu)) { + assert(kvm_arm_sve_supported(cs)); + cpu->kvm_init_features[0] |= 1 << KVM_ARM_VCPU_SVE; + } /* Do KVM_ARM_VCPU_INIT ioctl */ ret = kvm_arm_vcpu_init(cs); @@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs) return ret; } + if (cpu_isar_feature(aa64_sve, cpu)) { + ret = kvm_arm_vcpu_finalize(cs, KVM_ARM_VCPU_SVE); + if (ret) { + return ret; + } + } + /* * When KVM is in use, PSCI is emulated in-kernel and not by qemu. * Currently KVM has its own idea about MPIDR assignment, so we diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c index XXXXXXX..XXXXXXX 100644 --- a/tests/arm-cpu-features.c +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) assert_has_feature(qts, "host", "aarch64"); assert_has_feature(qts, "host", "pmu"); + assert_has_feature(qts, "max", "sve"); + assert_error(qts, "cortex-a15", "We cannot guarantee the CPU type 'cortex-a15' works " "with KVM on this host", NULL); } else { assert_has_not_feature(qts, "host", "aarch64"); assert_has_not_feature(qts, "host", "pmu"); + + assert_has_not_feature(qts, "max", "sve"); } qtest_quit(qts); -- 2.20.1 From: Andrew Jones <drjones@redhat.com> kvm_arm_create_scratch_host_vcpu() takes a struct kvm_vcpu_init parameter. Rather than just using it as an output parameter to pass back the preferred target, use it also as an input parameter, allowing a caller to pass a selected target if they wish and to also pass cpu features. If the caller doesn't want to select a target they can pass -1 for the target which indicates they want to use the preferred target and have it passed back like before. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Reviewed-by: Beata Michalska <beata.michalska@linaro.org> Message-id: 20191031142734.8590-8-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/kvm.c | 20 +++++++++++++++----- target/arm/kvm32.c | 6 +++++- target/arm/kvm64.c | 6 +++++- 3 files changed, 25 insertions(+), 7 deletions(-) diff --git a/target/arm/kvm.c b/target/arm/kvm.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm.c +++ b/target/arm/kvm.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_create_scratch_host_vcpu(const uint32_t *cpus_to_try, int *fdarray, struct kvm_vcpu_init *init) { - int ret, kvmfd = -1, vmfd = -1, cpufd = -1; + int ret = 0, kvmfd = -1, vmfd = -1, cpufd = -1; kvmfd = qemu_open("/dev/kvm", O_RDWR); if (kvmfd < 0) { @@ -XXX,XX +XXX,XX @@ bool kvm_arm_create_scratch_host_vcpu(const uint32_t *cpus_to_try, goto finish; } - ret = ioctl(vmfd, KVM_ARM_PREFERRED_TARGET, init); + if (init->target == -1) { + struct kvm_vcpu_init preferred; + + ret = ioctl(vmfd, KVM_ARM_PREFERRED_TARGET, &preferred); + if (!ret) { + init->target = preferred.target; + } + } if (ret >= 0) { ret = ioctl(cpufd, KVM_ARM_VCPU_INIT, init); if (ret < 0) { @@ -XXX,XX +XXX,XX @@ bool kvm_arm_create_scratch_host_vcpu(const uint32_t *cpus_to_try, * creating one kind of guest CPU which is its preferred * CPU type. */ + struct kvm_vcpu_init try; + while (*cpus_to_try != QEMU_KVM_ARM_TARGET_NONE) { - init->target = *cpus_to_try++; - memset(init->features, 0, sizeof(init->features)); - ret = ioctl(cpufd, KVM_ARM_VCPU_INIT, init); + try.target = *cpus_to_try++; + memcpy(try.features, init->features, sizeof(init->features)); + ret = ioctl(cpufd, KVM_ARM_VCPU_INIT, &try); if (ret >= 0) { break; } @@ -XXX,XX +XXX,XX @@ bool kvm_arm_create_scratch_host_vcpu(const uint32_t *cpus_to_try, if (ret < 0) { goto err; } + init->target = try.target; } else { /* Treat a NULL cpus_to_try argument the same as an empty * list, which means we will fail the call since this must diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm32.c +++ b/target/arm/kvm32.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) QEMU_KVM_ARM_TARGET_CORTEX_A15, QEMU_KVM_ARM_TARGET_NONE }; - struct kvm_vcpu_init init; + /* + * target = -1 informs kvm_arm_create_scratch_host_vcpu() + * to use the preferred target + */ + struct kvm_vcpu_init init = { .target = -1, }; if (!kvm_arm_create_scratch_host_vcpu(cpus_to_try, fdarray, &init)) { return false; diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) KVM_ARM_TARGET_CORTEX_A57, QEMU_KVM_ARM_TARGET_NONE }; - struct kvm_vcpu_init init; + /* + * target = -1 informs kvm_arm_create_scratch_host_vcpu() + * to use the preferred target + */ + struct kvm_vcpu_init init = { .target = -1, }; if (!kvm_arm_create_scratch_host_vcpu(cpus_to_try, fdarray, &init)) { return false; -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Extend the SVE vq map initialization and validation with KVM's supported vector lengths when KVM is enabled. In order to determine and select supported lengths we add two new KVM functions for getting and setting the KVM_REG_ARM64_SVE_VLS pseudo-register. This patch has been co-authored with Richard Henderson, who reworked the target/arm/cpu64.c changes in order to push all the validation and auto-enabling/disabling steps into the finalizer, resulting in a nice LOC reduction. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Message-id: 20191031142734.8590-9-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/kvm_arm.h | 12 +++ target/arm/cpu64.c | 176 ++++++++++++++++++++++++++++---------- target/arm/kvm64.c | 100 +++++++++++++++++++++- tests/arm-cpu-features.c | 104 +++++++++++++++++++++- docs/arm-cpu-features.rst | 45 +++++++--- 5 files changed, 379 insertions(+), 58 deletions(-) diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm_arm.h +++ b/target/arm/kvm_arm.h @@ -XXX,XX +XXX,XX @@ typedef struct ARMHostCPUFeatures { */ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf); +/** + * kvm_arm_sve_get_vls: + * @cs: CPUState + * @map: bitmap to fill in + * + * Get all the SVE vector lengths supported by the KVM host, setting + * the bits corresponding to their length in quadwords minus one + * (vq - 1) in @map up to ARM_MAX_VQ. + */ +void kvm_arm_sve_get_vls(CPUState *cs, unsigned long *map); + /** * kvm_arm_set_cpu_features_from_host: * @cpu: ARMCPU to set the features for @@ -XXX,XX +XXX,XX @@ static inline int kvm_arm_vgic_probe(void) static inline void kvm_arm_pmu_set_irq(CPUState *cs, int irq) {} static inline void kvm_arm_pmu_init(CPUState *cs) {} +static inline void kvm_arm_sve_get_vls(CPUState *cs, unsigned long *map) {} #endif static inline const char *gic_class_name(void) diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) * any of the above. Finally, if SVE is not disabled, then at least one * vector length must be enabled. */ + DECLARE_BITMAP(kvm_supported, ARM_MAX_VQ); DECLARE_BITMAP(tmp, ARM_MAX_VQ); uint32_t vq, max_vq = 0; + /* Collect the set of vector lengths supported by KVM. */ + bitmap_zero(kvm_supported, ARM_MAX_VQ); + if (kvm_enabled() && kvm_arm_sve_supported(CPU(cpu))) { + kvm_arm_sve_get_vls(CPU(cpu), kvm_supported); + } else if (kvm_enabled()) { + assert(!cpu_isar_feature(aa64_sve, cpu)); + } + /* * Process explicit sve<N> properties. * From the properties, sve_vq_map<N> implies sve_vq_init<N>. @@ -XXX,XX +XXX,XX @@ void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) return; } - /* Propagate enabled bits down through required powers-of-two. */ - for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { - if (!test_bit(vq - 1, cpu->sve_vq_init)) { - set_bit(vq - 1, cpu->sve_vq_map); + if (kvm_enabled()) { + /* + * For KVM we have to automatically enable all supported unitialized + * lengths, even when the smaller lengths are not all powers-of-two. + */ + bitmap_andnot(tmp, kvm_supported, cpu->sve_vq_init, max_vq); + bitmap_or(cpu->sve_vq_map, cpu->sve_vq_map, tmp, max_vq); + } else { + /* Propagate enabled bits down through required powers-of-two. */ + for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { + if (!test_bit(vq - 1, cpu->sve_vq_init)) { + set_bit(vq - 1, cpu->sve_vq_map); + } } } } else if (cpu->sve_max_vq == 0) { @@ -XXX,XX +XXX,XX @@ void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) return; } - /* Disabling a power-of-two disables all larger lengths. */ - if (test_bit(0, cpu->sve_vq_init)) { - error_setg(errp, "cannot disable sve128"); - error_append_hint(errp, "Disabling sve128 results in all vector " - "lengths being disabled.\n"); - error_append_hint(errp, "With SVE enabled, at least one vector " - "length must be enabled.\n"); - return; - } - for (vq = 2; vq <= ARM_MAX_VQ; vq <<= 1) { - if (test_bit(vq - 1, cpu->sve_vq_init)) { - break; + if (kvm_enabled()) { + /* Disabling a supported length disables all larger lengths. */ + for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { + if (test_bit(vq - 1, cpu->sve_vq_init) && + test_bit(vq - 1, kvm_supported)) { + break; + } } + max_vq = vq <= ARM_MAX_VQ ? vq - 1 : ARM_MAX_VQ; + bitmap_andnot(cpu->sve_vq_map, kvm_supported, + cpu->sve_vq_init, max_vq); + if (max_vq == 0 || bitmap_empty(cpu->sve_vq_map, max_vq)) { + error_setg(errp, "cannot disable sve%d", vq * 128); + error_append_hint(errp, "Disabling sve%d results in all " + "vector lengths being disabled.\n", + vq * 128); + error_append_hint(errp, "With SVE enabled, at least one " + "vector length must be enabled.\n"); + return; + } + } else { + /* Disabling a power-of-two disables all larger lengths. */ + if (test_bit(0, cpu->sve_vq_init)) { + error_setg(errp, "cannot disable sve128"); + error_append_hint(errp, "Disabling sve128 results in all " + "vector lengths being disabled.\n"); + error_append_hint(errp, "With SVE enabled, at least one " + "vector length must be enabled.\n"); + return; + } + for (vq = 2; vq <= ARM_MAX_VQ; vq <<= 1) { + if (test_bit(vq - 1, cpu->sve_vq_init)) { + break; + } + } + max_vq = vq <= ARM_MAX_VQ ? vq - 1 : ARM_MAX_VQ; + bitmap_complement(cpu->sve_vq_map, cpu->sve_vq_init, max_vq); } - max_vq = vq <= ARM_MAX_VQ ? vq - 1 : ARM_MAX_VQ; - bitmap_complement(cpu->sve_vq_map, cpu->sve_vq_init, max_vq); max_vq = find_last_bit(cpu->sve_vq_map, max_vq) + 1; } @@ -XXX,XX +XXX,XX @@ void arm_cpu_sve_finalize(ARMCPU *cpu, Error **errp) assert(max_vq != 0); bitmap_clear(cpu->sve_vq_map, max_vq, ARM_MAX_VQ - max_vq); - /* Ensure all required powers-of-two are enabled. */ - for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { - if (!test_bit(vq - 1, cpu->sve_vq_map)) { - error_setg(errp, "cannot disable sve%d", vq * 128); - error_append_hint(errp, "sve%d is required as it " - "is a power-of-two length smaller than " - "the maximum, sve%d\n", - vq * 128, max_vq * 128); + if (kvm_enabled()) { + /* Ensure the set of lengths matches what KVM supports. */ + bitmap_xor(tmp, cpu->sve_vq_map, kvm_supported, max_vq); + if (!bitmap_empty(tmp, max_vq)) { + vq = find_last_bit(tmp, max_vq) + 1; + if (test_bit(vq - 1, cpu->sve_vq_map)) { + if (cpu->sve_max_vq) { + error_setg(errp, "cannot set sve-max-vq=%d", + cpu->sve_max_vq); + error_append_hint(errp, "This KVM host does not support " + "the vector length %d-bits.\n", + vq * 128); + error_append_hint(errp, "It may not be possible to use " + "sve-max-vq with this KVM host. Try " + "using only sve<N> properties.\n"); + } else { + error_setg(errp, "cannot enable sve%d", vq * 128); + error_append_hint(errp, "This KVM host does not support " + "the vector length %d-bits.\n", + vq * 128); + } + } else { + error_setg(errp, "cannot disable sve%d", vq * 128); + error_append_hint(errp, "The KVM host requires all " + "supported vector lengths smaller " + "than %d bits to also be enabled.\n", + max_vq * 128); + } return; } + } else { + /* Ensure all required powers-of-two are enabled. */ + for (vq = pow2floor(max_vq); vq >= 1; vq >>= 1) { + if (!test_bit(vq - 1, cpu->sve_vq_map)) { + error_setg(errp, "cannot disable sve%d", vq * 128); + error_append_hint(errp, "sve%d is required as it " + "is a power-of-two length smaller than " + "the maximum, sve%d\n", + vq * 128, max_vq * 128); + return; + } + } } /* @@ -XXX,XX +XXX,XX @@ static void cpu_max_set_sve_max_vq(Object *obj, Visitor *v, const char *name, { ARMCPU *cpu = ARM_CPU(obj); Error *err = NULL; + uint32_t max_vq; - visit_type_uint32(v, name, &cpu->sve_max_vq, &err); - - if (!err && (cpu->sve_max_vq == 0 || cpu->sve_max_vq > ARM_MAX_VQ)) { - error_setg(&err, "unsupported SVE vector length"); - error_append_hint(&err, "Valid sve-max-vq in range [1-%d]\n", - ARM_MAX_VQ); + visit_type_uint32(v, name, &max_vq, &err); + if (err) { + error_propagate(errp, err); + return; } - error_propagate(errp, err); + + if (kvm_enabled() && !kvm_arm_sve_supported(CPU(cpu))) { + error_setg(errp, "cannot set sve-max-vq"); + error_append_hint(errp, "SVE not supported by KVM on this host\n"); + return; + } + + if (max_vq == 0 || max_vq > ARM_MAX_VQ) { + error_setg(errp, "unsupported SVE vector length"); + error_append_hint(errp, "Valid sve-max-vq in range [1-%d]\n", + ARM_MAX_VQ); + return; + } + + cpu->sve_max_vq = max_vq; } static void cpu_arm_get_sve_vq(Object *obj, Visitor *v, const char *name, @@ -XXX,XX +XXX,XX @@ static void cpu_arm_set_sve_vq(Object *obj, Visitor *v, const char *name, return; } + if (value && kvm_enabled() && !kvm_arm_sve_supported(CPU(cpu))) { + error_setg(errp, "cannot enable %s", name); + error_append_hint(errp, "SVE not supported by KVM on this host\n"); + return; + } + if (value) { set_bit(vq - 1, cpu->sve_vq_map); } else { @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) cpu->ctr = 0x80038003; /* 32 byte I and D cacheline size, VIPT icache */ cpu->dcz_blocksize = 7; /* 512 bytes */ #endif - - object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, - cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); - - for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { - char name[8]; - sprintf(name, "sve%d", vq * 128); - object_property_add(obj, name, "bool", cpu_arm_get_sve_vq, - cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); - } } object_property_add(obj, "sve", "bool", cpu_arm_get_sve, cpu_arm_set_sve, NULL, NULL, &error_fatal); + object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, + cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); + + for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { + char name[8]; + sprintf(name, "sve%d", vq * 128); + object_property_add(obj, name, "bool", cpu_arm_get_sve_vq, + cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); + } } struct ARMCPUInfo { diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_sve_supported(CPUState *cpu) return kvm_check_extension(s, KVM_CAP_ARM_SVE); } +QEMU_BUILD_BUG_ON(KVM_ARM64_SVE_VQ_MIN != 1); + +void kvm_arm_sve_get_vls(CPUState *cs, unsigned long *map) +{ + /* Only call this function if kvm_arm_sve_supported() returns true. */ + static uint64_t vls[KVM_ARM64_SVE_VLS_WORDS]; + static bool probed; + uint32_t vq = 0; + int i, j; + + bitmap_clear(map, 0, ARM_MAX_VQ); + + /* + * KVM ensures all host CPUs support the same set of vector lengths. + * So we only need to create the scratch VCPUs once and then cache + * the results. + */ + if (!probed) { + struct kvm_vcpu_init init = { + .target = -1, + .features[0] = (1 << KVM_ARM_VCPU_SVE), + }; + struct kvm_one_reg reg = { + .id = KVM_REG_ARM64_SVE_VLS, + .addr = (uint64_t)&vls[0], + }; + int fdarray[3], ret; + + probed = true; + + if (!kvm_arm_create_scratch_host_vcpu(NULL, fdarray, &init)) { + error_report("failed to create scratch VCPU with SVE enabled"); + abort(); + } + ret = ioctl(fdarray[2], KVM_GET_ONE_REG, &reg); + kvm_arm_destroy_scratch_host_vcpu(fdarray); + if (ret) { + error_report("failed to get KVM_REG_ARM64_SVE_VLS: %s", + strerror(errno)); + abort(); + } + + for (i = KVM_ARM64_SVE_VLS_WORDS - 1; i >= 0; --i) { + if (vls[i]) { + vq = 64 - clz64(vls[i]) + i * 64; + break; + } + } + if (vq > ARM_MAX_VQ) { + warn_report("KVM supports vector lengths larger than " + "QEMU can enable"); + } + } + + for (i = 0; i < KVM_ARM64_SVE_VLS_WORDS; ++i) { + if (!vls[i]) { + continue; + } + for (j = 1; j <= 64; ++j) { + vq = j + i * 64; + if (vq > ARM_MAX_VQ) { + return; + } + if (vls[i] & (1UL << (j - 1))) { + set_bit(vq - 1, map); + } + } + } +} + +static int kvm_arm_sve_set_vls(CPUState *cs) +{ + uint64_t vls[KVM_ARM64_SVE_VLS_WORDS] = {0}; + struct kvm_one_reg reg = { + .id = KVM_REG_ARM64_SVE_VLS, + .addr = (uint64_t)&vls[0], + }; + ARMCPU *cpu = ARM_CPU(cs); + uint32_t vq; + int i, j; + + assert(cpu->sve_max_vq <= KVM_ARM64_SVE_VQ_MAX); + + for (vq = 1; vq <= cpu->sve_max_vq; ++vq) { + if (test_bit(vq - 1, cpu->sve_vq_map)) { + i = (vq - 1) / 64; + j = (vq - 1) % 64; + vls[i] |= 1UL << j; + } + } + + return kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg); +} + #define ARM_CPU_ID_MPIDR 3, 0, 0, 0, 5 int kvm_arch_init_vcpu(CPUState *cs) @@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs) if (cpu->kvm_target == QEMU_KVM_ARM_TARGET_NONE || !object_dynamic_cast(OBJECT(cpu), TYPE_AARCH64_CPU)) { - fprintf(stderr, "KVM is not supported for this guest CPU type\n"); + error_report("KVM is not supported for this guest CPU type"); return -EINVAL; } @@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs) } if (cpu_isar_feature(aa64_sve, cpu)) { + ret = kvm_arm_sve_set_vls(cs); + if (ret) { + return ret; + } ret = kvm_arm_vcpu_finalize(cs, KVM_ARM_VCPU_SVE); if (ret) { return ret; diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c index XXXXXXX..XXXXXXX 100644 --- a/tests/arm-cpu-features.c +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ static QDict *resp_get_props(QDict *resp) return qdict; } +static bool resp_get_feature(QDict *resp, const char *feature) +{ + QDict *props; + + g_assert(resp); + g_assert(resp_has_props(resp)); + props = resp_get_props(resp); + g_assert(qdict_get(props, feature)); + return qdict_get_bool(props, feature); +} + #define assert_has_feature(qts, cpu_type, feature) \ ({ \ QDict *_resp = do_query_no_props(qts, cpu_type); \ @@ -XXX,XX +XXX,XX @@ static void sve_tests_sve_off(const void *data) qtest_quit(qts); } +static void sve_tests_sve_off_kvm(const void *data) +{ + QTestState *qts; + + qts = qtest_init(MACHINE_KVM "-cpu max,sve=off"); + + /* + * We don't know if this host supports SVE so we don't + * attempt to test enabling anything. We only test that + * everything is disabled (as it should be with sve=off) + * and that using sve<N>=off to explicitly disable vector + * lengths is OK too. + */ + assert_sve_vls(qts, "max", 0, NULL); + assert_sve_vls(qts, "max", 0, "{ 'sve128': false }"); + + qtest_quit(qts); +} + static void test_query_cpu_model_expansion(const void *data) { QTestState *qts; @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) } if (g_str_equal(qtest_get_arch(), "aarch64")) { + bool kvm_supports_sve; + char max_name[8], name[8]; + uint32_t max_vq, vq; + uint64_t vls; + QDict *resp; + char *error; + assert_has_feature(qts, "host", "aarch64"); assert_has_feature(qts, "host", "pmu"); - assert_has_feature(qts, "max", "sve"); - assert_error(qts, "cortex-a15", "We cannot guarantee the CPU type 'cortex-a15' works " "with KVM on this host", NULL); + + assert_has_feature(qts, "max", "sve"); + resp = do_query_no_props(qts, "max"); + kvm_supports_sve = resp_get_feature(resp, "sve"); + vls = resp_get_sve_vls(resp); + qobject_unref(resp); + + if (kvm_supports_sve) { + g_assert(vls != 0); + max_vq = 64 - __builtin_clzll(vls); + sprintf(max_name, "sve%d", max_vq * 128); + + /* Enabling a supported length is of course fine. */ + assert_sve_vls(qts, "max", vls, "{ %s: true }", max_name); + + /* Get the next supported length smaller than max-vq. */ + vq = 64 - __builtin_clzll(vls & ~BIT_ULL(max_vq - 1)); + if (vq) { + /* + * We have at least one length smaller than max-vq, + * so we can disable max-vq. + */ + assert_sve_vls(qts, "max", (vls & ~BIT_ULL(max_vq - 1)), + "{ %s: false }", max_name); + + /* + * Smaller, supported vector lengths cannot be disabled + * unless all larger, supported vector lengths are also + * disabled. + */ + sprintf(name, "sve%d", vq * 128); + error = g_strdup_printf("cannot disable %s", name); + assert_error(qts, "max", error, + "{ %s: true, %s: false }", + max_name, name); + g_free(error); + } + + /* + * The smallest, supported vector length is required, because + * we need at least one vector length enabled. + */ + vq = __builtin_ffsll(vls); + sprintf(name, "sve%d", vq * 128); + error = g_strdup_printf("cannot disable %s", name); + assert_error(qts, "max", error, "{ %s: false }", name); + g_free(error); + + /* Get an unsupported length. */ + for (vq = 1; vq <= max_vq; ++vq) { + if (!(vls & BIT_ULL(vq - 1))) { + break; + } + } + if (vq <= SVE_MAX_VQ) { + sprintf(name, "sve%d", vq * 128); + error = g_strdup_printf("cannot enable %s", name); + assert_error(qts, "max", error, "{ %s: true }", name); + g_free(error); + } + } else { + g_assert(vls == 0); + } } else { assert_has_not_feature(qts, "host", "aarch64"); assert_has_not_feature(qts, "host", "pmu"); @@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv) NULL, sve_tests_sve_max_vq_8); qtest_add_data_func("/arm/max/query-cpu-model-expansion/sve-off", NULL, sve_tests_sve_off); + qtest_add_data_func("/arm/kvm/query-cpu-model-expansion/sve-off", + NULL, sve_tests_sve_off_kvm); } return g_test_run(); diff --git a/docs/arm-cpu-features.rst b/docs/arm-cpu-features.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/arm-cpu-features.rst +++ b/docs/arm-cpu-features.rst @@ -XXX,XX +XXX,XX @@ SVE CPU Property Dependencies and Constraints 1) At least one vector length must be enabled when `sve` is enabled. - 2) If a vector length `N` is enabled, then all power-of-two vector - lengths smaller than `N` must also be enabled. E.g. if `sve512` - is enabled, then the 128-bit and 256-bit vector lengths must also - be enabled. + 2) If a vector length `N` is enabled, then, when KVM is enabled, all + smaller, host supported vector lengths must also be enabled. If + KVM is not enabled, then only all the smaller, power-of-two vector + lengths must be enabled. E.g. with KVM if the host supports all + vector lengths up to 512-bits (128, 256, 384, 512), then if `sve512` + is enabled, the 128-bit vector length, 256-bit vector length, and + 384-bit vector length must also be enabled. Without KVM, the 384-bit + vector length would not be required. + + 3) If KVM is enabled then only vector lengths that the host CPU type + support may be enabled. If SVE is not supported by the host, then + no `sve*` properties may be enabled. SVE CPU Property Parsing Semantics ---------------------------------- @@ -XXX,XX +XXX,XX @@ SVE CPU Property Parsing Semantics an error is generated. 2) If SVE is enabled (`sve=on`), but no `sve<N>` CPU properties are - provided, then all supported vector lengths are enabled, including - the non-power-of-two lengths. + provided, then all supported vector lengths are enabled, which when + KVM is not in use means including the non-power-of-two lengths, and, + when KVM is in use, it means all vector lengths supported by the host + processor. 3) If SVE is enabled, then an error is generated when attempting to disable the last enabled vector length (see constraint (1) of "SVE @@ -XXX,XX +XXX,XX @@ SVE CPU Property Parsing Semantics has been explicitly disabled, then an error is generated (see constraint (2) of "SVE CPU Property Dependencies and Constraints"). - 5) If one or more `sve<N>` CPU properties are set `off`, but no `sve<N>`, + 5) When KVM is enabled, if the host does not support SVE, then an error + is generated when attempting to enable any `sve*` properties (see + constraint (3) of "SVE CPU Property Dependencies and Constraints"). + + 6) When KVM is enabled, if the host does support SVE, then an error is + generated when attempting to enable any vector lengths not supported + by the host (see constraint (3) of "SVE CPU Property Dependencies and + Constraints"). + + 7) If one or more `sve<N>` CPU properties are set `off`, but no `sve<N>`, CPU properties are set `on`, then the specified vector lengths are disabled but the default for any unspecified lengths remains enabled. - Disabling a power-of-two vector length also disables all vector - lengths larger than the power-of-two length (see constraint (2) of - "SVE CPU Property Dependencies and Constraints"). + When KVM is not enabled, disabling a power-of-two vector length also + disables all vector lengths larger than the power-of-two length. + When KVM is enabled, then disabling any supported vector length also + disables all larger vector lengths (see constraint (2) of "SVE CPU + Property Dependencies and Constraints"). - 6) If one or more `sve<N>` CPU properties are set to `on`, then they + 8) If one or more `sve<N>` CPU properties are set to `on`, then they are enabled and all unspecified lengths default to disabled, except for the required lengths per constraint (2) of "SVE CPU Property Dependencies and Constraints", which will even be auto-enabled if they were not explicitly enabled. - 7) If SVE was disabled (`sve=off`), allowing all vector lengths to be + 9) If SVE was disabled (`sve=off`), allowing all vector lengths to be explicitly disabled (i.e. avoiding the error specified in (3) of "SVE CPU Property Parsing Semantics"), then if later an `sve=on` is provided an error will be generated. To avoid this error, one must -- 2.20.1 From: Andrew Jones <drjones@redhat.com> Allow cpu 'host' to enable SVE when it's available, unless the user chooses to disable it with the added 'sve=off' cpu property. Also give the user the ability to select vector lengths with the sve<N> properties. We don't adopt 'max' cpu's other sve property, sve-max-vq, because that property is difficult to use with KVM. That property assumes all vector lengths in the range from 1 up to and including the specified maximum length are supported, but there may be optional lengths not supported by the host in that range. With KVM one must be more specific when enabling vector lengths. Signed-off-by: Andrew Jones <drjones@redhat.com> Reviewed-by: Eric Auger <eric.auger@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com> Message-id: 20191031142734.8590-10-drjones@redhat.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 2 ++ target/arm/cpu.c | 3 +++ target/arm/cpu64.c | 33 +++++++++++++++++---------------- target/arm/kvm64.c | 14 +++++++++++++- tests/arm-cpu-features.c | 17 ++++++++--------- docs/arm-cpu-features.rst | 19 ++++++++++++------- 6 files changed, 55 insertions(+), 33 deletions(-) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ int aarch64_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void aarch64_sve_narrow_vq(CPUARMState *env, unsigned vq); void aarch64_sve_change_el(CPUARMState *env, int old_el, int new_el, bool el0_a64); +void aarch64_add_sve_properties(Object *obj); #else static inline void aarch64_sve_narrow_vq(CPUARMState *env, unsigned vq) { } static inline void aarch64_sve_change_el(CPUARMState *env, int o, int n, bool a) { } +static inline void aarch64_add_sve_properties(Object *obj) { } #endif #if !defined(CONFIG_TCG) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_host_initfn(Object *obj) ARMCPU *cpu = ARM_CPU(obj); kvm_arm_set_cpu_features_from_host(cpu); + if (arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) { + aarch64_add_sve_properties(obj); + } arm_cpu_post_init(obj); } diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void cpu_arm_set_sve(Object *obj, Visitor *v, const char *name, cpu->isar.id_aa64pfr0 = t; } +void aarch64_add_sve_properties(Object *obj) +{ + uint32_t vq; + + object_property_add(obj, "sve", "bool", cpu_arm_get_sve, + cpu_arm_set_sve, NULL, NULL, &error_fatal); + + for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { + char name[8]; + sprintf(name, "sve%d", vq * 128); + object_property_add(obj, name, "bool", cpu_arm_get_sve_vq, + cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); + } +} + /* -cpu max: if KVM is enabled, like -cpu host (best possible with this host); * otherwise, a CPU with as many features enabled as our emulation supports. * The version of '-cpu max' for qemu-system-arm is defined in cpu.c; @@ -XXX,XX +XXX,XX @@ static void cpu_arm_set_sve(Object *obj, Visitor *v, const char *name, static void aarch64_max_initfn(Object *obj) { ARMCPU *cpu = ARM_CPU(obj); - uint32_t vq; - uint64_t t; if (kvm_enabled()) { kvm_arm_set_cpu_features_from_host(cpu); - if (kvm_arm_sve_supported(CPU(cpu))) { - t = cpu->isar.id_aa64pfr0; - t = FIELD_DP64(t, ID_AA64PFR0, SVE, 1); - cpu->isar.id_aa64pfr0 = t; - } } else { + uint64_t t; uint32_t u; aarch64_a57_initfn(obj); @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) #endif } - object_property_add(obj, "sve", "bool", cpu_arm_get_sve, - cpu_arm_set_sve, NULL, NULL, &error_fatal); + aarch64_add_sve_properties(obj); object_property_add(obj, "sve-max-vq", "uint32", cpu_max_get_sve_max_vq, cpu_max_set_sve_max_vq, NULL, NULL, &error_fatal); - - for (vq = 1; vq <= ARM_MAX_VQ; ++vq) { - char name[8]; - sprintf(name, "sve%d", vq * 128); - object_property_add(obj, name, "bool", cpu_arm_get_sve_vq, - cpu_arm_set_sve_vq, NULL, NULL, &error_fatal); - } } struct ARMCPUInfo { diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) * and then query that CPU for the relevant ID registers. */ int fdarray[3]; + bool sve_supported; uint64_t features = 0; + uint64_t t; int err; /* Old kernels may not know about the PREFERRED_TARGET ioctl: however @@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf) ARM64_SYS_REG(3, 0, 0, 3, 2)); } + sve_supported = ioctl(fdarray[0], KVM_CHECK_EXTENSION, KVM_CAP_ARM_SVE) > 0; + kvm_arm_destroy_scratch_host_vcpu(fdarray); if (err < 0) { return false; } - /* We can assume any KVM supporting CPU is at least a v8 + /* Add feature bits that can't appear until after VCPU init. */ + if (sve_supported) { + t = ahcf->isar.id_aa64pfr0; + t = FIELD_DP64(t, ID_AA64PFR0, SVE, 1); + ahcf->isar.id_aa64pfr0 = t; + } + + /* + * We can assume any KVM supporting CPU is at least a v8 * with VFPv4+Neon; this in turn implies most of the other * feature bits. */ diff --git a/tests/arm-cpu-features.c b/tests/arm-cpu-features.c index XXXXXXX..XXXXXXX 100644 --- a/tests/arm-cpu-features.c +++ b/tests/arm-cpu-features.c @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) "We cannot guarantee the CPU type 'cortex-a15' works " "with KVM on this host", NULL); - assert_has_feature(qts, "max", "sve"); - resp = do_query_no_props(qts, "max"); + assert_has_feature(qts, "host", "sve"); + resp = do_query_no_props(qts, "host"); kvm_supports_sve = resp_get_feature(resp, "sve"); vls = resp_get_sve_vls(resp); qobject_unref(resp); @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) sprintf(max_name, "sve%d", max_vq * 128); /* Enabling a supported length is of course fine. */ - assert_sve_vls(qts, "max", vls, "{ %s: true }", max_name); + assert_sve_vls(qts, "host", vls, "{ %s: true }", max_name); /* Get the next supported length smaller than max-vq. */ vq = 64 - __builtin_clzll(vls & ~BIT_ULL(max_vq - 1)); @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) * We have at least one length smaller than max-vq, * so we can disable max-vq. */ - assert_sve_vls(qts, "max", (vls & ~BIT_ULL(max_vq - 1)), + assert_sve_vls(qts, "host", (vls & ~BIT_ULL(max_vq - 1)), "{ %s: false }", max_name); /* @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) */ sprintf(name, "sve%d", vq * 128); error = g_strdup_printf("cannot disable %s", name); - assert_error(qts, "max", error, + assert_error(qts, "host", error, "{ %s: true, %s: false }", max_name, name); g_free(error); @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) vq = __builtin_ffsll(vls); sprintf(name, "sve%d", vq * 128); error = g_strdup_printf("cannot disable %s", name); - assert_error(qts, "max", error, "{ %s: false }", name); + assert_error(qts, "host", error, "{ %s: false }", name); g_free(error); /* Get an unsupported length. */ @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) if (vq <= SVE_MAX_VQ) { sprintf(name, "sve%d", vq * 128); error = g_strdup_printf("cannot enable %s", name); - assert_error(qts, "max", error, "{ %s: true }", name); + assert_error(qts, "host", error, "{ %s: true }", name); g_free(error); } } else { @@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data) } else { assert_has_not_feature(qts, "host", "aarch64"); assert_has_not_feature(qts, "host", "pmu"); - - assert_has_not_feature(qts, "max", "sve"); + assert_has_not_feature(qts, "host", "sve"); } qtest_quit(qts); diff --git a/docs/arm-cpu-features.rst b/docs/arm-cpu-features.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/arm-cpu-features.rst +++ b/docs/arm-cpu-features.rst @@ -XXX,XX +XXX,XX @@ SVE CPU Property Examples $ qemu-system-aarch64 -M virt -cpu max - 3) Only enable the 128-bit vector length:: + 3) When KVM is enabled, implicitly enable all host CPU supported vector + lengths with the `host` CPU type:: + + $ qemu-system-aarch64 -M virt,accel=kvm -cpu host + + 4) Only enable the 128-bit vector length:: $ qemu-system-aarch64 -M virt -cpu max,sve128=on - 4) Disable the 512-bit vector length and all larger vector lengths, + 5) Disable the 512-bit vector length and all larger vector lengths, since 512 is a power-of-two. This results in all the smaller, uninitialized lengths (128, 256, and 384) defaulting to enabled:: $ qemu-system-aarch64 -M virt -cpu max,sve512=off - 5) Enable the 128-bit, 256-bit, and 512-bit vector lengths:: + 6) Enable the 128-bit, 256-bit, and 512-bit vector lengths:: $ qemu-system-aarch64 -M virt -cpu max,sve128=on,sve256=on,sve512=on - 6) The same as (5), but since the 128-bit and 256-bit vector + 7) The same as (6), but since the 128-bit and 256-bit vector lengths are required for the 512-bit vector length to be enabled, then allow them to be auto-enabled:: $ qemu-system-aarch64 -M virt -cpu max,sve512=on - 7) Do the same as (6), but by first disabling SVE and then re-enabling it:: + 8) Do the same as (7), but by first disabling SVE and then re-enabling it:: $ qemu-system-aarch64 -M virt -cpu max,sve=off,sve512=on,sve=on - 8) Force errors regarding the last vector length:: + 9) Force errors regarding the last vector length:: $ qemu-system-aarch64 -M virt -cpu max,sve128=off $ qemu-system-aarch64 -M virt -cpu max,sve=off,sve128=off,sve=on @@ -XXX,XX +XXX,XX @@ The examples in "SVE CPU Property Examples" exhibit many ways to select vector lengths which developers may find useful in order to avoid overly verbose command lines. However, the recommended way to select vector lengths is to explicitly enable each desired length. Therefore only -example's (1), (3), and (5) exhibit recommended uses of the properties. +example's (1), (4), and (6) exhibit recommended uses of the properties. -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Rebuild hflags when modifying CPUState at boot. Fixes: e979972a6a Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20191031040830.18800-2-edgar.iglesias@xilinx.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/boot.c | 1 + 1 file changed, 1 insertion(+) diff --git a/hw/arm/boot.c b/hw/arm/boot.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/boot.c +++ b/hw/arm/boot.c @@ -XXX,XX +XXX,XX @@ static void do_cpu_reset(void *opaque) info->secondary_cpu_reset_hook(cpu, info); } } + arm_rebuild_hflags(env); } } -- 2.20.1 From: Christophe Lyon <christophe.lyon@linaro.org> rt==15 is a special case when reading the flags: it means the destination is APSR. This patch avoids rejecting vmrs apsr_nzcv, fpscr as illegal instruction. Cc: qemu-stable@nongnu.org Signed-off-by: Christophe Lyon <christophe.lyon@linaro.org> Message-id: 20191025095711.10853-1-christophe.lyon@linaro.org [PMM: updated the comment] Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-vfp.inc.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/target/arm/translate-vfp.inc.c b/target/arm/translate-vfp.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.inc.c +++ b/target/arm/translate-vfp.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) if (arm_dc_feature(s, ARM_FEATURE_M)) { /* * The only M-profile VFP vmrs/vmsr sysreg is FPSCR. - * Writes to R15 are UNPREDICTABLE; we choose to undef. + * Accesses to R15 are UNPREDICTABLE; we choose to undef. + * (FPSCR -> r15 is a special case which writes to the PSR flags.) */ - if (a->rt == 15 || a->reg != ARM_VFP_FPSCR) { + if (a->rt == 15 && (!a->l || a->reg != ARM_VFP_FPSCR)) { return false; } } -- 2.20.1