target-arm queue; this one has a fair scattering of more miscellaneous things in it which I've sent out this week. I've shoved those in as well as it seemed the least-effort way of getting them into master; a few of them are dependencies on arm-related patches I have brewing. thanks -- PMM The following changes since commit 2702c2d3eb74e3908c0c5dbf3a71c8987595a86e: Merge remote-tracking branch 'remotes/stsquad/tags/pull-travis-updates-140618-1' into staging (2018-06-15 12:49:36 +0100) are available in the Git repository at: git://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20180615 for you to fetch changes up to 14120108f87b3f9e1beacdf0a6096e464e62bb65: target/arm: Allow ARMv6-M Thumb2 instructions (2018-06-15 15:23:34 +0100) ---------------------------------------------------------------- target-arm and miscellaneous queue: * fix KVM state save/restore for GICv3 priority registers for high IRQ numbers * hw/arm/mps2-tz: Put ethernet controller behind PPC * hw/sh/sh7750: Convert away from old_mmio * hw/m68k/mcf5206: Convert away from old_mmio * hw/block/pflash_cfi02: Convert away from old_mmio * hw/watchdog/wdt_i6300esb: Convert away from old_mmio * hw/input/pckbd: Convert away from old_mmio * hw/char/parallel: Convert away from old_mmio * armv7m: refactor to get rid of armv7m_init() function * arm: Don't crash if user tries to use a Cortex-M CPU without an NVIC * hw/core/or-irq: Support more than 16 inputs to an OR gate * cpu-defs.h: Document CPUIOTLBEntry 'addr' field * cputlb: Pass cpu_transaction_failed() the correct physaddr * CODING_STYLE: Define our preferred form for multiline comments * Add and use new stn_*_p() and ldn_*_p() memory access functions * target/arm: More parts of the upcoming SVE support * aspeed_scu: Implement RNG register * m25p80: add support for two bytes WRSR for Macronix chips * exec.c: Handle IOMMUs being in the path of TCG CPU memory accesses * target/arm: Allow ARMv6-M Thumb2 instructions ---------------------------------------------------------------- Cédric Le Goater (1): m25p80: add support for two bytes WRSR for Macronix chips Joel Stanley (1): aspeed_scu: Implement RNG register Julia Suvorova (1): target/arm: Allow ARMv6-M Thumb2 instructions Peter Maydell (21): hw/arm/mps2-tz: Put ethernet controller behind PPC hw/sh/sh7750: Convert away from old_mmio hw/m68k/mcf5206: Convert away from old_mmio hw/block/pflash_cfi02: Convert away from old_mmio hw/watchdog/wdt_i6300esb: Convert away from old_mmio hw/input/pckbd: Convert away from old_mmio hw/char/parallel: Convert away from old_mmio stellaris: Stop using armv7m_init() hw/arm/armv7m: Remove unused armv7m_init() function arm: Don't crash if user tries to use a Cortex-M CPU without an NVIC hw/core/or-irq: Support more than 16 inputs to an OR gate cpu-defs.h: Document CPUIOTLBEntry 'addr' field cputlb: Pass cpu_transaction_failed() the correct physaddr CODING_STYLE: Define our preferred form for multiline comments bswap: Add new stn_*_p() and ldn_*_p() memory access functions exec.c: Don't accidentally sign-extend 4-byte loads in subpage_read() exec.c: Use stn_p() and ldn_p() instead of explicit switches iommu: Add IOMMU index concept to IOMMU API iommu: Add IOMMU index argument to notifier APIs iommu: Add IOMMU index argument to translate method exec.c: Handle IOMMUs in address_space_translate_for_iotlb() Richard Henderson (18): target/arm: Extend vec_reg_offset to larger sizes target/arm: Implement SVE Permute - Unpredicated Group target/arm: Implement SVE Permute - Predicates Group target/arm: Implement SVE Permute - Interleaving Group target/arm: Implement SVE compress active elements target/arm: Implement SVE conditionally broadcast/extract element target/arm: Implement SVE copy to vector (predicated) target/arm: Implement SVE reverse within elements target/arm: Implement SVE vector splice (predicated) target/arm: Implement SVE Select Vectors Group target/arm: Implement SVE Integer Compare - Vectors Group target/arm: Implement SVE Integer Compare - Immediate Group target/arm: Implement SVE Partition Break Group target/arm: Implement SVE Predicate Count Group target/arm: Implement SVE Integer Compare - Scalars Group target/arm: Implement FDUP/DUP target/arm: Implement SVE Integer Wide Immediate - Unpredicated Group target/arm: Implement SVE Floating Point Arithmetic - Unpredicated Group Shannon Zhao (1): arm_gicv3_kvm: kvm_dist_get/put_priority: skip the registers banked by GICR_IPRIORITYR include/exec/cpu-all.h | 4 + include/exec/cpu-defs.h | 9 + include/exec/exec-all.h | 16 +- include/exec/memory.h | 65 +- include/hw/arm/arm.h | 8 +- include/hw/or-irq.h | 5 +- include/qemu/bswap.h | 52 ++ include/qom/cpu.h | 3 + target/arm/helper-sve.h | 294 +++++++++ target/arm/helper.h | 19 + target/arm/translate-a64.h | 26 +- accel/tcg/cputlb.c | 59 +- exec.c | 263 ++++---- hw/alpha/typhoon.c | 3 +- hw/arm/armv7m.c | 28 +- hw/arm/mps2-tz.c | 32 +- hw/arm/smmuv3.c | 2 +- hw/arm/stellaris.c | 12 +- hw/block/m25p80.c | 1 + hw/block/pflash_cfi02.c | 97 +-- hw/char/parallel.c | 50 +- hw/core/or-irq.c | 39 +- hw/dma/rc4030.c | 2 +- hw/i386/amd_iommu.c | 2 +- hw/i386/intel_iommu.c | 8 +- hw/input/pckbd.c | 14 +- hw/intc/arm_gicv3_kvm.c | 18 +- hw/intc/armv7m_nvic.c | 6 +- hw/m68k/mcf5206.c | 48 +- hw/misc/aspeed_scu.c | 20 + hw/ppc/spapr_iommu.c | 5 +- hw/s390x/s390-pci-bus.c | 2 +- hw/s390x/s390-pci-inst.c | 4 +- hw/sh4/sh7750.c | 44 +- hw/sparc/sun4m_iommu.c | 3 +- hw/sparc64/sun4u_iommu.c | 2 +- hw/vfio/common.c | 6 +- hw/virtio/vhost.c | 7 +- hw/watchdog/wdt_i6300esb.c | 48 +- memory.c | 33 +- target/arm/cpu.c | 18 + target/arm/sve_helper.c | 1250 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 1458 +++++++++++++++++++++++++++++++++++++++++++ target/arm/translate.c | 43 +- target/arm/vec_helper.c | 69 ++ CODING_STYLE | 17 + docs/devel/loads-stores.rst | 15 + target/arm/sve.decode | 248 ++++++++ 48 files changed, 4114 insertions(+), 363 deletions(-) From: Shannon Zhao <zhaoshenglong@huawei.com> While for_each_dist_irq_reg loop starts from GIC_INTERNAL, it forgot to offset the date array and index. This will overlap the GICR registers value and leave the last GIC_INTERNAL irq's registers out of update. Fixes: 367b9f527becdd20ddf116e17a3c0c2bbc486920 Cc: qemu-stable@nongnu.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/intc/arm_gicv3_kvm.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/hw/intc/arm_gicv3_kvm.c b/hw/intc/arm_gicv3_kvm.c index XXXXXXX..XXXXXXX 100644 --- a/hw/intc/arm_gicv3_kvm.c +++ b/hw/intc/arm_gicv3_kvm.c @@ -XXX,XX +XXX,XX @@ static void kvm_dist_get_priority(GICv3State *s, uint32_t offset, uint8_t *bmp) uint32_t reg, *field; int irq; - field = (uint32_t *)bmp; + /* For the KVM GICv3, affinity routing is always enabled, and the first 8 + * GICD_IPRIORITYR<n> registers are always RAZ/WI. The corresponding + * functionality is replaced by GICR_IPRIORITYR<n>. It doesn't need to + * sync them. So it needs to skip the field of GIC_INTERNAL irqs in bmp and + * offset. + */ + field = (uint32_t *)(bmp + GIC_INTERNAL); + offset += (GIC_INTERNAL * 8) / 8; for_each_dist_irq_reg(irq, s->num_irq, 8) { kvm_gicd_access(s, offset, &reg, false); *field = reg; @@ -XXX,XX +XXX,XX @@ static void kvm_dist_put_priority(GICv3State *s, uint32_t offset, uint8_t *bmp) uint32_t reg, *field; int irq; - field = (uint32_t *)bmp; + /* For the KVM GICv3, affinity routing is always enabled, and the first 8 + * GICD_IPRIORITYR<n> registers are always RAZ/WI. The corresponding + * functionality is replaced by GICR_IPRIORITYR<n>. It doesn't need to + * sync them. So it needs to skip the field of GIC_INTERNAL irqs in bmp and + * offset. + */ + field = (uint32_t *)(bmp + GIC_INTERNAL); + offset += (GIC_INTERNAL * 8) / 8; for_each_dist_irq_reg(irq, s->num_irq, 8) { reg = *field; kvm_gicd_access(s, offset, &reg, true); -- 2.17.1 The ethernet controller in the AN505 MPC FPGA image is behind the same AHB Peripheral Protection Controller that handles the graphics and GPIOs. (In the documentation this is clear in the block diagram but the ethernet controller was omitted from the table listing devices connected to the PPC.) The ethernet sits behind AHB PPCEXP0 interface 5. We had incorrectly claimed that this was a "gpio4", but there are only 4 GPIOs in this image. Correct the QEMU model to match the hardware. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20180515171446.10834-1-peter.maydell@linaro.org --- hw/arm/mps2-tz.c | 32 +++++++++++++++++++++++--------- 1 file changed, 23 insertions(+), 9 deletions(-) diff --git a/hw/arm/mps2-tz.c b/hw/arm/mps2-tz.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/mps2-tz.c +++ b/hw/arm/mps2-tz.c @@ -XXX,XX +XXX,XX @@ typedef struct { UnimplementedDeviceState spi[5]; UnimplementedDeviceState i2c[4]; UnimplementedDeviceState i2s_audio; - UnimplementedDeviceState gpio[5]; + UnimplementedDeviceState gpio[4]; UnimplementedDeviceState dma[4]; UnimplementedDeviceState gfx; CMSDKAPBUART uart[5]; SplitIRQ sec_resp_splitter; qemu_or_irq uart_irq_orgate; + DeviceState *lan9118; } MPS2TZMachineState; #define TYPE_MPS2TZ_MACHINE "mps2tz" @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_fpgaio(MPS2TZMachineState *mms, void *opaque, return sysbus_mmio_get_region(SYS_BUS_DEVICE(fpgaio), 0); } +static MemoryRegion *make_eth_dev(MPS2TZMachineState *mms, void *opaque, + const char *name, hwaddr size) +{ + SysBusDevice *s; + DeviceState *iotkitdev = DEVICE(&mms->iotkit); + NICInfo *nd = &nd_table[0]; + + /* In hardware this is a LAN9220; the LAN9118 is software compatible + * except that it doesn't support the checksum-offload feature. + */ + qemu_check_nic_model(nd, "lan9118"); + mms->lan9118 = qdev_create(NULL, "lan9118"); + qdev_set_nic_properties(mms->lan9118, nd); + qdev_init_nofail(mms->lan9118); + + s = SYS_BUS_DEVICE(mms->lan9118); + sysbus_connect_irq(s, 0, qdev_get_gpio_in_named(iotkitdev, "EXP_IRQ", 16)); + return sysbus_mmio_get_region(s, 0); +} + static void mps2tz_common_init(MachineState *machine) { MPS2TZMachineState *mms = MPS2TZ_MACHINE(machine); @@ -XXX,XX +XXX,XX @@ static void mps2tz_common_init(MachineState *machine) { "gpio1", make_unimp_dev, &mms->gpio[1], 0x40101000, 0x1000 }, { "gpio2", make_unimp_dev, &mms->gpio[2], 0x40102000, 0x1000 }, { "gpio3", make_unimp_dev, &mms->gpio[3], 0x40103000, 0x1000 }, - { "gpio4", make_unimp_dev, &mms->gpio[4], 0x40104000, 0x1000 }, + { "eth", make_eth_dev, NULL, 0x42000000, 0x100000 }, }, }, { .name = "ahb_ppcexp1", @@ -XXX,XX +XXX,XX @@ static void mps2tz_common_init(MachineState *machine) "cfg_sec_resp", 0)); } - /* In hardware this is a LAN9220; the LAN9118 is software compatible - * except that it doesn't support the checksum-offload feature. - * The ethernet controller is not behind a PPC. - */ - lan9118_init(&nd_table[0], 0x42000000, - qdev_get_gpio_in_named(iotkitdev, "EXP_IRQ", 16)); - create_unimplemented_device("FPGA NS PC", 0x48007000, 0x1000); armv7m_load_kernel(ARM_CPU(first_cpu), machine->kernel_filename, 0x400000); -- 2.17.1 Convert the sh7750 device away from using the old_mmio field of MemoryRegionOps. This device is used by the sh4 r2d board. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180601141223.26630-2-peter.maydell@linaro.org --- hw/sh4/sh7750.c | 44 ++++++++++++++++++++++++++++++++++++-------- 1 file changed, 36 insertions(+), 8 deletions(-) diff --git a/hw/sh4/sh7750.c b/hw/sh4/sh7750.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sh4/sh7750.c +++ b/hw/sh4/sh7750.c @@ -XXX,XX +XXX,XX @@ static void sh7750_mem_writel(void *opaque, hwaddr addr, } } +static uint64_t sh7750_mem_readfn(void *opaque, hwaddr addr, unsigned size) +{ + switch (size) { + case 1: + return sh7750_mem_readb(opaque, addr); + case 2: + return sh7750_mem_readw(opaque, addr); + case 4: + return sh7750_mem_readl(opaque, addr); + default: + g_assert_not_reached(); + } +} + +static void sh7750_mem_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) +{ + switch (size) { + case 1: + sh7750_mem_writeb(opaque, addr, value); + break; + case 2: + sh7750_mem_writew(opaque, addr, value); + break; + case 4: + sh7750_mem_writel(opaque, addr, value); + break; + default: + g_assert_not_reached(); + } +} + static const MemoryRegionOps sh7750_mem_ops = { - .old_mmio = { - .read = {sh7750_mem_readb, - sh7750_mem_readw, - sh7750_mem_readl }, - .write = {sh7750_mem_writeb, - sh7750_mem_writew, - sh7750_mem_writel }, - }, + .read = sh7750_mem_readfn, + .write = sh7750_mem_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, }; -- 2.17.1 Convert the mcf5206 device away from using the old_mmio field of MemoryRegionOps. This device is used by the an5206 board. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Acked-by: Thomas Huth <huth@tuxfamily.org> Message-id: 20180601141223.26630-3-peter.maydell@linaro.org --- hw/m68k/mcf5206.c | 48 +++++++++++++++++++++++++++++++++++------------ 1 file changed, 36 insertions(+), 12 deletions(-) diff --git a/hw/m68k/mcf5206.c b/hw/m68k/mcf5206.c index XXXXXXX..XXXXXXX 100644 --- a/hw/m68k/mcf5206.c +++ b/hw/m68k/mcf5206.c @@ -XXX,XX +XXX,XX @@ static void m5206_mbar_writel(void *opaque, hwaddr offset, m5206_mbar_write(s, offset, value, 4); } +static uint64_t m5206_mbar_readfn(void *opaque, hwaddr addr, unsigned size) +{ + switch (size) { + case 1: + return m5206_mbar_readb(opaque, addr); + case 2: + return m5206_mbar_readw(opaque, addr); + case 4: + return m5206_mbar_readl(opaque, addr); + default: + g_assert_not_reached(); + } +} + +static void m5206_mbar_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) +{ + switch (size) { + case 1: + m5206_mbar_writeb(opaque, addr, value); + break; + case 2: + m5206_mbar_writew(opaque, addr, value); + break; + case 4: + m5206_mbar_writel(opaque, addr, value); + break; + default: + g_assert_not_reached(); + } +} + static const MemoryRegionOps m5206_mbar_ops = { - .old_mmio = { - .read = { - m5206_mbar_readb, - m5206_mbar_readw, - m5206_mbar_readl, - }, - .write = { - m5206_mbar_writeb, - m5206_mbar_writew, - m5206_mbar_writel, - }, - }, + .read = m5206_mbar_readfn, + .write = m5206_mbar_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, }; -- 2.17.1 Convert the pflash_cfi02 device away from using the old_mmio field of MemoryRegionOps. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Acked-by: Max Reitz <mreitz@redhat.com> Message-id: 20180601141223.26630-4-peter.maydell@linaro.org --- hw/block/pflash_cfi02.c | 97 ++++++++--------------------------------- 1 file changed, 18 insertions(+), 79 deletions(-) diff --git a/hw/block/pflash_cfi02.c b/hw/block/pflash_cfi02.c index XXXXXXX..XXXXXXX 100644 --- a/hw/block/pflash_cfi02.c +++ b/hw/block/pflash_cfi02.c @@ -XXX,XX +XXX,XX @@ static void pflash_write (pflash_t *pfl, hwaddr offset, pfl->cmd = 0; } - -static uint32_t pflash_readb_be(void *opaque, hwaddr addr) +static uint64_t pflash_be_readfn(void *opaque, hwaddr addr, unsigned size) { - return pflash_read(opaque, addr, 1, 1); + return pflash_read(opaque, addr, size, 1); } -static uint32_t pflash_readb_le(void *opaque, hwaddr addr) +static void pflash_be_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) { - return pflash_read(opaque, addr, 1, 0); + pflash_write(opaque, addr, value, size, 1); } -static uint32_t pflash_readw_be(void *opaque, hwaddr addr) +static uint64_t pflash_le_readfn(void *opaque, hwaddr addr, unsigned size) { - pflash_t *pfl = opaque; - - return pflash_read(pfl, addr, 2, 1); + return pflash_read(opaque, addr, size, 0); } -static uint32_t pflash_readw_le(void *opaque, hwaddr addr) +static void pflash_le_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) { - pflash_t *pfl = opaque; - - return pflash_read(pfl, addr, 2, 0); -} - -static uint32_t pflash_readl_be(void *opaque, hwaddr addr) -{ - pflash_t *pfl = opaque; - - return pflash_read(pfl, addr, 4, 1); -} - -static uint32_t pflash_readl_le(void *opaque, hwaddr addr) -{ - pflash_t *pfl = opaque; - - return pflash_read(pfl, addr, 4, 0); -} - -static void pflash_writeb_be(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_write(opaque, addr, value, 1, 1); -} - -static void pflash_writeb_le(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_write(opaque, addr, value, 1, 0); -} - -static void pflash_writew_be(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_t *pfl = opaque; - - pflash_write(pfl, addr, value, 2, 1); -} - -static void pflash_writew_le(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_t *pfl = opaque; - - pflash_write(pfl, addr, value, 2, 0); -} - -static void pflash_writel_be(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_t *pfl = opaque; - - pflash_write(pfl, addr, value, 4, 1); -} - -static void pflash_writel_le(void *opaque, hwaddr addr, - uint32_t value) -{ - pflash_t *pfl = opaque; - - pflash_write(pfl, addr, value, 4, 0); + pflash_write(opaque, addr, value, size, 0); } static const MemoryRegionOps pflash_cfi02_ops_be = { - .old_mmio = { - .read = { pflash_readb_be, pflash_readw_be, pflash_readl_be, }, - .write = { pflash_writeb_be, pflash_writew_be, pflash_writel_be, }, - }, + .read = pflash_be_readfn, + .write = pflash_be_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, }; static const MemoryRegionOps pflash_cfi02_ops_le = { - .old_mmio = { - .read = { pflash_readb_le, pflash_readw_le, pflash_readl_le, }, - .write = { pflash_writeb_le, pflash_writew_le, pflash_writel_le, }, - }, + .read = pflash_le_readfn, + .write = pflash_le_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, }; -- 2.17.1 Convert the wdt_i6300esb device away from using the old_mmio field of MemoryRegionOps. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180601141223.26630-5-peter.maydell@linaro.org --- hw/watchdog/wdt_i6300esb.c | 48 ++++++++++++++++++++++++++++---------- 1 file changed, 36 insertions(+), 12 deletions(-) diff --git a/hw/watchdog/wdt_i6300esb.c b/hw/watchdog/wdt_i6300esb.c index XXXXXXX..XXXXXXX 100644 --- a/hw/watchdog/wdt_i6300esb.c +++ b/hw/watchdog/wdt_i6300esb.c @@ -XXX,XX +XXX,XX @@ static void i6300esb_mem_writel(void *vp, hwaddr addr, uint32_t val) } } +static uint64_t i6300esb_mem_readfn(void *opaque, hwaddr addr, unsigned size) +{ + switch (size) { + case 1: + return i6300esb_mem_readb(opaque, addr); + case 2: + return i6300esb_mem_readw(opaque, addr); + case 4: + return i6300esb_mem_readl(opaque, addr); + default: + g_assert_not_reached(); + } +} + +static void i6300esb_mem_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) +{ + switch (size) { + case 1: + i6300esb_mem_writeb(opaque, addr, value); + break; + case 2: + i6300esb_mem_writew(opaque, addr, value); + break; + case 4: + i6300esb_mem_writel(opaque, addr, value); + break; + default: + g_assert_not_reached(); + } +} + static const MemoryRegionOps i6300esb_ops = { - .old_mmio = { - .read = { - i6300esb_mem_readb, - i6300esb_mem_readw, - i6300esb_mem_readl, - }, - .write = { - i6300esb_mem_writeb, - i6300esb_mem_writew, - i6300esb_mem_writel, - }, - }, + .read = i6300esb_mem_readfn, + .write = i6300esb_mem_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_LITTLE_ENDIAN, }; -- 2.17.1 Convert the pckbd device away from using the old_mmio field of MemoryRegionOps. This change only affects the memory-mapped variant of the i8042, which is used by the Unicore32 'puv3' board and the MIPS Jazz boards 'magnum' and 'pica61'. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180601141223.26630-6-peter.maydell@linaro.org --- hw/input/pckbd.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/hw/input/pckbd.c b/hw/input/pckbd.c index XXXXXXX..XXXXXXX 100644 --- a/hw/input/pckbd.c +++ b/hw/input/pckbd.c @@ -XXX,XX +XXX,XX @@ static const VMStateDescription vmstate_kbd = { }; /* Memory mapped interface */ -static uint32_t kbd_mm_readb (void *opaque, hwaddr addr) +static uint64_t kbd_mm_readfn(void *opaque, hwaddr addr, unsigned size) { KBDState *s = opaque; @@ -XXX,XX +XXX,XX @@ static uint32_t kbd_mm_readb (void *opaque, hwaddr addr) return kbd_read_data(s, 0, 1) & 0xff; } -static void kbd_mm_writeb (void *opaque, hwaddr addr, uint32_t value) +static void kbd_mm_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) { KBDState *s = opaque; @@ -XXX,XX +XXX,XX @@ static void kbd_mm_writeb (void *opaque, hwaddr addr, uint32_t value) kbd_write_data(s, 0, value & 0xff, 1); } + static const MemoryRegionOps i8042_mmio_ops = { + .read = kbd_mm_readfn, + .write = kbd_mm_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, - .old_mmio = { - .read = { kbd_mm_readb, kbd_mm_readb, kbd_mm_readb }, - .write = { kbd_mm_writeb, kbd_mm_writeb, kbd_mm_writeb }, - }, }; void i8042_mm_init(qemu_irq kbd_irq, qemu_irq mouse_irq, -- 2.17.1 Convert the parallel device away from using the old_mmio field of MemoryRegionOps. This change only affects the memory-mapped variant, which is used by the MIPS Jazz boards 'magnum' and 'pica61'. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180601141223.26630-7-peter.maydell@linaro.org --- hw/char/parallel.c | 50 ++++++++++------------------------------------ 1 file changed, 11 insertions(+), 39 deletions(-) diff --git a/hw/char/parallel.c b/hw/char/parallel.c index XXXXXXX..XXXXXXX 100644 --- a/hw/char/parallel.c +++ b/hw/char/parallel.c @@ -XXX,XX +XXX,XX @@ static void parallel_isa_realizefn(DeviceState *dev, Error **errp) } /* Memory mapped interface */ -static uint32_t parallel_mm_readb (void *opaque, hwaddr addr) +static uint64_t parallel_mm_readfn(void *opaque, hwaddr addr, unsigned size) { ParallelState *s = opaque; - return parallel_ioport_read_sw(s, addr >> s->it_shift) & 0xFF; + return parallel_ioport_read_sw(s, addr >> s->it_shift) & + MAKE_64BIT_MASK(0, size * 8); } -static void parallel_mm_writeb (void *opaque, - hwaddr addr, uint32_t value) +static void parallel_mm_writefn(void *opaque, hwaddr addr, + uint64_t value, unsigned size) { ParallelState *s = opaque; - parallel_ioport_write_sw(s, addr >> s->it_shift, value & 0xFF); -} - -static uint32_t parallel_mm_readw (void *opaque, hwaddr addr) -{ - ParallelState *s = opaque; - - return parallel_ioport_read_sw(s, addr >> s->it_shift) & 0xFFFF; -} - -static void parallel_mm_writew (void *opaque, - hwaddr addr, uint32_t value) -{ - ParallelState *s = opaque; - - parallel_ioport_write_sw(s, addr >> s->it_shift, value & 0xFFFF); -} - -static uint32_t parallel_mm_readl (void *opaque, hwaddr addr) -{ - ParallelState *s = opaque; - - return parallel_ioport_read_sw(s, addr >> s->it_shift); -} - -static void parallel_mm_writel (void *opaque, - hwaddr addr, uint32_t value) -{ - ParallelState *s = opaque; - - parallel_ioport_write_sw(s, addr >> s->it_shift, value); + parallel_ioport_write_sw(s, addr >> s->it_shift, + value & MAKE_64BIT_MASK(0, size * 8)); } static const MemoryRegionOps parallel_mm_ops = { - .old_mmio = { - .read = { parallel_mm_readb, parallel_mm_readw, parallel_mm_readl }, - .write = { parallel_mm_writeb, parallel_mm_writew, parallel_mm_writel }, - }, + .read = parallel_mm_readfn, + .write = parallel_mm_writefn, + .valid.min_access_size = 1, + .valid.max_access_size = 4, .endianness = DEVICE_NATIVE_ENDIAN, }; -- 2.17.1 The stellaris board is still using the legacy armv7m_init() function, which predates conversion of the ARMv7M into a proper QOM container object. Make the board code directly create the ARMv7M object instead. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20180601144328.23817-2-peter.maydell@linaro.org --- hw/arm/stellaris.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/hw/arm/stellaris.c b/hw/arm/stellaris.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/stellaris.c +++ b/hw/arm/stellaris.c @@ -XXX,XX +XXX,XX @@ #include "qemu/log.h" #include "exec/address-spaces.h" #include "sysemu/sysemu.h" +#include "hw/arm/armv7m.h" #include "hw/char/pl011.h" #include "hw/misc/unimp.h" #include "cpu.h" @@ -XXX,XX +XXX,XX @@ static void stellaris_init(MachineState *ms, stellaris_board_info *board) &error_fatal); memory_region_add_subregion(system_memory, 0x20000000, sram); - nvic = armv7m_init(system_memory, flash_size, NUM_IRQ_LINES, - ms->kernel_filename, ms->cpu_type); + nvic = qdev_create(NULL, TYPE_ARMV7M); + qdev_prop_set_uint32(nvic, "num-irq", NUM_IRQ_LINES); + qdev_prop_set_string(nvic, "cpu-type", ms->cpu_type); + object_property_set_link(OBJECT(nvic), OBJECT(get_system_memory()), + "memory", &error_abort); + /* This will exit with an error if the user passed us a bad cpu_type */ + qdev_init_nofail(nvic); qdev_connect_gpio_out_named(nvic, "SYSRESETREQ", 0, qemu_allocate_irq(&do_sys_reset, NULL, 0)); @@ -XXX,XX +XXX,XX @@ static void stellaris_init(MachineState *ms, stellaris_board_info *board) create_unimplemented_device("analogue-comparator", 0x4003c000, 0x1000); create_unimplemented_device("hibernation", 0x400fc000, 0x1000); create_unimplemented_device("flash-control", 0x400fd000, 0x1000); + + armv7m_load_kernel(ARM_CPU(first_cpu), ms->kernel_filename, flash_size); } /* FIXME: Figure out how to generate these from stellaris_boards. */ -- 2.17.1 Remove the now-unused armv7m_init() function. This was a legacy from before we properly QOMified ARMv7M, and it has some flaws: * it combines work that needs to be done by an SoC object (creating and initializing the TYPE_ARMV7M object) with work that needs to be done by the board model (setting the system up to load the ELF file specified with -kernel) * TYPE_ARMV7M creation failure is fatal, but an SoC object wants to arrange to propagate the failure outward * it uses allocate-and-create via qdev_create() whereas the current preferred style for SoC objects is to do creation in-place Board and SoC models can instead do the two jobs this function was doing themselves, in the right places and with whatever their preferred style/error handling is. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20180601144328.23817-3-peter.maydell@linaro.org --- include/hw/arm/arm.h | 8 ++------ hw/arm/armv7m.c | 21 --------------------- 2 files changed, 2 insertions(+), 27 deletions(-) diff --git a/include/hw/arm/arm.h b/include/hw/arm/arm.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/arm.h +++ b/include/hw/arm/arm.h @@ -XXX,XX +XXX,XX @@ typedef enum { ARM_ENDIANNESS_BE32, } arm_endianness; -/* armv7m.c */ -DeviceState *armv7m_init(MemoryRegion *system_memory, int mem_size, int num_irq, - const char *kernel_filename, const char *cpu_type); /** * armv7m_load_kernel: * @cpu: CPU @@ -XXX,XX +XXX,XX @@ DeviceState *armv7m_init(MemoryRegion *system_memory, int mem_size, int num_irq, * @mem_size: mem_size: maximum image size to load * * Load the guest image for an ARMv7M system. This must be called by - * any ARMv7M board, either directly or via armv7m_init(). (This is - * necessary to ensure that the CPU resets correctly on system reset, - * as well as for kernel loading.) + * any ARMv7M board. (This is necessary to ensure that the CPU resets + * correctly on system reset, as well as for kernel loading.) */ void armv7m_load_kernel(ARMCPU *cpu, const char *kernel_filename, int mem_size); diff --git a/hw/arm/armv7m.c b/hw/arm/armv7m.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/armv7m.c +++ b/hw/arm/armv7m.c @@ -XXX,XX +XXX,XX @@ static void armv7m_reset(void *opaque) cpu_reset(CPU(cpu)); } -/* Init CPU and memory for a v7-M based board. - mem_size is in bytes. - Returns the ARMv7M device. */ - -DeviceState *armv7m_init(MemoryRegion *system_memory, int mem_size, int num_irq, - const char *kernel_filename, const char *cpu_type) -{ - DeviceState *armv7m; - - armv7m = qdev_create(NULL, TYPE_ARMV7M); - qdev_prop_set_uint32(armv7m, "num-irq", num_irq); - qdev_prop_set_string(armv7m, "cpu-type", cpu_type); - object_property_set_link(OBJECT(armv7m), OBJECT(get_system_memory()), - "memory", &error_abort); - /* This will exit with an error if the user passed us a bad cpu_type */ - qdev_init_nofail(armv7m); - - armv7m_load_kernel(ARM_CPU(first_cpu), kernel_filename, mem_size); - return armv7m; -} - void armv7m_load_kernel(ARMCPU *cpu, const char *kernel_filename, int mem_size) { int image_size; -- 2.17.1 The Cortex-M CPU and its NVIC are two intimately intertwined parts of the same hardware; it is not possible to use one without the other. Unfortunately a lot of our board models don't do any sanity checking on the CPU type the user asks for, so a command line like qemu-system-arm -M versatilepb -cpu cortex-m3 will create an M3 without an NVIC, and coredump immediately. In the other direction, trying a non-M-profile CPU in an M-profile board won't blow up, but doesn't do anything useful either: qemu-system-arm -M lm3s6965evb -cpu arm926 Add some checking in the NVIC and CPU realize functions that the user isn't trying to use an NVIC without an M-profile CPU or an M-profile CPU without an NVIC, so we can produce a helpful error message rather than a core dump. Fixes: https://bugs.launchpad.net/qemu/+bug/1766896 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20180601160355.15393-1-peter.maydell@linaro.org --- hw/arm/armv7m.c | 7 ++++++- hw/intc/armv7m_nvic.c | 6 +++++- target/arm/cpu.c | 18 ++++++++++++++++++ 3 files changed, 29 insertions(+), 2 deletions(-) diff --git a/hw/arm/armv7m.c b/hw/arm/armv7m.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/armv7m.c +++ b/hw/arm/armv7m.c @@ -XXX,XX +XXX,XX @@ static void armv7m_realize(DeviceState *dev, Error **errp) return; } } + + /* Tell the CPU where the NVIC is; it will fail realize if it doesn't + * have one. + */ + s->cpu->env.nvic = &s->nvic; + object_property_set_bool(OBJECT(s->cpu), true, "realized", &err); if (err != NULL) { error_propagate(errp, err); @@ -XXX,XX +XXX,XX @@ static void armv7m_realize(DeviceState *dev, Error **errp) sbd = SYS_BUS_DEVICE(&s->nvic); sysbus_connect_irq(sbd, 0, qdev_get_gpio_in(DEVICE(s->cpu), ARM_CPU_IRQ)); - s->cpu->env.nvic = &s->nvic; memory_region_add_subregion(&s->container, 0xe000e000, sysbus_mmio_get_region(sbd, 0)); diff --git a/hw/intc/armv7m_nvic.c b/hw/intc/armv7m_nvic.c index XXXXXXX..XXXXXXX 100644 --- a/hw/intc/armv7m_nvic.c +++ b/hw/intc/armv7m_nvic.c @@ -XXX,XX +XXX,XX @@ static void armv7m_nvic_realize(DeviceState *dev, Error **errp) int regionlen; s->cpu = ARM_CPU(qemu_get_cpu(0)); - assert(s->cpu); + + if (!s->cpu || !arm_feature(&s->cpu->env, ARM_FEATURE_M)) { + error_setg(errp, "The NVIC can only be used with a Cortex-M CPU"); + return; + } if (s->num_irq > NVIC_MAX_IRQ) { error_setg(errp, "num-irq %d exceeds NVIC maximum", s->num_irq); 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) return; } +#ifndef CONFIG_USER_ONLY + /* The NVIC and M-profile CPU are two halves of a single piece of + * hardware; trying to use one without the other is a command line + * error and will result in segfaults if not caught here. + */ + if (arm_feature(env, ARM_FEATURE_M)) { + if (!env->nvic) { + error_setg(errp, "This board cannot be used with Cortex-M CPUs"); + return; + } + } else { + if (env->nvic) { + error_setg(errp, "This board can only be used with Cortex-M CPUs"); + return; + } + } +#endif + cpu_exec_realizefn(cs, &local_err); if (local_err != NULL) { error_propagate(errp, local_err); -- 2.17.1 For the IoTKit MPC support, we need to wire together the interrupt outputs of 17 MPCs; this exceeds the current value of MAX_OR_LINES. Increase MAX_OR_LINES to 32 (which should be enough for anyone). The tricky part is retaining the migration compatibility for existing OR gates; we add a subsection which is only used for larger OR gates, and define it such that we can freely increase MAX_OR_LINES in future (or even move to a dynamically allocated levels[] array without an upper size limit) without breaking compatibility. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180604152941.20374-10-peter.maydell@linaro.org --- include/hw/or-irq.h | 5 ++++- hw/core/or-irq.c | 39 +++++++++++++++++++++++++++++++++++++-- 2 files changed, 41 insertions(+), 3 deletions(-) diff --git a/include/hw/or-irq.h b/include/hw/or-irq.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/or-irq.h +++ b/include/hw/or-irq.h @@ -XXX,XX +XXX,XX @@ #define TYPE_OR_IRQ "or-irq" -#define MAX_OR_LINES 16 +/* This can safely be increased if necessary without breaking + * migration compatibility (as long as it remains greater than 15). + */ +#define MAX_OR_LINES 32 typedef struct OrIRQState qemu_or_irq; diff --git a/hw/core/or-irq.c b/hw/core/or-irq.c index XXXXXXX..XXXXXXX 100644 --- a/hw/core/or-irq.c +++ b/hw/core/or-irq.c @@ -XXX,XX +XXX,XX @@ static void or_irq_init(Object *obj) qdev_init_gpio_out(DEVICE(obj), &s->out_irq, 1); } +/* The original version of this device had a fixed 16 entries in its + * VMState array; devices with more inputs than this need to + * migrate the extra lines via a subsection. + * The subsection migrates as much of the levels[] array as is needed + * (including repeating the first 16 elements), to avoid the awkwardness + * of splitting it in two to meet the requirements of VMSTATE_VARRAY_UINT16. + */ +#define OLD_MAX_OR_LINES 16 +#if MAX_OR_LINES < OLD_MAX_OR_LINES +#error MAX_OR_LINES must be at least 16 for migration compatibility +#endif + +static bool vmstate_extras_needed(void *opaque) +{ + qemu_or_irq *s = OR_IRQ(opaque); + + return s->num_lines >= OLD_MAX_OR_LINES; +} + +static const VMStateDescription vmstate_or_irq_extras = { + .name = "or-irq-extras", + .version_id = 1, + .minimum_version_id = 1, + .needed = vmstate_extras_needed, + .fields = (VMStateField[]) { + VMSTATE_VARRAY_UINT16_UNSAFE(levels, qemu_or_irq, num_lines, 0, + vmstate_info_bool, bool), + VMSTATE_END_OF_LIST(), + }, +}; + static const VMStateDescription vmstate_or_irq = { .name = TYPE_OR_IRQ, .version_id = 1, .minimum_version_id = 1, .fields = (VMStateField[]) { - VMSTATE_BOOL_ARRAY(levels, qemu_or_irq, MAX_OR_LINES), + VMSTATE_BOOL_SUB_ARRAY(levels, qemu_or_irq, 0, OLD_MAX_OR_LINES), VMSTATE_END_OF_LIST(), - } + }, + .subsections = (const VMStateDescription*[]) { + &vmstate_or_irq_extras, + NULL + }, }; static Property or_irq_properties[] = { -- 2.17.1 The 'addr' field in the CPUIOTLBEntry struct has a rather non-obvious use; add a comment documenting it (reverse-engineered from what the code that sets it is doing). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180611125633.32755-2-peter.maydell@linaro.org --- include/exec/cpu-defs.h | 9 +++++++++ accel/tcg/cputlb.c | 12 ++++++++++++ 2 files changed, 21 insertions(+) diff --git a/include/exec/cpu-defs.h b/include/exec/cpu-defs.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/cpu-defs.h +++ b/include/exec/cpu-defs.h @@ -XXX,XX +XXX,XX @@ QEMU_BUILD_BUG_ON(sizeof(CPUTLBEntry) != (1 << CPU_TLB_ENTRY_BITS)); * structs into one.) */ typedef struct CPUIOTLBEntry { + /* + * @addr contains: + * - in the lower TARGET_PAGE_BITS, a physical section number + * - with the lower TARGET_PAGE_BITS masked off, an offset which + * must be added to the virtual address to obtain: + * + the ram_addr_t of the target RAM (if the physical section + * number is PHYS_SECTION_NOTDIRTY or PHYS_SECTION_ROM) + * + the offset within the target MemoryRegion (otherwise) + */ hwaddr addr; MemTxAttrs attrs; } CPUIOTLBEntry; diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cputlb.c +++ b/accel/tcg/cputlb.c @@ -XXX,XX +XXX,XX @@ void tlb_set_page_with_attrs(CPUState *cpu, target_ulong vaddr, env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index]; /* refill the tlb */ + /* + * At this point iotlb contains a physical section number in the lower + * TARGET_PAGE_BITS, and either + * + the ram_addr_t of the page base of the target RAM (if NOTDIRTY or ROM) + * + the offset within section->mr of the page base (otherwise) + * We subtract the vaddr (which is page aligned and thus won't + * disturb the low bits) to give an offset which can be added to the + * (non-page-aligned) vaddr of the eventual memory access to get + * the MemoryRegion offset for the access. Note that the vaddr we + * subtract here is that of the page base, and not the same as the + * vaddr we add back in io_readx()/io_writex()/get_page_addr_code(). + */ env->iotlb[mmu_idx][index].addr = iotlb - vaddr; env->iotlb[mmu_idx][index].attrs = attrs; -- 2.17.1 The API for cpu_transaction_failed() says that it takes the physical address for the failed transaction. However we were actually passing it the offset within the target MemoryRegion. We don't currently have any target CPU implementations of this hook that require the physical address; fix this bug so we don't get confused if we ever do add one. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180611125633.32755-3-peter.maydell@linaro.org --- include/exec/exec-all.h | 13 ++++++++++-- accel/tcg/cputlb.c | 44 +++++++++++++++++++++++++++++------------ exec.c | 5 +++-- 3 files changed, 45 insertions(+), 17 deletions(-) diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/exec-all.h +++ b/include/exec/exec-all.h @@ -XXX,XX +XXX,XX @@ void tb_lock_reset(void); #if !defined(CONFIG_USER_ONLY) -struct MemoryRegion *iotlb_to_region(CPUState *cpu, - hwaddr index, MemTxAttrs attrs); +/** + * iotlb_to_section: + * @cpu: CPU performing the access + * @index: TCG CPU IOTLB entry + * + * Given a TCG CPU IOTLB entry, return the MemoryRegionSection that + * it refers to. @index will have been initially created and returned + * by memory_region_section_get_iotlb(). + */ +struct MemoryRegionSection *iotlb_to_section(CPUState *cpu, + hwaddr index, MemTxAttrs attrs); void tlb_fill(CPUState *cpu, target_ulong addr, int size, MMUAccessType access_type, int mmu_idx, uintptr_t retaddr); diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cputlb.c +++ b/accel/tcg/cputlb.c @@ -XXX,XX +XXX,XX @@ static uint64_t io_readx(CPUArchState *env, CPUIOTLBEntry *iotlbentry, target_ulong addr, uintptr_t retaddr, int size) { CPUState *cpu = ENV_GET_CPU(env); - hwaddr physaddr = iotlbentry->addr; - MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs); + hwaddr mr_offset; + MemoryRegionSection *section; + MemoryRegion *mr; uint64_t val; bool locked = false; MemTxResult r; - physaddr = (physaddr & TARGET_PAGE_MASK) + addr; + section = iotlb_to_section(cpu, iotlbentry->addr, iotlbentry->attrs); + mr = section->mr; + mr_offset = (iotlbentry->addr & TARGET_PAGE_MASK) + addr; cpu->mem_io_pc = retaddr; if (mr != &io_mem_rom && mr != &io_mem_notdirty && !cpu->can_do_io) { cpu_io_recompile(cpu, retaddr); @@ -XXX,XX +XXX,XX @@ static uint64_t io_readx(CPUArchState *env, CPUIOTLBEntry *iotlbentry, qemu_mutex_lock_iothread(); locked = true; } - r = memory_region_dispatch_read(mr, physaddr, + r = memory_region_dispatch_read(mr, mr_offset, &val, size, iotlbentry->attrs); if (r != MEMTX_OK) { + hwaddr physaddr = mr_offset + + section->offset_within_address_space - + section->offset_within_region; + cpu_transaction_failed(cpu, physaddr, addr, size, MMU_DATA_LOAD, mmu_idx, iotlbentry->attrs, r, retaddr); } @@ -XXX,XX +XXX,XX @@ static void io_writex(CPUArchState *env, CPUIOTLBEntry *iotlbentry, uintptr_t retaddr, int size) { CPUState *cpu = ENV_GET_CPU(env); - hwaddr physaddr = iotlbentry->addr; - MemoryRegion *mr = iotlb_to_region(cpu, physaddr, iotlbentry->attrs); + hwaddr mr_offset; + MemoryRegionSection *section; + MemoryRegion *mr; bool locked = false; MemTxResult r; - physaddr = (physaddr & TARGET_PAGE_MASK) + addr; + section = iotlb_to_section(cpu, iotlbentry->addr, iotlbentry->attrs); + mr = section->mr; + mr_offset = (iotlbentry->addr & TARGET_PAGE_MASK) + addr; if (mr != &io_mem_rom && mr != &io_mem_notdirty && !cpu->can_do_io) { cpu_io_recompile(cpu, retaddr); } @@ -XXX,XX +XXX,XX @@ static void io_writex(CPUArchState *env, CPUIOTLBEntry *iotlbentry, qemu_mutex_lock_iothread(); locked = true; } - r = memory_region_dispatch_write(mr, physaddr, + r = memory_region_dispatch_write(mr, mr_offset, val, size, iotlbentry->attrs); if (r != MEMTX_OK) { + hwaddr physaddr = mr_offset + + section->offset_within_address_space - + section->offset_within_region; + cpu_transaction_failed(cpu, physaddr, addr, size, MMU_DATA_STORE, mmu_idx, iotlbentry->attrs, r, retaddr); } @@ -XXX,XX +XXX,XX @@ static bool victim_tlb_hit(CPUArchState *env, size_t mmu_idx, size_t index, */ tb_page_addr_t get_page_addr_code(CPUArchState *env, target_ulong addr) { - int mmu_idx, index, pd; + int mmu_idx, index; void *p; MemoryRegion *mr; + MemoryRegionSection *section; CPUState *cpu = ENV_GET_CPU(env); CPUIOTLBEntry *iotlbentry; - hwaddr physaddr; + hwaddr physaddr, mr_offset; index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1); mmu_idx = cpu_mmu_index(env, true); @@ -XXX,XX +XXX,XX @@ tb_page_addr_t get_page_addr_code(CPUArchState *env, target_ulong addr) } } iotlbentry = &env->iotlb[mmu_idx][index]; - pd = iotlbentry->addr & ~TARGET_PAGE_MASK; - mr = iotlb_to_region(cpu, pd, iotlbentry->attrs); + section = iotlb_to_section(cpu, iotlbentry->addr, iotlbentry->attrs); + mr = section->mr; if (memory_region_is_unassigned(mr)) { qemu_mutex_lock_iothread(); if (memory_region_request_mmio_ptr(mr, addr)) { @@ -XXX,XX +XXX,XX @@ tb_page_addr_t get_page_addr_code(CPUArchState *env, target_ulong addr) * and use the MemTXResult it produced). However it is the * simplest place we have currently available for the check. */ - physaddr = (iotlbentry->addr & TARGET_PAGE_MASK) + addr; + mr_offset = (iotlbentry->addr & TARGET_PAGE_MASK) + addr; + physaddr = mr_offset + + section->offset_within_address_space - + section->offset_within_region; cpu_transaction_failed(cpu, physaddr, addr, 0, MMU_INST_FETCH, mmu_idx, iotlbentry->attrs, MEMTX_DECODE_ERROR, 0); diff --git a/exec.c b/exec.c index XXXXXXX..XXXXXXX 100644 --- a/exec.c +++ b/exec.c @@ -XXX,XX +XXX,XX @@ static const MemoryRegionOps readonly_mem_ops = { }, }; -MemoryRegion *iotlb_to_region(CPUState *cpu, hwaddr index, MemTxAttrs attrs) +MemoryRegionSection *iotlb_to_section(CPUState *cpu, + hwaddr index, MemTxAttrs attrs) { int asidx = cpu_asidx_from_attrs(cpu, attrs); CPUAddressSpace *cpuas = &cpu->cpu_ases[asidx]; AddressSpaceDispatch *d = atomic_rcu_read(&cpuas->memory_dispatch); MemoryRegionSection *sections = d->map.sections; - return sections[index & ~TARGET_PAGE_MASK].mr; + return &sections[index & ~TARGET_PAGE_MASK]; } static void io_mem_init(void) -- 2.17.1 The codebase has a bit of a mix of different multiline comment styles. State a preference for the Linux kernel style: /* * Star on the left for each line. * Leading slash-star and trailing star-slash * each go on a line of their own. */ Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Cornelia Huck <cohuck@redhat.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> Reviewed-by: John Snow <jsnow@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20180611141716.3813-1-peter.maydell@linaro.org --- CODING_STYLE | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/CODING_STYLE b/CODING_STYLE index XXXXXXX..XXXXXXX 100644 --- a/CODING_STYLE +++ b/CODING_STYLE @@ -XXX,XX +XXX,XX @@ We use traditional C-style /* */ comments and avoid // comments. Rationale: The // form is valid in C99, so this is purely a matter of consistency of style. The checkpatch script will warn you about this. +Multiline comment blocks should have a row of stars on the left, +and the initial /* and terminating */ both on their own lines: + /* + * like + * this + */ +This is the same format required by the Linux kernel coding style. + +(Some of the existing comments in the codebase use the GNU Coding +Standards form which does not have stars on the left, or other +variations; avoid these when writing new comments, but don't worry +about converting to the preferred form unless you're editing that +comment anyway.) + +Rationale: Consistency, and ease of visually picking out a multiline +comment from the surrounding code. + 8. trace-events style 8.1 0x prefix -- 2.17.1 There's a common pattern in QEMU where a function needs to perform a data load or store of an N byte integer in a particular endianness. At the moment this is handled by doing a switch() on the size and calling the appropriate ld*_p or st*_p function for each size. Provide a new family of functions ldn_*_p() and stn_*_p() which take the size as an argument and do the switch() themselves. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180611171007.4165-2-peter.maydell@linaro.org --- include/exec/cpu-all.h | 4 +++ include/qemu/bswap.h | 52 +++++++++++++++++++++++++++++++++++++ docs/devel/loads-stores.rst | 15 +++++++++++ 3 files changed, 71 insertions(+) diff --git a/include/exec/cpu-all.h b/include/exec/cpu-all.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/cpu-all.h +++ b/include/exec/cpu-all.h @@ -XXX,XX +XXX,XX @@ static inline void tswap64s(uint64_t *s) #define stq_p(p, v) stq_be_p(p, v) #define stfl_p(p, v) stfl_be_p(p, v) #define stfq_p(p, v) stfq_be_p(p, v) +#define ldn_p(p, sz) ldn_be_p(p, sz) +#define stn_p(p, sz, v) stn_be_p(p, sz, v) #else #define lduw_p(p) lduw_le_p(p) #define ldsw_p(p) ldsw_le_p(p) @@ -XXX,XX +XXX,XX @@ static inline void tswap64s(uint64_t *s) #define stq_p(p, v) stq_le_p(p, v) #define stfl_p(p, v) stfl_le_p(p, v) #define stfq_p(p, v) stfq_le_p(p, v) +#define ldn_p(p, sz) ldn_le_p(p, sz) +#define stn_p(p, sz, v) stn_le_p(p, sz, v) #endif /* MMU memory access macros */ diff --git a/include/qemu/bswap.h b/include/qemu/bswap.h index XXXXXXX..XXXXXXX 100644 --- a/include/qemu/bswap.h +++ b/include/qemu/bswap.h @@ -XXX,XX +XXX,XX @@ typedef union { * For accessors that take a guest address rather than a * host address, see the cpu_{ld,st}_* accessors defined in * cpu_ldst.h. + * + * For cases where the size to be used is not fixed at compile time, + * there are + * stn{endian}_p(ptr, sz, val) + * which stores @val to @ptr as an @endian-order number @sz bytes in size + * and + * ldn{endian}_p(ptr, sz) + * which loads @sz bytes from @ptr as an unsigned @endian-order number + * and returns it in a uint64_t. */ static inline int ldub_p(const void *ptr) @@ -XXX,XX +XXX,XX @@ static inline unsigned long leul_to_cpu(unsigned long v) #endif } +/* Store v to p as a sz byte value in host order */ +#define DO_STN_LDN_P(END) \ + static inline void stn_## END ## _p(void *ptr, int sz, uint64_t v) \ + { \ + switch (sz) { \ + case 1: \ + stb_p(ptr, v); \ + break; \ + case 2: \ + stw_ ## END ## _p(ptr, v); \ + break; \ + case 4: \ + stl_ ## END ## _p(ptr, v); \ + break; \ + case 8: \ + stq_ ## END ## _p(ptr, v); \ + break; \ + default: \ + g_assert_not_reached(); \ + } \ + } \ + static inline uint64_t ldn_## END ## _p(const void *ptr, int sz) \ + { \ + switch (sz) { \ + case 1: \ + return ldub_p(ptr); \ + case 2: \ + return lduw_ ## END ## _p(ptr); \ + case 4: \ + return (uint32_t)ldl_ ## END ## _p(ptr); \ + case 8: \ + return ldq_ ## END ## _p(ptr); \ + default: \ + g_assert_not_reached(); \ + } \ + } + +DO_STN_LDN_P(he) +DO_STN_LDN_P(le) +DO_STN_LDN_P(be) + +#undef DO_STN_LDN_P + #undef le_bswap #undef be_bswap #undef le_bswaps diff --git a/docs/devel/loads-stores.rst b/docs/devel/loads-stores.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/devel/loads-stores.rst +++ b/docs/devel/loads-stores.rst @@ -XXX,XX +XXX,XX @@ The ``_{endian}`` infix is omitted for target-endian accesses. The target endian accessors are only available to source files which are built per-target. +There are also functions which take the size as an argument: + +load: ``ldn{endian}_p(ptr, sz)`` + +which performs an unsigned load of ``sz`` bytes from ``ptr`` +as an ``{endian}`` order value and returns it in a uint64_t. + +store: ``stn{endian}_p(ptr, sz, val)`` + +which stores ``val`` to ``ptr`` as an ``{endian}`` order value +of size ``sz`` bytes. + + Regexes for git grep - ``\<ldf\?[us]\?[bwlq]\(_[hbl]e\)\?_p\>`` - ``\<stf\?[bwlq]\(_[hbl]e\)\?_p\>`` + - ``\<ldn_\([hbl]e\)?_p\>`` + - ``\<stn_\([hbl]e\)?_p\>`` ``cpu_{ld,st}_*`` ~~~~~~~~~~~~~~~~~ -- 2.17.1 In subpage_read() we perform a load of the data into a local buffer which we then access using ldub_p(), lduw_p(), ldl_p() or ldq_p() depending on its size, storing the result into the uint64_t *data. Since ldl_p() returns an 'int', this means that for the 4-byte case we will sign-extend the data, whereas for 1 and 2 byte reads we zero-extend it. This ought not to matter since the caller will likely ignore values in the high bytes of the data, but add a cast so that we're consistent. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180611171007.4165-3-peter.maydell@linaro.org --- exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/exec.c b/exec.c index XXXXXXX..XXXXXXX 100644 --- a/exec.c +++ b/exec.c @@ -XXX,XX +XXX,XX @@ static MemTxResult subpage_read(void *opaque, hwaddr addr, uint64_t *data, *data = lduw_p(buf); return MEMTX_OK; case 4: - *data = ldl_p(buf); + *data = (uint32_t)ldl_p(buf); return MEMTX_OK; case 8: *data = ldq_p(buf); -- 2.17.1 Now we have stn_p() and ldn_p() we can use them in various functions in exec.c that used to have their own switch-on-size code. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180611171007.4165-4-peter.maydell@linaro.org --- exec.c | 112 +++++---------------------------------------------------- 1 file changed, 8 insertions(+), 104 deletions(-) diff --git a/exec.c b/exec.c index XXXXXXX..XXXXXXX 100644 --- a/exec.c +++ b/exec.c @@ -XXX,XX +XXX,XX @@ static void notdirty_mem_write(void *opaque, hwaddr ram_addr, memory_notdirty_write_prepare(&ndi, current_cpu, current_cpu->mem_io_vaddr, ram_addr, size); - switch (size) { - case 1: - stb_p(qemu_map_ram_ptr(NULL, ram_addr), val); - break; - case 2: - stw_p(qemu_map_ram_ptr(NULL, ram_addr), val); - break; - case 4: - stl_p(qemu_map_ram_ptr(NULL, ram_addr), val); - break; - case 8: - stq_p(qemu_map_ram_ptr(NULL, ram_addr), val); - break; - default: - abort(); - } + stn_p(qemu_map_ram_ptr(NULL, ram_addr), size, val); memory_notdirty_write_complete(&ndi); } @@ -XXX,XX +XXX,XX @@ static MemTxResult subpage_read(void *opaque, hwaddr addr, uint64_t *data, if (res) { return res; } - switch (len) { - case 1: - *data = ldub_p(buf); - return MEMTX_OK; - case 2: - *data = lduw_p(buf); - return MEMTX_OK; - case 4: - *data = (uint32_t)ldl_p(buf); - return MEMTX_OK; - case 8: - *data = ldq_p(buf); - return MEMTX_OK; - default: - abort(); - } + *data = ldn_p(buf, len); + return MEMTX_OK; } static MemTxResult subpage_write(void *opaque, hwaddr addr, @@ -XXX,XX +XXX,XX @@ static MemTxResult subpage_write(void *opaque, hwaddr addr, " value %"PRIx64"\n", __func__, subpage, len, addr, value); #endif - switch (len) { - case 1: - stb_p(buf, value); - break; - case 2: - stw_p(buf, value); - break; - case 4: - stl_p(buf, value); - break; - case 8: - stq_p(buf, value); - break; - default: - abort(); - } + stn_p(buf, len, value); return flatview_write(subpage->fv, addr + subpage->base, attrs, buf, len); } @@ -XXX,XX +XXX,XX @@ static MemTxResult flatview_write_continue(FlatView *fv, hwaddr addr, l = memory_access_size(mr, l, addr1); /* XXX: could force current_cpu to NULL to avoid potential bugs */ - switch (l) { - case 8: - /* 64 bit write access */ - val = ldq_p(buf); - result |= memory_region_dispatch_write(mr, addr1, val, 8, - attrs); - break; - case 4: - /* 32 bit write access */ - val = (uint32_t)ldl_p(buf); - result |= memory_region_dispatch_write(mr, addr1, val, 4, - attrs); - break; - case 2: - /* 16 bit write access */ - val = lduw_p(buf); - result |= memory_region_dispatch_write(mr, addr1, val, 2, - attrs); - break; - case 1: - /* 8 bit write access */ - val = ldub_p(buf); - result |= memory_region_dispatch_write(mr, addr1, val, 1, - attrs); - break; - default: - abort(); - } + val = ldn_p(buf, l); + result |= memory_region_dispatch_write(mr, addr1, val, l, attrs); } else { /* RAM case */ ptr = qemu_ram_ptr_length(mr->ram_block, addr1, &l, false); @@ -XXX,XX +XXX,XX @@ MemTxResult flatview_read_continue(FlatView *fv, hwaddr addr, /* I/O case */ release_lock |= prepare_mmio_access(mr); l = memory_access_size(mr, l, addr1); - switch (l) { - case 8: - /* 64 bit read access */ - result |= memory_region_dispatch_read(mr, addr1, &val, 8, - attrs); - stq_p(buf, val); - break; - case 4: - /* 32 bit read access */ - result |= memory_region_dispatch_read(mr, addr1, &val, 4, - attrs); - stl_p(buf, val); - break; - case 2: - /* 16 bit read access */ - result |= memory_region_dispatch_read(mr, addr1, &val, 2, - attrs); - stw_p(buf, val); - break; - case 1: - /* 8 bit read access */ - result |= memory_region_dispatch_read(mr, addr1, &val, 1, - attrs); - stb_p(buf, val); - break; - default: - abort(); - } + result |= memory_region_dispatch_read(mr, addr1, &val, l, attrs); + stn_p(buf, l, val); } else { /* RAM case */ ptr = qemu_ram_ptr_length(mr->ram_block, addr1, &l, false); -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Rearrange the arithmetic so that we are agnostic about the total size of the vector and the size of the element. This will allow us to index up to the 32nd byte and with 16-byte elements. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180613015641.5667-2-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 26 +++++++++++++++++--------- 1 file changed, 17 insertions(+), 9 deletions(-) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline void assert_fp_access_checked(DisasContext *s) static inline int vec_reg_offset(DisasContext *s, int regno, int element, TCGMemOp size) { - int offs = 0; + int element_size = 1 << size; + int offs = element * element_size; #ifdef HOST_WORDS_BIGENDIAN /* This is complicated slightly because vfp.zregs[n].d[0] is - * still the low half and vfp.zregs[n].d[1] the high half - * of the 128 bit vector, even on big endian systems. - * Calculate the offset assuming a fully bigendian 128 bits, - * then XOR to account for the order of the two 64 bit halves. + * still the lowest and vfp.zregs[n].d[15] the highest of the + * 256 byte vector, even on big endian systems. + * + * Calculate the offset assuming fully little-endian, + * then XOR to account for the order of the 8-byte units. + * + * For 16 byte elements, the two 8 byte halves will not form a + * host int128 if the host is bigendian, since they're in the + * wrong order. However the only 16 byte operation we have is + * a move, so we can ignore this for the moment. More complicated + * operations will have to special case loading and storing from + * the zregs array. */ - offs += (16 - ((element + 1) * (1 << size))); - offs ^= 8; -#else - offs += element * (1 << size); + if (element_size < 8) { + offs ^= 8 - element_size; + } #endif offs += offsetof(CPUARMState, vfp.zregs[regno]); assert_fp_access_checked(s); -- 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: 20180613015641.5667-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 23 +++++++ target/arm/sve_helper.c | 114 +++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 133 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 27 ++++++++ 4 files changed, 297 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_cpy_z_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) DEF_HELPER_FLAGS_4(sve_ext, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_insr_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_insr_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_insr_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_insr_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_3(sve_rev_b, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_rev_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_rev_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_rev_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_tbl_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_tbl_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_tbl_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_tbl_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_sunpk_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_sunpk_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_sunpk_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_uunpk_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_uunpk_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_uunpk_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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_ext)(void *vd, void *vn, void *vm, uint32_t desc) memcpy(vd + n_siz, &tmp, n_ofs); } } + +#define DO_INSR(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, uint64_t val, uint32_t desc) \ +{ \ + intptr_t opr_sz = simd_oprsz(desc); \ + swap_memmove(vd + sizeof(TYPE), vn, opr_sz - sizeof(TYPE)); \ + *(TYPE *)(vd + H(0)) = val; \ +} + +DO_INSR(sve_insr_b, uint8_t, H1) +DO_INSR(sve_insr_h, uint16_t, H1_2) +DO_INSR(sve_insr_s, uint32_t, H1_4) +DO_INSR(sve_insr_d, uint64_t, ) + +#undef DO_INSR + +void HELPER(sve_rev_b)(void *vd, void *vn, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc); + for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) { + uint64_t f = *(uint64_t *)(vn + i); + uint64_t b = *(uint64_t *)(vn + j); + *(uint64_t *)(vd + i) = bswap64(b); + *(uint64_t *)(vd + j) = bswap64(f); + } +} + +static inline uint64_t hswap64(uint64_t h) +{ + uint64_t m = 0x0000ffff0000ffffull; + h = rol64(h, 32); + return ((h & m) << 16) | ((h >> 16) & m); +} + +void HELPER(sve_rev_h)(void *vd, void *vn, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc); + for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) { + uint64_t f = *(uint64_t *)(vn + i); + uint64_t b = *(uint64_t *)(vn + j); + *(uint64_t *)(vd + i) = hswap64(b); + *(uint64_t *)(vd + j) = hswap64(f); + } +} + +void HELPER(sve_rev_s)(void *vd, void *vn, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc); + for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) { + uint64_t f = *(uint64_t *)(vn + i); + uint64_t b = *(uint64_t *)(vn + j); + *(uint64_t *)(vd + i) = rol64(b, 32); + *(uint64_t *)(vd + j) = rol64(f, 32); + } +} + +void HELPER(sve_rev_d)(void *vd, void *vn, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc); + for (i = 0, j = opr_sz - 8; i < opr_sz / 2; i += 8, j -= 8) { + uint64_t f = *(uint64_t *)(vn + i); + uint64_t b = *(uint64_t *)(vn + j); + *(uint64_t *)(vd + i) = b; + *(uint64_t *)(vd + j) = f; + } +} + +#define DO_TBL(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, uint32_t desc) \ +{ \ + intptr_t i, opr_sz = simd_oprsz(desc); \ + uintptr_t elem = opr_sz / sizeof(TYPE); \ + TYPE *d = vd, *n = vn, *m = vm; \ + ARMVectorReg tmp; \ + if (unlikely(vd == vn)) { \ + n = memcpy(&tmp, vn, opr_sz); \ + } \ + for (i = 0; i < elem; i++) { \ + TYPE j = m[H(i)]; \ + d[H(i)] = j < elem ? n[H(j)] : 0; \ + } \ +} + +DO_TBL(sve_tbl_b, uint8_t, H1) +DO_TBL(sve_tbl_h, uint16_t, H2) +DO_TBL(sve_tbl_s, uint32_t, H4) +DO_TBL(sve_tbl_d, uint64_t, ) + +#undef TBL + +#define DO_UNPK(NAME, TYPED, TYPES, HD, HS) \ +void HELPER(NAME)(void *vd, void *vn, uint32_t desc) \ +{ \ + intptr_t i, opr_sz = simd_oprsz(desc); \ + TYPED *d = vd; \ + TYPES *n = vn; \ + ARMVectorReg tmp; \ + if (unlikely(vn - vd < opr_sz)) { \ + n = memcpy(&tmp, n, opr_sz / 2); \ + } \ + for (i = 0; i < opr_sz / sizeof(TYPED); i++) { \ + d[HD(i)] = n[HS(i)]; \ + } \ +} + +DO_UNPK(sve_sunpk_h, int16_t, int8_t, H2, H1) +DO_UNPK(sve_sunpk_s, int32_t, int16_t, H4, H2) +DO_UNPK(sve_sunpk_d, int64_t, int32_t, , H4) + +DO_UNPK(sve_uunpk_h, uint16_t, uint8_t, H2, H1) +DO_UNPK(sve_uunpk_s, uint32_t, uint16_t, H4, H2) +DO_UNPK(sve_uunpk_d, uint64_t, uint32_t, , H4) + +#undef DO_UNPK 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_EXT(DisasContext *s, arg_EXT *a, uint32_t insn) return true; } +/* + *** SVE Permute - Unpredicated Group + */ + +static bool trans_DUP_s(DisasContext *s, arg_DUP_s *a, uint32_t insn) +{ + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_dup_i64(a->esz, vec_full_reg_offset(s, a->rd), + vsz, vsz, cpu_reg_sp(s, a->rn)); + } + return true; +} + +static bool trans_DUP_x(DisasContext *s, arg_DUP_x *a, uint32_t insn) +{ + if ((a->imm & 0x1f) == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + unsigned dofs = vec_full_reg_offset(s, a->rd); + unsigned esz, index; + + esz = ctz32(a->imm); + index = a->imm >> (esz + 1); + + if ((index << esz) < vsz) { + unsigned nofs = vec_reg_offset(s, a->rn, index, esz); + tcg_gen_gvec_dup_mem(esz, dofs, nofs, vsz, vsz); + } else { + tcg_gen_gvec_dup64i(dofs, vsz, vsz, 0); + } + } + return true; +} + +static void do_insr_i64(DisasContext *s, arg_rrr_esz *a, TCGv_i64 val) +{ + typedef void gen_insr(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv_i32); + static gen_insr * const fns[4] = { + gen_helper_sve_insr_b, gen_helper_sve_insr_h, + gen_helper_sve_insr_s, gen_helper_sve_insr_d, + }; + unsigned vsz = vec_full_reg_size(s); + TCGv_i32 desc = tcg_const_i32(simd_desc(vsz, vsz, 0)); + TCGv_ptr t_zd = tcg_temp_new_ptr(); + TCGv_ptr t_zn = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(t_zd, cpu_env, vec_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(t_zn, cpu_env, vec_full_reg_offset(s, a->rn)); + + fns[a->esz](t_zd, t_zn, val, desc); + + tcg_temp_free_ptr(t_zd); + tcg_temp_free_ptr(t_zn); + tcg_temp_free_i32(desc); +} + +static bool trans_INSR_f(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + TCGv_i64 t = tcg_temp_new_i64(); + tcg_gen_ld_i64(t, cpu_env, vec_reg_offset(s, a->rm, 0, MO_64)); + do_insr_i64(s, a, t); + tcg_temp_free_i64(t); + } + return true; +} + +static bool trans_INSR_r(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_insr_i64(s, a, cpu_reg(s, a->rm)); + } + return true; +} + +static bool trans_REV_v(DisasContext *s, arg_rr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_2 * const fns[4] = { + gen_helper_sve_rev_b, gen_helper_sve_rev_h, + gen_helper_sve_rev_s, gen_helper_sve_rev_d + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_2_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vsz, vsz, 0, fns[a->esz]); + } + return true; +} + +static bool trans_TBL(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[4] = { + gen_helper_sve_tbl_b, gen_helper_sve_tbl_h, + gen_helper_sve_tbl_s, gen_helper_sve_tbl_d + }; + + 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->esz]); + } + return true; +} + +static bool trans_UNPK(DisasContext *s, arg_UNPK *a, uint32_t insn) +{ + static gen_helper_gvec_2 * const fns[4][2] = { + { NULL, NULL }, + { gen_helper_sve_sunpk_h, gen_helper_sve_uunpk_h }, + { gen_helper_sve_sunpk_s, gen_helper_sve_uunpk_s }, + { gen_helper_sve_sunpk_d, gen_helper_sve_uunpk_d }, + }; + + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_2_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn) + + (a->h ? vsz / 2 : 0), + vsz, vsz, 0, fns[a->esz][a->u]); + } + return true; +} + /* *** 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 @@ %imm4_16_p1 16:4 !function=plus1 %imm6_22_5 22:1 5:5 +%imm7_22_16 22:2 16:5 %imm8_16_10 16:5 10:3 %imm9_16_10 16:s6 10:3 @@ -XXX,XX +XXX,XX @@ # Three operand, vector element size @rd_rn_rm ........ esz:2 . rm:5 ... ... rn:5 rd:5 &rrr_esz +@rdn_rm ........ esz:2 ...... ...... rm:5 rd:5 \ + &rrr_esz rn=%reg_movprfx # Three operand with "memory" size, aka immediate left shift @rd_rn_msz_rm ........ ... rm:5 .... imm:2 rn:5 rd:5 &rrri @@ -XXX,XX +XXX,XX @@ CPY_z_i 00000101 .. 01 .... 00 . ........ ..... @rdn_pg4 imm=%sh8_i8s EXT 00000101 001 ..... 000 ... rm:5 rd:5 \ &rrri rn=%reg_movprfx imm=%imm8_16_10 +### SVE Permute - Unpredicated Group + +# SVE broadcast general register +DUP_s 00000101 .. 1 00000 001110 ..... ..... @rd_rn + +# SVE broadcast indexed element +DUP_x 00000101 .. 1 ..... 001000 rn:5 rd:5 \ + &rri imm=%imm7_22_16 + +# SVE insert SIMD&FP scalar register +INSR_f 00000101 .. 1 10100 001110 ..... ..... @rdn_rm + +# SVE insert general register +INSR_r 00000101 .. 1 00100 001110 ..... ..... @rdn_rm + +# SVE reverse vector elements +REV_v 00000101 .. 1 11000 001110 ..... ..... @rd_rn + +# SVE vector table lookup +TBL 00000101 .. 1 ..... 001100 ..... ..... @rd_rn_rm + +# SVE unpack vector elements +UNPK 00000101 esz:2 1100 u:1 h:1 001110 rn:5 rd:5 + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 6 + target/arm/sve_helper.c | 290 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 120 +++++++++++++++ target/arm/sve.decode | 18 +++ 4 files changed, 434 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_uunpk_h, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_uunpk_s, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_uunpk_d, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_zip_p, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_uzp_p, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_trn_p, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_rev_p, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(sve_punpk_p, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ DO_UNPK(sve_uunpk_s, uint32_t, uint16_t, H4, H2) DO_UNPK(sve_uunpk_d, uint64_t, uint32_t, , H4) #undef DO_UNPK + +/* Mask of bits included in the even numbered predicates of width esz. + * We also use this for expand_bits/compress_bits, and so extend the + * same pattern out to 16-bit units. + */ +static const uint64_t even_bit_esz_masks[5] = { + 0x5555555555555555ull, + 0x3333333333333333ull, + 0x0f0f0f0f0f0f0f0full, + 0x00ff00ff00ff00ffull, + 0x0000ffff0000ffffull, +}; + +/* Zero-extend units of 2**N bits to units of 2**(N+1) bits. + * For N==0, this corresponds to the operation that in qemu/bitops.h + * we call half_shuffle64; this algorithm is from Hacker's Delight, + * section 7-2 Shuffling Bits. + */ +static uint64_t expand_bits(uint64_t x, int n) +{ + int i; + + x &= 0xffffffffu; + for (i = 4; i >= n; i--) { + int sh = 1 << i; + x = ((x << sh) | x) & even_bit_esz_masks[i]; + } + return x; +} + +/* Compress units of 2**(N+1) bits to units of 2**N bits. + * For N==0, this corresponds to the operation that in qemu/bitops.h + * we call half_unshuffle64; this algorithm is from Hacker's Delight, + * section 7-2 Shuffling Bits, where it is called an inverse half shuffle. + */ +static uint64_t compress_bits(uint64_t x, int n) +{ + int i; + + for (i = n; i <= 4; i++) { + int sh = 1 << i; + x &= even_bit_esz_masks[i]; + x = (x >> sh) | x; + } + return x & 0xffffffffu; +} + +void HELPER(sve_zip_p)(void *vd, void *vn, void *vm, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + int esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + intptr_t high = extract32(pred_desc, SIMD_DATA_SHIFT + 2, 1); + uint64_t *d = vd; + intptr_t i; + + if (oprsz <= 8) { + uint64_t nn = *(uint64_t *)vn; + uint64_t mm = *(uint64_t *)vm; + int half = 4 * oprsz; + + nn = extract64(nn, high * half, half); + mm = extract64(mm, high * half, half); + nn = expand_bits(nn, esz); + mm = expand_bits(mm, esz); + d[0] = nn + (mm << (1 << esz)); + } else { + ARMPredicateReg tmp_n, tmp_m; + + /* We produce output faster than we consume input. + Therefore we must be mindful of possible overlap. */ + if ((vn - vd) < (uintptr_t)oprsz) { + vn = memcpy(&tmp_n, vn, oprsz); + } + if ((vm - vd) < (uintptr_t)oprsz) { + vm = memcpy(&tmp_m, vm, oprsz); + } + if (high) { + high = oprsz >> 1; + } + + if ((high & 3) == 0) { + uint32_t *n = vn, *m = vm; + high >>= 2; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); i++) { + uint64_t nn = n[H4(high + i)]; + uint64_t mm = m[H4(high + i)]; + + nn = expand_bits(nn, esz); + mm = expand_bits(mm, esz); + d[i] = nn + (mm << (1 << esz)); + } + } else { + uint8_t *n = vn, *m = vm; + uint16_t *d16 = vd; + + for (i = 0; i < oprsz / 2; i++) { + uint16_t nn = n[H1(high + i)]; + uint16_t mm = m[H1(high + i)]; + + nn = expand_bits(nn, esz); + mm = expand_bits(mm, esz); + d16[H2(i)] = nn + (mm << (1 << esz)); + } + } + } +} + +void HELPER(sve_uzp_p)(void *vd, void *vn, void *vm, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + int esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + int odd = extract32(pred_desc, SIMD_DATA_SHIFT + 2, 1) << esz; + uint64_t *d = vd, *n = vn, *m = vm; + uint64_t l, h; + intptr_t i; + + if (oprsz <= 8) { + l = compress_bits(n[0] >> odd, esz); + h = compress_bits(m[0] >> odd, esz); + d[0] = extract64(l + (h << (4 * oprsz)), 0, 8 * oprsz); + } else { + ARMPredicateReg tmp_m; + intptr_t oprsz_16 = oprsz / 16; + + if ((vm - vd) < (uintptr_t)oprsz) { + m = memcpy(&tmp_m, vm, oprsz); + } + + for (i = 0; i < oprsz_16; i++) { + l = n[2 * i + 0]; + h = n[2 * i + 1]; + l = compress_bits(l >> odd, esz); + h = compress_bits(h >> odd, esz); + d[i] = l + (h << 32); + } + + /* For VL which is not a power of 2, the results from M do not + align nicely with the uint64_t for D. Put the aligned results + from M into TMP_M and then copy it into place afterward. */ + if (oprsz & 15) { + d[i] = compress_bits(n[2 * i] >> odd, esz); + + for (i = 0; i < oprsz_16; i++) { + l = m[2 * i + 0]; + h = m[2 * i + 1]; + l = compress_bits(l >> odd, esz); + h = compress_bits(h >> odd, esz); + tmp_m.p[i] = l + (h << 32); + } + tmp_m.p[i] = compress_bits(m[2 * i] >> odd, esz); + + swap_memmove(vd + oprsz / 2, &tmp_m, oprsz / 2); + } else { + for (i = 0; i < oprsz_16; i++) { + l = m[2 * i + 0]; + h = m[2 * i + 1]; + l = compress_bits(l >> odd, esz); + h = compress_bits(h >> odd, esz); + d[oprsz_16 + i] = l + (h << 32); + } + } + } +} + +void HELPER(sve_trn_p)(void *vd, void *vn, void *vm, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + uintptr_t esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + bool odd = extract32(pred_desc, SIMD_DATA_SHIFT + 2, 1); + uint64_t *d = vd, *n = vn, *m = vm; + uint64_t mask; + int shr, shl; + intptr_t i; + + shl = 1 << esz; + shr = 0; + mask = even_bit_esz_masks[esz]; + if (odd) { + mask <<= shl; + shr = shl; + shl = 0; + } + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); i++) { + uint64_t nn = (n[i] & mask) >> shr; + uint64_t mm = (m[i] & mask) << shl; + d[i] = nn + mm; + } +} + +/* Reverse units of 2**N bits. */ +static uint64_t reverse_bits_64(uint64_t x, int n) +{ + int i, sh; + + x = bswap64(x); + for (i = 2, sh = 4; i >= n; i--, sh >>= 1) { + uint64_t mask = even_bit_esz_masks[i]; + x = ((x & mask) << sh) | ((x >> sh) & mask); + } + return x; +} + +static uint8_t reverse_bits_8(uint8_t x, int n) +{ + static const uint8_t mask[3] = { 0x55, 0x33, 0x0f }; + int i, sh; + + for (i = 2, sh = 4; i >= n; i--, sh >>= 1) { + x = ((x & mask[i]) << sh) | ((x >> sh) & mask[i]); + } + return x; +} + +void HELPER(sve_rev_p)(void *vd, void *vn, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + int esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + intptr_t i, oprsz_2 = oprsz / 2; + + if (oprsz <= 8) { + uint64_t l = *(uint64_t *)vn; + l = reverse_bits_64(l << (64 - 8 * oprsz), esz); + *(uint64_t *)vd = l; + } else if ((oprsz & 15) == 0) { + for (i = 0; i < oprsz_2; i += 8) { + intptr_t ih = oprsz - 8 - i; + uint64_t l = reverse_bits_64(*(uint64_t *)(vn + i), esz); + uint64_t h = reverse_bits_64(*(uint64_t *)(vn + ih), esz); + *(uint64_t *)(vd + i) = h; + *(uint64_t *)(vd + ih) = l; + } + } else { + for (i = 0; i < oprsz_2; i += 1) { + intptr_t il = H1(i); + intptr_t ih = H1(oprsz - 1 - i); + uint8_t l = reverse_bits_8(*(uint8_t *)(vn + il), esz); + uint8_t h = reverse_bits_8(*(uint8_t *)(vn + ih), esz); + *(uint8_t *)(vd + il) = h; + *(uint8_t *)(vd + ih) = l; + } + } +} + +void HELPER(sve_punpk_p)(void *vd, void *vn, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + intptr_t high = extract32(pred_desc, SIMD_DATA_SHIFT + 2, 1); + uint64_t *d = vd; + intptr_t i; + + if (oprsz <= 8) { + uint64_t nn = *(uint64_t *)vn; + int half = 4 * oprsz; + + nn = extract64(nn, high * half, half); + nn = expand_bits(nn, 0); + d[0] = nn; + } else { + ARMPredicateReg tmp_n; + + /* We produce output faster than we consume input. + Therefore we must be mindful of possible overlap. */ + if ((vn - vd) < (uintptr_t)oprsz) { + vn = memcpy(&tmp_n, vn, oprsz); + } + if (high) { + high = oprsz >> 1; + } + + if ((high & 3) == 0) { + uint32_t *n = vn; + high >>= 2; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); i++) { + uint64_t nn = n[H4(high + i)]; + d[i] = expand_bits(nn, 0); + } + } else { + uint16_t *d16 = vd; + uint8_t *n = vn; + + for (i = 0; i < oprsz / 2; i++) { + uint16_t nn = n[H1(high + i)]; + d16[H2(i)] = expand_bits(nn, 0); + } + } + } +} 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_UNPK(DisasContext *s, arg_UNPK *a, uint32_t insn) return true; } +/* + *** SVE Permute - Predicates Group + */ + +static bool do_perm_pred3(DisasContext *s, arg_rrr_esz *a, bool high_odd, + gen_helper_gvec_3 *fn) +{ + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = pred_full_reg_size(s); + + /* Predicate sizes may be smaller and cannot use simd_desc. + We cannot round up, as we do elsewhere, because we need + the exact size for ZIP2 and REV. We retain the style for + the other helpers for consistency. */ + TCGv_ptr t_d = tcg_temp_new_ptr(); + TCGv_ptr t_n = tcg_temp_new_ptr(); + TCGv_ptr t_m = tcg_temp_new_ptr(); + TCGv_i32 t_desc; + int desc; + + desc = vsz - 2; + desc = deposit32(desc, SIMD_DATA_SHIFT, 2, a->esz); + desc = deposit32(desc, SIMD_DATA_SHIFT + 2, 2, high_odd); + + tcg_gen_addi_ptr(t_d, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(t_n, cpu_env, pred_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(t_m, cpu_env, pred_full_reg_offset(s, a->rm)); + t_desc = tcg_const_i32(desc); + + fn(t_d, t_n, t_m, t_desc); + + tcg_temp_free_ptr(t_d); + tcg_temp_free_ptr(t_n); + tcg_temp_free_ptr(t_m); + tcg_temp_free_i32(t_desc); + return true; +} + +static bool do_perm_pred2(DisasContext *s, arg_rr_esz *a, bool high_odd, + gen_helper_gvec_2 *fn) +{ + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = pred_full_reg_size(s); + TCGv_ptr t_d = tcg_temp_new_ptr(); + TCGv_ptr t_n = tcg_temp_new_ptr(); + TCGv_i32 t_desc; + int desc; + + tcg_gen_addi_ptr(t_d, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(t_n, cpu_env, pred_full_reg_offset(s, a->rn)); + + /* Predicate sizes may be smaller and cannot use simd_desc. + We cannot round up, as we do elsewhere, because we need + the exact size for ZIP2 and REV. We retain the style for + the other helpers for consistency. */ + + desc = vsz - 2; + desc = deposit32(desc, SIMD_DATA_SHIFT, 2, a->esz); + desc = deposit32(desc, SIMD_DATA_SHIFT + 2, 2, high_odd); + t_desc = tcg_const_i32(desc); + + fn(t_d, t_n, t_desc); + + tcg_temp_free_i32(t_desc); + tcg_temp_free_ptr(t_d); + tcg_temp_free_ptr(t_n); + return true; +} + +static bool trans_ZIP1_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 0, gen_helper_sve_zip_p); +} + +static bool trans_ZIP2_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 1, gen_helper_sve_zip_p); +} + +static bool trans_UZP1_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 0, gen_helper_sve_uzp_p); +} + +static bool trans_UZP2_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 1, gen_helper_sve_uzp_p); +} + +static bool trans_TRN1_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 0, gen_helper_sve_trn_p); +} + +static bool trans_TRN2_p(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_perm_pred3(s, a, 1, gen_helper_sve_trn_p); +} + +static bool trans_REV_p(DisasContext *s, arg_rr_esz *a, uint32_t insn) +{ + return do_perm_pred2(s, a, 0, gen_helper_sve_rev_p); +} + +static bool trans_PUNPKLO(DisasContext *s, arg_PUNPKLO *a, uint32_t insn) +{ + return do_perm_pred2(s, a, 0, gen_helper_sve_punpk_p); +} + +static bool trans_PUNPKHI(DisasContext *s, arg_PUNPKHI *a, uint32_t insn) +{ + return do_perm_pred2(s, a, 1, gen_helper_sve_punpk_p); +} + /* *** 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 @@ # Three operand, vector element size @rd_rn_rm ........ esz:2 . rm:5 ... ... rn:5 rd:5 &rrr_esz +@pd_pn_pm ........ esz:2 .. rm:4 ....... rn:4 . rd:4 &rrr_esz @rdn_rm ........ esz:2 ...... ...... rm:5 rd:5 \ &rrr_esz rn=%reg_movprfx @@ -XXX,XX +XXX,XX @@ TBL 00000101 .. 1 ..... 001100 ..... ..... @rd_rn_rm # SVE unpack vector elements UNPK 00000101 esz:2 1100 u:1 h:1 001110 rn:5 rd:5 +### SVE Permute - Predicates Group + +# SVE permute predicate elements +ZIP1_p 00000101 .. 10 .... 010 000 0 .... 0 .... @pd_pn_pm +ZIP2_p 00000101 .. 10 .... 010 001 0 .... 0 .... @pd_pn_pm +UZP1_p 00000101 .. 10 .... 010 010 0 .... 0 .... @pd_pn_pm +UZP2_p 00000101 .. 10 .... 010 011 0 .... 0 .... @pd_pn_pm +TRN1_p 00000101 .. 10 .... 010 100 0 .... 0 .... @pd_pn_pm +TRN2_p 00000101 .. 10 .... 010 101 0 .... 0 .... @pd_pn_pm + +# SVE reverse predicate elements +REV_p 00000101 .. 11 0100 010 000 0 .... 0 .... @pd_pn + +# SVE unpack predicate elements +PUNPKLO 00000101 00 11 0000 010 000 0 .... 0 .... @pd_pn_e0 +PUNPKHI 00000101 00 11 0001 010 000 0 .... 0 .... @pd_pn_e0 + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 15 ++++++++ target/arm/sve_helper.c | 72 ++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 75 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 10 +++++ 4 files changed, 172 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_trn_p, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_rev_p, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_punpk_p, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_zip_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_zip_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_zip_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_zip_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_uzp_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_uzp_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_uzp_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_uzp_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_trn_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_trn_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_trn_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_trn_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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_punpk_p)(void *vd, void *vn, uint32_t pred_desc) } } } + +#define DO_ZIP(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, uint32_t desc) \ +{ \ + intptr_t oprsz = simd_oprsz(desc); \ + intptr_t i, oprsz_2 = oprsz / 2; \ + ARMVectorReg tmp_n, tmp_m; \ + /* We produce output faster than we consume input. \ + Therefore we must be mindful of possible overlap. */ \ + if (unlikely((vn - vd) < (uintptr_t)oprsz)) { \ + vn = memcpy(&tmp_n, vn, oprsz_2); \ + } \ + if (unlikely((vm - vd) < (uintptr_t)oprsz)) { \ + vm = memcpy(&tmp_m, vm, oprsz_2); \ + } \ + for (i = 0; i < oprsz_2; i += sizeof(TYPE)) { \ + *(TYPE *)(vd + H(2 * i + 0)) = *(TYPE *)(vn + H(i)); \ + *(TYPE *)(vd + H(2 * i + sizeof(TYPE))) = *(TYPE *)(vm + H(i)); \ + } \ +} + +DO_ZIP(sve_zip_b, uint8_t, H1) +DO_ZIP(sve_zip_h, uint16_t, H1_2) +DO_ZIP(sve_zip_s, uint32_t, H1_4) +DO_ZIP(sve_zip_d, uint64_t, ) + +#define DO_UZP(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, uint32_t desc) \ +{ \ + intptr_t oprsz = simd_oprsz(desc); \ + intptr_t oprsz_2 = oprsz / 2; \ + intptr_t odd_ofs = simd_data(desc); \ + intptr_t i; \ + ARMVectorReg tmp_m; \ + if (unlikely((vm - vd) < (uintptr_t)oprsz)) { \ + vm = memcpy(&tmp_m, vm, oprsz); \ + } \ + for (i = 0; i < oprsz_2; i += sizeof(TYPE)) { \ + *(TYPE *)(vd + H(i)) = *(TYPE *)(vn + H(2 * i + odd_ofs)); \ + } \ + for (i = 0; i < oprsz_2; i += sizeof(TYPE)) { \ + *(TYPE *)(vd + H(oprsz_2 + i)) = *(TYPE *)(vm + H(2 * i + odd_ofs)); \ + } \ +} + +DO_UZP(sve_uzp_b, uint8_t, H1) +DO_UZP(sve_uzp_h, uint16_t, H1_2) +DO_UZP(sve_uzp_s, uint32_t, H1_4) +DO_UZP(sve_uzp_d, uint64_t, ) + +#define DO_TRN(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, uint32_t desc) \ +{ \ + intptr_t oprsz = simd_oprsz(desc); \ + intptr_t odd_ofs = simd_data(desc); \ + intptr_t i; \ + for (i = 0; i < oprsz; i += 2 * sizeof(TYPE)) { \ + TYPE ae = *(TYPE *)(vn + H(i + odd_ofs)); \ + TYPE be = *(TYPE *)(vm + H(i + odd_ofs)); \ + *(TYPE *)(vd + H(i + 0)) = ae; \ + *(TYPE *)(vd + H(i + sizeof(TYPE))) = be; \ + } \ +} + +DO_TRN(sve_trn_b, uint8_t, H1) +DO_TRN(sve_trn_h, uint16_t, H1_2) +DO_TRN(sve_trn_s, uint32_t, H1_4) +DO_TRN(sve_trn_d, uint64_t, ) + +#undef DO_ZIP +#undef DO_UZP +#undef DO_TRN 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_PUNPKHI(DisasContext *s, arg_PUNPKHI *a, uint32_t insn) return do_perm_pred2(s, a, 1, gen_helper_sve_punpk_p); } +/* + *** SVE Permute - Interleaving Group + */ + +static bool do_zip(DisasContext *s, arg_rrr_esz *a, bool high) +{ + static gen_helper_gvec_3 * const fns[4] = { + gen_helper_sve_zip_b, gen_helper_sve_zip_h, + gen_helper_sve_zip_s, gen_helper_sve_zip_d, + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + unsigned high_ofs = high ? vsz / 2 : 0; + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn) + high_ofs, + vec_full_reg_offset(s, a->rm) + high_ofs, + vsz, vsz, 0, fns[a->esz]); + } + return true; +} + +static bool do_zzz_data_ool(DisasContext *s, arg_rrr_esz *a, int data, + gen_helper_gvec_3 *fn) +{ + 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, data, fn); + } + return true; +} + +static bool trans_ZIP1_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zip(s, a, false); +} + +static bool trans_ZIP2_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zip(s, a, true); +} + +static gen_helper_gvec_3 * const uzp_fns[4] = { + gen_helper_sve_uzp_b, gen_helper_sve_uzp_h, + gen_helper_sve_uzp_s, gen_helper_sve_uzp_d, +}; + +static bool trans_UZP1_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zzz_data_ool(s, a, 0, uzp_fns[a->esz]); +} + +static bool trans_UZP2_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zzz_data_ool(s, a, 1 << a->esz, uzp_fns[a->esz]); +} + +static gen_helper_gvec_3 * const trn_fns[4] = { + gen_helper_sve_trn_b, gen_helper_sve_trn_h, + gen_helper_sve_trn_s, gen_helper_sve_trn_d, +}; + +static bool trans_TRN1_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zzz_data_ool(s, a, 0, trn_fns[a->esz]); +} + +static bool trans_TRN2_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) +{ + return do_zzz_data_ool(s, a, 1 << a->esz, trn_fns[a->esz]); +} + /* *** 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 @@ REV_p 00000101 .. 11 0100 010 000 0 .... 0 .... @pd_pn PUNPKLO 00000101 00 11 0000 010 000 0 .... 0 .... @pd_pn_e0 PUNPKHI 00000101 00 11 0001 010 000 0 .... 0 .... @pd_pn_e0 +### SVE Permute - Interleaving Group + +# SVE permute vector elements +ZIP1_z 00000101 .. 1 ..... 011 000 ..... ..... @rd_rn_rm +ZIP2_z 00000101 .. 1 ..... 011 001 ..... ..... @rd_rn_rm +UZP1_z 00000101 .. 1 ..... 011 010 ..... ..... @rd_rn_rm +UZP2_z 00000101 .. 1 ..... 011 011 ..... ..... @rd_rn_rm +TRN1_z 00000101 .. 1 ..... 011 100 ..... ..... @rd_rn_rm +TRN2_z 00000101 .. 1 ..... 011 101 ..... ..... @rd_rn_rm + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 3 +++ target/arm/sve_helper.c | 34 ++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 12 ++++++++++++ target/arm/sve.decode | 6 ++++++ 4 files changed, 55 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_trn_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_trn_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_trn_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_compact_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_compact_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ DO_TRN(sve_trn_d, uint64_t, ) #undef DO_ZIP #undef DO_UZP #undef DO_TRN + +void HELPER(sve_compact_s)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc) / 4; + uint32_t *d = vd, *n = vn; + uint8_t *pg = vg; + + for (i = j = 0; i < opr_sz; i++) { + if (pg[H1(i / 2)] & (i & 1 ? 0x10 : 0x01)) { + d[H4(j)] = n[H4(i)]; + j++; + } + } + for (; j < opr_sz; j++) { + d[H4(j)] = 0; + } +} + +void HELPER(sve_compact_d)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, j, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + + for (i = j = 0; i < opr_sz; i++) { + if (pg[H1(i)] & 1) { + d[j] = n[i]; + j++; + } + } + for (; j < opr_sz; j++) { + d[j] = 0; + } +} 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_TRN2_z(DisasContext *s, arg_rrr_esz *a, uint32_t insn) return do_zzz_data_ool(s, a, 1 << a->esz, trn_fns[a->esz]); } +/* + *** SVE Permute Vector - Predicated Group + */ + +static bool trans_COMPACT(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[4] = { + NULL, NULL, gen_helper_sve_compact_s, gen_helper_sve_compact_d + }; + return do_zpz_ool(s, a, fns[a->esz]); +} + /* *** 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 @@ UZP2_z 00000101 .. 1 ..... 011 011 ..... ..... @rd_rn_rm TRN1_z 00000101 .. 1 ..... 011 100 ..... ..... @rd_rn_rm TRN2_z 00000101 .. 1 ..... 011 101 ..... ..... @rd_rn_rm +### SVE Permute - Predicated Group + +# SVE compress active elements +# Note esz >= 2 +COMPACT 00000101 .. 100001 100 ... ..... ..... @rd_pg_rn + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 + target/arm/sve_helper.c | 12 ++ target/arm/translate-sve.c | 328 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 20 +++ 4 files changed, 362 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_trn_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_compact_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_compact_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_2(sve_last_active_element, TCG_CALL_NO_RWG, s32, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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_compact_d)(void *vd, void *vn, void *vg, uint32_t desc) d[j] = 0; } } + +/* Similar to the ARM LastActiveElement pseudocode function, except the + * result is multiplied by the element size. This includes the not found + * indication; e.g. not found for esz=3 is -8. + */ +int32_t HELPER(sve_last_active_element)(void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + intptr_t esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + + return last_active_element(vg, DIV_ROUND_UP(oprsz, 8), esz); +} 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_COMPACT(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_zpz_ool(s, a, fns[a->esz]); } +/* Call the helper that computes the ARM LastActiveElement pseudocode + * function, scaled by the element size. This includes the not found + * indication; e.g. not found for esz=3 is -8. + */ +static void find_last_active(DisasContext *s, TCGv_i32 ret, int esz, int pg) +{ + /* Predicate sizes may be smaller and cannot use simd_desc. We cannot + * round up, as we do elsewhere, because we need the exact size. + */ + TCGv_ptr t_p = tcg_temp_new_ptr(); + TCGv_i32 t_desc; + unsigned vsz = pred_full_reg_size(s); + unsigned desc; + + desc = vsz - 2; + desc = deposit32(desc, SIMD_DATA_SHIFT, 2, esz); + + tcg_gen_addi_ptr(t_p, cpu_env, pred_full_reg_offset(s, pg)); + t_desc = tcg_const_i32(desc); + + gen_helper_sve_last_active_element(ret, t_p, t_desc); + + tcg_temp_free_i32(t_desc); + tcg_temp_free_ptr(t_p); +} + +/* Increment LAST to the offset of the next element in the vector, + * wrapping around to 0. + */ +static void incr_last_active(DisasContext *s, TCGv_i32 last, int esz) +{ + unsigned vsz = vec_full_reg_size(s); + + tcg_gen_addi_i32(last, last, 1 << esz); + if (is_power_of_2(vsz)) { + tcg_gen_andi_i32(last, last, vsz - 1); + } else { + TCGv_i32 max = tcg_const_i32(vsz); + TCGv_i32 zero = tcg_const_i32(0); + tcg_gen_movcond_i32(TCG_COND_GEU, last, last, max, zero, last); + tcg_temp_free_i32(max); + tcg_temp_free_i32(zero); + } +} + +/* If LAST < 0, set LAST to the offset of the last element in the vector. */ +static void wrap_last_active(DisasContext *s, TCGv_i32 last, int esz) +{ + unsigned vsz = vec_full_reg_size(s); + + if (is_power_of_2(vsz)) { + tcg_gen_andi_i32(last, last, vsz - 1); + } else { + TCGv_i32 max = tcg_const_i32(vsz - (1 << esz)); + TCGv_i32 zero = tcg_const_i32(0); + tcg_gen_movcond_i32(TCG_COND_LT, last, last, zero, max, last); + tcg_temp_free_i32(max); + tcg_temp_free_i32(zero); + } +} + +/* Load an unsigned element of ESZ from BASE+OFS. */ +static TCGv_i64 load_esz(TCGv_ptr base, int ofs, int esz) +{ + TCGv_i64 r = tcg_temp_new_i64(); + + switch (esz) { + case 0: + tcg_gen_ld8u_i64(r, base, ofs); + break; + case 1: + tcg_gen_ld16u_i64(r, base, ofs); + break; + case 2: + tcg_gen_ld32u_i64(r, base, ofs); + break; + case 3: + tcg_gen_ld_i64(r, base, ofs); + break; + default: + g_assert_not_reached(); + } + return r; +} + +/* Load an unsigned element of ESZ from RM[LAST]. */ +static TCGv_i64 load_last_active(DisasContext *s, TCGv_i32 last, + int rm, int esz) +{ + TCGv_ptr p = tcg_temp_new_ptr(); + TCGv_i64 r; + + /* Convert offset into vector into offset into ENV. + * The final adjustment for the vector register base + * is added via constant offset to the load. + */ +#ifdef HOST_WORDS_BIGENDIAN + /* Adjust for element ordering. See vec_reg_offset. */ + if (esz < 3) { + tcg_gen_xori_i32(last, last, 8 - (1 << esz)); + } +#endif + tcg_gen_ext_i32_ptr(p, last); + tcg_gen_add_ptr(p, p, cpu_env); + + r = load_esz(p, vec_full_reg_offset(s, rm), esz); + tcg_temp_free_ptr(p); + + return r; +} + +/* Compute CLAST for a Zreg. */ +static bool do_clast_vector(DisasContext *s, arg_rprr_esz *a, bool before) +{ + TCGv_i32 last; + TCGLabel *over; + TCGv_i64 ele; + unsigned vsz, esz = a->esz; + + if (!sve_access_check(s)) { + return true; + } + + last = tcg_temp_local_new_i32(); + over = gen_new_label(); + + find_last_active(s, last, esz, a->pg); + + /* There is of course no movcond for a 2048-bit vector, + * so we must branch over the actual store. + */ + tcg_gen_brcondi_i32(TCG_COND_LT, last, 0, over); + + if (!before) { + incr_last_active(s, last, esz); + } + + ele = load_last_active(s, last, a->rm, esz); + tcg_temp_free_i32(last); + + vsz = vec_full_reg_size(s); + tcg_gen_gvec_dup_i64(esz, vec_full_reg_offset(s, a->rd), vsz, vsz, ele); + tcg_temp_free_i64(ele); + + /* If this insn used MOVPRFX, we may need a second move. */ + if (a->rd != a->rn) { + TCGLabel *done = gen_new_label(); + tcg_gen_br(done); + + gen_set_label(over); + do_mov_z(s, a->rd, a->rn); + + gen_set_label(done); + } else { + gen_set_label(over); + } + return true; +} + +static bool trans_CLASTA_z(DisasContext *s, arg_rprr_esz *a, uint32_t insn) +{ + return do_clast_vector(s, a, false); +} + +static bool trans_CLASTB_z(DisasContext *s, arg_rprr_esz *a, uint32_t insn) +{ + return do_clast_vector(s, a, true); +} + +/* Compute CLAST for a scalar. */ +static void do_clast_scalar(DisasContext *s, int esz, int pg, int rm, + bool before, TCGv_i64 reg_val) +{ + TCGv_i32 last = tcg_temp_new_i32(); + TCGv_i64 ele, cmp, zero; + + find_last_active(s, last, esz, pg); + + /* Extend the original value of last prior to incrementing. */ + cmp = tcg_temp_new_i64(); + tcg_gen_ext_i32_i64(cmp, last); + + if (!before) { + incr_last_active(s, last, esz); + } + + /* The conceit here is that while last < 0 indicates not found, after + * adjusting for cpu_env->vfp.zregs[rm], it is still a valid address + * from which we can load garbage. We then discard the garbage with + * a conditional move. + */ + ele = load_last_active(s, last, rm, esz); + tcg_temp_free_i32(last); + + zero = tcg_const_i64(0); + tcg_gen_movcond_i64(TCG_COND_GE, reg_val, cmp, zero, ele, reg_val); + + tcg_temp_free_i64(zero); + tcg_temp_free_i64(cmp); + tcg_temp_free_i64(ele); +} + +/* Compute CLAST for a Vreg. */ +static bool do_clast_fp(DisasContext *s, arg_rpr_esz *a, bool before) +{ + if (sve_access_check(s)) { + int esz = a->esz; + int ofs = vec_reg_offset(s, a->rd, 0, esz); + TCGv_i64 reg = load_esz(cpu_env, ofs, esz); + + do_clast_scalar(s, esz, a->pg, a->rn, before, reg); + write_fp_dreg(s, a->rd, reg); + tcg_temp_free_i64(reg); + } + return true; +} + +static bool trans_CLASTA_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_clast_fp(s, a, false); +} + +static bool trans_CLASTB_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_clast_fp(s, a, true); +} + +/* Compute CLAST for a Xreg. */ +static bool do_clast_general(DisasContext *s, arg_rpr_esz *a, bool before) +{ + TCGv_i64 reg; + + if (!sve_access_check(s)) { + return true; + } + + reg = cpu_reg(s, a->rd); + switch (a->esz) { + case 0: + tcg_gen_ext8u_i64(reg, reg); + break; + case 1: + tcg_gen_ext16u_i64(reg, reg); + break; + case 2: + tcg_gen_ext32u_i64(reg, reg); + break; + case 3: + break; + default: + g_assert_not_reached(); + } + + do_clast_scalar(s, a->esz, a->pg, a->rn, before, reg); + return true; +} + +static bool trans_CLASTA_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_clast_general(s, a, false); +} + +static bool trans_CLASTB_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_clast_general(s, a, true); +} + +/* Compute LAST for a scalar. */ +static TCGv_i64 do_last_scalar(DisasContext *s, int esz, + int pg, int rm, bool before) +{ + TCGv_i32 last = tcg_temp_new_i32(); + TCGv_i64 ret; + + find_last_active(s, last, esz, pg); + if (before) { + wrap_last_active(s, last, esz); + } else { + incr_last_active(s, last, esz); + } + + ret = load_last_active(s, last, rm, esz); + tcg_temp_free_i32(last); + return ret; +} + +/* Compute LAST for a Vreg. */ +static bool do_last_fp(DisasContext *s, arg_rpr_esz *a, bool before) +{ + if (sve_access_check(s)) { + TCGv_i64 val = do_last_scalar(s, a->esz, a->pg, a->rn, before); + write_fp_dreg(s, a->rd, val); + tcg_temp_free_i64(val); + } + return true; +} + +static bool trans_LASTA_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_last_fp(s, a, false); +} + +static bool trans_LASTB_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_last_fp(s, a, true); +} + +/* Compute LAST for a Xreg. */ +static bool do_last_general(DisasContext *s, arg_rpr_esz *a, bool before) +{ + if (sve_access_check(s)) { + TCGv_i64 val = do_last_scalar(s, a->esz, a->pg, a->rn, before); + tcg_gen_mov_i64(cpu_reg(s, a->rd), val); + tcg_temp_free_i64(val); + } + return true; +} + +static bool trans_LASTA_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_last_general(s, a, false); +} + +static bool trans_LASTB_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_last_general(s, a, true); +} + /* *** 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 @@ TRN2_z 00000101 .. 1 ..... 011 101 ..... ..... @rd_rn_rm # Note esz >= 2 COMPACT 00000101 .. 100001 100 ... ..... ..... @rd_pg_rn +# SVE conditionally broadcast element to vector +CLASTA_z 00000101 .. 10100 0 100 ... ..... ..... @rdn_pg_rm +CLASTB_z 00000101 .. 10100 1 100 ... ..... ..... @rdn_pg_rm + +# SVE conditionally copy element to SIMD&FP scalar +CLASTA_v 00000101 .. 10101 0 100 ... ..... ..... @rd_pg_rn +CLASTB_v 00000101 .. 10101 1 100 ... ..... ..... @rd_pg_rn + +# SVE conditionally copy element to general register +CLASTA_r 00000101 .. 11000 0 101 ... ..... ..... @rd_pg_rn +CLASTB_r 00000101 .. 11000 1 101 ... ..... ..... @rd_pg_rn + +# SVE copy element to SIMD&FP scalar register +LASTA_v 00000101 .. 10001 0 100 ... ..... ..... @rd_pg_rn +LASTB_v 00000101 .. 10001 1 100 ... ..... ..... @rd_pg_rn + +# SVE copy element to general register +LASTA_r 00000101 .. 10000 0 101 ... ..... ..... @rd_pg_rn +LASTB_r 00000101 .. 10000 1 101 ... ..... ..... @rd_pg_rn + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 19 +++++++++++++++++++ target/arm/sve.decode | 6 ++++++ 2 files changed, 25 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_LASTB_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_last_general(s, a, true); } +static bool trans_CPY_m_r(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_cpy_m(s, a->esz, a->rd, a->rd, a->pg, cpu_reg_sp(s, a->rn)); + } + return true; +} + +static bool trans_CPY_m_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + int ofs = vec_reg_offset(s, a->rn, 0, a->esz); + TCGv_i64 t = load_esz(cpu_env, ofs, a->esz); + do_cpy_m(s, a->esz, a->rd, a->rd, a->pg, t); + tcg_temp_free_i64(t); + } + return true; +} + /* *** 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 @@ LASTB_v 00000101 .. 10001 1 100 ... ..... ..... @rd_pg_rn LASTA_r 00000101 .. 10000 0 101 ... ..... ..... @rd_pg_rn LASTB_r 00000101 .. 10000 1 101 ... ..... ..... @rd_pg_rn +# SVE copy element from SIMD&FP scalar register +CPY_m_v 00000101 .. 100000 100 ... ..... ..... @rd_pg_rn + +# SVE copy element from general register to vector (predicated) +CPY_m_r 00000101 .. 101000 101 ... ..... ..... @rd_pg_rn + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 14 +++++++++++++ target/arm/sve_helper.c | 41 +++++++++++++++++++++++++++++++------- target/arm/translate-sve.c | 38 +++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 7 +++++++ 4 files changed, 93 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_4(sve_compact_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_2(sve_last_active_element, TCG_CALL_NO_RWG, s32, ptr, i32) +DEF_HELPER_FLAGS_4(sve_revb_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_revb_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_revb_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_revh_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_revh_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_revw_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_rbit_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_rbit_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_rbit_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_rbit_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ static inline uint64_t expand_pred_s(uint8_t byte) return word[byte & 0x11]; } +/* Swap 16-bit words within a 32-bit word. */ +static inline uint32_t hswap32(uint32_t h) +{ + return rol32(h, 16); +} + +/* Swap 16-bit words within a 64-bit word. */ +static inline uint64_t hswap64(uint64_t h) +{ + uint64_t m = 0x0000ffff0000ffffull; + h = rol64(h, 32); + return ((h & m) << 16) | ((h >> 16) & m); +} + +/* Swap 32-bit words within a 64-bit word. */ +static inline uint64_t wswap64(uint64_t h) +{ + return rol64(h, 32); +} + #define LOGICAL_PPPP(NAME, FUNC) \ void HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) \ { \ @@ -XXX,XX +XXX,XX @@ DO_ZPZ(sve_neg_h, uint16_t, H1_2, DO_NEG) DO_ZPZ(sve_neg_s, uint32_t, H1_4, DO_NEG) DO_ZPZ_D(sve_neg_d, uint64_t, DO_NEG) +DO_ZPZ(sve_revb_h, uint16_t, H1_2, bswap16) +DO_ZPZ(sve_revb_s, uint32_t, H1_4, bswap32) +DO_ZPZ_D(sve_revb_d, uint64_t, bswap64) + +DO_ZPZ(sve_revh_s, uint32_t, H1_4, hswap32) +DO_ZPZ_D(sve_revh_d, uint64_t, hswap64) + +DO_ZPZ_D(sve_revw_d, uint64_t, wswap64) + +DO_ZPZ(sve_rbit_b, uint8_t, H1, revbit8) +DO_ZPZ(sve_rbit_h, uint16_t, H1_2, revbit16) +DO_ZPZ(sve_rbit_s, uint32_t, H1_4, revbit32) +DO_ZPZ_D(sve_rbit_d, uint64_t, revbit64) + /* Three-operand expander, unpredicated, in which the third operand is "wide". */ #define DO_ZZW(NAME, TYPE, TYPEW, H, OP) \ @@ -XXX,XX +XXX,XX @@ void HELPER(sve_rev_b)(void *vd, void *vn, uint32_t desc) } } -static inline uint64_t hswap64(uint64_t h) -{ - uint64_t m = 0x0000ffff0000ffffull; - h = rol64(h, 32); - return ((h & m) << 16) | ((h >> 16) & m); -} - void HELPER(sve_rev_h)(void *vd, void *vn, uint32_t desc) { intptr_t i, j, opr_sz = simd_oprsz(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_CPY_m_v(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return true; } +static bool trans_REVB(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[4] = { + NULL, + gen_helper_sve_revb_h, + gen_helper_sve_revb_s, + gen_helper_sve_revb_d, + }; + return do_zpz_ool(s, a, fns[a->esz]); +} + +static bool trans_REVH(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[4] = { + NULL, + NULL, + gen_helper_sve_revh_s, + gen_helper_sve_revh_d, + }; + return do_zpz_ool(s, a, fns[a->esz]); +} + +static bool trans_REVW(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + return do_zpz_ool(s, a, a->esz == 3 ? gen_helper_sve_revw_d : NULL); +} + +static bool trans_RBIT(DisasContext *s, arg_rpr_esz *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[4] = { + gen_helper_sve_rbit_b, + gen_helper_sve_rbit_h, + gen_helper_sve_rbit_s, + gen_helper_sve_rbit_d, + }; + return do_zpz_ool(s, a, fns[a->esz]); +} + /* *** 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 @@ CPY_m_v 00000101 .. 100000 100 ... ..... ..... @rd_pg_rn # SVE copy element from general register to vector (predicated) CPY_m_r 00000101 .. 101000 101 ... ..... ..... @rd_pg_rn +# SVE reverse within elements +# Note esz >= operation size +REVB 00000101 .. 1001 00 100 ... ..... ..... @rd_pg_rn +REVH 00000101 .. 1001 01 100 ... ..... ..... @rd_pg_rn +REVW 00000101 .. 1001 10 100 ... ..... ..... @rd_pg_rn +RBIT 00000101 .. 1001 11 100 ... ..... ..... @rd_pg_rn + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-10-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 ++ target/arm/sve_helper.c | 37 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 13 +++++++++++++ target/arm/sve.decode | 3 +++ 4 files changed, 55 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_rbit_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_splice, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ int32_t HELPER(sve_last_active_element)(void *vg, uint32_t pred_desc) return last_active_element(vg, DIV_ROUND_UP(oprsz, 8), esz); } + +void HELPER(sve_splice)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) +{ + intptr_t opr_sz = simd_oprsz(desc) / 8; + int esz = simd_data(desc); + uint64_t pg, first_g, last_g, len, mask = pred_esz_masks[esz]; + intptr_t i, first_i, last_i; + ARMVectorReg tmp; + + first_i = last_i = 0; + first_g = last_g = 0; + + /* Find the extent of the active elements within VG. */ + for (i = QEMU_ALIGN_UP(opr_sz, 8) - 8; i >= 0; i -= 8) { + pg = *(uint64_t *)(vg + i) & mask; + if (pg) { + if (last_g == 0) { + last_g = pg; + last_i = i; + } + first_g = pg; + first_i = i; + } + } + + len = 0; + if (first_g != 0) { + first_i = first_i * 8 + ctz64(first_g); + last_i = last_i * 8 + 63 - clz64(last_g); + len = last_i - first_i + (1 << esz); + if (vd == vm) { + vm = memcpy(&tmp, vm, opr_sz * 8); + } + swap_memmove(vd, vn + first_i, len); + } + swap_memmove(vd + len, vm, opr_sz * 8 - len); +} 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_RBIT(DisasContext *s, arg_rpr_esz *a, uint32_t insn) return do_zpz_ool(s, a, fns[a->esz]); } +static bool trans_SPLICE(DisasContext *s, arg_rprr_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_4_ool(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), + vsz, vsz, a->esz, gen_helper_sve_splice); + } + return true; +} + /* *** 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 @@ REVH 00000101 .. 1001 01 100 ... ..... ..... @rd_pg_rn REVW 00000101 .. 1001 10 100 ... ..... ..... @rd_pg_rn RBIT 00000101 .. 1001 11 100 ... ..... ..... @rd_pg_rn +# SVE vector splice (predicated) +SPLICE 00000101 .. 101 100 100 ... ..... ..... @rdn_pg_rm + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-11-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 9 +++++++ target/arm/sve_helper.c | 55 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 2 ++ target/arm/sve.decode | 6 +++++ 4 files changed, 72 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_lsl_zpzz_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_lsl_zpzz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_b, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_asr_zpzw_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_asr_zpzw_h, 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(sve_splice)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) } swap_memmove(vd + len, vm, opr_sz * 8 - len); } + +void HELPER(sve_sel_zpzz_b)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + uint64_t nn = n[i], mm = m[i]; + uint64_t pp = expand_pred_b(pg[H1(i)]); + d[i] = (nn & pp) | (mm & ~pp); + } +} + +void HELPER(sve_sel_zpzz_h)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + uint64_t nn = n[i], mm = m[i]; + uint64_t pp = expand_pred_h(pg[H1(i)]); + d[i] = (nn & pp) | (mm & ~pp); + } +} + +void HELPER(sve_sel_zpzz_s)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + uint64_t nn = n[i], mm = m[i]; + uint64_t pp = expand_pred_s(pg[H1(i)]); + d[i] = (nn & pp) | (mm & ~pp); + } +} + +void HELPER(sve_sel_zpzz_d)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + uint64_t nn = n[i], mm = m[i]; + d[i] = (pg[H1(i)] & 1 ? nn : mm); + } +} 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_UDIV_zpzz(DisasContext *s, arg_rprr_esz *a, uint32_t insn) return do_zpzz_ool(s, a, fns[a->esz]); } +DO_ZPZZ(SEL, sel) + #undef DO_ZPZZ /* 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_esz rn=%reg_movprfx @rdm_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 \ &rprr_esz rm=%reg_movprfx +@rd_pg4_rn_rm ........ esz:2 . rm:5 .. pg:4 rn:5 rd:5 &rprr_esz # Three register operand, with governing predicate, vector element size @rda_pg_rn_rm ........ esz:2 . rm:5 ... pg:3 rn:5 rd:5 \ @@ -XXX,XX +XXX,XX @@ RBIT 00000101 .. 1001 11 100 ... ..... ..... @rd_pg_rn # SVE vector splice (predicated) SPLICE 00000101 .. 101 100 100 ... ..... ..... @rdn_pg_rm +### SVE Select Vectors Group + +# SVE select vector elements (predicated) +SEL_zpzz 00000101 .. 1 ..... 11 .... ..... ..... @rd_pg4_rn_rm + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-12-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 115 +++++++++++++++++++++++ target/arm/sve_helper.c | 187 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 91 ++++++++++++++++++ target/arm/sve.decode | 24 +++++ 4 files changed, 417 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_rbit_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_splice, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzz_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzz_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzz_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzz_d, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmple_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplt_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplo_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpls_ppzw_b, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmple_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplt_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplo_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpls_ppzw_h, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_cmpeq_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpne_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpge_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpgt_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphi_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmphs_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmple_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplt_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmplo_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_cmpls_ppzw_s, TCG_CALL_NO_RWG, + i32, ptr, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ static uint32_t iter_predtest_fwd(uint64_t d, uint64_t g, uint32_t flags) return flags; } +/* This is an iterative function, called for each Pd and Pg word + * moving backward. + */ +static uint32_t iter_predtest_bwd(uint64_t d, uint64_t g, uint32_t flags) +{ + if (likely(g)) { + /* Compute C from first (i.e last) !(D & G). + Use bit 2 to signal first G bit seen. */ + if (!(flags & 4)) { + flags += 4 - 1; /* add bit 2, subtract C from PREDTEST_INIT */ + flags |= (d & pow2floor(g)) == 0; + } + + /* Accumulate Z from each D & G. */ + flags |= ((d & g) != 0) << 1; + + /* Compute N from last (i.e first) D & G. Replace previous. */ + flags = deposit32(flags, 31, 1, (d & (g & -g)) != 0); + } + return flags; +} + /* The same for a single word predicate. */ uint32_t HELPER(sve_predtest1)(uint64_t d, uint64_t g) { @@ -XXX,XX +XXX,XX @@ void HELPER(sve_sel_zpzz_d)(void *vd, void *vn, void *vm, d[i] = (pg[H1(i)] & 1 ? nn : mm); } } + +/* Two operand comparison controlled by a predicate. + * ??? It is very tempting to want to be able to expand this inline + * with x86 instructions, e.g. + * + * vcmpeqw zm, zn, %ymm0 + * vpmovmskb %ymm0, %eax + * and $0x5555, %eax + * and pg, %eax + * + * or even aarch64, e.g. + * + * // mask = 4000 1000 0400 0100 0040 0010 0004 0001 + * cmeq v0.8h, zn, zm + * and v0.8h, v0.8h, mask + * addv h0, v0.8h + * and v0.8b, pg + * + * However, coming up with an abstraction that allows vector inputs and + * a scalar output, and also handles the byte-ordering of sub-uint64_t + * scalar outputs, is tricky. + */ +#define DO_CMP_PPZZ(NAME, TYPE, OP, H, MASK) \ +uint32_t HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) \ +{ \ + intptr_t opr_sz = simd_oprsz(desc); \ + uint32_t flags = PREDTEST_INIT; \ + intptr_t i = opr_sz; \ + do { \ + uint64_t out = 0, pg; \ + do { \ + i -= sizeof(TYPE), out <<= sizeof(TYPE); \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + TYPE mm = *(TYPE *)(vm + H(i)); \ + out |= nn OP mm; \ + } while (i & 63); \ + pg = *(uint64_t *)(vg + (i >> 3)) & MASK; \ + out &= pg; \ + *(uint64_t *)(vd + (i >> 3)) = out; \ + flags = iter_predtest_bwd(out, pg, flags); \ + } while (i > 0); \ + return flags; \ +} + +#define DO_CMP_PPZZ_B(NAME, TYPE, OP) \ + DO_CMP_PPZZ(NAME, TYPE, OP, H1, 0xffffffffffffffffull) +#define DO_CMP_PPZZ_H(NAME, TYPE, OP) \ + DO_CMP_PPZZ(NAME, TYPE, OP, H1_2, 0x5555555555555555ull) +#define DO_CMP_PPZZ_S(NAME, TYPE, OP) \ + DO_CMP_PPZZ(NAME, TYPE, OP, H1_4, 0x1111111111111111ull) +#define DO_CMP_PPZZ_D(NAME, TYPE, OP) \ + DO_CMP_PPZZ(NAME, TYPE, OP, , 0x0101010101010101ull) + +DO_CMP_PPZZ_B(sve_cmpeq_ppzz_b, uint8_t, ==) +DO_CMP_PPZZ_H(sve_cmpeq_ppzz_h, uint16_t, ==) +DO_CMP_PPZZ_S(sve_cmpeq_ppzz_s, uint32_t, ==) +DO_CMP_PPZZ_D(sve_cmpeq_ppzz_d, uint64_t, ==) + +DO_CMP_PPZZ_B(sve_cmpne_ppzz_b, uint8_t, !=) +DO_CMP_PPZZ_H(sve_cmpne_ppzz_h, uint16_t, !=) +DO_CMP_PPZZ_S(sve_cmpne_ppzz_s, uint32_t, !=) +DO_CMP_PPZZ_D(sve_cmpne_ppzz_d, uint64_t, !=) + +DO_CMP_PPZZ_B(sve_cmpgt_ppzz_b, int8_t, >) +DO_CMP_PPZZ_H(sve_cmpgt_ppzz_h, int16_t, >) +DO_CMP_PPZZ_S(sve_cmpgt_ppzz_s, int32_t, >) +DO_CMP_PPZZ_D(sve_cmpgt_ppzz_d, int64_t, >) + +DO_CMP_PPZZ_B(sve_cmpge_ppzz_b, int8_t, >=) +DO_CMP_PPZZ_H(sve_cmpge_ppzz_h, int16_t, >=) +DO_CMP_PPZZ_S(sve_cmpge_ppzz_s, int32_t, >=) +DO_CMP_PPZZ_D(sve_cmpge_ppzz_d, int64_t, >=) + +DO_CMP_PPZZ_B(sve_cmphi_ppzz_b, uint8_t, >) +DO_CMP_PPZZ_H(sve_cmphi_ppzz_h, uint16_t, >) +DO_CMP_PPZZ_S(sve_cmphi_ppzz_s, uint32_t, >) +DO_CMP_PPZZ_D(sve_cmphi_ppzz_d, uint64_t, >) + +DO_CMP_PPZZ_B(sve_cmphs_ppzz_b, uint8_t, >=) +DO_CMP_PPZZ_H(sve_cmphs_ppzz_h, uint16_t, >=) +DO_CMP_PPZZ_S(sve_cmphs_ppzz_s, uint32_t, >=) +DO_CMP_PPZZ_D(sve_cmphs_ppzz_d, uint64_t, >=) + +#undef DO_CMP_PPZZ_B +#undef DO_CMP_PPZZ_H +#undef DO_CMP_PPZZ_S +#undef DO_CMP_PPZZ_D +#undef DO_CMP_PPZZ + +/* Similar, but the second source is "wide". */ +#define DO_CMP_PPZW(NAME, TYPE, TYPEW, OP, H, MASK) \ +uint32_t HELPER(NAME)(void *vd, void *vn, void *vm, void *vg, uint32_t desc) \ +{ \ + intptr_t opr_sz = simd_oprsz(desc); \ + uint32_t flags = PREDTEST_INIT; \ + intptr_t i = opr_sz; \ + do { \ + uint64_t out = 0, pg; \ + do { \ + TYPEW mm = *(TYPEW *)(vm + i - 8); \ + do { \ + i -= sizeof(TYPE), out <<= sizeof(TYPE); \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + out |= nn OP mm; \ + } while (i & 7); \ + } while (i & 63); \ + pg = *(uint64_t *)(vg + (i >> 3)) & MASK; \ + out &= pg; \ + *(uint64_t *)(vd + (i >> 3)) = out; \ + flags = iter_predtest_bwd(out, pg, flags); \ + } while (i > 0); \ + return flags; \ +} + +#define DO_CMP_PPZW_B(NAME, TYPE, TYPEW, OP) \ + DO_CMP_PPZW(NAME, TYPE, TYPEW, OP, H1, 0xffffffffffffffffull) +#define DO_CMP_PPZW_H(NAME, TYPE, TYPEW, OP) \ + DO_CMP_PPZW(NAME, TYPE, TYPEW, OP, H1_2, 0x5555555555555555ull) +#define DO_CMP_PPZW_S(NAME, TYPE, TYPEW, OP) \ + DO_CMP_PPZW(NAME, TYPE, TYPEW, OP, H1_4, 0x1111111111111111ull) + +DO_CMP_PPZW_B(sve_cmpeq_ppzw_b, uint8_t, uint64_t, ==) +DO_CMP_PPZW_H(sve_cmpeq_ppzw_h, uint16_t, uint64_t, ==) +DO_CMP_PPZW_S(sve_cmpeq_ppzw_s, uint32_t, uint64_t, ==) + +DO_CMP_PPZW_B(sve_cmpne_ppzw_b, uint8_t, uint64_t, !=) +DO_CMP_PPZW_H(sve_cmpne_ppzw_h, uint16_t, uint64_t, !=) +DO_CMP_PPZW_S(sve_cmpne_ppzw_s, uint32_t, uint64_t, !=) + +DO_CMP_PPZW_B(sve_cmpgt_ppzw_b, int8_t, int64_t, >) +DO_CMP_PPZW_H(sve_cmpgt_ppzw_h, int16_t, int64_t, >) +DO_CMP_PPZW_S(sve_cmpgt_ppzw_s, int32_t, int64_t, >) + +DO_CMP_PPZW_B(sve_cmpge_ppzw_b, int8_t, int64_t, >=) +DO_CMP_PPZW_H(sve_cmpge_ppzw_h, int16_t, int64_t, >=) +DO_CMP_PPZW_S(sve_cmpge_ppzw_s, int32_t, int64_t, >=) + +DO_CMP_PPZW_B(sve_cmphi_ppzw_b, uint8_t, uint64_t, >) +DO_CMP_PPZW_H(sve_cmphi_ppzw_h, uint16_t, uint64_t, >) +DO_CMP_PPZW_S(sve_cmphi_ppzw_s, uint32_t, uint64_t, >) + +DO_CMP_PPZW_B(sve_cmphs_ppzw_b, uint8_t, uint64_t, >=) +DO_CMP_PPZW_H(sve_cmphs_ppzw_h, uint16_t, uint64_t, >=) +DO_CMP_PPZW_S(sve_cmphs_ppzw_s, uint32_t, uint64_t, >=) + +DO_CMP_PPZW_B(sve_cmplt_ppzw_b, int8_t, int64_t, <) +DO_CMP_PPZW_H(sve_cmplt_ppzw_h, int16_t, int64_t, <) +DO_CMP_PPZW_S(sve_cmplt_ppzw_s, int32_t, int64_t, <) + +DO_CMP_PPZW_B(sve_cmple_ppzw_b, int8_t, int64_t, <=) +DO_CMP_PPZW_H(sve_cmple_ppzw_h, int16_t, int64_t, <=) +DO_CMP_PPZW_S(sve_cmple_ppzw_s, int32_t, int64_t, <=) + +DO_CMP_PPZW_B(sve_cmplo_ppzw_b, uint8_t, uint64_t, <) +DO_CMP_PPZW_H(sve_cmplo_ppzw_h, uint16_t, uint64_t, <) +DO_CMP_PPZW_S(sve_cmplo_ppzw_s, uint32_t, uint64_t, <) + +DO_CMP_PPZW_B(sve_cmpls_ppzw_b, uint8_t, uint64_t, <=) +DO_CMP_PPZW_H(sve_cmpls_ppzw_h, uint16_t, uint64_t, <=) +DO_CMP_PPZW_S(sve_cmpls_ppzw_s, uint32_t, uint64_t, <=) + +#undef DO_CMP_PPZW_B +#undef DO_CMP_PPZW_H +#undef DO_CMP_PPZW_S +#undef DO_CMP_PPZW 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 "trace-tcg.h" #include "translate-a64.h" + +typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr, + TCGv_ptr, TCGv_ptr, TCGv_i32); + /* * Helpers for extracting complex instruction fields. */ @@ -XXX,XX +XXX,XX @@ static bool trans_SPLICE(DisasContext *s, arg_rprr_esz *a, uint32_t insn) return true; } +/* + *** SVE Integer Compare - Vectors Group + */ + +static bool do_ppzz_flags(DisasContext *s, arg_rprr_esz *a, + gen_helper_gvec_flags_4 *gen_fn) +{ + TCGv_ptr pd, zn, zm, pg; + unsigned vsz; + TCGv_i32 t; + + if (gen_fn == NULL) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + vsz = vec_full_reg_size(s); + t = tcg_const_i32(simd_desc(vsz, vsz, 0)); + pd = tcg_temp_new_ptr(); + zn = tcg_temp_new_ptr(); + zm = tcg_temp_new_ptr(); + pg = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(pd, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(zn, cpu_env, vec_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(zm, cpu_env, vec_full_reg_offset(s, a->rm)); + tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg)); + + gen_fn(t, pd, zn, zm, pg, t); + + tcg_temp_free_ptr(pd); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(zm); + tcg_temp_free_ptr(pg); + + do_pred_flags(t); + + tcg_temp_free_i32(t); + return true; +} + +#define DO_PPZZ(NAME, name) \ +static bool trans_##NAME##_ppzz(DisasContext *s, arg_rprr_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_gvec_flags_4 * const fns[4] = { \ + gen_helper_sve_##name##_ppzz_b, gen_helper_sve_##name##_ppzz_h, \ + gen_helper_sve_##name##_ppzz_s, gen_helper_sve_##name##_ppzz_d, \ + }; \ + return do_ppzz_flags(s, a, fns[a->esz]); \ +} + +DO_PPZZ(CMPEQ, cmpeq) +DO_PPZZ(CMPNE, cmpne) +DO_PPZZ(CMPGT, cmpgt) +DO_PPZZ(CMPGE, cmpge) +DO_PPZZ(CMPHI, cmphi) +DO_PPZZ(CMPHS, cmphs) + +#undef DO_PPZZ + +#define DO_PPZW(NAME, name) \ +static bool trans_##NAME##_ppzw(DisasContext *s, arg_rprr_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_gvec_flags_4 * const fns[4] = { \ + gen_helper_sve_##name##_ppzw_b, gen_helper_sve_##name##_ppzw_h, \ + gen_helper_sve_##name##_ppzw_s, NULL \ + }; \ + return do_ppzz_flags(s, a, fns[a->esz]); \ +} + +DO_PPZW(CMPEQ, cmpeq) +DO_PPZW(CMPNE, cmpne) +DO_PPZW(CMPGT, cmpgt) +DO_PPZW(CMPGE, cmpge) +DO_PPZW(CMPHI, cmphi) +DO_PPZW(CMPHS, cmphs) +DO_PPZW(CMPLT, cmplt) +DO_PPZW(CMPLE, cmple) +DO_PPZW(CMPLO, cmplo) +DO_PPZW(CMPLS, cmpls) + +#undef DO_PPZW + /* *** 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 @@ @rdm_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 \ &rprr_esz rm=%reg_movprfx @rd_pg4_rn_rm ........ esz:2 . rm:5 .. pg:4 rn:5 rd:5 &rprr_esz +@pd_pg_rn_rm ........ esz:2 . rm:5 ... pg:3 rn:5 . rd:4 &rprr_esz # Three register operand, with governing predicate, vector element size @rda_pg_rn_rm ........ esz:2 . rm:5 ... pg:3 rn:5 rd:5 \ @@ -XXX,XX +XXX,XX @@ SPLICE 00000101 .. 101 100 100 ... ..... ..... @rdn_pg_rm # SVE select vector elements (predicated) SEL_zpzz 00000101 .. 1 ..... 11 .... ..... ..... @rd_pg4_rn_rm +### SVE Integer Compare - Vectors Group + +# SVE integer compare_vectors +CMPHS_ppzz 00100100 .. 0 ..... 000 ... ..... 0 .... @pd_pg_rn_rm +CMPHI_ppzz 00100100 .. 0 ..... 000 ... ..... 1 .... @pd_pg_rn_rm +CMPGE_ppzz 00100100 .. 0 ..... 100 ... ..... 0 .... @pd_pg_rn_rm +CMPGT_ppzz 00100100 .. 0 ..... 100 ... ..... 1 .... @pd_pg_rn_rm +CMPEQ_ppzz 00100100 .. 0 ..... 101 ... ..... 0 .... @pd_pg_rn_rm +CMPNE_ppzz 00100100 .. 0 ..... 101 ... ..... 1 .... @pd_pg_rn_rm + +# SVE integer compare with wide elements +# Note these require esz != 3. +CMPEQ_ppzw 00100100 .. 0 ..... 001 ... ..... 0 .... @pd_pg_rn_rm +CMPNE_ppzw 00100100 .. 0 ..... 001 ... ..... 1 .... @pd_pg_rn_rm +CMPGE_ppzw 00100100 .. 0 ..... 010 ... ..... 0 .... @pd_pg_rn_rm +CMPGT_ppzw 00100100 .. 0 ..... 010 ... ..... 1 .... @pd_pg_rn_rm +CMPLT_ppzw 00100100 .. 0 ..... 011 ... ..... 0 .... @pd_pg_rn_rm +CMPLE_ppzw 00100100 .. 0 ..... 011 ... ..... 1 .... @pd_pg_rn_rm +CMPHS_ppzw 00100100 .. 0 ..... 110 ... ..... 0 .... @pd_pg_rn_rm +CMPHI_ppzw 00100100 .. 0 ..... 110 ... ..... 1 .... @pd_pg_rn_rm +CMPLO_ppzw 00100100 .. 0 ..... 111 ... ..... 0 .... @pd_pg_rn_rm +CMPLS_ppzw 00100100 .. 0 ..... 111 ... ..... 1 .... @pd_pg_rn_rm + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-13-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 44 +++++++++++++++++++ target/arm/sve_helper.c | 88 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 66 ++++++++++++++++++++++++++++ target/arm/sve.decode | 23 ++++++++++ 4 files changed, 221 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_cmplo_ppzw_s, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve_cmpls_ppzw_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpeq_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpne_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpgt_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpge_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplt_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmple_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphs_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphi_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplo_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpls_ppzi_b, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_cmpeq_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpne_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpgt_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpge_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplt_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmple_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphs_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphi_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplo_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpls_ppzi_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_cmpeq_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpne_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpgt_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpge_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplt_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmple_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphs_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphi_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplo_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpls_ppzi_s, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_cmpeq_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpne_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpgt_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpge_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplt_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmple_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphs_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmphi_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmplo_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_cmpls_ppzi_d, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(sve_and_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_bic_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_eor_pppp, 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 @@ DO_CMP_PPZW_S(sve_cmpls_ppzw_s, uint32_t, uint64_t, <=) #undef DO_CMP_PPZW_H #undef DO_CMP_PPZW_S #undef DO_CMP_PPZW + +/* Similar, but the second source is immediate. */ +#define DO_CMP_PPZI(NAME, TYPE, OP, H, MASK) \ +uint32_t HELPER(NAME)(void *vd, void *vn, void *vg, uint32_t desc) \ +{ \ + intptr_t opr_sz = simd_oprsz(desc); \ + uint32_t flags = PREDTEST_INIT; \ + TYPE mm = simd_data(desc); \ + intptr_t i = opr_sz; \ + do { \ + uint64_t out = 0, pg; \ + do { \ + i -= sizeof(TYPE), out <<= sizeof(TYPE); \ + TYPE nn = *(TYPE *)(vn + H(i)); \ + out |= nn OP mm; \ + } while (i & 63); \ + pg = *(uint64_t *)(vg + (i >> 3)) & MASK; \ + out &= pg; \ + *(uint64_t *)(vd + (i >> 3)) = out; \ + flags = iter_predtest_bwd(out, pg, flags); \ + } while (i > 0); \ + return flags; \ +} + +#define DO_CMP_PPZI_B(NAME, TYPE, OP) \ + DO_CMP_PPZI(NAME, TYPE, OP, H1, 0xffffffffffffffffull) +#define DO_CMP_PPZI_H(NAME, TYPE, OP) \ + DO_CMP_PPZI(NAME, TYPE, OP, H1_2, 0x5555555555555555ull) +#define DO_CMP_PPZI_S(NAME, TYPE, OP) \ + DO_CMP_PPZI(NAME, TYPE, OP, H1_4, 0x1111111111111111ull) +#define DO_CMP_PPZI_D(NAME, TYPE, OP) \ + DO_CMP_PPZI(NAME, TYPE, OP, , 0x0101010101010101ull) + +DO_CMP_PPZI_B(sve_cmpeq_ppzi_b, uint8_t, ==) +DO_CMP_PPZI_H(sve_cmpeq_ppzi_h, uint16_t, ==) +DO_CMP_PPZI_S(sve_cmpeq_ppzi_s, uint32_t, ==) +DO_CMP_PPZI_D(sve_cmpeq_ppzi_d, uint64_t, ==) + +DO_CMP_PPZI_B(sve_cmpne_ppzi_b, uint8_t, !=) +DO_CMP_PPZI_H(sve_cmpne_ppzi_h, uint16_t, !=) +DO_CMP_PPZI_S(sve_cmpne_ppzi_s, uint32_t, !=) +DO_CMP_PPZI_D(sve_cmpne_ppzi_d, uint64_t, !=) + +DO_CMP_PPZI_B(sve_cmpgt_ppzi_b, int8_t, >) +DO_CMP_PPZI_H(sve_cmpgt_ppzi_h, int16_t, >) +DO_CMP_PPZI_S(sve_cmpgt_ppzi_s, int32_t, >) +DO_CMP_PPZI_D(sve_cmpgt_ppzi_d, int64_t, >) + +DO_CMP_PPZI_B(sve_cmpge_ppzi_b, int8_t, >=) +DO_CMP_PPZI_H(sve_cmpge_ppzi_h, int16_t, >=) +DO_CMP_PPZI_S(sve_cmpge_ppzi_s, int32_t, >=) +DO_CMP_PPZI_D(sve_cmpge_ppzi_d, int64_t, >=) + +DO_CMP_PPZI_B(sve_cmphi_ppzi_b, uint8_t, >) +DO_CMP_PPZI_H(sve_cmphi_ppzi_h, uint16_t, >) +DO_CMP_PPZI_S(sve_cmphi_ppzi_s, uint32_t, >) +DO_CMP_PPZI_D(sve_cmphi_ppzi_d, uint64_t, >) + +DO_CMP_PPZI_B(sve_cmphs_ppzi_b, uint8_t, >=) +DO_CMP_PPZI_H(sve_cmphs_ppzi_h, uint16_t, >=) +DO_CMP_PPZI_S(sve_cmphs_ppzi_s, uint32_t, >=) +DO_CMP_PPZI_D(sve_cmphs_ppzi_d, uint64_t, >=) + +DO_CMP_PPZI_B(sve_cmplt_ppzi_b, int8_t, <) +DO_CMP_PPZI_H(sve_cmplt_ppzi_h, int16_t, <) +DO_CMP_PPZI_S(sve_cmplt_ppzi_s, int32_t, <) +DO_CMP_PPZI_D(sve_cmplt_ppzi_d, int64_t, <) + +DO_CMP_PPZI_B(sve_cmple_ppzi_b, int8_t, <=) +DO_CMP_PPZI_H(sve_cmple_ppzi_h, int16_t, <=) +DO_CMP_PPZI_S(sve_cmple_ppzi_s, int32_t, <=) +DO_CMP_PPZI_D(sve_cmple_ppzi_d, int64_t, <=) + +DO_CMP_PPZI_B(sve_cmplo_ppzi_b, uint8_t, <) +DO_CMP_PPZI_H(sve_cmplo_ppzi_h, uint16_t, <) +DO_CMP_PPZI_S(sve_cmplo_ppzi_s, uint32_t, <) +DO_CMP_PPZI_D(sve_cmplo_ppzi_d, uint64_t, <) + +DO_CMP_PPZI_B(sve_cmpls_ppzi_b, uint8_t, <=) +DO_CMP_PPZI_H(sve_cmpls_ppzi_h, uint16_t, <=) +DO_CMP_PPZI_S(sve_cmpls_ppzi_s, uint32_t, <=) +DO_CMP_PPZI_D(sve_cmpls_ppzi_d, uint64_t, <=) + +#undef DO_CMP_PPZI_B +#undef DO_CMP_PPZI_H +#undef DO_CMP_PPZI_S +#undef DO_CMP_PPZI_D +#undef DO_CMP_PPZI 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 "translate-a64.h" +typedef void gen_helper_gvec_flags_3(TCGv_i32, TCGv_ptr, TCGv_ptr, + TCGv_ptr, TCGv_i32); typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); @@ -XXX,XX +XXX,XX @@ DO_PPZW(CMPLS, cmpls) #undef DO_PPZW +/* + *** SVE Integer Compare - Immediate Groups + */ + +static bool do_ppzi_flags(DisasContext *s, arg_rpri_esz *a, + gen_helper_gvec_flags_3 *gen_fn) +{ + TCGv_ptr pd, zn, pg; + unsigned vsz; + TCGv_i32 t; + + if (gen_fn == NULL) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + vsz = vec_full_reg_size(s); + t = tcg_const_i32(simd_desc(vsz, vsz, a->imm)); + pd = tcg_temp_new_ptr(); + zn = tcg_temp_new_ptr(); + pg = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(pd, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(zn, cpu_env, vec_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(pg, cpu_env, pred_full_reg_offset(s, a->pg)); + + gen_fn(t, pd, zn, pg, t); + + tcg_temp_free_ptr(pd); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pg); + + do_pred_flags(t); + + tcg_temp_free_i32(t); + return true; +} + +#define DO_PPZI(NAME, name) \ +static bool trans_##NAME##_ppzi(DisasContext *s, arg_rpri_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_gvec_flags_3 * const fns[4] = { \ + gen_helper_sve_##name##_ppzi_b, gen_helper_sve_##name##_ppzi_h, \ + gen_helper_sve_##name##_ppzi_s, gen_helper_sve_##name##_ppzi_d, \ + }; \ + return do_ppzi_flags(s, a, fns[a->esz]); \ +} + +DO_PPZI(CMPEQ, cmpeq) +DO_PPZI(CMPNE, cmpne) +DO_PPZI(CMPGT, cmpgt) +DO_PPZI(CMPGE, cmpge) +DO_PPZI(CMPHI, cmphi) +DO_PPZI(CMPHS, cmphs) +DO_PPZI(CMPLT, cmplt) +DO_PPZI(CMPLE, cmple) +DO_PPZI(CMPLO, cmplo) +DO_PPZI(CMPLS, cmpls) + +#undef DO_PPZI + /* *** 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 @@ @rdn_dbm ........ .. .... dbm:13 rd:5 \ &rr_dbm rn=%reg_movprfx +# Predicate output, vector and immediate input, +# controlling predicate, element size. +@pd_pg_rn_i7 ........ esz:2 . imm:7 . pg:3 rn:5 . rd:4 &rpri_esz +@pd_pg_rn_i5 ........ esz:2 . imm:s5 ... pg:3 rn:5 . rd:4 &rpri_esz + # Basic Load/Store with 9-bit immediate offset @pd_rn_i9 ........ ........ ...... rn:5 . rd:4 \ &rri imm=%imm9_16_10 @@ -XXX,XX +XXX,XX @@ CMPHI_ppzw 00100100 .. 0 ..... 110 ... ..... 1 .... @pd_pg_rn_rm CMPLO_ppzw 00100100 .. 0 ..... 111 ... ..... 0 .... @pd_pg_rn_rm CMPLS_ppzw 00100100 .. 0 ..... 111 ... ..... 1 .... @pd_pg_rn_rm +### SVE Integer Compare - Unsigned Immediate Group + +# SVE integer compare with unsigned immediate +CMPHS_ppzi 00100100 .. 1 ....... 0 ... ..... 0 .... @pd_pg_rn_i7 +CMPHI_ppzi 00100100 .. 1 ....... 0 ... ..... 1 .... @pd_pg_rn_i7 +CMPLO_ppzi 00100100 .. 1 ....... 1 ... ..... 0 .... @pd_pg_rn_i7 +CMPLS_ppzi 00100100 .. 1 ....... 1 ... ..... 1 .... @pd_pg_rn_i7 + +### SVE Integer Compare - Signed Immediate Group + +# SVE integer compare with signed immediate +CMPGE_ppzi 00100101 .. 0 ..... 000 ... ..... 0 .... @pd_pg_rn_i5 +CMPGT_ppzi 00100101 .. 0 ..... 000 ... ..... 1 .... @pd_pg_rn_i5 +CMPLT_ppzi 00100101 .. 0 ..... 001 ... ..... 0 .... @pd_pg_rn_i5 +CMPLE_ppzi 00100101 .. 0 ..... 001 ... ..... 1 .... @pd_pg_rn_i5 +CMPEQ_ppzi 00100101 .. 0 ..... 100 ... ..... 0 .... @pd_pg_rn_i5 +CMPNE_ppzi 00100101 .. 0 ..... 100 ... ..... 1 .... @pd_pg_rn_i5 + ### SVE Predicate Logical Operations Group # SVE predicate logical operations -- 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: 20180613015641.5667-14-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 18 +++ target/arm/sve_helper.c | 248 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 106 ++++++++++++++++ target/arm/sve.decode | 19 +++ 4 files changed, 391 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_orn_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_nor_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_nand_pppp, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sve_brkpa, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_brkpb, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_brkpas, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_brkpbs, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_brka_z, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkb_z, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brka_m, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkb_m, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_brkas_z, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkbs_z, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkas_m, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkbs_m, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve_brkn, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sve_brkns, TCG_CALL_NO_RWG, i32, 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 @@ DO_CMP_PPZI_D(sve_cmpls_ppzi_d, uint64_t, <=) #undef DO_CMP_PPZI_S #undef DO_CMP_PPZI_D #undef DO_CMP_PPZI + +/* Similar to the ARM LastActive pseudocode function. */ +static bool last_active_pred(void *vd, void *vg, intptr_t oprsz) +{ + intptr_t i; + + for (i = QEMU_ALIGN_UP(oprsz, 8) - 8; i >= 0; i -= 8) { + uint64_t pg = *(uint64_t *)(vg + i); + if (pg) { + return (pow2floor(pg) & *(uint64_t *)(vd + i)) != 0; + } + } + return 0; +} + +/* Compute a mask into RETB that is true for all G, up to and including + * (if after) or excluding (if !after) the first G & N. + * Return true if BRK found. + */ +static bool compute_brk(uint64_t *retb, uint64_t n, uint64_t g, + bool brk, bool after) +{ + uint64_t b; + + if (brk) { + b = 0; + } else if ((g & n) == 0) { + /* For all G, no N are set; break not found. */ + b = g; + } else { + /* Break somewhere in N. Locate it. */ + b = g & n; /* guard true, pred true */ + b = b & -b; /* first such */ + if (after) { + b = b | (b - 1); /* break after same */ + } else { + b = b - 1; /* break before same */ + } + brk = true; + } + + *retb = b; + return brk; +} + +/* Compute a zeroing BRK. */ +static void compute_brk_z(uint64_t *d, uint64_t *n, uint64_t *g, + intptr_t oprsz, bool after) +{ + bool brk = false; + intptr_t i; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); ++i) { + uint64_t this_b, this_g = g[i]; + + brk = compute_brk(&this_b, n[i], this_g, brk, after); + d[i] = this_b & this_g; + } +} + +/* Likewise, but also compute flags. */ +static uint32_t compute_brks_z(uint64_t *d, uint64_t *n, uint64_t *g, + intptr_t oprsz, bool after) +{ + uint32_t flags = PREDTEST_INIT; + bool brk = false; + intptr_t i; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); ++i) { + uint64_t this_b, this_d, this_g = g[i]; + + brk = compute_brk(&this_b, n[i], this_g, brk, after); + d[i] = this_d = this_b & this_g; + flags = iter_predtest_fwd(this_d, this_g, flags); + } + return flags; +} + +/* Compute a merging BRK. */ +static void compute_brk_m(uint64_t *d, uint64_t *n, uint64_t *g, + intptr_t oprsz, bool after) +{ + bool brk = false; + intptr_t i; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); ++i) { + uint64_t this_b, this_g = g[i]; + + brk = compute_brk(&this_b, n[i], this_g, brk, after); + d[i] = (this_b & this_g) | (d[i] & ~this_g); + } +} + +/* Likewise, but also compute flags. */ +static uint32_t compute_brks_m(uint64_t *d, uint64_t *n, uint64_t *g, + intptr_t oprsz, bool after) +{ + uint32_t flags = PREDTEST_INIT; + bool brk = false; + intptr_t i; + + for (i = 0; i < oprsz / 8; ++i) { + uint64_t this_b, this_d = d[i], this_g = g[i]; + + brk = compute_brk(&this_b, n[i], this_g, brk, after); + d[i] = this_d = (this_b & this_g) | (this_d & ~this_g); + flags = iter_predtest_fwd(this_d, this_g, flags); + } + return flags; +} + +static uint32_t do_zero(ARMPredicateReg *d, intptr_t oprsz) +{ + /* It is quicker to zero the whole predicate than loop on OPRSZ. + * The compiler should turn this into 4 64-bit integer stores. + */ + memset(d, 0, sizeof(ARMPredicateReg)); + return PREDTEST_INIT; +} + +void HELPER(sve_brkpa)(void *vd, void *vn, void *vm, void *vg, + uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + if (last_active_pred(vn, vg, oprsz)) { + compute_brk_z(vd, vm, vg, oprsz, true); + } else { + do_zero(vd, oprsz); + } +} + +uint32_t HELPER(sve_brkpas)(void *vd, void *vn, void *vm, void *vg, + uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + if (last_active_pred(vn, vg, oprsz)) { + return compute_brks_z(vd, vm, vg, oprsz, true); + } else { + return do_zero(vd, oprsz); + } +} + +void HELPER(sve_brkpb)(void *vd, void *vn, void *vm, void *vg, + uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + if (last_active_pred(vn, vg, oprsz)) { + compute_brk_z(vd, vm, vg, oprsz, false); + } else { + do_zero(vd, oprsz); + } +} + +uint32_t HELPER(sve_brkpbs)(void *vd, void *vn, void *vm, void *vg, + uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + if (last_active_pred(vn, vg, oprsz)) { + return compute_brks_z(vd, vm, vg, oprsz, false); + } else { + return do_zero(vd, oprsz); + } +} + +void HELPER(sve_brka_z)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + compute_brk_z(vd, vn, vg, oprsz, true); +} + +uint32_t HELPER(sve_brkas_z)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + return compute_brks_z(vd, vn, vg, oprsz, true); +} + +void HELPER(sve_brkb_z)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + compute_brk_z(vd, vn, vg, oprsz, false); +} + +uint32_t HELPER(sve_brkbs_z)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + return compute_brks_z(vd, vn, vg, oprsz, false); +} + +void HELPER(sve_brka_m)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + compute_brk_m(vd, vn, vg, oprsz, true); +} + +uint32_t HELPER(sve_brkas_m)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + return compute_brks_m(vd, vn, vg, oprsz, true); +} + +void HELPER(sve_brkb_m)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + compute_brk_m(vd, vn, vg, oprsz, false); +} + +uint32_t HELPER(sve_brkbs_m)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + return compute_brks_m(vd, vn, vg, oprsz, false); +} + +void HELPER(sve_brkn)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + + if (!last_active_pred(vn, vg, oprsz)) { + do_zero(vd, oprsz); + } +} + +/* As if PredTest(Ones(PL), D, esz). */ +static uint32_t predtest_ones(ARMPredicateReg *d, intptr_t oprsz, + uint64_t esz_mask) +{ + uint32_t flags = PREDTEST_INIT; + intptr_t i; + + for (i = 0; i < oprsz / 8; i++) { + flags = iter_predtest_fwd(d->p[i], esz_mask, flags); + } + if (oprsz & 7) { + uint64_t mask = ~(-1ULL << (8 * (oprsz & 7))); + flags = iter_predtest_fwd(d->p[i], esz_mask & mask, flags); + } + return flags; +} + +uint32_t HELPER(sve_brkns)(void *vd, void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + + if (last_active_pred(vn, vg, oprsz)) { + return predtest_ones(vd, oprsz, -1); + } else { + return do_zero(vd, oprsz); + } +} 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_PPZI(CMPLS, cmpls) #undef DO_PPZI +/* + *** SVE Partition Break Group + */ + +static bool do_brk3(DisasContext *s, arg_rprr_s *a, + gen_helper_gvec_4 *fn, gen_helper_gvec_flags_4 *fn_s) +{ + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = pred_full_reg_size(s); + + /* Predicate sizes may be smaller and cannot use simd_desc. */ + TCGv_ptr d = tcg_temp_new_ptr(); + TCGv_ptr n = tcg_temp_new_ptr(); + TCGv_ptr m = tcg_temp_new_ptr(); + TCGv_ptr g = tcg_temp_new_ptr(); + TCGv_i32 t = tcg_const_i32(vsz - 2); + + tcg_gen_addi_ptr(d, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(n, cpu_env, pred_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(m, cpu_env, pred_full_reg_offset(s, a->rm)); + tcg_gen_addi_ptr(g, cpu_env, pred_full_reg_offset(s, a->pg)); + + if (a->s) { + fn_s(t, d, n, m, g, t); + do_pred_flags(t); + } else { + fn(d, n, m, g, t); + } + tcg_temp_free_ptr(d); + tcg_temp_free_ptr(n); + tcg_temp_free_ptr(m); + tcg_temp_free_ptr(g); + tcg_temp_free_i32(t); + return true; +} + +static bool do_brk2(DisasContext *s, arg_rpr_s *a, + gen_helper_gvec_3 *fn, gen_helper_gvec_flags_3 *fn_s) +{ + if (!sve_access_check(s)) { + return true; + } + + unsigned vsz = pred_full_reg_size(s); + + /* Predicate sizes may be smaller and cannot use simd_desc. */ + TCGv_ptr d = tcg_temp_new_ptr(); + TCGv_ptr n = tcg_temp_new_ptr(); + TCGv_ptr g = tcg_temp_new_ptr(); + TCGv_i32 t = tcg_const_i32(vsz - 2); + + tcg_gen_addi_ptr(d, cpu_env, pred_full_reg_offset(s, a->rd)); + tcg_gen_addi_ptr(n, cpu_env, pred_full_reg_offset(s, a->rn)); + tcg_gen_addi_ptr(g, cpu_env, pred_full_reg_offset(s, a->pg)); + + if (a->s) { + fn_s(t, d, n, g, t); + do_pred_flags(t); + } else { + fn(d, n, g, t); + } + tcg_temp_free_ptr(d); + tcg_temp_free_ptr(n); + tcg_temp_free_ptr(g); + tcg_temp_free_i32(t); + return true; +} + +static bool trans_BRKPA(DisasContext *s, arg_rprr_s *a, uint32_t insn) +{ + return do_brk3(s, a, gen_helper_sve_brkpa, gen_helper_sve_brkpas); +} + +static bool trans_BRKPB(DisasContext *s, arg_rprr_s *a, uint32_t insn) +{ + return do_brk3(s, a, gen_helper_sve_brkpb, gen_helper_sve_brkpbs); +} + +static bool trans_BRKA_m(DisasContext *s, arg_rpr_s *a, uint32_t insn) +{ + return do_brk2(s, a, gen_helper_sve_brka_m, gen_helper_sve_brkas_m); +} + +static bool trans_BRKB_m(DisasContext *s, arg_rpr_s *a, uint32_t insn) +{ + return do_brk2(s, a, gen_helper_sve_brkb_m, gen_helper_sve_brkbs_m); +} + +static bool trans_BRKA_z(DisasContext *s, arg_rpr_s *a, uint32_t insn) +{ + return do_brk2(s, a, gen_helper_sve_brka_z, gen_helper_sve_brkas_z); +} + +static bool trans_BRKB_z(DisasContext *s, arg_rpr_s *a, uint32_t insn) +{ + return do_brk2(s, a, gen_helper_sve_brkb_z, gen_helper_sve_brkbs_z); +} + +static bool trans_BRKN(DisasContext *s, arg_rpr_s *a, uint32_t insn) +{ + return do_brk2(s, a, gen_helper_sve_brkn, gen_helper_sve_brkns); +} + /* *** 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 @@ &rri_esz rd rn imm esz &rrr_esz rd rn rm esz &rpr_esz rd pg rn esz +&rpr_s rd pg rn s &rprr_s rd pg rn rm s &rprr_esz rd pg rn rm esz &rprrr_esz rd pg rn rm ra esz @@ -XXX,XX +XXX,XX @@ @pd_pn ........ esz:2 .. .... ....... rn:4 . rd:4 &rr_esz @rd_rn ........ esz:2 ...... ...... rn:5 rd:5 &rr_esz +# Two operand with governing predicate, flags setting +@pd_pg_pn_s ........ . s:1 ...... .. pg:4 . rn:4 . rd:4 &rpr_s + # Three operand with unused vector element size @rd_rn_rm_e0 ........ ... rm:5 ... ... rn:5 rd:5 &rrr_esz esz=0 @@ -XXX,XX +XXX,XX @@ PFIRST 00100101 01 011 000 11000 00 .... 0 .... @pd_pn_e0 # SVE predicate next active PNEXT 00100101 .. 011 001 11000 10 .... 0 .... @pd_pn +### SVE Partition Break Group + +# SVE propagate break from previous partition +BRKPA 00100101 0. 00 .... 11 .... 0 .... 0 .... @pd_pg_pn_pm_s +BRKPB 00100101 0. 00 .... 11 .... 0 .... 1 .... @pd_pg_pn_pm_s + +# SVE partition break condition +BRKA_z 00100101 0. 01000001 .... 0 .... 0 .... @pd_pg_pn_s +BRKB_z 00100101 1. 01000001 .... 0 .... 0 .... @pd_pg_pn_s +BRKA_m 00100101 0. 01000001 .... 0 .... 1 .... @pd_pg_pn_s +BRKB_m 00100101 1. 01000001 .... 0 .... 1 .... @pd_pg_pn_s + +# SVE propagate break to next partition +BRKN 00100101 0. 01100001 .... 0 .... 0 .... @pd_pg_pn_s + ### 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: 20180613015641.5667-15-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 + target/arm/sve_helper.c | 14 ++++ target/arm/translate-sve.c | 133 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 27 ++++++++ 4 files changed, 176 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_brkbs_m, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_brkn, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_brkns, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_cntp, TCG_CALL_NO_RWG, i64, 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 @@ uint32_t HELPER(sve_brkns)(void *vd, void *vn, void *vg, uint32_t pred_desc) return do_zero(vd, oprsz); } } + +uint64_t HELPER(sve_cntp)(void *vn, void *vg, uint32_t pred_desc) +{ + intptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + intptr_t esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + uint64_t *n = vn, *g = vg, sum = 0, mask = pred_esz_masks[esz]; + intptr_t i; + + for (i = 0; i < DIV_ROUND_UP(oprsz, 8); ++i) { + uint64_t t = n[i] & g[i] & mask; + sum += ctpop64(t); + } + return sum; +} 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 "translate-a64.h" +typedef void GVecGen2sFn(unsigned, uint32_t, uint32_t, + TCGv_i64, uint32_t, uint32_t); + typedef void gen_helper_gvec_flags_3(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void gen_helper_gvec_flags_4(TCGv_i32, TCGv_ptr, TCGv_ptr, @@ -XXX,XX +XXX,XX @@ static bool trans_BRKN(DisasContext *s, arg_rpr_s *a, uint32_t insn) return do_brk2(s, a, gen_helper_sve_brkn, gen_helper_sve_brkns); } +/* + *** SVE Predicate Count Group + */ + +static void do_cntp(DisasContext *s, TCGv_i64 val, int esz, int pn, int pg) +{ + unsigned psz = pred_full_reg_size(s); + + if (psz <= 8) { + uint64_t psz_mask; + + tcg_gen_ld_i64(val, cpu_env, pred_full_reg_offset(s, pn)); + if (pn != pg) { + TCGv_i64 g = tcg_temp_new_i64(); + tcg_gen_ld_i64(g, cpu_env, pred_full_reg_offset(s, pg)); + tcg_gen_and_i64(val, val, g); + tcg_temp_free_i64(g); + } + + /* Reduce the pred_esz_masks value simply to reduce the + * size of the code generated here. + */ + psz_mask = MAKE_64BIT_MASK(0, psz * 8); + tcg_gen_andi_i64(val, val, pred_esz_masks[esz] & psz_mask); + + tcg_gen_ctpop_i64(val, val); + } else { + TCGv_ptr t_pn = tcg_temp_new_ptr(); + TCGv_ptr t_pg = tcg_temp_new_ptr(); + unsigned desc; + TCGv_i32 t_desc; + + desc = psz - 2; + desc = deposit32(desc, SIMD_DATA_SHIFT, 2, esz); + + tcg_gen_addi_ptr(t_pn, cpu_env, pred_full_reg_offset(s, pn)); + tcg_gen_addi_ptr(t_pg, cpu_env, pred_full_reg_offset(s, pg)); + t_desc = tcg_const_i32(desc); + + gen_helper_sve_cntp(val, t_pn, t_pg, t_desc); + tcg_temp_free_ptr(t_pn); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i32(t_desc); + } +} + +static bool trans_CNTP(DisasContext *s, arg_CNTP *a, uint32_t insn) +{ + if (sve_access_check(s)) { + do_cntp(s, cpu_reg(s, a->rd), a->esz, a->rn, a->pg); + } + return true; +} + +static bool trans_INCDECP_r(DisasContext *s, arg_incdec_pred *a, + uint32_t insn) +{ + if (sve_access_check(s)) { + TCGv_i64 reg = cpu_reg(s, a->rd); + TCGv_i64 val = tcg_temp_new_i64(); + + do_cntp(s, val, a->esz, a->pg, a->pg); + if (a->d) { + tcg_gen_sub_i64(reg, reg, val); + } else { + tcg_gen_add_i64(reg, reg, val); + } + tcg_temp_free_i64(val); + } + return true; +} + +static bool trans_INCDECP_z(DisasContext *s, arg_incdec2_pred *a, + uint32_t insn) +{ + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_i64 val = tcg_temp_new_i64(); + GVecGen2sFn *gvec_fn = a->d ? tcg_gen_gvec_subs : tcg_gen_gvec_adds; + + do_cntp(s, val, a->esz, a->pg, a->pg); + gvec_fn(a->esz, vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), val, vsz, vsz); + } + return true; +} + +static bool trans_SINCDECP_r_32(DisasContext *s, arg_incdec_pred *a, + uint32_t insn) +{ + if (sve_access_check(s)) { + TCGv_i64 reg = cpu_reg(s, a->rd); + TCGv_i64 val = tcg_temp_new_i64(); + + do_cntp(s, val, a->esz, a->pg, a->pg); + do_sat_addsub_32(reg, val, a->u, a->d); + } + return true; +} + +static bool trans_SINCDECP_r_64(DisasContext *s, arg_incdec_pred *a, + uint32_t insn) +{ + if (sve_access_check(s)) { + TCGv_i64 reg = cpu_reg(s, a->rd); + TCGv_i64 val = tcg_temp_new_i64(); + + do_cntp(s, val, a->esz, a->pg, a->pg); + do_sat_addsub_64(reg, val, a->u, a->d); + } + return true; +} + +static bool trans_SINCDECP_z(DisasContext *s, arg_incdec2_pred *a, + uint32_t insn) +{ + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + TCGv_i64 val = tcg_temp_new_i64(); + do_cntp(s, val, a->esz, a->pg, a->pg); + do_sat_addsub_vec(s, a->esz, a->rd, a->rn, val, a->u, a->d); + } + return true; +} + /* *** 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 @@ &ptrue rd esz pat s &incdec_cnt rd pat esz imm d u &incdec2_cnt rd rn pat esz imm d u +&incdec_pred rd pg esz d u +&incdec2_pred rd rn pg esz d u ########################################################################### # Named instruction formats. These are generally used to @@ -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 # Two register operands with a 6-bit signed immediate. @rd_rn_i6 ........ ... rn:5 ..... imm:s6 rd:5 &rri @@ -XXX,XX +XXX,XX @@ @incdec2_cnt ........ esz:2 .. .... ...... pat:5 rd:5 \ &incdec2_cnt imm=%imm4_16_p1 rn=%reg_movprfx +# One register, predicate. +# User must fill in U and D. +@incdec_pred ........ esz:2 .... .. ..... .. pg:4 rd:5 &incdec_pred +@incdec2_pred ........ esz:2 .... .. ..... .. pg:4 rd:5 \ + &incdec2_pred rn=%reg_movprfx + ########################################################################### # Instruction patterns. Grouped according to the SVE encodingindex.xhtml. @@ -XXX,XX +XXX,XX @@ BRKB_m 00100101 1. 01000001 .... 0 .... 1 .... @pd_pg_pn_s # SVE propagate break to next partition BRKN 00100101 0. 01100001 .... 0 .... 0 .... @pd_pg_pn_s +### SVE Predicate Count Group + +# SVE predicate count +CNTP 00100101 .. 100 000 10 .... 0 .... ..... @rd_pg4_pn + +# SVE inc/dec register by predicate count +INCDECP_r 00100101 .. 10110 d:1 10001 00 .... ..... @incdec_pred u=1 + +# SVE inc/dec vector by predicate count +INCDECP_z 00100101 .. 10110 d:1 10000 00 .... ..... @incdec2_pred u=1 + +# SVE saturating inc/dec register by predicate count +SINCDECP_r_32 00100101 .. 1010 d:1 u:1 10001 00 .... ..... @incdec_pred +SINCDECP_r_64 00100101 .. 1010 d:1 u:1 10001 10 .... ..... @incdec_pred + +# SVE saturating inc/dec vector by predicate count +SINCDECP_z 00100101 .. 1010 d:1 u:1 10000 00 .... ..... @incdec2_pred + ### SVE Memory - 32-bit Gather and Unsized Contiguous Group # SVE load predicate register -- 2.17.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20180613015641.5667-16-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 + target/arm/sve_helper.c | 31 ++++++++++++ target/arm/translate-sve.c | 99 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 8 +++ 4 files changed, 140 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_brkn, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_brkns, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_cntp, TCG_CALL_NO_RWG, i64, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(sve_while, TCG_CALL_NO_RWG, i32, ptr, i32, 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 @@ uint64_t HELPER(sve_cntp)(void *vn, void *vg, uint32_t pred_desc) } return sum; } + +uint32_t HELPER(sve_while)(void *vd, uint32_t count, uint32_t pred_desc) +{ + uintptr_t oprsz = extract32(pred_desc, 0, SIMD_OPRSZ_BITS) + 2; + intptr_t esz = extract32(pred_desc, SIMD_DATA_SHIFT, 2); + uint64_t esz_mask = pred_esz_masks[esz]; + ARMPredicateReg *d = vd; + uint32_t flags; + intptr_t i; + + /* Begin with a zero predicate register. */ + flags = do_zero(d, oprsz); + if (count == 0) { + return flags; + } + + /* Scale from predicate element count to bits. */ + count <<= esz; + /* Bound to the bits in the predicate. */ + count = MIN(count, oprsz * 8); + + /* Set all of the requested bits. */ + for (i = 0; i < count / 64; ++i) { + d->p[i] = esz_mask; + } + if (count & 63) { + d->p[i] = MAKE_64BIT_MASK(0, count & 63) & esz_mask; + } + + return predtest_ones(d, oprsz, esz_mask); +} 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_SINCDECP_z(DisasContext *s, arg_incdec2_pred *a, return true; } +/* + *** SVE Integer Compare Scalars Group + */ + +static bool trans_CTERM(DisasContext *s, arg_CTERM *a, uint32_t insn) +{ + if (!sve_access_check(s)) { + return true; + } + + TCGCond cond = (a->ne ? TCG_COND_NE : TCG_COND_EQ); + TCGv_i64 rn = read_cpu_reg(s, a->rn, a->sf); + TCGv_i64 rm = read_cpu_reg(s, a->rm, a->sf); + TCGv_i64 cmp = tcg_temp_new_i64(); + + tcg_gen_setcond_i64(cond, cmp, rn, rm); + tcg_gen_extrl_i64_i32(cpu_NF, cmp); + tcg_temp_free_i64(cmp); + + /* VF = !NF & !CF. */ + tcg_gen_xori_i32(cpu_VF, cpu_NF, 1); + tcg_gen_andc_i32(cpu_VF, cpu_VF, cpu_CF); + + /* Both NF and VF actually look at bit 31. */ + tcg_gen_neg_i32(cpu_NF, cpu_NF); + tcg_gen_neg_i32(cpu_VF, cpu_VF); + return true; +} + +static bool trans_WHILE(DisasContext *s, arg_WHILE *a, uint32_t insn) +{ + if (!sve_access_check(s)) { + return true; + } + + TCGv_i64 op0 = read_cpu_reg(s, a->rn, 1); + TCGv_i64 op1 = read_cpu_reg(s, a->rm, 1); + TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); + TCGv_i32 t2, t3; + TCGv_ptr ptr; + unsigned desc, vsz = vec_full_reg_size(s); + TCGCond cond; + + if (!a->sf) { + if (a->u) { + tcg_gen_ext32u_i64(op0, op0); + tcg_gen_ext32u_i64(op1, op1); + } else { + tcg_gen_ext32s_i64(op0, op0); + tcg_gen_ext32s_i64(op1, op1); + } + } + + /* For the helper, compress the different conditions into a computation + * of how many iterations for which the condition is true. + * + * This is slightly complicated by 0 <= UINT64_MAX, which is nominally + * 2**64 iterations, overflowing to 0. Of course, predicate registers + * aren't that large, so any value >= predicate size is sufficient. + */ + tcg_gen_sub_i64(t0, op1, op0); + + /* t0 = MIN(op1 - op0, vsz). */ + tcg_gen_movi_i64(t1, vsz); + tcg_gen_umin_i64(t0, t0, t1); + if (a->eq) { + /* Equality means one more iteration. */ + tcg_gen_addi_i64(t0, t0, 1); + } + + /* t0 = (condition true ? t0 : 0). */ + cond = (a->u + ? (a->eq ? TCG_COND_LEU : TCG_COND_LTU) + : (a->eq ? TCG_COND_LE : TCG_COND_LT)); + tcg_gen_movi_i64(t1, 0); + tcg_gen_movcond_i64(cond, t0, op0, op1, t0, t1); + + t2 = tcg_temp_new_i32(); + tcg_gen_extrl_i64_i32(t2, t0); + tcg_temp_free_i64(t0); + tcg_temp_free_i64(t1); + + desc = (vsz / 8) - 2; + desc = deposit32(desc, SIMD_DATA_SHIFT, 2, a->esz); + t3 = tcg_const_i32(desc); + + ptr = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(ptr, cpu_env, pred_full_reg_offset(s, a->rd)); + + gen_helper_sve_while(t2, ptr, t2, t3); + do_pred_flags(t2); + + tcg_temp_free_ptr(ptr); + tcg_temp_free_i32(t2); + tcg_temp_free_i32(t3); + return true; +} + /* *** 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 @@ SINCDECP_r_64 00100101 .. 1010 d:1 u:1 10001 10 .... ..... @incdec_pred # SVE saturating inc/dec vector by predicate count SINCDECP_z 00100101 .. 1010 d:1 u:1 10000 00 .... ..... @incdec2_pred +### SVE Integer Compare - Scalars Group + +# SVE conditionally terminate scalars +CTERM 00100101 1 sf:1 1 rm:5 001000 rn:5 ne:1 0000 + +# SVE integer compare scalar count and limit +WHILE 00100101 esz:2 1 rm:5 000 sf:1 u:1 1 rn:5 eq:1 rd:4 + ### 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: 20180613015641.5667-17-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-sve.c | 37 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 8 ++++++++ 2 files changed, 45 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_WHILE(DisasContext *s, arg_WHILE *a, uint32_t insn) return true; } +/* + *** SVE Integer Wide Immediate - Unpredicated Group + */ + +static bool trans_FDUP(DisasContext *s, arg_FDUP *a, uint32_t insn) +{ + if (a->esz == 0) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + int dofs = vec_full_reg_offset(s, a->rd); + uint64_t imm; + + /* Decode the VFP immediate. */ + imm = vfp_expand_imm(a->esz, a->imm); + imm = dup_const(a->esz, imm); + + tcg_gen_gvec_dup64i(dofs, vsz, vsz, imm); + } + return true; +} + +static bool trans_DUP_i(DisasContext *s, arg_DUP_i *a, uint32_t insn) +{ + if (a->esz == 0 && extract32(insn, 13, 1)) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + int dofs = vec_full_reg_offset(s, a->rd); + + tcg_gen_gvec_dup64i(dofs, vsz, vsz, dup_const(a->esz, a->imm)); + } + return true; +} + /* *** 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 @@ CTERM 00100101 1 sf:1 1 rm:5 001000 rn:5 ne:1 0000 # SVE integer compare scalar count and limit WHILE 00100101 esz:2 1 rm:5 000 sf:1 u:1 1 rn:5 eq:1 rd:4 +### SVE Integer Wide Immediate - Unpredicated Group + +# SVE broadcast floating-point immediate (unpredicated) +FDUP 00100101 esz:2 111 00 1110 imm:8 rd:5 + +# SVE broadcast integer immediate (unpredicated) +DUP_i 00100101 esz:2 111 00 011 . ........ rd:5 imm=%sh8_i8s + ### 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: 20180613015641.5667-18-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 25 +++++++ target/arm/sve_helper.c | 41 +++++++++++ target/arm/translate-sve.c | 144 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 26 +++++++ 4 files changed, 236 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_brkns, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_cntp, TCG_CALL_NO_RWG, i64, ptr, ptr, i32) DEF_HELPER_FLAGS_3(sve_while, TCG_CALL_NO_RWG, i32, ptr, i32, i32) + +DEF_HELPER_FLAGS_4(sve_subri_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_subri_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_subri_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_subri_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_4(sve_smaxi_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smaxi_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smaxi_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smaxi_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_4(sve_smini_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smini_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smini_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_smini_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_4(sve_umaxi_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umaxi_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umaxi_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umaxi_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_4(sve_umini_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umini_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umini_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(sve_umini_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, 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_VPZ_D(sve_uminv_d, uint64_t, uint64_t, -1, DO_MIN) #undef DO_VPZ #undef DO_VPZ_D +/* Two vector operand, one scalar operand, unpredicated. */ +#define DO_ZZI(NAME, TYPE, OP) \ +void HELPER(NAME)(void *vd, void *vn, uint64_t s64, uint32_t desc) \ +{ \ + intptr_t i, opr_sz = simd_oprsz(desc) / sizeof(TYPE); \ + TYPE s = s64, *d = vd, *n = vn; \ + for (i = 0; i < opr_sz; ++i) { \ + d[i] = OP(n[i], s); \ + } \ +} + +#define DO_SUBR(X, Y) (Y - X) + +DO_ZZI(sve_subri_b, uint8_t, DO_SUBR) +DO_ZZI(sve_subri_h, uint16_t, DO_SUBR) +DO_ZZI(sve_subri_s, uint32_t, DO_SUBR) +DO_ZZI(sve_subri_d, uint64_t, DO_SUBR) + +DO_ZZI(sve_smaxi_b, int8_t, DO_MAX) +DO_ZZI(sve_smaxi_h, int16_t, DO_MAX) +DO_ZZI(sve_smaxi_s, int32_t, DO_MAX) +DO_ZZI(sve_smaxi_d, int64_t, DO_MAX) + +DO_ZZI(sve_smini_b, int8_t, DO_MIN) +DO_ZZI(sve_smini_h, int16_t, DO_MIN) +DO_ZZI(sve_smini_s, int32_t, DO_MIN) +DO_ZZI(sve_smini_d, int64_t, DO_MIN) + +DO_ZZI(sve_umaxi_b, uint8_t, DO_MAX) +DO_ZZI(sve_umaxi_h, uint16_t, DO_MAX) +DO_ZZI(sve_umaxi_s, uint32_t, DO_MAX) +DO_ZZI(sve_umaxi_d, uint64_t, DO_MAX) + +DO_ZZI(sve_umini_b, uint8_t, DO_MIN) +DO_ZZI(sve_umini_h, uint16_t, DO_MIN) +DO_ZZI(sve_umini_s, uint32_t, DO_MIN) +DO_ZZI(sve_umini_d, uint64_t, DO_MIN) + +#undef DO_ZZI + #undef DO_AND #undef DO_ORR #undef DO_EOR @@ -XXX,XX +XXX,XX @@ DO_VPZ_D(sve_uminv_d, uint64_t, uint64_t, -1, DO_MIN) #undef DO_ASR #undef DO_LSR #undef DO_LSL +#undef DO_SUBR /* Similar to the ARM LastActiveElement pseudocode function, except the result is multiplied by the element size. This includes the not found 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 inline int expand_imm_sh8s(int x) return (int8_t)x << (x & 0x100 ? 8 : 0); } +static inline int expand_imm_sh8u(int x) +{ + return (uint8_t)x << (x & 0x100 ? 8 : 0); +} + /* * Include the generated decoder. */ @@ -XXX,XX +XXX,XX @@ static bool trans_DUP_i(DisasContext *s, arg_DUP_i *a, uint32_t insn) return true; } +static bool trans_ADD_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + if (a->esz == 0 && extract32(insn, 13, 1)) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_addi(a->esz, vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), a->imm, vsz, vsz); + } + return true; +} + +static bool trans_SUB_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + a->imm = -a->imm; + return trans_ADD_zzi(s, a, insn); +} + +static bool trans_SUBR_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + static const GVecGen2s op[4] = { + { .fni8 = tcg_gen_vec_sub8_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_sve_subri_b, + .opc = INDEX_op_sub_vec, + .vece = MO_8, + .scalar_first = true }, + { .fni8 = tcg_gen_vec_sub16_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_sve_subri_h, + .opc = INDEX_op_sub_vec, + .vece = MO_16, + .scalar_first = true }, + { .fni4 = tcg_gen_sub_i32, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_sve_subri_s, + .opc = INDEX_op_sub_vec, + .vece = MO_32, + .scalar_first = true }, + { .fni8 = tcg_gen_sub_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_sve_subri_d, + .opc = INDEX_op_sub_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64, + .scalar_first = true } + }; + + if (a->esz == 0 && extract32(insn, 13, 1)) { + return false; + } + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_i64 c = tcg_const_i64(a->imm); + tcg_gen_gvec_2s(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vsz, vsz, c, &op[a->esz]); + tcg_temp_free_i64(c); + } + return true; +} + +static bool trans_MUL_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_muli(a->esz, vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), a->imm, vsz, vsz); + } + return true; +} + +static bool do_zzi_sat(DisasContext *s, arg_rri_esz *a, uint32_t insn, + bool u, bool d) +{ + if (a->esz == 0 && extract32(insn, 13, 1)) { + return false; + } + if (sve_access_check(s)) { + TCGv_i64 val = tcg_const_i64(a->imm); + do_sat_addsub_vec(s, a->esz, a->rd, a->rn, val, u, d); + tcg_temp_free_i64(val); + } + return true; +} + +static bool trans_SQADD_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + return do_zzi_sat(s, a, insn, false, false); +} + +static bool trans_UQADD_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + return do_zzi_sat(s, a, insn, true, false); +} + +static bool trans_SQSUB_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + return do_zzi_sat(s, a, insn, false, true); +} + +static bool trans_UQSUB_zzi(DisasContext *s, arg_rri_esz *a, uint32_t insn) +{ + return do_zzi_sat(s, a, insn, true, true); +} + +static bool do_zzi_ool(DisasContext *s, arg_rri_esz *a, gen_helper_gvec_2i *fn) +{ + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + TCGv_i64 c = tcg_const_i64(a->imm); + + tcg_gen_gvec_2i_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + c, vsz, vsz, 0, fn); + tcg_temp_free_i64(c); + } + return true; +} + +#define DO_ZZI(NAME, name) \ +static bool trans_##NAME##_zzi(DisasContext *s, arg_rri_esz *a, \ + uint32_t insn) \ +{ \ + static gen_helper_gvec_2i * const fns[4] = { \ + gen_helper_sve_##name##i_b, gen_helper_sve_##name##i_h, \ + gen_helper_sve_##name##i_s, gen_helper_sve_##name##i_d, \ + }; \ + return do_zzi_ool(s, a, fns[a->esz]); \ +} + +DO_ZZI(SMAX, smax) +DO_ZZI(UMAX, umax) +DO_ZZI(SMIN, smin) +DO_ZZI(UMIN, umin) + +#undef DO_ZZI + /* *** 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 @@ # Signed 8-bit immediate, optionally shifted left by 8. %sh8_i8s 5:9 !function=expand_imm_sh8s +# Unsigned 8-bit immediate, optionally shifted left by 8. +%sh8_i8u 5:9 !function=expand_imm_sh8u # Either a copy of rd (at bit 0), or a different source # as propagated via the MOVPRFX instruction. @@ -XXX,XX +XXX,XX @@ @pd_pn_pm ........ esz:2 .. rm:4 ....... rn:4 . rd:4 &rrr_esz @rdn_rm ........ esz:2 ...... ...... rm:5 rd:5 \ &rrr_esz rn=%reg_movprfx +@rdn_sh_i8u ........ esz:2 ...... ...... ..... rd:5 \ + &rri_esz rn=%reg_movprfx imm=%sh8_i8u +@rdn_i8u ........ esz:2 ...... ... imm:8 rd:5 \ + &rri_esz rn=%reg_movprfx +@rdn_i8s ........ esz:2 ...... ... imm:s8 rd:5 \ + &rri_esz rn=%reg_movprfx # Three operand with "memory" size, aka immediate left shift @rd_rn_msz_rm ........ ... rm:5 .... imm:2 rn:5 rd:5 &rrri @@ -XXX,XX +XXX,XX @@ FDUP 00100101 esz:2 111 00 1110 imm:8 rd:5 # SVE broadcast integer immediate (unpredicated) DUP_i 00100101 esz:2 111 00 011 . ........ rd:5 imm=%sh8_i8s +# SVE integer add/subtract immediate (unpredicated) +ADD_zzi 00100101 .. 100 000 11 . ........ ..... @rdn_sh_i8u +SUB_zzi 00100101 .. 100 001 11 . ........ ..... @rdn_sh_i8u +SUBR_zzi 00100101 .. 100 011 11 . ........ ..... @rdn_sh_i8u +SQADD_zzi 00100101 .. 100 100 11 . ........ ..... @rdn_sh_i8u +UQADD_zzi 00100101 .. 100 101 11 . ........ ..... @rdn_sh_i8u +SQSUB_zzi 00100101 .. 100 110 11 . ........ ..... @rdn_sh_i8u +UQSUB_zzi 00100101 .. 100 111 11 . ........ ..... @rdn_sh_i8u + +# SVE integer min/max immediate (unpredicated) +SMAX_zzi 00100101 .. 101 000 110 ........ ..... @rdn_i8s +UMAX_zzi 00100101 .. 101 001 110 ........ ..... @rdn_i8u +SMIN_zzi 00100101 .. 101 010 110 ........ ..... @rdn_i8s +UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u + +# SVE integer multiply immediate (unpredicated) +MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s + ### 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: 20180613015641.5667-19-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 14 ++++++++ target/arm/helper.h | 19 +++++++++++ target/arm/translate-sve.c | 42 +++++++++++++++++++++++ target/arm/vec_helper.c | 69 ++++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 10 ++++++ 5 files changed, 154 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_umini_b, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) DEF_HELPER_FLAGS_4(sve_umini_h, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) DEF_HELPER_FLAGS_4(sve_umini_s, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) DEF_HELPER_FLAGS_4(sve_umini_d, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + +DEF_HELPER_FLAGS_5(gvec_recps_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_recps_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_recps_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(gvec_rsqrts_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +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) 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_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) + +DEF_HELPER_FLAGS_5(gvec_fsub_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fsub_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fsub_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(gvec_fmul_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fmul_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fmul_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(gvec_ftsmul_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_ftsmul_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_ftsmul_d, TCG_CALL_NO_RWG, + void, 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 Arithmetic - Unpredicated Group + */ + +static bool do_zzz_fp(DisasContext *s, arg_rrr_esz *a, + gen_helper_gvec_3_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_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, 0, fn); + tcg_temp_free_ptr(status); + } + return true; +} + + +#define DO_FP3(NAME, name) \ +static bool trans_##NAME(DisasContext *s, arg_rrr_esz *a, uint32_t insn) \ +{ \ + static gen_helper_gvec_3_ptr * const fns[4] = { \ + NULL, gen_helper_gvec_##name##_h, \ + gen_helper_gvec_##name##_s, gen_helper_gvec_##name##_d \ + }; \ + return do_zzz_fp(s, a, fns[a->esz]); \ +} + +DO_FP3(FADD_zzz, fadd) +DO_FP3(FSUB_zzz, fsub) +DO_FP3(FMUL_zzz, fmul) +DO_FP3(FTSMUL, ftsmul) +DO_FP3(FRECPS, recps) +DO_FP3(FRSQRTS, rsqrts) + +#undef DO_FP3 + /* *** SVE Memory - 32-bit Gather and Unsized Contiguous 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)); } + +/* Floating-point trigonometric starting value. + * See the ARM ARM pseudocode function FPTrigSMul. + */ +static float16 float16_ftsmul(float16 op1, uint16_t op2, float_status *stat) +{ + float16 result = float16_mul(op1, op1, stat); + if (!float16_is_any_nan(result)) { + result = float16_set_sign(result, op2 & 1); + } + return result; +} + +static float32 float32_ftsmul(float32 op1, uint32_t op2, float_status *stat) +{ + float32 result = float32_mul(op1, op1, stat); + if (!float32_is_any_nan(result)) { + result = float32_set_sign(result, op2 & 1); + } + return result; +} + +static float64 float64_ftsmul(float64 op1, uint64_t op2, float_status *stat) +{ + float64 result = float64_mul(op1, op1, stat); + if (!float64_is_any_nan(result)) { + result = float64_set_sign(result, op2 & 1); + } + return result; +} + +#define DO_3OP(NAME, FUNC, TYPE) \ +void HELPER(NAME)(void *vd, void *vn, void *vm, void *stat, uint32_t desc) \ +{ \ + intptr_t i, oprsz = simd_oprsz(desc); \ + TYPE *d = vd, *n = vn, *m = vm; \ + for (i = 0; i < oprsz / sizeof(TYPE); i++) { \ + d[i] = FUNC(n[i], m[i], stat); \ + } \ +} + +DO_3OP(gvec_fadd_h, float16_add, float16) +DO_3OP(gvec_fadd_s, float32_add, float32) +DO_3OP(gvec_fadd_d, float64_add, float64) + +DO_3OP(gvec_fsub_h, float16_sub, float16) +DO_3OP(gvec_fsub_s, float32_sub, float32) +DO_3OP(gvec_fsub_d, float64_sub, float64) + +DO_3OP(gvec_fmul_h, float16_mul, float16) +DO_3OP(gvec_fmul_s, float32_mul, float32) +DO_3OP(gvec_fmul_d, float64_mul, float64) + +DO_3OP(gvec_ftsmul_h, float16_ftsmul, float16) +DO_3OP(gvec_ftsmul_s, float32_ftsmul, float32) +DO_3OP(gvec_ftsmul_d, float64_ftsmul, float64) + +#ifdef TARGET_AARCH64 + +DO_3OP(gvec_recps_h, helper_recpsf_f16, float16) +DO_3OP(gvec_recps_s, helper_recpsf_f32, float32) +DO_3OP(gvec_recps_d, helper_recpsf_f64, float64) + +DO_3OP(gvec_rsqrts_h, helper_rsqrtsf_f16, float16) +DO_3OP(gvec_rsqrts_s, helper_rsqrtsf_f32, float32) +DO_3OP(gvec_rsqrts_d, helper_rsqrtsf_f64, float64) + +#endif +#undef DO_3OP 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 Arithmetic - Unpredicated Group + +# SVE floating-point arithmetic (unpredicated) +FADD_zzz 01100101 .. 0 ..... 000 000 ..... ..... @rd_rn_rm +FSUB_zzz 01100101 .. 0 ..... 000 001 ..... ..... @rd_rn_rm +FMUL_zzz 01100101 .. 0 ..... 000 010 ..... ..... @rd_rn_rm +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 Memory - 32-bit Gather and Unsized Contiguous Group # SVE load predicate register -- 2.17.1 From: Joel Stanley <joel@jms.id.au> The ASPEED SoCs contain a single register that returns random data when read. This models that register so that guests can use it. The random number data register has a corresponding control register, however it returns data regardless of the state of the enabled bit, so the model follows this behaviour. When the qcrypto call fails we exit as the guest uses the random number device to feed it's entropy pool, which is used for cryptographic purposes. Reviewed-by: Cédric Le Goater <clg@kaod.org> Signed-off-by: Joel Stanley <joel@jms.id.au> Message-id: 20180613114836.9265-1-joel@jms.id.au Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/misc/aspeed_scu.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/hw/misc/aspeed_scu.c b/hw/misc/aspeed_scu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/misc/aspeed_scu.c +++ b/hw/misc/aspeed_scu.c @@ -XXX,XX +XXX,XX @@ #include "qapi/visitor.h" #include "qemu/bitops.h" #include "qemu/log.h" +#include "crypto/random.h" #include "trace.h" #define TO_REG(offset) ((offset) >> 2) @@ -XXX,XX +XXX,XX @@ static const uint32_t ast2500_a1_resets[ASPEED_SCU_NR_REGS] = { [BMC_DEV_ID] = 0x00002402U }; +static uint32_t aspeed_scu_get_random(void) +{ + Error *err = NULL; + uint32_t num; + + if (qcrypto_random_bytes((uint8_t *)&num, sizeof(num), &err)) { + error_report_err(err); + exit(1); + } + + return num; +} + static uint64_t aspeed_scu_read(void *opaque, hwaddr offset, unsigned size) { AspeedSCUState *s = ASPEED_SCU(opaque); @@ -XXX,XX +XXX,XX @@ static uint64_t aspeed_scu_read(void *opaque, hwaddr offset, unsigned size) } switch (reg) { + case RNG_DATA: + /* On hardware, RNG_DATA works regardless of + * the state of the enable bit in RNG_CTRL + */ + s->regs[RNG_DATA] = aspeed_scu_get_random(); + break; case WAKEUP_EN: qemu_log_mask(LOG_GUEST_ERROR, "%s: Read of write-only offset 0x%" HWADDR_PRIx "\n", -- 2.17.1 From: Cédric Le Goater <clg@kaod.org> On Macronix chips, two bytes can written to the WRSR. First byte will configure the status register and the second the configuration register. It is important to save the configuration value as it contains the dummy cycle setting when using dual or quad IO mode. Signed-off-by: Cédric Le Goater <clg@kaod.org> Acked-by: Alistair Francis <alistair.francis@wdc.com> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/block/m25p80.c | 1 + 1 file changed, 1 insertion(+) diff --git a/hw/block/m25p80.c b/hw/block/m25p80.c index XXXXXXX..XXXXXXX 100644 --- a/hw/block/m25p80.c +++ b/hw/block/m25p80.c @@ -XXX,XX +XXX,XX @@ static void complete_collecting_data(Flash *s) case MAN_MACRONIX: s->quad_enable = extract32(s->data[0], 6, 1); if (s->len > 1) { + s->volatile_cfg = s->data[1]; s->four_bytes_address_mode = extract32(s->data[1], 5, 1); } break; -- 2.17.1 If an IOMMU supports mappings that care about the memory transaction attributes, then it no longer has a unique address -> output mapping, but more than one. We can represent these using an IOMMU index, analogous to TCG's mmu indexes. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180604152941.20374-2-peter.maydell@linaro.org --- include/exec/memory.h | 55 +++++++++++++++++++++++++++++++++++++++++++ memory.c | 23 ++++++++++++++++++ 2 files changed, 78 insertions(+) diff --git a/include/exec/memory.h b/include/exec/memory.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/memory.h +++ b/include/exec/memory.h @@ -XXX,XX +XXX,XX @@ enum IOMMUMemoryRegionAttr { * to report whenever mappings are changed, by calling * memory_region_notify_iommu() (or, if necessary, by calling * memory_region_notify_one() for each registered notifier). + * + * Conceptually an IOMMU provides a mapping from input address + * to an output TLB entry. If the IOMMU is aware of memory transaction + * attributes and the output TLB entry depends on the transaction + * attributes, we represent this using IOMMU indexes. Each index + * selects a particular translation table that the IOMMU has: + * @attrs_to_index returns the IOMMU index for a set of transaction attributes + * @translate takes an input address and an IOMMU index + * and the mapping returned can only depend on the input address and the + * IOMMU index. + * + * Most IOMMUs don't care about the transaction attributes and support + * only a single IOMMU index. A more complex IOMMU might have one index + * for secure transactions and one for non-secure transactions. */ typedef struct IOMMUMemoryRegionClass { /* private */ @@ -XXX,XX +XXX,XX @@ typedef struct IOMMUMemoryRegionClass { */ int (*get_attr)(IOMMUMemoryRegion *iommu, enum IOMMUMemoryRegionAttr attr, void *data); + + /* Return the IOMMU index to use for a given set of transaction attributes. + * + * Optional method: if an IOMMU only supports a single IOMMU index then + * the default implementation of memory_region_iommu_attrs_to_index() + * will return 0. + * + * The indexes supported by an IOMMU must be contiguous, starting at 0. + * + * @iommu: the IOMMUMemoryRegion + * @attrs: memory transaction attributes + */ + int (*attrs_to_index)(IOMMUMemoryRegion *iommu, MemTxAttrs attrs); + + /* Return the number of IOMMU indexes this IOMMU supports. + * + * Optional method: if this method is not provided, then + * memory_region_iommu_num_indexes() will return 1, indicating that + * only a single IOMMU index is supported. + * + * @iommu: the IOMMUMemoryRegion + */ + int (*num_indexes)(IOMMUMemoryRegion *iommu); } IOMMUMemoryRegionClass; typedef struct CoalescedMemoryRange CoalescedMemoryRange; @@ -XXX,XX +XXX,XX @@ int memory_region_iommu_get_attr(IOMMUMemoryRegion *iommu_mr, enum IOMMUMemoryRegionAttr attr, void *data); +/** + * memory_region_iommu_attrs_to_index: return the IOMMU index to + * use for translations with the given memory transaction attributes. + * + * @iommu_mr: the memory region + * @attrs: the memory transaction attributes + */ +int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion *iommu_mr, + MemTxAttrs attrs); + +/** + * memory_region_iommu_num_indexes: return the total number of IOMMU + * indexes that this IOMMU supports. + * + * @iommu_mr: the memory region + */ +int memory_region_iommu_num_indexes(IOMMUMemoryRegion *iommu_mr); + /** * memory_region_name: get a memory region's name * diff --git a/memory.c b/memory.c index XXXXXXX..XXXXXXX 100644 --- a/memory.c +++ b/memory.c @@ -XXX,XX +XXX,XX @@ int memory_region_iommu_get_attr(IOMMUMemoryRegion *iommu_mr, return imrc->get_attr(iommu_mr, attr, data); } +int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion *iommu_mr, + MemTxAttrs attrs) +{ + IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_GET_CLASS(iommu_mr); + + if (!imrc->attrs_to_index) { + return 0; + } + + return imrc->attrs_to_index(iommu_mr, attrs); +} + +int memory_region_iommu_num_indexes(IOMMUMemoryRegion *iommu_mr) +{ + IOMMUMemoryRegionClass *imrc = IOMMU_MEMORY_REGION_GET_CLASS(iommu_mr); + + if (!imrc->num_indexes) { + return 1; + } + + return imrc->num_indexes(iommu_mr); +} + void memory_region_set_log(MemoryRegion *mr, bool log, unsigned client) { uint8_t mask = 1 << client; -- 2.17.1 Add support for multiple IOMMU indexes to the IOMMU notifier APIs. When initializing a notifier with iommu_notifier_init(), the caller must pass the IOMMU index that it is interested in. When a change happens, the IOMMU implementation must pass memory_region_notify_iommu() the IOMMU index that has changed and that notifiers must be called for. IOMMUs which support only a single index don't need to change. Callers which only really support working with IOMMUs with a single index can use the result of passing MEMTXATTRS_UNSPECIFIED to memory_region_iommu_attrs_to_index(). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180604152941.20374-3-peter.maydell@linaro.org --- include/exec/memory.h | 7 ++++++- hw/i386/intel_iommu.c | 6 +++--- hw/ppc/spapr_iommu.c | 2 +- hw/s390x/s390-pci-inst.c | 4 ++-- hw/vfio/common.c | 6 +++++- hw/virtio/vhost.c | 7 ++++++- memory.c | 8 +++++++- 7 files changed, 30 insertions(+), 10 deletions(-) diff --git a/include/exec/memory.h b/include/exec/memory.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/memory.h +++ b/include/exec/memory.h @@ -XXX,XX +XXX,XX @@ struct IOMMUNotifier { /* Notify for address space range start <= addr <= end */ hwaddr start; hwaddr end; + int iommu_idx; QLIST_ENTRY(IOMMUNotifier) node; }; typedef struct IOMMUNotifier IOMMUNotifier; static inline void iommu_notifier_init(IOMMUNotifier *n, IOMMUNotify fn, IOMMUNotifierFlag flags, - hwaddr start, hwaddr end) + hwaddr start, hwaddr end, + int iommu_idx) { n->notify = fn; n->notifier_flags = flags; n->start = start; n->end = end; + n->iommu_idx = iommu_idx; } /* @@ -XXX,XX +XXX,XX @@ uint64_t memory_region_iommu_get_min_page_size(IOMMUMemoryRegion *iommu_mr); * should be notified with an UNMAP followed by a MAP. * * @iommu_mr: the memory region that was changed + * @iommu_idx: the IOMMU index for the translation table which has changed * @entry: the new entry in the IOMMU translation table. The entry * replaces all old entries for the same virtual I/O address range. * Deleted entries have .@perm == 0. */ void memory_region_notify_iommu(IOMMUMemoryRegion *iommu_mr, + int iommu_idx, IOMMUTLBEntry entry); /** diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/i386/intel_iommu.c +++ b/hw/i386/intel_iommu.c @@ -XXX,XX +XXX,XX @@ static int vtd_dev_to_context_entry(IntelIOMMUState *s, uint8_t bus_num, static int vtd_sync_shadow_page_hook(IOMMUTLBEntry *entry, void *private) { - memory_region_notify_iommu((IOMMUMemoryRegion *)private, *entry); + memory_region_notify_iommu((IOMMUMemoryRegion *)private, 0, *entry); return 0; } @@ -XXX,XX +XXX,XX @@ static void vtd_iotlb_page_invalidate_notify(IntelIOMMUState *s, .addr_mask = size - 1, .perm = IOMMU_NONE, }; - memory_region_notify_iommu(&vtd_as->iommu, entry); + memory_region_notify_iommu(&vtd_as->iommu, 0, entry); } } } @@ -XXX,XX +XXX,XX @@ static bool vtd_process_device_iotlb_desc(IntelIOMMUState *s, entry.iova = addr; entry.perm = IOMMU_NONE; entry.translated_addr = 0; - memory_region_notify_iommu(&vtd_dev_as->iommu, entry); + memory_region_notify_iommu(&vtd_dev_as->iommu, 0, entry); done: return true; diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/ppc/spapr_iommu.c +++ b/hw/ppc/spapr_iommu.c @@ -XXX,XX +XXX,XX @@ static target_ulong put_tce_emu(sPAPRTCETable *tcet, target_ulong ioba, entry.translated_addr = tce & page_mask; entry.addr_mask = ~page_mask; entry.perm = spapr_tce_iommu_access_flags(tce); - memory_region_notify_iommu(&tcet->iommu, entry); + memory_region_notify_iommu(&tcet->iommu, 0, entry); return H_SUCCESS; } diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c index XXXXXXX..XXXXXXX 100644 --- a/hw/s390x/s390-pci-inst.c +++ b/hw/s390x/s390-pci-inst.c @@ -XXX,XX +XXX,XX @@ static void s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry) } notify.perm = IOMMU_NONE; - memory_region_notify_iommu(&iommu->iommu_mr, notify); + memory_region_notify_iommu(&iommu->iommu_mr, 0, notify); notify.perm = entry->perm; } @@ -XXX,XX +XXX,XX @@ static void s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry) g_hash_table_replace(iommu->iotlb, &cache->iova, cache); } - memory_region_notify_iommu(&iommu->iommu_mr, notify); + memory_region_notify_iommu(&iommu->iommu_mr, 0, notify); } int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra) diff --git a/hw/vfio/common.c b/hw/vfio/common.c index XXXXXXX..XXXXXXX 100644 --- a/hw/vfio/common.c +++ b/hw/vfio/common.c @@ -XXX,XX +XXX,XX @@ static void vfio_listener_region_add(MemoryListener *listener, if (memory_region_is_iommu(section->mr)) { VFIOGuestIOMMU *giommu; IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr); + int iommu_idx; trace_vfio_listener_region_add_iommu(iova, end); /* @@ -XXX,XX +XXX,XX @@ static void vfio_listener_region_add(MemoryListener *listener, llend = int128_add(int128_make64(section->offset_within_region), section->size); llend = int128_sub(llend, int128_one()); + iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr, + MEMTXATTRS_UNSPECIFIED); iommu_notifier_init(&giommu->n, vfio_iommu_map_notify, IOMMU_NOTIFIER_ALL, section->offset_within_region, - int128_get64(llend)); + int128_get64(llend), + iommu_idx); QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next); memory_region_register_iommu_notifier(section->mr, &giommu->n); diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c index XXXXXXX..XXXXXXX 100644 --- a/hw/virtio/vhost.c +++ b/hw/virtio/vhost.c @@ -XXX,XX +XXX,XX @@ static void vhost_iommu_region_add(MemoryListener *listener, iommu_listener); struct vhost_iommu *iommu; Int128 end; + int iommu_idx; + IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr); if (!memory_region_is_iommu(section->mr)) { return; @@ -XXX,XX +XXX,XX @@ static void vhost_iommu_region_add(MemoryListener *listener, end = int128_add(int128_make64(section->offset_within_region), section->size); end = int128_sub(end, int128_one()); + iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr, + MEMTXATTRS_UNSPECIFIED); iommu_notifier_init(&iommu->n, vhost_iommu_unmap_notify, IOMMU_NOTIFIER_UNMAP, section->offset_within_region, - int128_get64(end)); + int128_get64(end), + iommu_idx); iommu->mr = section->mr; iommu->iommu_offset = section->offset_within_address_space - section->offset_within_region; diff --git a/memory.c b/memory.c index XXXXXXX..XXXXXXX 100644 --- a/memory.c +++ b/memory.c @@ -XXX,XX +XXX,XX @@ void memory_region_register_iommu_notifier(MemoryRegion *mr, iommu_mr = IOMMU_MEMORY_REGION(mr); assert(n->notifier_flags != IOMMU_NOTIFIER_NONE); assert(n->start <= n->end); + assert(n->iommu_idx >= 0 && + n->iommu_idx < memory_region_iommu_num_indexes(iommu_mr)); + QLIST_INSERT_HEAD(&iommu_mr->iommu_notify, n, node); memory_region_update_iommu_notify_flags(iommu_mr); } @@ -XXX,XX +XXX,XX @@ void memory_region_notify_one(IOMMUNotifier *notifier, } void memory_region_notify_iommu(IOMMUMemoryRegion *iommu_mr, + int iommu_idx, IOMMUTLBEntry entry) { IOMMUNotifier *iommu_notifier; @@ -XXX,XX +XXX,XX @@ void memory_region_notify_iommu(IOMMUMemoryRegion *iommu_mr, assert(memory_region_is_iommu(MEMORY_REGION(iommu_mr))); IOMMU_NOTIFIER_FOREACH(iommu_notifier, iommu_mr) { - memory_region_notify_one(iommu_notifier, &entry); + if (iommu_notifier->iommu_idx == iommu_idx) { + memory_region_notify_one(iommu_notifier, &entry); + } } } -- 2.17.1 Add an IOMMU index argument to the translate method of IOMMUs. Since all of our current IOMMU implementations support only a single IOMMU index, this has no effect on the behaviour. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180604152941.20374-4-peter.maydell@linaro.org --- include/exec/memory.h | 3 ++- exec.c | 11 +++++++++-- hw/alpha/typhoon.c | 3 ++- hw/arm/smmuv3.c | 2 +- hw/dma/rc4030.c | 2 +- hw/i386/amd_iommu.c | 2 +- hw/i386/intel_iommu.c | 2 +- hw/ppc/spapr_iommu.c | 3 ++- hw/s390x/s390-pci-bus.c | 2 +- hw/sparc/sun4m_iommu.c | 3 ++- hw/sparc64/sun4u_iommu.c | 2 +- memory.c | 2 +- 12 files changed, 24 insertions(+), 13 deletions(-) diff --git a/include/exec/memory.h b/include/exec/memory.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/memory.h +++ b/include/exec/memory.h @@ -XXX,XX +XXX,XX @@ typedef struct IOMMUMemoryRegionClass { * @iommu: the IOMMUMemoryRegion * @hwaddr: address to be translated within the memory region * @flag: requested access permissions + * @iommu_idx: IOMMU index for the translation */ IOMMUTLBEntry (*translate)(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag); + IOMMUAccessFlags flag, int iommu_idx); /* Returns minimum supported page size in bytes. * If this method is not provided then the minimum is assumed to * be TARGET_PAGE_SIZE. diff --git a/exec.c b/exec.c index XXXXXXX..XXXXXXX 100644 --- a/exec.c +++ b/exec.c @@ -XXX,XX +XXX,XX @@ static MemoryRegionSection address_space_translate_iommu(IOMMUMemoryRegion *iomm do { hwaddr addr = *xlat; IOMMUMemoryRegionClass *imrc = memory_region_get_iommu_class_nocheck(iommu_mr); - IOMMUTLBEntry iotlb = imrc->translate(iommu_mr, addr, is_write ? - IOMMU_WO : IOMMU_RO); + int iommu_idx = 0; + IOMMUTLBEntry iotlb; + + if (imrc->attrs_to_index) { + iommu_idx = imrc->attrs_to_index(iommu_mr, attrs); + } + + iotlb = imrc->translate(iommu_mr, addr, is_write ? + IOMMU_WO : IOMMU_RO, iommu_idx); if (!(iotlb.perm & (1 << is_write))) { goto unassigned; diff --git a/hw/alpha/typhoon.c b/hw/alpha/typhoon.c index XXXXXXX..XXXXXXX 100644 --- a/hw/alpha/typhoon.c +++ b/hw/alpha/typhoon.c @@ -XXX,XX +XXX,XX @@ static bool window_translate(TyphoonWindow *win, hwaddr addr, Pchip and generate a machine check interrupt. */ static IOMMUTLBEntry typhoon_translate_iommu(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, + int iommu_idx) { TyphoonPchip *pchip = container_of(iommu, TyphoonPchip, iommu); IOMMUTLBEntry ret; diff --git a/hw/arm/smmuv3.c b/hw/arm/smmuv3.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/smmuv3.c +++ b/hw/arm/smmuv3.c @@ -XXX,XX +XXX,XX @@ static int smmuv3_decode_config(IOMMUMemoryRegion *mr, SMMUTransCfg *cfg, } static IOMMUTLBEntry smmuv3_translate(IOMMUMemoryRegion *mr, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { SMMUDevice *sdev = container_of(mr, SMMUDevice, iommu); SMMUv3State *s = sdev->smmu; diff --git a/hw/dma/rc4030.c b/hw/dma/rc4030.c index XXXXXXX..XXXXXXX 100644 --- a/hw/dma/rc4030.c +++ b/hw/dma/rc4030.c @@ -XXX,XX +XXX,XX @@ static const MemoryRegionOps jazzio_ops = { }; static IOMMUTLBEntry rc4030_dma_translate(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { rc4030State *s = container_of(iommu, rc4030State, dma_mr); IOMMUTLBEntry ret = { diff --git a/hw/i386/amd_iommu.c b/hw/i386/amd_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/i386/amd_iommu.c +++ b/hw/i386/amd_iommu.c @@ -XXX,XX +XXX,XX @@ static inline bool amdvi_is_interrupt_addr(hwaddr addr) } static IOMMUTLBEntry amdvi_translate(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { AMDVIAddressSpace *as = container_of(iommu, AMDVIAddressSpace, iommu); AMDVIState *s = as->iommu_state; diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/i386/intel_iommu.c +++ b/hw/i386/intel_iommu.c @@ -XXX,XX +XXX,XX @@ static void vtd_mem_write(void *opaque, hwaddr addr, } static IOMMUTLBEntry vtd_iommu_translate(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu); IntelIOMMUState *s = vtd_as->iommu_state; diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/ppc/spapr_iommu.c +++ b/hw/ppc/spapr_iommu.c @@ -XXX,XX +XXX,XX @@ static void spapr_tce_free_table(uint64_t *table, int fd, uint32_t nb_table) /* Called from RCU critical section */ static IOMMUTLBEntry spapr_tce_translate_iommu(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, + int iommu_idx) { sPAPRTCETable *tcet = container_of(iommu, sPAPRTCETable, iommu); uint64_t tce; diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c index XXXXXXX..XXXXXXX 100644 --- a/hw/s390x/s390-pci-bus.c +++ b/hw/s390x/s390-pci-bus.c @@ -XXX,XX +XXX,XX @@ uint16_t s390_guest_io_table_walk(uint64_t g_iota, hwaddr addr, } static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { S390PCIIOMMU *iommu = container_of(mr, S390PCIIOMMU, iommu_mr); S390IOTLBEntry *entry; diff --git a/hw/sparc/sun4m_iommu.c b/hw/sparc/sun4m_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sparc/sun4m_iommu.c +++ b/hw/sparc/sun4m_iommu.c @@ -XXX,XX +XXX,XX @@ static void iommu_bad_addr(IOMMUState *s, hwaddr addr, /* Called from RCU critical section */ static IOMMUTLBEntry sun4m_translate_iommu(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flags) + IOMMUAccessFlags flags, + int iommu_idx) { IOMMUState *is = container_of(iommu, IOMMUState, iommu); hwaddr page, pa; diff --git a/hw/sparc64/sun4u_iommu.c b/hw/sparc64/sun4u_iommu.c index XXXXXXX..XXXXXXX 100644 --- a/hw/sparc64/sun4u_iommu.c +++ b/hw/sparc64/sun4u_iommu.c @@ -XXX,XX +XXX,XX @@ /* Called from RCU critical section */ static IOMMUTLBEntry sun4u_translate_iommu(IOMMUMemoryRegion *iommu, hwaddr addr, - IOMMUAccessFlags flag) + IOMMUAccessFlags flag, int iommu_idx) { IOMMUState *is = container_of(iommu, IOMMUState, iommu); hwaddr baseaddr, offset; diff --git a/memory.c b/memory.c index XXXXXXX..XXXXXXX 100644 --- a/memory.c +++ b/memory.c @@ -XXX,XX +XXX,XX @@ void memory_region_iommu_replay(IOMMUMemoryRegion *iommu_mr, IOMMUNotifier *n) granularity = memory_region_iommu_get_min_page_size(iommu_mr); for (addr = 0; addr < memory_region_size(mr); addr += granularity) { - iotlb = imrc->translate(iommu_mr, addr, IOMMU_NONE); + iotlb = imrc->translate(iommu_mr, addr, IOMMU_NONE, n->iommu_idx); if (iotlb.perm != IOMMU_NONE) { n->notify(n, &iotlb); } -- 2.17.1 Currently we don't support board configurations that put an IOMMU in the path of the CPU's memory transactions, and instead just assert() if the memory region fonud in address_space_translate_for_iotlb() is an IOMMUMemoryRegion. Remove this limitation by having the function handle IOMMUs. This is mostly straightforward, but we must make sure we have a notifier registered for every IOMMU that a transaction has passed through, so that we can flush the TLB appropriately when any of the IOMMUs change their mappings. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20180604152941.20374-5-peter.maydell@linaro.org --- include/exec/exec-all.h | 3 +- include/qom/cpu.h | 3 + accel/tcg/cputlb.c | 3 +- exec.c | 135 +++++++++++++++++++++++++++++++++++++++- 4 files changed, 140 insertions(+), 4 deletions(-) diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/exec-all.h +++ b/include/exec/exec-all.h @@ -XXX,XX +XXX,XX @@ void tb_flush_jmp_cache(CPUState *cpu, target_ulong addr); MemoryRegionSection * address_space_translate_for_iotlb(CPUState *cpu, int asidx, hwaddr addr, - hwaddr *xlat, hwaddr *plen); + hwaddr *xlat, hwaddr *plen, + MemTxAttrs attrs, int *prot); hwaddr memory_region_section_get_iotlb(CPUState *cpu, MemoryRegionSection *section, target_ulong vaddr, diff --git a/include/qom/cpu.h b/include/qom/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/include/qom/cpu.h +++ b/include/qom/cpu.h @@ -XXX,XX +XXX,XX @@ struct CPUState { uint16_t pending_tlb_flush; int hvf_fd; + + /* track IOMMUs whose translations we've cached in the TCG TLB */ + GArray *iommu_notifiers; }; QTAILQ_HEAD(CPUTailQ, CPUState); diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cputlb.c +++ b/accel/tcg/cputlb.c @@ -XXX,XX +XXX,XX @@ void tlb_set_page_with_attrs(CPUState *cpu, target_ulong vaddr, } sz = size; - section = address_space_translate_for_iotlb(cpu, asidx, paddr, &xlat, &sz); + section = address_space_translate_for_iotlb(cpu, asidx, paddr, &xlat, &sz, + attrs, &prot); assert(sz >= TARGET_PAGE_SIZE); tlb_debug("vaddr=" TARGET_FMT_lx " paddr=0x" TARGET_FMT_plx diff --git a/exec.c b/exec.c index XXXXXXX..XXXXXXX 100644 --- a/exec.c +++ b/exec.c @@ -XXX,XX +XXX,XX @@ MemoryRegion *flatview_translate(FlatView *fv, hwaddr addr, hwaddr *xlat, return mr; } +typedef struct TCGIOMMUNotifier { + IOMMUNotifier n; + MemoryRegion *mr; + CPUState *cpu; + int iommu_idx; + bool active; +} TCGIOMMUNotifier; + +static void tcg_iommu_unmap_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb) +{ + TCGIOMMUNotifier *notifier = container_of(n, TCGIOMMUNotifier, n); + + if (!notifier->active) { + return; + } + tlb_flush(notifier->cpu); + notifier->active = false; + /* We leave the notifier struct on the list to avoid reallocating it later. + * Generally the number of IOMMUs a CPU deals with will be small. + * In any case we can't unregister the iommu notifier from a notify + * callback. + */ +} + +static void tcg_register_iommu_notifier(CPUState *cpu, + IOMMUMemoryRegion *iommu_mr, + int iommu_idx) +{ + /* Make sure this CPU has an IOMMU notifier registered for this + * IOMMU/IOMMU index combination, so that we can flush its TLB + * when the IOMMU tells us the mappings we've cached have changed. + */ + MemoryRegion *mr = MEMORY_REGION(iommu_mr); + TCGIOMMUNotifier *notifier; + int i; + + for (i = 0; i < cpu->iommu_notifiers->len; i++) { + notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i); + if (notifier->mr == mr && notifier->iommu_idx == iommu_idx) { + break; + } + } + if (i == cpu->iommu_notifiers->len) { + /* Not found, add a new entry at the end of the array */ + cpu->iommu_notifiers = g_array_set_size(cpu->iommu_notifiers, i + 1); + notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i); + + notifier->mr = mr; + notifier->iommu_idx = iommu_idx; + notifier->cpu = cpu; + /* Rather than trying to register interest in the specific part + * of the iommu's address space that we've accessed and then + * expand it later as subsequent accesses touch more of it, we + * just register interest in the whole thing, on the assumption + * that iommu reconfiguration will be rare. + */ + iommu_notifier_init(&notifier->n, + tcg_iommu_unmap_notify, + IOMMU_NOTIFIER_UNMAP, + 0, + HWADDR_MAX, + iommu_idx); + memory_region_register_iommu_notifier(notifier->mr, &notifier->n); + } + + if (!notifier->active) { + notifier->active = true; + } +} + +static void tcg_iommu_free_notifier_list(CPUState *cpu) +{ + /* Destroy the CPU's notifier list */ + int i; + TCGIOMMUNotifier *notifier; + + for (i = 0; i < cpu->iommu_notifiers->len; i++) { + notifier = &g_array_index(cpu->iommu_notifiers, TCGIOMMUNotifier, i); + memory_region_unregister_iommu_notifier(notifier->mr, &notifier->n); + } + g_array_free(cpu->iommu_notifiers, true); +} + /* Called from RCU critical section */ MemoryRegionSection * address_space_translate_for_iotlb(CPUState *cpu, int asidx, hwaddr addr, - hwaddr *xlat, hwaddr *plen) + hwaddr *xlat, hwaddr *plen, + MemTxAttrs attrs, int *prot) { MemoryRegionSection *section; + IOMMUMemoryRegion *iommu_mr; + IOMMUMemoryRegionClass *imrc; + IOMMUTLBEntry iotlb; + int iommu_idx; AddressSpaceDispatch *d = atomic_rcu_read(&cpu->cpu_ases[asidx].memory_dispatch); - section = address_space_translate_internal(d, addr, xlat, plen, false); + for (;;) { + section = address_space_translate_internal(d, addr, &addr, plen, false); + + iommu_mr = memory_region_get_iommu(section->mr); + if (!iommu_mr) { + break; + } + + imrc = memory_region_get_iommu_class_nocheck(iommu_mr); + + iommu_idx = imrc->attrs_to_index(iommu_mr, attrs); + tcg_register_iommu_notifier(cpu, iommu_mr, iommu_idx); + /* We need all the permissions, so pass IOMMU_NONE so the IOMMU + * doesn't short-cut its translation table walk. + */ + iotlb = imrc->translate(iommu_mr, addr, IOMMU_NONE, iommu_idx); + addr = ((iotlb.translated_addr & ~iotlb.addr_mask) + | (addr & iotlb.addr_mask)); + /* Update the caller's prot bits to remove permissions the IOMMU + * is giving us a failure response for. If we get down to no + * permissions left at all we can give up now. + */ + if (!(iotlb.perm & IOMMU_RO)) { + *prot &= ~(PAGE_READ | PAGE_EXEC); + } + if (!(iotlb.perm & IOMMU_WO)) { + *prot &= ~PAGE_WRITE; + } + + if (!*prot) { + goto translate_fail; + } + + d = flatview_to_dispatch(address_space_to_flatview(iotlb.target_as)); + } assert(!memory_region_is_iommu(section->mr)); + *xlat = addr; return section; + +translate_fail: + return &d->map.sections[PHYS_SECTION_UNASSIGNED]; } #endif @@ -XXX,XX +XXX,XX @@ void cpu_exec_unrealizefn(CPUState *cpu) if (qdev_get_vmsd(DEVICE(cpu)) == NULL) { vmstate_unregister(NULL, &vmstate_cpu_common, cpu); } +#ifndef CONFIG_USER_ONLY + tcg_iommu_free_notifier_list(cpu); +#endif } Property cpu_common_props[] = { @@ -XXX,XX +XXX,XX @@ void cpu_exec_realizefn(CPUState *cpu, Error **errp) if (cc->vmsd != NULL) { vmstate_register(NULL, cpu->cpu_index, cc->vmsd, cpu); } + + cpu->iommu_notifiers = g_array_new(false, true, sizeof(TCGIOMMUNotifier)); #endif } -- 2.17.1 From: Julia Suvorova <jusual@mail.ru> ARMv6-M supports 6 Thumb2 instructions. This patch checks for these instructions and allows their execution. Like Thumb2 cores, ARMv6-M always interprets BL instruction as 32-bit. This patch is required for future Cortex-M0 support. Signed-off-by: Julia Suvorova <jusual@mail.ru> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20180612204632.28780-1-jusual@mail.ru [PMM: move armv6m_insn[] and armv6m_mask[] closer to point of use, and mark 'const'. Check for M-and-not-v7 rather than M-and-6.] Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 43 +++++++++++++++++++++++++++++++++++++----- 1 file changed, 38 insertions(+), 5 deletions(-) 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 bool thumb_insn_is_16bit(DisasContext *s, uint32_t insn) * end up actually treating this as two 16-bit insns, though, * if it's half of a bl/blx pair that might span a page boundary. */ - if (arm_dc_feature(s, ARM_FEATURE_THUMB2)) { + if (arm_dc_feature(s, ARM_FEATURE_THUMB2) || + arm_dc_feature(s, ARM_FEATURE_M)) { /* Thumb2 cores (including all M profile ones) always treat * 32-bit insns as 32-bit. */ @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) int conds; int logic_cc; - /* The only 32 bit insn that's allowed for Thumb1 is the combined - * BL/BLX prefix and suffix. + /* + * ARMv6-M supports a limited subset of Thumb2 instructions. + * Other Thumb1 architectures allow only 32-bit + * combined BL/BLX prefix and suffix. */ - if ((insn & 0xf800e800) != 0xf000e800) { + if (arm_dc_feature(s, ARM_FEATURE_M) && + !arm_dc_feature(s, ARM_FEATURE_V7)) { + int i; + bool found = false; + const uint32_t armv6m_insn[] = {0xf3808000 /* msr */, + 0xf3b08040 /* dsb */, + 0xf3b08050 /* dmb */, + 0xf3b08060 /* isb */, + 0xf3e08000 /* mrs */, + 0xf000d000 /* bl */}; + const uint32_t armv6m_mask[] = {0xffe0d000, + 0xfff0d0f0, + 0xfff0d0f0, + 0xfff0d0f0, + 0xffe0d000, + 0xf800d000}; + + for (i = 0; i < ARRAY_SIZE(armv6m_insn); i++) { + if ((insn & armv6m_mask[i]) == armv6m_insn[i]) { + found = true; + break; + } + } + if (!found) { + goto illegal_op; + } + } else if ((insn & 0xf800e800) != 0xf000e800) { ARCH(6T2); } @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; case 3: /* Special control operations. */ - ARCH(7); + if (!arm_dc_feature(s, ARM_FEATURE_V7) && + !(arm_dc_feature(s, ARM_FEATURE_V6) && + arm_dc_feature(s, ARM_FEATURE_M))) { + goto illegal_op; + } op = (insn >> 4) & 0xf; switch (op) { case 2: /* clrex */ -- 2.17.1 Most of this is the Neon decodetree patches, followed by Edgar's versal cleanups. thanks -- PMM The following changes since commit 2ef486e76d64436be90f7359a3071fb2a56ce835: Merge remote-tracking branch 'remotes/marcel/tags/rdma-pull-request' into staging (2020-05-03 14:12:56 +0100) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20200504 for you to fetch changes up to 9aefc6cf9b73f66062d2f914a0136756e7a28211: target/arm: Move gen_ function typedefs to translate.h (2020-05-04 12:59:26 +0100) ---------------------------------------------------------------- target-arm queue: * Start of conversion of Neon insns to decodetree * versal board: support SD and RTC * Implement ARMv8.2-TTS2UXN * Make VQDMULL undefined when U=1 * Some minor code cleanups ---------------------------------------------------------------- Edgar E. Iglesias (11): hw/arm: versal: Remove inclusion of arm_gicv3_common.h hw/arm: versal: Move misplaced comment hw/arm: versal-virt: Fix typo xlnx-ve -> xlnx-versal hw/arm: versal: Embed the UARTs into the SoC type hw/arm: versal: Embed the GEMs into the SoC type hw/arm: versal: Embed the ADMAs into the SoC type hw/arm: versal: Embed the APUs into the SoC type hw/arm: versal: Add support for SD hw/arm: versal: Add support for the RTC hw/arm: versal-virt: Add support for SD hw/arm: versal-virt: Add support for the RTC Fredrik Strupe (1): target/arm: Make VQDMULL undefined when U=1 Peter Maydell (25): target/arm: Don't use a TLB for ARMMMUIdx_Stage2 target/arm: Use enum constant in get_phys_addr_lpae() call target/arm: Add new 's1_is_el0' argument to get_phys_addr_lpae() target/arm: Implement ARMv8.2-TTS2UXN target/arm: Use correct variable for setting 'max' cpu's ID_AA64DFR0 target/arm/translate-vfp.inc.c: Remove duplicate simd_r32 check target/arm: Don't allow Thumb Neon insns without FEATURE_NEON target/arm: Add stubs for AArch32 Neon decodetree target/arm: Convert VCMLA (vector) to decodetree target/arm: Convert VCADD (vector) to decodetree target/arm: Convert V[US]DOT (vector) to decodetree target/arm: Convert VFM[AS]L (vector) to decodetree target/arm: Convert VCMLA (scalar) to decodetree target/arm: Convert V[US]DOT (scalar) to decodetree target/arm: Convert VFM[AS]L (scalar) to decodetree target/arm: Convert Neon load/store multiple structures to decodetree target/arm: Convert Neon 'load single structure to all lanes' to decodetree target/arm: Convert Neon 'load/store single structure' to decodetree target/arm: Convert Neon 3-reg-same VADD/VSUB to decodetree target/arm: Convert Neon 3-reg-same logic ops to decodetree target/arm: Convert Neon 3-reg-same VMAX/VMIN to decodetree target/arm: Convert Neon 3-reg-same comparisons to decodetree target/arm: Convert Neon 3-reg-same VQADD/VQSUB to decodetree target/arm: Convert Neon 3-reg-same VMUL, VMLA, VMLS, VSHL to decodetree target/arm: Move gen_ function typedefs to translate.h Philippe Mathieu-Daudé (2): hw/arm/mps2-tz: Use TYPE_IOTKIT instead of hardcoded string target/arm: Use uint64_t for midr field in CPU state struct include/hw/arm/xlnx-versal.h | 31 +- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 38 ++- target/arm/translate-a64.h | 9 - target/arm/translate.h | 26 ++ target/arm/neon-dp.decode | 86 +++++ target/arm/neon-ls.decode | 52 +++ target/arm/neon-shared.decode | 66 ++++ hw/arm/mps2-tz.c | 2 +- hw/arm/xlnx-versal-virt.c | 74 ++++- hw/arm/xlnx-versal.c | 115 +++++-- target/arm/cpu.c | 3 +- target/arm/cpu64.c | 8 +- target/arm/helper.c | 183 ++++------ target/arm/translate-a64.c | 17 - target/arm/translate-neon.inc.c | 714 +++++++++++++++++++++++++++++++++++++++ target/arm/translate-vfp.inc.c | 6 - target/arm/translate.c | 716 +++------------------------------------- target/arm/Makefile.objs | 18 + 19 files changed, 1302 insertions(+), 864 deletions(-) create mode 100644 target/arm/neon-dp.decode create mode 100644 target/arm/neon-ls.decode create mode 100644 target/arm/neon-shared.decode create mode 100644 target/arm/translate-neon.inc.c From: Fredrik Strupe <fredrik@strupe.net> According to Arm ARM, VQDMULL is only valid when U=0, while having U=1 is unallocated. Signed-off-by: Fredrik Strupe <fredrik@strupe.net> Fixes: 695272dcb976 ("target-arm: Handle UNDEF cases for Neon 3-regs-different-widths") Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) 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) {0, 0, 0, 0}, /* VMLSL */ {0, 0, 0, 9}, /* VQDMLSL */ {0, 0, 0, 0}, /* Integer VMULL */ - {0, 0, 0, 1}, /* VQDMULL */ + {0, 0, 0, 9}, /* VQDMULL */ {0, 0, 0, 0xa}, /* Polynomial VMULL */ {0, 0, 0, 7}, /* Reserved: always UNDEF */ }; -- 2.20.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> By using the TYPE_* definitions for devices, we can: - quickly find where devices are used with 'git-grep' - easily rename a device (one-line change). Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200428154650.21991-1-f4bug@amsat.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/mps2-tz.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/mps2-tz.c b/hw/arm/mps2-tz.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/mps2-tz.c +++ b/hw/arm/mps2-tz.c @@ -XXX,XX +XXX,XX @@ static void mps2tz_common_init(MachineState *machine) exit(EXIT_FAILURE); } - sysbus_init_child_obj(OBJECT(machine), "iotkit", &mms->iotkit, + sysbus_init_child_obj(OBJECT(machine), TYPE_IOTKIT, &mms->iotkit, sizeof(mms->iotkit), mmc->armsse_type); iotkitdev = DEVICE(&mms->iotkit); object_property_set_link(OBJECT(&mms->iotkit), OBJECT(system_memory), -- 2.20.1 We define ARMMMUIdx_Stage2 as being an MMU index which uses a QEMU TLB. However we never actually use the TLB -- all stage 2 lookups are done by direct calls to get_phys_addr_lpae() followed by a physical address load via address_space_ld*(). Remove Stage2 from the list of ARM MMU indexes which correspond to real core MMU indexes, and instead put it in the set of "NOTLB" ARM MMU indexes. This allows us to drop NB_MMU_MODES to 11. It also means we can safely add support for the ARMv8.3-TTS2UXN extension, which adds permission bits to the stage 2 descriptors which define execute permission separatel for EL0 and EL1; supporting that while keeping Stage2 in a QEMU TLB would require us to use separate TLBs for "Stage2 for an EL0 access" and "Stage2 for an EL1 access", which is a lot of extra complication given we aren't even using the QEMU TLB. In the process of updating the comment on our MMU index use, fix a couple of other minor errors: * NS EL2 EL2&0 was missing from the list in the comment * some text hadn't been updated from when we bumped NB_MMU_MODES above 8 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-2-peter.maydell@linaro.org --- target/arm/cpu-param.h | 2 +- target/arm/cpu.h | 21 +++++--- target/arm/helper.c | 112 ++++------------------------------------- 3 files changed, 27 insertions(+), 108 deletions(-) diff --git a/target/arm/cpu-param.h b/target/arm/cpu-param.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu-param.h +++ b/target/arm/cpu-param.h @@ -XXX,XX +XXX,XX @@ # define TARGET_PAGE_BITS_MIN 10 #endif -#define NB_MMU_MODES 12 +#define NB_MMU_MODES 11 #endif 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 @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * handling via the TLB. The only way to do a stage 1 translation without * the immediate stage 2 translation is via the ATS or AT system insns, * which can be slow-pathed and always do a page table walk. + * The only use of stage 2 translations is either as part of an s1+2 + * lookup or when loading the descriptors during a stage 1 page table walk, + * and in both those cases we don't use the TLB. * 4. we can also safely fold together the "32 bit EL3" and "64 bit EL3" * translation regimes, because they map reasonably well to each other * and they can't both be active at the same time. @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * NS EL1 EL1&0 stage 1+2 (aka NS PL1) * NS EL1 EL1&0 stage 1+2 +PAN * NS EL0 EL2&0 + * NS EL2 EL2&0 * NS EL2 EL2&0 +PAN * NS EL2 (aka NS PL2) * S EL0 EL1&0 (aka S PL0) * S EL1 EL1&0 (not used if EL3 is 32 bit) * S EL1 EL1&0 +PAN * S EL3 (aka S PL1) - * NS EL1&0 stage 2 * - * for a total of 12 different mmu_idx. + * for a total of 11 different mmu_idx. * * R profile CPUs have an MPU, but can use the same set of MMU indexes * as A profile. They only need to distinguish NS EL0 and NS EL1 (and @@ -XXX,XX +XXX,XX @@ bool write_cpustate_to_list(ARMCPU *cpu, bool kvm_sync); * are not quite the same -- different CPU types (most notably M profile * vs A/R profile) would like to use MMU indexes with different semantics, * but since we don't ever need to use all of those in a single CPU we - * can avoid setting NB_MMU_MODES to more than 8. The lower bits of + * can avoid having to set NB_MMU_MODES to "total number of A profile MMU + * modes + total number of M profile MMU modes". The lower bits of * ARMMMUIdx are the core TLB mmu index, and the higher bits are always * the same for any particular CPU. * Variables of type ARMMUIdx are always full values, and the core @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdx { ARMMMUIdx_SE10_1_PAN = 9 | ARM_MMU_IDX_A, ARMMMUIdx_SE3 = 10 | ARM_MMU_IDX_A, - ARMMMUIdx_Stage2 = 11 | ARM_MMU_IDX_A, - /* * These are not allocated TLBs and are used only for AT system * instructions or for the first stage of an S12 page table walk. @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdx { ARMMMUIdx_Stage1_E0 = 0 | ARM_MMU_IDX_NOTLB, ARMMMUIdx_Stage1_E1 = 1 | ARM_MMU_IDX_NOTLB, ARMMMUIdx_Stage1_E1_PAN = 2 | ARM_MMU_IDX_NOTLB, + /* + * Not allocated a TLB: used only for second stage of an S12 page + * table walk, or for descriptor loads during first stage of an S1 + * page table walk. Note that if we ever want to have a TLB for this + * then various TLB flush insns which currently are no-ops or flush + * only stage 1 MMU indexes will need to change to flush stage 2. + */ + ARMMMUIdx_Stage2 = 3 | ARM_MMU_IDX_NOTLB, /* * M-profile. @@ -XXX,XX +XXX,XX @@ typedef enum ARMMMUIdxBit { TO_CORE_BIT(SE10_1), TO_CORE_BIT(SE10_1_PAN), TO_CORE_BIT(SE3), - TO_CORE_BIT(Stage2), TO_CORE_BIT(MUser), TO_CORE_BIT(MPriv), 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 tlbiall_nsnh_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); } static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, @@ -XXX,XX +XXX,XX @@ static void tlbiall_nsnh_is_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx_all_cpus_synced(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); } -static void tlbiipas2_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - /* Invalidate by IPA. This has to invalidate any structures that - * contain only stage 2 translation information, but does not need - * to apply to structures that contain combined stage 1 and stage 2 - * translation information. - * This must NOP if EL2 isn't implemented or SCR_EL3.NS is zero. - */ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 40); - - tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_Stage2); -} - -static void tlbiipas2_is_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 40); - - tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, - ARMMMUIdxBit_Stage2); -} static void tlbiall_hyp_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value) @@ -XXX,XX +XXX,XX @@ static void vttbr_write(CPUARMState *env, const ARMCPRegInfo *ri, tlb_flush_by_mmuidx(cs, ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2); + ARMMMUIdxBit_E10_0); raw_write(env, ri, value); } } @@ -XXX,XX +XXX,XX @@ static int alle1_tlbmask(CPUARMState *env) return ARMMMUIdxBit_SE10_1 | ARMMMUIdxBit_SE10_1_PAN | ARMMMUIdxBit_SE10_0; - } else if (arm_feature(env, ARM_FEATURE_EL2)) { - return ARMMMUIdxBit_E10_1 | - ARMMMUIdxBit_E10_1_PAN | - ARMMMUIdxBit_E10_0 | - ARMMMUIdxBit_Stage2; } else { return ARMMMUIdxBit_E10_1 | ARMMMUIdxBit_E10_1_PAN | @@ -XXX,XX +XXX,XX @@ static void tlbi_aa64_vae3is_write(CPUARMState *env, const ARMCPRegInfo *ri, ARMMMUIdxBit_SE3); } -static void tlbi_aa64_ipas2e1_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - /* Invalidate by IPA. This has to invalidate any structures that - * contain only stage 2 translation information, but does not need - * to apply to structures that contain combined stage 1 and stage 2 - * translation information. - * This must NOP if EL2 isn't implemented or SCR_EL3.NS is zero. - */ - ARMCPU *cpu = env_archcpu(env); - CPUState *cs = CPU(cpu); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 48); - - tlb_flush_page_by_mmuidx(cs, pageaddr, ARMMMUIdxBit_Stage2); -} - -static void tlbi_aa64_ipas2e1is_write(CPUARMState *env, const ARMCPRegInfo *ri, - uint64_t value) -{ - CPUState *cs = env_cpu(env); - uint64_t pageaddr; - - if (!arm_feature(env, ARM_FEATURE_EL2) || !(env->cp15.scr_el3 & SCR_NS)) { - return; - } - - pageaddr = sextract64(value << 12, 0, 48); - - tlb_flush_page_by_mmuidx_all_cpus_synced(cs, pageaddr, - ARMMMUIdxBit_Stage2); -} - static CPAccessResult aa64_zva_access(CPUARMState *env, const ARMCPRegInfo *ri, bool isread) { @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_vae1_write }, { .name = "TLBI_IPAS2E1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1is_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_IPAS2LE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1is_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_ALLE1IS", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 3, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbi_aa64_alle1is_write }, { .name = "TLBI_IPAS2E1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_IPAS2LE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .access = PL2_W, .type = ARM_CP_NO_RAW, - .writefn = tlbi_aa64_ipas2e1_write }, + .access = PL2_W, .type = ARM_CP_NOP }, { .name = "TLBI_ALLE1", .state = ARM_CP_STATE_AA64, .opc0 = 1, .opc1 = 4, .crn = 8, .crm = 7, .opc2 = 4, .access = PL2_W, .type = ARM_CP_NO_RAW, @@ -XXX,XX +XXX,XX @@ static const ARMCPRegInfo v8_cp_reginfo[] = { .writefn = tlbimva_hyp_is_write }, { .name = "TLBIIPAS2", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 1, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2IS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 1, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_is_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2L", .cp = 15, .opc1 = 4, .crn = 8, .crm = 4, .opc2 = 5, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_write }, + .type = ARM_CP_NOP, .access = PL2_W }, { .name = "TLBIIPAS2LIS", .cp = 15, .opc1 = 4, .crn = 8, .crm = 0, .opc2 = 5, - .type = ARM_CP_NO_RAW, .access = PL2_W, - .writefn = tlbiipas2_is_write }, + .type = ARM_CP_NOP, .access = PL2_W }, /* 32 bit cache operations */ { .name = "ICIALLUIS", .cp = 15, .opc1 = 0, .crn = 7, .crm = 1, .opc2 = 0, .type = ARM_CP_NOP, .access = PL1_W, .accessfn = aa64_cacheop_pou_access }, -- 2.20.1 The access_type argument to get_phys_addr_lpae() is an MMUAccessType; use the enum constant MMU_DATA_LOAD rather than a literal 0 when we call it in S1_ptw_translate(). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-3-peter.maydell@linaro.org --- target/arm/helper.c | 5 +++-- 1 file changed, 3 insertions(+), 2 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 hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, pcacheattrs = &cacheattrs; } - ret = get_phys_addr_lpae(env, addr, 0, ARMMMUIdx_Stage2, &s2pa, - &txattrs, &s2prot, &s2size, fi, pcacheattrs); + ret = get_phys_addr_lpae(env, addr, MMU_DATA_LOAD, ARMMMUIdx_Stage2, + &s2pa, &txattrs, &s2prot, &s2size, fi, + pcacheattrs); if (ret) { assert(fi->type != ARMFault_None); fi->s2addr = addr; -- 2.20.1 For ARMv8.2-TTS2UXN, the stage 2 page table walk wants to know whether the stage 1 access is for EL0 or not, because whether exec permission is given can depend on whether this is an EL0 or EL1 access. Add a new argument to get_phys_addr_lpae() so the call sites can pass this information in. Since get_phys_addr_lpae() doesn't already have a doc comment, add one so we have a place to put the documentation of the semantics of the new s1_is_el0 argument. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-4-peter.maydell@linaro.org --- target/arm/helper.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 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 bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, MMUAccessType access_type, ARMMMUIdx mmu_idx, + bool s1_is_el0, hwaddr *phys_ptr, MemTxAttrs *txattrs, int *prot, target_ulong *page_size_ptr, ARMMMUFaultInfo *fi, ARMCacheAttrs *cacheattrs); @@ -XXX,XX +XXX,XX @@ static hwaddr S1_ptw_translate(CPUARMState *env, ARMMMUIdx mmu_idx, } ret = get_phys_addr_lpae(env, addr, MMU_DATA_LOAD, ARMMMUIdx_Stage2, + false, &s2pa, &txattrs, &s2prot, &s2size, fi, pcacheattrs); if (ret) { @@ -XXX,XX +XXX,XX @@ static ARMVAParameters aa32_va_parameters(CPUARMState *env, uint32_t va, }; } +/** + * get_phys_addr_lpae: perform one stage of page table walk, LPAE format + * + * Returns false if the translation was successful. Otherwise, phys_ptr, attrs, + * prot and page_size may not be filled in, and the populated fsr value provides + * information on why the translation aborted, in the format of a long-format + * DFSR/IFSR fault register, with the following caveats: + * * the WnR bit is never set (the caller must do this). + * + * @env: CPUARMState + * @address: virtual address to get physical address for + * @access_type: MMU_DATA_LOAD, MMU_DATA_STORE or MMU_INST_FETCH + * @mmu_idx: MMU index indicating required translation regime + * @s1_is_el0: if @mmu_idx is ARMMMUIdx_Stage2 (so this is a stage 2 page table + * walk), must be true if this is stage 2 of a stage 1+2 walk for an + * EL0 access). If @mmu_idx is anything else, @s1_is_el0 is ignored. + * @phys_ptr: set to the physical address corresponding to the virtual address + * @attrs: set to the memory transaction attributes to use + * @prot: set to the permissions for the page containing phys_ptr + * @page_size_ptr: set to the size of the page containing phys_ptr + * @fi: set to fault info if the translation fails + * @cacheattrs: (if non-NULL) set to the cacheability/shareability attributes + */ static bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, MMUAccessType access_type, ARMMMUIdx mmu_idx, + bool s1_is_el0, hwaddr *phys_ptr, MemTxAttrs *txattrs, int *prot, target_ulong *page_size_ptr, ARMMMUFaultInfo *fi, ARMCacheAttrs *cacheattrs) @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, /* S1 is done. Now do S2 translation. */ ret = get_phys_addr_lpae(env, ipa, access_type, ARMMMUIdx_Stage2, + mmu_idx == ARMMMUIdx_E10_0, phys_ptr, attrs, &s2_prot, page_size, fi, cacheattrs != NULL ? &cacheattrs2 : NULL); @@ -XXX,XX +XXX,XX @@ bool get_phys_addr(CPUARMState *env, target_ulong address, } if (regime_using_lpae_format(env, mmu_idx)) { - return get_phys_addr_lpae(env, address, access_type, mmu_idx, + return get_phys_addr_lpae(env, address, access_type, mmu_idx, false, phys_ptr, attrs, prot, page_size, fi, cacheattrs); } else if (regime_sctlr(env, mmu_idx) & SCTLR_XP) { -- 2.20.1 The ARMv8.2-TTS2UXN feature extends the XN field in stage 2 translation table descriptors from just bit [54] to bits [54:53], allowing stage 2 to control execution permissions separately for EL0 and EL1. Implement the new semantics of the XN field and enable the feature for our 'max' CPU. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200330210400.11724-5-peter.maydell@linaro.org --- target/arm/cpu.h | 15 +++++++++++++++ target/arm/cpu.c | 1 + target/arm/cpu64.c | 2 ++ target/arm/helper.c | 37 +++++++++++++++++++++++++++++++------ 4 files changed, 49 insertions(+), 6 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 @@ static inline bool isar_feature_aa32_ccidx(const ARMISARegisters *id) return FIELD_EX32(id->id_mmfr4, ID_MMFR4, CCIDX) != 0; } +static inline bool isar_feature_aa32_tts2uxn(const ARMISARegisters *id) +{ + return FIELD_EX32(id->id_mmfr4, ID_MMFR4, XNX) != 0; +} + /* * 64-bit feature tests via id registers. */ @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_aa64_ccidx(const ARMISARegisters *id) return FIELD_EX64(id->id_aa64mmfr2, ID_AA64MMFR2, CCIDX) != 0; } +static inline bool isar_feature_aa64_tts2uxn(const ARMISARegisters *id) +{ + return FIELD_EX64(id->id_aa64mmfr1, ID_AA64MMFR1, XNX) != 0; +} + /* * Feature tests for "does this exist in either 32-bit or 64-bit?" */ @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_any_ccidx(const ARMISARegisters *id) return isar_feature_aa64_ccidx(id) || isar_feature_aa32_ccidx(id); } +static inline bool isar_feature_any_tts2uxn(const ARMISARegisters *id) +{ + return isar_feature_aa64_tts2uxn(id) || isar_feature_aa32_tts2uxn(id); +} + /* * Forward to the above feature tests given an ARMCPU pointer. */ 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) t = FIELD_DP32(t, ID_MMFR4, HPDS, 1); /* AA32HPD */ t = FIELD_DP32(t, ID_MMFR4, AC2, 1); /* ACTLR2, HACTLR2 */ t = FIELD_DP32(t, ID_MMFR4, CNP, 1); /* TTCNP */ + t = FIELD_DP32(t, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = t; } #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) t = FIELD_DP64(t, ID_AA64MMFR1, VH, 1); t = FIELD_DP64(t, ID_AA64MMFR1, PAN, 2); /* ATS1E1 */ t = FIELD_DP64(t, ID_AA64MMFR1, VMIDBITS, 2); /* VMID16 */ + t = FIELD_DP64(t, ID_AA64MMFR1, XNX, 1); /* TTS2UXN */ cpu->isar.id_aa64mmfr1 = t; t = cpu->isar.id_aa64mmfr2; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) u = FIELD_DP32(u, ID_MMFR4, HPDS, 1); /* AA32HPD */ u = FIELD_DP32(u, ID_MMFR4, AC2, 1); /* ACTLR2, HACTLR2 */ u = FIELD_DP32(u, ID_MMFR4, CNP, 1); /* TTCNP */ + u = FIELD_DP32(u, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = u; u = cpu->isar.id_aa64dfr0; 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 @@ simple_ap_to_rw_prot(CPUARMState *env, ARMMMUIdx mmu_idx, int ap) * * @env: CPUARMState * @s2ap: The 2-bit stage2 access permissions (S2AP) - * @xn: XN (execute-never) bit + * @xn: XN (execute-never) bits + * @s1_is_el0: true if this is S2 of an S1+2 walk for EL0 */ -static int get_S2prot(CPUARMState *env, int s2ap, int xn) +static int get_S2prot(CPUARMState *env, int s2ap, int xn, bool s1_is_el0) { int prot = 0; @@ -XXX,XX +XXX,XX @@ static int get_S2prot(CPUARMState *env, int s2ap, int xn) if (s2ap & 2) { prot |= PAGE_WRITE; } - if (!xn) { - if (arm_el_is_aa64(env, 2) || prot & PAGE_READ) { + + if (cpu_isar_feature(any_tts2uxn, env_archcpu(env))) { + switch (xn) { + case 0: prot |= PAGE_EXEC; + break; + case 1: + if (s1_is_el0) { + prot |= PAGE_EXEC; + } + break; + case 2: + break; + case 3: + if (!s1_is_el0) { + prot |= PAGE_EXEC; + } + break; + default: + g_assert_not_reached(); + } + } else { + if (!extract32(xn, 1, 1)) { + if (arm_el_is_aa64(env, 2) || prot & PAGE_READ) { + prot |= PAGE_EXEC; + } } } return prot; @@ -XXX,XX +XXX,XX @@ static bool get_phys_addr_lpae(CPUARMState *env, target_ulong address, } ap = extract32(attrs, 4, 2); - xn = extract32(attrs, 12, 1); if (mmu_idx == ARMMMUIdx_Stage2) { ns = true; - *prot = get_S2prot(env, ap, xn); + xn = extract32(attrs, 11, 2); + *prot = get_S2prot(env, ap, xn, s1_is_el0); } else { ns = extract32(attrs, 3, 1); + xn = extract32(attrs, 12, 1); pxn = extract32(attrs, 11, 1); *prot = get_S1prot(env, mmu_idx, aarch64, ap, ns, xn, pxn); } -- 2.20.1 In aarch64_max_initfn() we update both 32-bit and 64-bit ID registers. The intended pattern is that for 64-bit ID registers we use FIELD_DP64 and the uint64_t 't' register, while 32-bit ID registers use FIELD_DP32 and the uint32_t 'u' register. For ID_AA64DFR0 we accidentally used 'u', meaning that the top 32 bits of this 64-bit ID register would end up always zero. Luckily at the moment that's what they should be anyway, so this bug has no visible effects. Use the right-sized variable. Fixes: 3bec78447a958d481991 Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Laurent Desnogues <laurent.desnogues@gmail.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200423110915.10527-1-peter.maydell@linaro.org --- target/arm/cpu64.c | 6 +++--- 1 file changed, 3 insertions(+), 3 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) u = FIELD_DP32(u, ID_MMFR4, XNX, 1); /* TTS2UXN */ cpu->isar.id_mmfr4 = u; - u = cpu->isar.id_aa64dfr0; - u = FIELD_DP64(u, ID_AA64DFR0, PMUVER, 5); /* v8.4-PMU */ - cpu->isar.id_aa64dfr0 = u; + t = cpu->isar.id_aa64dfr0; + t = FIELD_DP64(t, ID_AA64DFR0, PMUVER, 5); /* v8.4-PMU */ + cpu->isar.id_aa64dfr0 = t; u = cpu->isar.id_dfr0; u = FIELD_DP32(u, ID_DFR0, PERFMON, 5); /* v8.4-PMU */ -- 2.20.1 From: Philippe Mathieu-Daudé <f4bug@amsat.org> MIDR_EL1 is a 64-bit system register with the top 32-bit being RES0. Represent it in QEMU's ARMCPU struct with a uint64_t, not a uint32_t. This fixes an error when compiling with -Werror=conversion because we were manipulating the register value using a local uint64_t variable: target/arm/cpu64.c: In function ‘aarch64_max_initfn’: target/arm/cpu64.c:628:21: error: conversion from ‘uint64_t’ {aka ‘long unsigned int’} to ‘uint32_t’ {aka ‘unsigned int’} may change value [-Werror=conversion] 628 | cpu->midr = t; | ^ and future-proofs us against a possible future architecture change using some of the top 32 bits. Suggested-by: Laurent Desnogues <laurent.desnogues@gmail.com> Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Laurent Desnogues <laurent.desnogues@gmail.com> Message-id: 20200428172634.29707-1-f4bug@amsat.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 2 +- target/arm/cpu.c | 2 +- 2 files changed, 2 insertions(+), 2 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 { uint64_t id_aa64dfr0; uint64_t id_aa64dfr1; } isar; - uint32_t midr; + uint64_t midr; uint32_t revidr; uint32_t reset_fpsid; uint32_t ctr; 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 const ARMCPUInfo arm_cpus[] = { static Property arm_cpu_properties[] = { DEFINE_PROP_BOOL("start-powered-off", ARMCPU, start_powered_off, false), DEFINE_PROP_UINT32("psci-conduit", ARMCPU, psci_conduit, 0), - DEFINE_PROP_UINT32("midr", ARMCPU, midr, 0), + DEFINE_PROP_UINT64("midr", ARMCPU, midr, 0), DEFINE_PROP_UINT64("mp-affinity", ARMCPU, mp_affinity, ARM64_AFFINITY_INVALID), DEFINE_PROP_INT32("node-id", ARMCPU, node_id, CPU_UNSET_NUMA_NODE_ID), -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Remove inclusion of arm_gicv3_common.h, this already gets included via xlnx-versal.h. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-2-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal.c | 1 - 1 file changed, 1 deletion(-) diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "kvm_arm.h" #include "hw/misc/unimp.h" -#include "hw/intc/arm_gicv3_common.h" #include "hw/arm/xlnx-versal.h" #include "hw/char/pl011.h" -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Move misplaced comment. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-3-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) obj = object_new(XLNX_VERSAL_ACPU_TYPE); if (!obj) { - /* Secondary CPUs start in PSCI powered-down state */ error_report("Unable to create apu.cpu[%d] of type %s", i, XLNX_VERSAL_ACPU_TYPE); exit(EXIT_FAILURE); @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) object_property_set_int(obj, s->cfg.psci_conduit, "psci-conduit", &error_abort); if (i) { + /* Secondary CPUs start in PSCI powered-down state */ object_property_set_bool(obj, true, "start-powered-off", &error_abort); } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Fix typo xlnx-ve -> xlnx-versal. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-4-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) psci_conduit = QEMU_PSCI_CONDUIT_SMC; } - sysbus_init_child_obj(OBJECT(machine), "xlnx-ve", &s->soc, + sysbus_init_child_obj(OBJECT(machine), "xlnx-versal", &s->soc, sizeof(s->soc), TYPE_XLNX_VERSAL); object_property_set_link(OBJECT(&s->soc), OBJECT(machine->ram), "ddr", &error_abort); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the UARTs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-5-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 12 ++++++------ 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/sysbus.h" #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" +#include "hw/char/pl011.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { MemoryRegion mr_ocm; struct { - SysBusDevice *uart[XLNX_VERSAL_NR_UARTS]; + PL011State uart[XLNX_VERSAL_NR_UARTS]; SysBusDevice *gem[XLNX_VERSAL_NR_GEMS]; SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; } iou; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ #include "kvm_arm.h" #include "hw/misc/unimp.h" #include "hw/arm/xlnx-versal.h" -#include "hw/char/pl011.h" #define XLNX_VERSAL_ACPU_TYPE ARM_CPU_TYPE_NAME("cortex-a72") #define GEM_REVISION 0x40070106 @@ -XXX,XX +XXX,XX @@ static void versal_create_uarts(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, TYPE_PL011); - s->lpd.iou.uart[i] = SYS_BUS_DEVICE(dev); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.uart[i], sizeof(s->lpd.iou.uart[i]), + TYPE_PL011); + dev = DEVICE(&s->lpd.iou.uart[i]); qdev_prop_set_chr(dev, "chardev", serial_hd(i)); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.uart[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, addrs[i], mr); - sysbus_connect_irq(s->lpd.iou.uart[i], 0, pic[irqs[i]]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[irqs[i]]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the GEMs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-6-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 15 ++++++++------- 2 files changed, 10 insertions(+), 8 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" +#include "hw/net/cadence_gem.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { PL011State uart[XLNX_VERSAL_NR_UARTS]; - SysBusDevice *gem[XLNX_VERSAL_NR_GEMS]; + CadenceGEMState gem[XLNX_VERSAL_NR_GEMS]; SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; } iou; } lpd; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_gems(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, "cadence_gem"); - s->lpd.iou.gem[i] = SYS_BUS_DEVICE(dev); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.gem[i], sizeof(s->lpd.iou.gem[i]), + TYPE_CADENCE_GEM); + dev = DEVICE(&s->lpd.iou.gem[i]); if (nd->used) { qemu_check_nic_model(nd, "cadence_gem"); qdev_set_nic_properties(dev, nd); } - object_property_set_int(OBJECT(s->lpd.iou.gem[i]), + object_property_set_int(OBJECT(dev), 2, "num-priority-queues", &error_abort); - object_property_set_link(OBJECT(s->lpd.iou.gem[i]), + object_property_set_link(OBJECT(dev), OBJECT(&s->mr_ps), "dma", &error_abort); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.gem[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, addrs[i], mr); - sysbus_connect_irq(s->lpd.iou.gem[i], 0, pic[irqs[i]]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[irqs[i]]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the ADMAs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-7-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 3 ++- hw/arm/xlnx-versal.c | 14 +++++++------- 2 files changed, 9 insertions(+), 8 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/arm/boot.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" +#include "hw/dma/xlnx-zdma.h" #include "hw/net/cadence_gem.h" #define TYPE_XLNX_VERSAL "xlnx-versal" @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { PL011State uart[XLNX_VERSAL_NR_UARTS]; CadenceGEMState gem[XLNX_VERSAL_NR_GEMS]; - SysBusDevice *adma[XLNX_VERSAL_NR_ADMAS]; + XlnxZDMA adma[XLNX_VERSAL_NR_ADMAS]; } iou; } lpd; diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_admas(Versal *s, qemu_irq *pic) DeviceState *dev; MemoryRegion *mr; - dev = qdev_create(NULL, "xlnx.zdma"); - s->lpd.iou.adma[i] = SYS_BUS_DEVICE(dev); - object_property_set_int(OBJECT(s->lpd.iou.adma[i]), 128, "bus-width", - &error_abort); - object_property_add_child(OBJECT(s), name, OBJECT(dev), &error_fatal); + sysbus_init_child_obj(OBJECT(s), name, + &s->lpd.iou.adma[i], sizeof(s->lpd.iou.adma[i]), + TYPE_XLNX_ZDMA); + dev = DEVICE(&s->lpd.iou.adma[i]); + object_property_set_int(OBJECT(dev), 128, "bus-width", &error_abort); qdev_init_nofail(dev); - mr = sysbus_mmio_get_region(s->lpd.iou.adma[i], 0); + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); memory_region_add_subregion(&s->mr_ps, MM_ADMA_CH0 + i * MM_ADMA_CH0_SIZE, mr); - sysbus_connect_irq(s->lpd.iou.adma[i], 0, pic[VERSAL_ADMA_IRQ_0 + i]); + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[VERSAL_ADMA_IRQ_0 + i]); g_free(name); } } -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Embed the APUs into the SoC type. Suggested-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-8-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 2 +- hw/arm/xlnx-versal-virt.c | 4 ++-- hw/arm/xlnx-versal.c | 19 +++++-------------- 3 files changed, 8 insertions(+), 17 deletions(-) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { struct { MemoryRegion mr; - ARMCPU *cpu[XLNX_VERSAL_NR_ACPUS]; + ARMCPU cpu[XLNX_VERSAL_NR_ACPUS]; GICv3State gic; } apu; } fpd; diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) s->binfo.get_dtb = versal_virt_get_dtb; s->binfo.modify_dtb = versal_virt_modify_dtb; if (machine->kernel_filename) { - arm_load_kernel(s->soc.fpd.apu.cpu[0], machine, &s->binfo); + arm_load_kernel(&s->soc.fpd.apu.cpu[0], machine, &s->binfo); } else { - AddressSpace *as = arm_boot_address_space(s->soc.fpd.apu.cpu[0], + AddressSpace *as = arm_boot_address_space(&s->soc.fpd.apu.cpu[0], &s->binfo); /* Some boot-loaders (e.g u-boot) don't like blobs at address 0 (NULL). * Offset things by 4K. */ diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) for (i = 0; i < ARRAY_SIZE(s->fpd.apu.cpu); i++) { Object *obj; - char *name; - - obj = object_new(XLNX_VERSAL_ACPU_TYPE); - if (!obj) { - error_report("Unable to create apu.cpu[%d] of type %s", - i, XLNX_VERSAL_ACPU_TYPE); - exit(EXIT_FAILURE); - } - - name = g_strdup_printf("apu-cpu[%d]", i); - object_property_add_child(OBJECT(s), name, obj, &error_fatal); - g_free(name); + object_initialize_child(OBJECT(s), "apu-cpu[*]", + &s->fpd.apu.cpu[i], sizeof(s->fpd.apu.cpu[i]), + XLNX_VERSAL_ACPU_TYPE, &error_abort, NULL); + obj = OBJECT(&s->fpd.apu.cpu[i]); object_property_set_int(obj, s->cfg.psci_conduit, "psci-conduit", &error_abort); if (i) { @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_cpus(Versal *s) object_property_set_link(obj, OBJECT(&s->fpd.apu.mr), "memory", &error_abort); object_property_set_bool(obj, true, "realized", &error_fatal); - s->fpd.apu.cpu[i] = ARM_CPU(obj); } } @@ -XXX,XX +XXX,XX @@ static void versal_create_apu_gic(Versal *s, qemu_irq *pic) } for (i = 0; i < nr_apu_cpus; i++) { - DeviceState *cpudev = DEVICE(s->fpd.apu.cpu[i]); + DeviceState *cpudev = DEVICE(&s->fpd.apu.cpu[i]); int ppibase = XLNX_VERSAL_NR_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS; qemu_irq maint_irq; int ti; -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for SD. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-9-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 12 ++++++++++++ hw/arm/xlnx-versal.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 43 insertions(+) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/sysbus.h" #include "hw/arm/boot.h" +#include "hw/sd/sdhci.h" #include "hw/intc/arm_gicv3.h" #include "hw/char/pl011.h" #include "hw/dma/xlnx-zdma.h" @@ -XXX,XX +XXX,XX @@ #define XLNX_VERSAL_NR_UARTS 2 #define XLNX_VERSAL_NR_GEMS 2 #define XLNX_VERSAL_NR_ADMAS 8 +#define XLNX_VERSAL_NR_SDS 2 #define XLNX_VERSAL_NR_IRQS 192 typedef struct Versal { @@ -XXX,XX +XXX,XX @@ typedef struct Versal { } iou; } lpd; + /* The Platform Management Controller subsystem. */ + struct { + struct { + SDHCIState sd[XLNX_VERSAL_NR_SDS]; + } iou; + } pmc; + struct { MemoryRegion *mr_ddr; uint32_t psci_conduit; @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define VERSAL_GEM1_IRQ_0 58 #define VERSAL_GEM1_WAKE_IRQ_0 59 #define VERSAL_ADMA_IRQ_0 60 +#define VERSAL_SD0_IRQ_0 126 /* Architecturally reserved IRQs suitable for virtualization. */ #define VERSAL_RSVD_IRQ_FIRST 111 @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define MM_FPD_CRF 0xfd1a0000U #define MM_FPD_CRF_SIZE 0x140000 +#define MM_PMC_SD0 0xf1040000U +#define MM_PMC_SD0_SIZE 0x10000 #define MM_PMC_CRP 0xf1260000U #define MM_PMC_CRP_SIZE 0x10000 #endif diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_admas(Versal *s, qemu_irq *pic) } } +#define SDHCI_CAPABILITIES 0x280737ec6481 /* Same as on ZynqMP. */ +static void versal_create_sds(Versal *s, qemu_irq *pic) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(s->pmc.iou.sd); i++) { + DeviceState *dev; + MemoryRegion *mr; + + sysbus_init_child_obj(OBJECT(s), "sd[*]", + &s->pmc.iou.sd[i], sizeof(s->pmc.iou.sd[i]), + TYPE_SYSBUS_SDHCI); + dev = DEVICE(&s->pmc.iou.sd[i]); + + object_property_set_uint(OBJECT(dev), + 3, "sd-spec-version", &error_fatal); + object_property_set_uint(OBJECT(dev), SDHCI_CAPABILITIES, "capareg", + &error_fatal); + object_property_set_uint(OBJECT(dev), UHS_I, "uhs", &error_fatal); + qdev_init_nofail(dev); + + mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0); + memory_region_add_subregion(&s->mr_ps, + MM_PMC_SD0 + i * MM_PMC_SD0_SIZE, mr); + + sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, + pic[VERSAL_SD0_IRQ_0 + i * 2]); + } +} + /* This takes the board allocated linear DDR memory and creates aliases * for each split DDR range/aperture on the Versal address map. */ @@ -XXX,XX +XXX,XX @@ static void versal_realize(DeviceState *dev, Error **errp) versal_create_uarts(s, pic); versal_create_gems(s, pic); versal_create_admas(s, pic); + versal_create_sds(s, pic); versal_map_ddr(s); versal_unimp(s); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> hw/arm: versal: Add support for the RTC. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-10-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- include/hw/arm/xlnx-versal.h | 8 ++++++++ hw/arm/xlnx-versal.c | 21 +++++++++++++++++++++ 2 files changed, 29 insertions(+) diff --git a/include/hw/arm/xlnx-versal.h b/include/hw/arm/xlnx-versal.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/xlnx-versal.h +++ b/include/hw/arm/xlnx-versal.h @@ -XXX,XX +XXX,XX @@ #include "hw/char/pl011.h" #include "hw/dma/xlnx-zdma.h" #include "hw/net/cadence_gem.h" +#include "hw/rtc/xlnx-zynqmp-rtc.h" #define TYPE_XLNX_VERSAL "xlnx-versal" #define XLNX_VERSAL(obj) OBJECT_CHECK(Versal, (obj), TYPE_XLNX_VERSAL) @@ -XXX,XX +XXX,XX @@ typedef struct Versal { struct { SDHCIState sd[XLNX_VERSAL_NR_SDS]; } iou; + + XlnxZynqMPRTC rtc; } pmc; struct { @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define VERSAL_GEM1_IRQ_0 58 #define VERSAL_GEM1_WAKE_IRQ_0 59 #define VERSAL_ADMA_IRQ_0 60 +#define VERSAL_RTC_APB_ERR_IRQ 121 #define VERSAL_SD0_IRQ_0 126 +#define VERSAL_RTC_ALARM_IRQ 142 +#define VERSAL_RTC_SECONDS_IRQ 143 /* Architecturally reserved IRQs suitable for virtualization. */ #define VERSAL_RSVD_IRQ_FIRST 111 @@ -XXX,XX +XXX,XX @@ typedef struct Versal { #define MM_PMC_SD0_SIZE 0x10000 #define MM_PMC_CRP 0xf1260000U #define MM_PMC_CRP_SIZE 0x10000 +#define MM_PMC_RTC 0xf12a0000 +#define MM_PMC_RTC_SIZE 0x10000 #endif diff --git a/hw/arm/xlnx-versal.c b/hw/arm/xlnx-versal.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal.c +++ b/hw/arm/xlnx-versal.c @@ -XXX,XX +XXX,XX @@ static void versal_create_sds(Versal *s, qemu_irq *pic) } } +static void versal_create_rtc(Versal *s, qemu_irq *pic) +{ + SysBusDevice *sbd; + MemoryRegion *mr; + + sysbus_init_child_obj(OBJECT(s), "rtc", &s->pmc.rtc, sizeof(s->pmc.rtc), + TYPE_XLNX_ZYNQMP_RTC); + sbd = SYS_BUS_DEVICE(&s->pmc.rtc); + qdev_init_nofail(DEVICE(sbd)); + + mr = sysbus_mmio_get_region(sbd, 0); + memory_region_add_subregion(&s->mr_ps, MM_PMC_RTC, mr); + + /* + * TODO: Connect the ALARM and SECONDS interrupts once our RTC model + * supports them. + */ + sysbus_connect_irq(sbd, 1, pic[VERSAL_RTC_APB_ERR_IRQ]); +} + /* This takes the board allocated linear DDR memory and creates aliases * for each split DDR range/aperture on the Versal address map. */ @@ -XXX,XX +XXX,XX @@ static void versal_realize(DeviceState *dev, Error **errp) versal_create_gems(s, pic); versal_create_admas(s, pic); versal_create_sds(s, pic); + versal_create_rtc(s, pic); versal_map_ddr(s); versal_unimp(s); -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for SD. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-11-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 46 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/sysbus-fdt.h" #include "hw/arm/fdt.h" #include "cpu.h" +#include "hw/qdev-properties.h" #include "hw/arm/xlnx-versal.h" #define TYPE_XLNX_VERSAL_VIRT_MACHINE MACHINE_TYPE_NAME("xlnx-versal-virt") @@ -XXX,XX +XXX,XX @@ static void fdt_add_zdma_nodes(VersalVirt *s) } } +static void fdt_add_sd_nodes(VersalVirt *s) +{ + const char clocknames[] = "clk_xin\0clk_ahb"; + const char compat[] = "arasan,sdhci-8.9a"; + int i; + + for (i = ARRAY_SIZE(s->soc.pmc.iou.sd) - 1; i >= 0; i--) { + uint64_t addr = MM_PMC_SD0 + MM_PMC_SD0_SIZE * i; + char *name = g_strdup_printf("/sdhci@%" PRIx64, addr); + + qemu_fdt_add_subnode(s->fdt, name); + + qemu_fdt_setprop_cells(s->fdt, name, "clocks", + s->phandle.clk_25Mhz, s->phandle.clk_25Mhz); + qemu_fdt_setprop(s->fdt, name, "clock-names", + clocknames, sizeof(clocknames)); + qemu_fdt_setprop_cells(s->fdt, name, "interrupts", + GIC_FDT_IRQ_TYPE_SPI, VERSAL_SD0_IRQ_0 + i * 2, + GIC_FDT_IRQ_FLAGS_LEVEL_HI); + qemu_fdt_setprop_sized_cells(s->fdt, name, "reg", + 2, addr, 2, MM_PMC_SD0_SIZE); + qemu_fdt_setprop(s->fdt, name, "compatible", compat, sizeof(compat)); + g_free(name); + } +} + static void fdt_nop_memory_nodes(void *fdt, Error **errp) { Error *err = NULL; @@ -XXX,XX +XXX,XX @@ static void create_virtio_regions(VersalVirt *s) } } +static void sd_plugin_card(SDHCIState *sd, DriveInfo *di) +{ + BlockBackend *blk = di ? blk_by_legacy_dinfo(di) : NULL; + DeviceState *card; + + card = qdev_create(qdev_get_child_bus(DEVICE(sd), "sd-bus"), TYPE_SD_CARD); + object_property_add_child(OBJECT(sd), "card[*]", OBJECT(card), + &error_fatal); + qdev_prop_set_drive(card, "drive", blk, &error_fatal); + object_property_set_bool(OBJECT(card), true, "realized", &error_fatal); +} + static void versal_virt_init(MachineState *machine) { VersalVirt *s = XLNX_VERSAL_VIRT_MACHINE(machine); int psci_conduit = QEMU_PSCI_CONDUIT_DISABLED; + int i; /* * If the user provides an Operating System to be loaded, we expect them @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) fdt_add_gic_nodes(s); fdt_add_timer_nodes(s); fdt_add_zdma_nodes(s); + fdt_add_sd_nodes(s); fdt_add_cpu_nodes(s, psci_conduit); fdt_add_clk_node(s, "/clk125", 125000000, s->phandle.clk_125Mhz); fdt_add_clk_node(s, "/clk25", 25000000, s->phandle.clk_25Mhz); @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) memory_region_add_subregion_overlap(get_system_memory(), 0, &s->soc.fpd.apu.mr, 0); + /* Plugin SD cards. */ + for (i = 0; i < ARRAY_SIZE(s->soc.pmc.iou.sd); i++) { + sd_plugin_card(&s->soc.pmc.iou.sd[i], drive_get_next(IF_SD)); + } + s->binfo.ram_size = machine->ram_size; s->binfo.loader_start = 0x0; s->binfo.get_dtb = versal_virt_get_dtb; -- 2.20.1 From: "Edgar E. Iglesias" <edgar.iglesias@xilinx.com> Add support for the RTC. Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com> Reviewed-by: Alistair Francis <alistair.francis@wdc.com> Reviewed-by: Luc Michel <luc.michel@greensocs.com> Message-id: 20200427181649.26851-12-edgar.iglesias@gmail.com Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- hw/arm/xlnx-versal-virt.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) diff --git a/hw/arm/xlnx-versal-virt.c b/hw/arm/xlnx-versal-virt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/xlnx-versal-virt.c +++ b/hw/arm/xlnx-versal-virt.c @@ -XXX,XX +XXX,XX @@ static void fdt_add_sd_nodes(VersalVirt *s) } } +static void fdt_add_rtc_node(VersalVirt *s) +{ + const char compat[] = "xlnx,zynqmp-rtc"; + const char interrupt_names[] = "alarm\0sec"; + char *name = g_strdup_printf("/rtc@%x", MM_PMC_RTC); + + qemu_fdt_add_subnode(s->fdt, name); + + qemu_fdt_setprop_cells(s->fdt, name, "interrupts", + GIC_FDT_IRQ_TYPE_SPI, VERSAL_RTC_ALARM_IRQ, + GIC_FDT_IRQ_FLAGS_LEVEL_HI, + GIC_FDT_IRQ_TYPE_SPI, VERSAL_RTC_SECONDS_IRQ, + GIC_FDT_IRQ_FLAGS_LEVEL_HI); + qemu_fdt_setprop(s->fdt, name, "interrupt-names", + interrupt_names, sizeof(interrupt_names)); + qemu_fdt_setprop_sized_cells(s->fdt, name, "reg", + 2, MM_PMC_RTC, 2, MM_PMC_RTC_SIZE); + qemu_fdt_setprop(s->fdt, name, "compatible", compat, sizeof(compat)); + g_free(name); +} + static void fdt_nop_memory_nodes(void *fdt, Error **errp) { Error *err = NULL; @@ -XXX,XX +XXX,XX @@ static void versal_virt_init(MachineState *machine) fdt_add_timer_nodes(s); fdt_add_zdma_nodes(s); fdt_add_sd_nodes(s); + fdt_add_rtc_node(s); fdt_add_cpu_nodes(s, psci_conduit); fdt_add_clk_node(s, "/clk125", 125000000, s->phandle.clk_125Mhz); fdt_add_clk_node(s, "/clk25", 25000000, s->phandle.clk_25Mhz); -- 2.20.1 Somewhere along theline we accidentally added a duplicate "using D16-D31 when they don't exist" check to do_vfm_dp() (probably an artifact of a patchseries rebase). Remove it. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200430181003.21682-2-peter.maydell@linaro.org --- target/arm/translate-vfp.inc.c | 6 ------ 1 file changed, 6 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 do_vfm_dp(DisasContext *s, arg_VFMA_dp *a, bool neg_n, bool neg_d) return false; } - /* UNDEF accesses to D16-D31 if they don't exist. */ - if (!dc_isar_feature(aa32_simd_r32, s) && - ((a->vd | a->vn | a->vm) & 0x10)) { - return false; - } - if (!vfp_access_check(s)) { return true; } -- 2.20.1 We were accidentally permitting decode of Thumb Neon insns even if the CPU didn't have the FEATURE_NEON bit set, because the feature check was being done before the call to disas_neon_data_insn() and disas_neon_ls_insn() in the Arm decoder but was omitted from the Thumb decoder. Push the feature bit check down into the called functions so it is done for both Arm and Thumb encodings. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Message-id: 20200430181003.21682-3-peter.maydell@linaro.org --- target/arm/translate.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) 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_ls_insn(DisasContext *s, uint32_t insn) TCGv_i32 tmp2; TCGv_i64 tmp64; + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return 1; + } + /* FIXME: this access check should not take precedence over UNDEF * for invalid encodings; we will generate incorrect syndrome information * for attempts to execute invalid vfp/neon encodings with FP disabled. @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) TCGv_ptr ptr1, ptr2, ptr3; TCGv_i64 tmp64; + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return 1; + } + /* FIXME: this access check should not take precedence over UNDEF * for invalid encodings; we will generate incorrect syndrome information * for attempts to execute invalid vfp/neon encodings with FP disabled. @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) if (((insn >> 25) & 7) == 1) { /* NEON Data processing. */ - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - goto illegal_op; - } - if (disas_neon_data_insn(s, insn)) { goto illegal_op; } @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } if ((insn & 0x0f100000) == 0x04000000) { /* NEON load/store. */ - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - goto illegal_op; - } - if (disas_neon_ls_insn(s, insn)) { goto illegal_op; } -- 2.20.1 Add the infrastructure for building and invoking a decodetree decoder for the AArch32 Neon encodings. At the moment the new decoder covers nothing, so we always fall back to the existing hand-written decode. We follow the same pattern we did for the VFP decodetree conversion (commit 78e138bc1f672c145ef6ace74617d and following): code that deals with Neon will be moving gradually out to translate-neon.vfp.inc, which we #include into translate.c. In order to share the decode files between A32 and T32, we split Neon into 3 parts: * data-processing * load-store * 'shared' encodings The first two groups of instructions have similar but not identical A32 and T32 encodings, so we need to manually transform the T32 encoding into the A32 one before calling the decoder; the third group covers the Neon instructions which are identical in A32 and T32. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-4-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 29 ++++++++++++++++++++++++++ target/arm/neon-ls.decode | 29 ++++++++++++++++++++++++++ target/arm/neon-shared.decode | 27 +++++++++++++++++++++++++ target/arm/translate-neon.inc.c | 32 +++++++++++++++++++++++++++++ target/arm/translate.c | 36 +++++++++++++++++++++++++++++++-- target/arm/Makefile.objs | 18 +++++++++++++++++ 6 files changed, 169 insertions(+), 2 deletions(-) create mode 100644 target/arm/neon-dp.decode create mode 100644 target/arm/neon-ls.decode create mode 100644 target/arm/neon-shared.decode create mode 100644 target/arm/translate-neon.inc.c diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon data-processing instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon data processing instructions where the T32 encoding +# is a simple transformation of the A32 encoding. +# More specifically, this file covers instructions where the A32 encoding is +# 0b1111_001p_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq +# and the T32 encoding is +# 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq +# This file works on the A32 encoding only; calling code for T32 has to +# transform the insn into the A32 version first. diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon load/store instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon load/store instructions where the T32 encoding +# is a simple transformation of the A32 encoding. +# More specifically, this file covers instructions where the A32 encoding is +# 0b1111_0100_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx +# and the T32 encoding is +# 0b1111_1001_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx +# This file works on the A32 encoding only; calling code for T32 has to +# transform the insn into the A32 version first. diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ +# AArch32 Neon instruction descriptions +# +# Copyright (c) 2020 Linaro, Ltd +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. + +# +# This file is processed by scripts/decodetree.py +# + +# Encodings for Neon instructions whose encoding is the same for +# both A32 and T32. + +# More specifically, this covers: +# 2reg scalar ext: 0b1111_1110_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx +# 3same ext: 0b1111_110x_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ +/* + * ARM translation: AArch32 Neon instructions + * + * Copyright (c) 2003 Fabrice Bellard + * Copyright (c) 2005-2007 CodeSourcery + * Copyright (c) 2007 OpenedHand, Ltd. + * Copyright (c) 2020 Linaro, Ltd. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +/* + * This file is intended to be included from translate.c; it uses + * some macros and definitions provided by that file. + * It might be possible to convert it to a standalone .c file eventually. + */ + +/* Include the generated Neon decoder */ +#include "decode-neon-dp.inc.c" +#include "decode-neon-ls.inc.c" +#include "decode-neon-shared.inc.c" 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 TCGv_ptr vfp_reg_ptr(bool dp, int reg) #define ARM_CP_RW_BIT (1 << 20) -/* Include the VFP decoder */ +/* Include the VFP and Neon decoders */ #include "translate-vfp.inc.c" +#include "translate-neon.inc.c" static inline void iwmmxt_load_reg(TCGv_i64 var, int reg) { @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) /* Unconditional instructions. */ /* TODO: Perhaps merge these into one decodetree output file. */ if (disas_a32_uncond(s, insn) || - disas_vfp_uncond(s, insn)) { + disas_vfp_uncond(s, insn) || + disas_neon_dp(s, insn) || + disas_neon_ls(s, insn) || + disas_neon_shared(s, insn)) { return; } /* fall back to legacy decoder */ @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) ARCH(6T2); } + if ((insn & 0xef000000) == 0xef000000) { + /* + * T32 encodings 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq + * transform into + * A32 encodings 0b1111_001p_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq + */ + uint32_t a32_insn = (insn & 0xe2ffffff) | + ((insn & (1 << 28)) >> 4) | (1 << 28); + + if (disas_neon_dp(s, a32_insn)) { + return; + } + } + + if ((insn & 0xff100000) == 0xf9000000) { + /* + * T32 encodings 0b1111_1001_ppp0_qqqq_qqqq_qqqq_qqqq_qqqq + * transform into + * A32 encodings 0b1111_0100_ppp0_qqqq_qqqq_qqqq_qqqq_qqqq + */ + uint32_t a32_insn = (insn & 0x00ffffff) | 0xf4000000; + + if (disas_neon_ls(s, a32_insn)) { + return; + } + } + /* * TODO: Perhaps merge these into one decodetree output file. * Note disas_vfp is written for a32 with cond field in the @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) */ if (disas_t32(s, insn) || disas_vfp_uncond(s, insn) || + disas_neon_shared(s, insn) || ((insn >> 28) == 0xe && disas_vfp(s, insn))) { return; } diff --git a/target/arm/Makefile.objs b/target/arm/Makefile.objs index XXXXXXX..XXXXXXX 100644 --- a/target/arm/Makefile.objs +++ b/target/arm/Makefile.objs @@ -XXX,XX +XXX,XX @@ target/arm/decode-sve.inc.c: $(SRC_PATH)/target/arm/sve.decode $(DECODETREE) $(PYTHON) $(DECODETREE) --decode disas_sve -o $@ $<,\ "GEN", $(TARGET_DIR)$@) +target/arm/decode-neon-shared.inc.c: $(SRC_PATH)/target/arm/neon-shared.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_shared -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + +target/arm/decode-neon-dp.inc.c: $(SRC_PATH)/target/arm/neon-dp.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_dp -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + +target/arm/decode-neon-ls.inc.c: $(SRC_PATH)/target/arm/neon-ls.decode $(DECODETREE) + $(call quiet-command,\ + $(PYTHON) $(DECODETREE) --static-decode disas_neon_ls -o $@ $<,\ + "GEN", $(TARGET_DIR)$@) + target/arm/decode-vfp.inc.c: $(SRC_PATH)/target/arm/vfp.decode $(DECODETREE) $(call quiet-command,\ $(PYTHON) $(DECODETREE) --static-decode disas_vfp -o $@ $<,\ @@ -XXX,XX +XXX,XX @@ target/arm/decode-t16.inc.c: $(SRC_PATH)/target/arm/t16.decode $(DECODETREE) "GEN", $(TARGET_DIR)$@) target/arm/translate-sve.o: target/arm/decode-sve.inc.c +target/arm/translate.o: target/arm/decode-neon-shared.inc.c +target/arm/translate.o: target/arm/decode-neon-dp.inc.c +target/arm/translate.o: target/arm/decode-neon-ls.inc.c target/arm/translate.o: target/arm/decode-vfp.inc.c target/arm/translate.o: target/arm/decode-vfp-uncond.inc.c target/arm/translate.o: target/arm/decode-a32.inc.c -- 2.20.1 Convert the VCMLA (vector) insns in the 3same extension group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-5-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 11 ++++++++++ target/arm/translate-neon.inc.c | 37 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 11 +--------- 3 files changed, 49 insertions(+), 10 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ # More specifically, this covers: # 2reg scalar ext: 0b1111_1110_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx # 3same ext: 0b1111_110x_xxxx_xxxx_xxxx_1x0x_xxxx_xxxx + +# VFP/Neon register fields; same as vfp.decode +%vm_dp 5:1 0:4 +%vm_sp 0:4 5:1 +%vn_dp 7:1 16:4 +%vn_sp 16:4 7:1 +%vd_dp 22:1 12:4 +%vd_sp 12:4 22:1 + +VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ #include "decode-neon-dp.inc.c" #include "decode-neon-ls.inc.c" #include "decode-neon-shared.inc.c" + +static bool trans_VCMLA(DisasContext *s, arg_VCMLA *a) +{ + int opr_sz; + TCGv_ptr fpst; + gen_helper_gvec_3_ptr *fn_gvec_ptr; + + if (!dc_isar_feature(aa32_vcma, s) + || (!a->size && !dc_isar_feature(aa32_fp16_arith, s))) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + fn_gvec_ptr = a->size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, a->rot, + fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} 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) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfe200f10) == 0xfc200800) { - /* VCMLA -- 1111 110R R.1S .... .... 1000 ...0 .... */ - int size = extract32(insn, 20, 1); - data = extract32(insn, 23, 2); /* rot */ - if (!dc_isar_feature(aa32_vcma, s) - || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { - return 1; - } - fn_gvec_ptr = size ? gen_helper_gvec_fcmlas : gen_helper_gvec_fcmlah; - } else if ((insn & 0xfea00f10) == 0xfc800800) { + if ((insn & 0xfea00f10) == 0xfc800800) { /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ int size = extract32(insn, 20, 1); data = extract32(insn, 24, 1); /* rot */ -- 2.20.1 Convert the VCADD (vector) insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-6-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 3 +++ target/arm/translate-neon.inc.c | 37 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 11 +--------- 3 files changed, 41 insertions(+), 10 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCMLA(DisasContext *s, arg_VCMLA *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VCADD(DisasContext *s, arg_VCADD *a) +{ + int opr_sz; + TCGv_ptr fpst; + gen_helper_gvec_3_ptr *fn_gvec_ptr; + + if (!dc_isar_feature(aa32_vcma, s) + || (!a->size && !dc_isar_feature(aa32_fp16_arith, s))) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + fn_gvec_ptr = a->size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, a->rot, + fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} 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) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfea00f10) == 0xfc800800) { - /* VCADD -- 1111 110R 1.0S .... .... 1000 ...0 .... */ - int size = extract32(insn, 20, 1); - data = extract32(insn, 24, 1); /* rot */ - if (!dc_isar_feature(aa32_vcma, s) - || (!size && !dc_isar_feature(aa32_fp16_arith, s))) { - return 1; - } - fn_gvec_ptr = size ? gen_helper_gvec_fcadds : gen_helper_gvec_fcaddh; - } else if ((insn & 0xfeb00f00) == 0xfc200d00) { + if ((insn & 0xfeb00f00) == 0xfc200d00) { /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ bool u = extract32(insn, 4, 1); if (!dc_isar_feature(aa32_dp, s)) { -- 2.20.1 Convert the V[US]DOT (vector) insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-7-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 4 ++++ target/arm/translate-neon.inc.c | 32 ++++++++++++++++++++++++++++++++ target/arm/translate.c | 9 +-------- 3 files changed, 37 insertions(+), 8 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA 1111 110 rot:2 . 1 size:1 .... .... 1000 . q:1 . 0 .... \ VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +# VUDOT and VSDOT +VDOT 1111 110 00 . 10 .... .... 1101 . q:1 . u:1 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCADD(DisasContext *s, arg_VCADD *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VDOT(DisasContext *s, arg_VDOT *a) +{ + int opr_sz; + gen_helper_gvec_3 *fn_gvec; + + if (!dc_isar_feature(aa32_dp, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + fn_gvec = a->u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; + tcg_gen_gvec_3_ool(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + opr_sz, opr_sz, 0, fn_gvec); + return true; +} 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) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xfeb00f00) == 0xfc200d00) { - /* V[US]DOT -- 1111 1100 0.10 .... .... 1101 .Q.U .... */ - bool u = extract32(insn, 4, 1); - if (!dc_isar_feature(aa32_dp, s)) { - return 1; - } - fn_gvec = u ? gen_helper_gvec_udot_b : gen_helper_gvec_sdot_b; - } else if ((insn & 0xff300f10) == 0xfc200810) { + if ((insn & 0xff300f10) == 0xfc200810) { /* VFM[AS]L -- 1111 1100 S.10 .... .... 1000 .Q.1 .... */ int is_s = extract32(insn, 23, 1); if (!dc_isar_feature(aa32_fhm, s)) { -- 2.20.1 Convert the VFM[AS]L (vector) insns to decodetree. This is the last insn in the legacy decoder for the 3same_ext group, so we can delete the legacy decoder function for the group entirely. Note that in disas_thumb2_insn() the parts of this encoding space where the decodetree decoder returns false will correctly be directed to illegal_op by the "(insn & (1 << 28))" check so they won't fall into disas_coproc_insn() by mistake. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-8-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 6 +++ target/arm/translate-neon.inc.c | 31 +++++++++++ target/arm/translate.c | 92 +-------------------------------- 3 files changed, 38 insertions(+), 91 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCADD 1111 110 rot:1 1 . 0 size:1 .... .... 1000 . q:1 . 0 .... \ # VUDOT and VSDOT VDOT 1111 110 00 . 10 .... .... 1101 . q:1 . u:1 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +# VFM[AS]L +VFML 1111 110 0 s:1 . 10 .... .... 1000 . 0 . 1 .... \ + vm=%vm_sp vn=%vn_sp vd=%vd_dp q=0 +VFML 1111 110 0 s:1 . 10 .... .... 1000 . 1 . 1 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp q=1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VDOT(DisasContext *s, arg_VDOT *a) opr_sz, opr_sz, 0, fn_gvec); return true; } + +static bool trans_VFML(DisasContext *s, arg_VFML *a) +{ + int opr_sz; + + if (!dc_isar_feature(aa32_fhm, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + (a->vd & 0x10)) { + return false; + } + + if (a->vd & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(a->q, a->vn), + vfp_reg_offset(a->q, a->vm), + cpu_env, opr_sz, opr_sz, a->s, /* is_2 == 0 */ + gen_helper_gvec_fmlal_a32); + return true; +} 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) return 0; } -/* Advanced SIMD three registers of the same length extension. - * 31 25 23 22 20 16 12 11 10 9 8 3 0 - * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ - * | 1 1 1 1 1 1 0 | op1 | D | op2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | - * +---------------+-----+---+-----+----+----+---+----+---+----+---------+----+ - */ -static int disas_neon_insn_3same_ext(DisasContext *s, uint32_t insn) -{ - 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; - int off_rn, off_rm; - bool is_long = false, q = extract32(insn, 6, 1); - bool ptr_is_env = false; - - if ((insn & 0xff300f10) == 0xfc200810) { - /* VFM[AS]L -- 1111 1100 S.10 .... .... 1000 .Q.1 .... */ - int is_s = extract32(insn, 23, 1); - if (!dc_isar_feature(aa32_fhm, s)) { - return 1; - } - is_long = true; - data = is_s; /* is_2 == 0 */ - fn_gvec_ptr = gen_helper_gvec_fmlal_a32; - ptr_is_env = true; - } else { - return 1; - } - - VFP_DREG_D(rd, insn); - if (rd & q) { - return 1; - } - if (q || !is_long) { - VFP_DREG_N(rn, insn); - VFP_DREG_M(rm, insn); - if ((rn | rm) & q & !is_long) { - return 1; - } - off_rn = vfp_reg_offset(1, rn); - off_rm = vfp_reg_offset(1, rm); - } else { - rn = VFP_SREG_N(insn); - rm = VFP_SREG_M(insn); - off_rn = vfp_reg_offset(0, rn); - off_rm = vfp_reg_offset(0, rm); - } - - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - if (!s->vfp_enabled) { - return 1; - } - - opr_sz = (1 + q) * 8; - if (fn_gvec_ptr) { - TCGv_ptr ptr; - if (ptr_is_env) { - ptr = cpu_env; - } else { - ptr = get_fpstatus_ptr(1); - } - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, - opr_sz, opr_sz, data, fn_gvec_ptr); - if (!ptr_is_env) { - tcg_temp_free_ptr(ptr); - } - } else { - tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, - opr_sz, opr_sz, data, fn_gvec); - } - return 0; -} - /* Advanced SIMD two registers and a scalar extension. * 31 24 23 22 20 16 12 11 10 9 8 3 0 * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } } } - } else if ((insn & 0x0e000a00) == 0x0c000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - if (disas_neon_insn_3same_ext(s, insn)) { - goto illegal_op; - } - return; } else if ((insn & 0x0f000a00) == 0x0e000800 && arm_dc_feature(s, ARM_FEATURE_V8)) { if (disas_neon_insn_2reg_scalar_ext(s, insn)) { @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; } - if ((insn & 0xfe000a00) == 0xfc000800 + if ((insn & 0xff000a00) == 0xfe000800 && arm_dc_feature(s, ARM_FEATURE_V8)) { /* The Thumb2 and ARM encodings are identical. */ - if (disas_neon_insn_3same_ext(s, insn)) { - goto illegal_op; - } - } else if ((insn & 0xff000a00) == 0xfe000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - /* The Thumb2 and ARM encodings are identical. */ if (disas_neon_insn_2reg_scalar_ext(s, insn)) { goto illegal_op; } -- 2.20.1 Convert VCMLA (scalar) in the 2reg-scalar-ext group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-9-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 5 +++++ target/arm/translate-neon.inc.c | 40 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 26 +-------------------- 3 files changed, 46 insertions(+), 25 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VFML 1111 110 0 s:1 . 10 .... .... 1000 . 0 . 1 .... \ vm=%vm_sp vn=%vn_sp vd=%vd_dp q=0 VFML 1111 110 0 s:1 . 10 .... .... 1000 . 1 . 1 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp q=1 + +VCMLA_scalar 1111 1110 0 . rot:2 .... .... 1000 . q:1 index:1 0 vm:4 \ + vn=%vn_dp vd=%vd_dp size=0 +VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp size=1 index=0 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VFML(DisasContext *s, arg_VFML *a) gen_helper_gvec_fmlal_a32); return true; } + +static bool trans_VCMLA_scalar(DisasContext *s, arg_VCMLA_scalar *a) +{ + gen_helper_gvec_3_ptr *fn_gvec_ptr; + int opr_sz; + TCGv_ptr fpst; + + if (!dc_isar_feature(aa32_vcma, s)) { + return false; + } + if (a->size == 0 && !dc_isar_feature(aa32_fp16_arith, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vd | a->vn) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn_gvec_ptr = (a->size ? gen_helper_gvec_fcmlas_idx + : gen_helper_gvec_fcmlah_idx); + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->vm), + fpst, opr_sz, opr_sz, + (a->index << 2) | a->rot, fn_gvec_ptr); + tcg_temp_free_ptr(fpst); + return true; +} 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_2reg_scalar_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xff000f10) == 0xfe000800) { - /* VCMLA (indexed) -- 1111 1110 S.RR .... .... 1000 ...0 .... */ - int rot = extract32(insn, 20, 2); - int size = extract32(insn, 23, 1); - int index; - - if (!dc_isar_feature(aa32_vcma, s)) { - return 1; - } - if (size == 0) { - if (!dc_isar_feature(aa32_fp16_arith, s)) { - 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 if ((insn & 0xffb00f00) == 0xfe200d00) { + if ((insn & 0xffb00f00) == 0xfe200d00) { /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ int u = extract32(insn, 4, 1); -- 2.20.1 Convert the V[US]DOT (scalar) insns in the 2reg-scalar-ext group to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-10-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 3 +++ target/arm/translate-neon.inc.c | 35 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 13 +----------- 3 files changed, 39 insertions(+), 12 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA_scalar 1111 1110 0 . rot:2 .... .... 1000 . q:1 index:1 0 vm:4 \ vn=%vn_dp vd=%vd_dp size=0 VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ vm=%vm_dp vn=%vn_dp vd=%vd_dp size=1 index=0 + +VDOT_scalar 1111 1110 0 . 10 .... .... 1101 . q:1 index:1 u:1 rm:4 \ + vm=%vm_dp vn=%vn_dp vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCMLA_scalar(DisasContext *s, arg_VCMLA_scalar *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VDOT_scalar(DisasContext *s, arg_VDOT_scalar *a) +{ + gen_helper_gvec_3 *fn_gvec; + int opr_sz; + TCGv_ptr fpst; + + if (!dc_isar_feature(aa32_dp, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn) & 0x10)) { + return false; + } + + if ((a->vd | a->vn) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn_gvec = a->u ? gen_helper_gvec_udot_idx_b : gen_helper_gvec_sdot_idx_b; + opr_sz = (1 + a->q) * 8; + fpst = get_fpstatus_ptr(1); + tcg_gen_gvec_3_ool(vfp_reg_offset(1, a->vd), + vfp_reg_offset(1, a->vn), + vfp_reg_offset(1, a->rm), + opr_sz, opr_sz, a->index, fn_gvec); + tcg_temp_free_ptr(fpst); + return true; +} 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_2reg_scalar_ext(DisasContext *s, uint32_t insn) bool is_long = false, q = extract32(insn, 6, 1); bool ptr_is_env = false; - if ((insn & 0xffb00f00) == 0xfe200d00) { - /* V[US]DOT -- 1111 1110 0.10 .... .... 1101 .Q.U .... */ - int u = extract32(insn, 4, 1); - - if (!dc_isar_feature(aa32_dp, s)) { - 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 if ((insn & 0xffa00f10) == 0xfe000810) { + if ((insn & 0xffa00f10) == 0xfe000810) { /* VFM[AS]L -- 1111 1110 0.0S .... .... 1000 .Q.1 .... */ int is_s = extract32(insn, 20, 1); int vm20 = extract32(insn, 0, 3); -- 2.20.1 Convert the VFM[AS]L (scalar) insns in the 2reg-scalar-ext group to decodetree. These are the last ones in the group so we can remove all the legacy decode for the group. Note that in disas_thumb2_insn() the parts of this encoding space where the decodetree decoder returns false will correctly be directed to illegal_op by the "(insn & (1 << 28))" check so they won't fall into disas_coproc_insn() by mistake. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-11-peter.maydell@linaro.org --- target/arm/neon-shared.decode | 7 +++ target/arm/translate-neon.inc.c | 32 ++++++++++ target/arm/translate.c | 107 +------------------------------- 3 files changed, 40 insertions(+), 106 deletions(-) diff --git a/target/arm/neon-shared.decode b/target/arm/neon-shared.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-shared.decode +++ b/target/arm/neon-shared.decode @@ -XXX,XX +XXX,XX @@ VCMLA_scalar 1111 1110 1 . rot:2 .... .... 1000 . q:1 . 0 .... \ VDOT_scalar 1111 1110 0 . 10 .... .... 1101 . q:1 index:1 u:1 rm:4 \ vm=%vm_dp vn=%vn_dp vd=%vd_dp + +%vfml_scalar_q0_rm 0:3 5:1 +%vfml_scalar_q1_index 5:1 3:1 +VFML_scalar 1111 1110 0 . 0 s:1 .... .... 1000 . 0 . 1 index:1 ... \ + rm=%vfml_scalar_q0_rm vn=%vn_sp vd=%vd_dp q=0 +VFML_scalar 1111 1110 0 . 0 s:1 .... .... 1000 . 1 . 1 . rm:3 \ + index=%vfml_scalar_q1_index vn=%vn_dp vd=%vd_dp q=1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VDOT_scalar(DisasContext *s, arg_VDOT_scalar *a) tcg_temp_free_ptr(fpst); return true; } + +static bool trans_VFML_scalar(DisasContext *s, arg_VFML_scalar *a) +{ + int opr_sz; + + if (!dc_isar_feature(aa32_fhm, s)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd & 0x10) || (a->q && (a->vn & 0x10)))) { + return false; + } + + if (a->vd & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + opr_sz = (1 + a->q) * 8; + tcg_gen_gvec_3_ptr(vfp_reg_offset(1, a->vd), + vfp_reg_offset(a->q, a->vn), + vfp_reg_offset(a->q, a->rm), + cpu_env, opr_sz, opr_sz, + (a->index << 2) | a->s, /* is_2 == 0 */ + gen_helper_gvec_fmlal_idx_a32); + return true; +} 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_dsp_insn(DisasContext *s, uint32_t insn) } #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n)) -#define VFP_SREG(insn, bigbit, smallbit) \ - ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1)) #define VFP_DREG(reg, insn, bigbit, smallbit) do { \ if (dc_isar_feature(aa32_simd_r32, s)) { \ reg = (((insn) >> (bigbit)) & 0x0f) \ @@ -XXX,XX +XXX,XX @@ static int disas_dsp_insn(DisasContext *s, uint32_t insn) reg = ((insn) >> (bigbit)) & 0x0f; \ }} while (0) -#define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22) #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22) -#define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7) #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7) -#define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5) #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5) static void gen_neon_dup_low16(TCGv_i32 var) @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) return 0; } -/* Advanced SIMD two registers and a scalar extension. - * 31 24 23 22 20 16 12 11 10 9 8 3 0 - * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ - * | 1 1 1 1 1 1 1 0 | o1 | D | o2 | Vn | Vd | 1 | o3 | 0 | o4 | N Q M U | Vm | - * +-----------------+----+---+----+----+----+---+----+---+----+---------+----+ - * - */ - -static int disas_neon_insn_2reg_scalar_ext(DisasContext *s, uint32_t insn) -{ - gen_helper_gvec_3 *fn_gvec = NULL; - gen_helper_gvec_3_ptr *fn_gvec_ptr = NULL; - int rd, rn, rm, opr_sz, data; - int off_rn, off_rm; - bool is_long = false, q = extract32(insn, 6, 1); - bool ptr_is_env = false; - - if ((insn & 0xffa00f10) == 0xfe000810) { - /* VFM[AS]L -- 1111 1110 0.0S .... .... 1000 .Q.1 .... */ - int is_s = extract32(insn, 20, 1); - int vm20 = extract32(insn, 0, 3); - int vm3 = extract32(insn, 3, 1); - int m = extract32(insn, 5, 1); - int index; - - if (!dc_isar_feature(aa32_fhm, s)) { - return 1; - } - if (q) { - rm = vm20; - index = m * 2 + vm3; - } else { - rm = vm20 * 2 + m; - index = vm3; - } - is_long = true; - data = (index << 2) | is_s; /* is_2 == 0 */ - fn_gvec_ptr = gen_helper_gvec_fmlal_idx_a32; - ptr_is_env = true; - } else { - return 1; - } - - VFP_DREG_D(rd, insn); - if (rd & q) { - return 1; - } - if (q || !is_long) { - VFP_DREG_N(rn, insn); - if (rn & q & !is_long) { - return 1; - } - off_rn = vfp_reg_offset(1, rn); - off_rm = vfp_reg_offset(1, rm); - } else { - rn = VFP_SREG_N(insn); - off_rn = vfp_reg_offset(0, rn); - off_rm = vfp_reg_offset(0, rm); - } - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - if (!s->vfp_enabled) { - return 1; - } - - opr_sz = (1 + q) * 8; - if (fn_gvec_ptr) { - TCGv_ptr ptr; - if (ptr_is_env) { - ptr = cpu_env; - } else { - ptr = get_fpstatus_ptr(1); - } - tcg_gen_gvec_3_ptr(vfp_reg_offset(1, rd), off_rn, off_rm, ptr, - opr_sz, opr_sz, data, fn_gvec_ptr); - if (!ptr_is_env) { - tcg_temp_free_ptr(ptr); - } - } else { - tcg_gen_gvec_3_ool(vfp_reg_offset(1, rd), off_rn, off_rm, - opr_sz, opr_sz, data, fn_gvec); - } - return 0; -} - static int disas_coproc_insn(DisasContext *s, uint32_t insn) { int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2; @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } } } - } else if ((insn & 0x0f000a00) == 0x0e000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - if (disas_neon_insn_2reg_scalar_ext(s, insn)) { - goto illegal_op; - } - return; } goto illegal_op; } @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; } - if ((insn & 0xff000a00) == 0xfe000800 - && arm_dc_feature(s, ARM_FEATURE_V8)) { - /* The Thumb2 and ARM encodings are identical. */ - if (disas_neon_insn_2reg_scalar_ext(s, insn)) { - goto illegal_op; - } - } else if (((insn >> 24) & 3) == 3) { + if (((insn >> 24) & 3) == 3) { /* Translate into the equivalent ARM encoding. */ insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4) | (1 << 28); if (disas_neon_data_insn(s, insn)) { -- 2.20.1 Convert the Neon "load/store multiple structures" insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-12-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 7 ++ target/arm/translate-neon.inc.c | 124 ++++++++++++++++++++++++++++++++ target/arm/translate.c | 91 +---------------------- 3 files changed, 133 insertions(+), 89 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ # 0b1111_1001_xxx0_xxxx_xxxx_xxxx_xxxx_xxxx # This file works on the A32 encoding only; calling code for T32 has to # transform the insn into the A32 version first. + +%vd_dp 22:1 12:4 + +# Neon load/store multiple structures + +VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ + vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VFML_scalar(DisasContext *s, arg_VFML_scalar *a) gen_helper_gvec_fmlal_idx_a32); return true; } + +static struct { + int nregs; + int interleave; + int spacing; +} const neon_ls_element_type[11] = { + {1, 4, 1}, + {1, 4, 2}, + {4, 1, 1}, + {2, 2, 2}, + {1, 3, 1}, + {1, 3, 2}, + {3, 1, 1}, + {1, 1, 1}, + {1, 2, 1}, + {1, 2, 2}, + {2, 1, 1} +}; + +static void gen_neon_ldst_base_update(DisasContext *s, int rm, int rn, + int stride) +{ + if (rm != 15) { + TCGv_i32 base; + + base = load_reg(s, rn); + if (rm == 13) { + tcg_gen_addi_i32(base, base, stride); + } else { + TCGv_i32 index; + index = load_reg(s, rm); + tcg_gen_add_i32(base, base, index); + tcg_temp_free_i32(index); + } + store_reg(s, rn, base); + } +} + +static bool trans_VLDST_multiple(DisasContext *s, arg_VLDST_multiple *a) +{ + /* Neon load/store multiple structures */ + int nregs, interleave, spacing, reg, n; + MemOp endian = s->be_data; + int mmu_idx = get_mem_index(s); + int size = a->size; + TCGv_i64 tmp64; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + if (a->itype > 10) { + return false; + } + /* Catch UNDEF cases for bad values of align field */ + switch (a->itype & 0xc) { + case 4: + if (a->align >= 2) { + return false; + } + break; + case 8: + if (a->align == 3) { + return false; + } + break; + default: + break; + } + nregs = neon_ls_element_type[a->itype].nregs; + interleave = neon_ls_element_type[a->itype].interleave; + spacing = neon_ls_element_type[a->itype].spacing; + if (size == 3 && (interleave | spacing) != 1) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + /* For our purposes, bytes are always little-endian. */ + if (size == 0) { + endian = MO_LE; + } + /* + * Consecutive little-endian elements from a single register + * can be promoted to a larger little-endian operation. + */ + if (interleave == 1 && endian == MO_LE) { + size = 3; + } + tmp64 = tcg_temp_new_i64(); + addr = tcg_temp_new_i32(); + tmp = tcg_const_i32(1 << size); + load_reg_var(s, addr, a->rn); + for (reg = 0; reg < nregs; reg++) { + for (n = 0; n < 8 >> size; n++) { + int xs; + for (xs = 0; xs < interleave; xs++) { + int tt = a->vd + reg + spacing * xs; + + if (a->l) { + gen_aa32_ld_i64(s, tmp64, addr, mmu_idx, endian | size); + neon_store_element64(tt, n, size, tmp64); + } else { + neon_load_element64(tmp64, tt, n, size); + gen_aa32_st_i64(s, tmp64, addr, mmu_idx, endian | size); + } + tcg_gen_add_i32(addr, addr, tmp); + } + } + } + tcg_temp_free_i32(addr); + tcg_temp_free_i32(tmp); + tcg_temp_free_i64(tmp64); + + gen_neon_ldst_base_update(s, a->rm, a->rn, nregs * interleave * 8); + return true; +} 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 void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1) } -static struct { - int nregs; - int interleave; - int spacing; -} const neon_ls_element_type[11] = { - {1, 4, 1}, - {1, 4, 2}, - {4, 1, 1}, - {2, 2, 2}, - {1, 3, 1}, - {1, 3, 2}, - {3, 1, 1}, - {1, 1, 1}, - {1, 2, 1}, - {1, 2, 2}, - {2, 1, 1} -}; - /* Translate a NEON load/store element instruction. Return nonzero if the instruction is invalid. */ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) { int rd, rn, rm; - int op; int nregs; - int interleave; - int spacing; int stride; int size; int reg; int load; - int n; int vec_size; - int mmu_idx; - MemOp endian; TCGv_i32 addr; TCGv_i32 tmp; - TCGv_i32 tmp2; - TCGv_i64 tmp64; if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { return 1; @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) rn = (insn >> 16) & 0xf; rm = insn & 0xf; load = (insn & (1 << 21)) != 0; - endian = s->be_data; - mmu_idx = get_mem_index(s); if ((insn & (1 << 23)) == 0) { - /* Load store all elements. */ - op = (insn >> 8) & 0xf; - size = (insn >> 6) & 3; - if (op > 10) - return 1; - /* Catch UNDEF cases for bad values of align field */ - switch (op & 0xc) { - case 4: - if (((insn >> 5) & 1) == 1) { - return 1; - } - break; - case 8: - if (((insn >> 4) & 3) == 3) { - return 1; - } - break; - default: - break; - } - nregs = neon_ls_element_type[op].nregs; - interleave = neon_ls_element_type[op].interleave; - spacing = neon_ls_element_type[op].spacing; - if (size == 3 && (interleave | spacing) != 1) { - return 1; - } - /* For our purposes, bytes are always little-endian. */ - if (size == 0) { - endian = MO_LE; - } - /* Consecutive little-endian elements from a single register - * can be promoted to a larger little-endian operation. - */ - if (interleave == 1 && endian == MO_LE) { - size = 3; - } - tmp64 = tcg_temp_new_i64(); - addr = tcg_temp_new_i32(); - tmp2 = tcg_const_i32(1 << size); - load_reg_var(s, addr, rn); - for (reg = 0; reg < nregs; reg++) { - for (n = 0; n < 8 >> size; n++) { - int xs; - for (xs = 0; xs < interleave; xs++) { - int tt = rd + reg + spacing * xs; - - if (load) { - gen_aa32_ld_i64(s, tmp64, addr, mmu_idx, endian | size); - neon_store_element64(tt, n, size, tmp64); - } else { - neon_load_element64(tmp64, tt, n, size); - gen_aa32_st_i64(s, tmp64, addr, mmu_idx, endian | size); - } - tcg_gen_add_i32(addr, addr, tmp2); - } - } - } - tcg_temp_free_i32(addr); - tcg_temp_free_i32(tmp2); - tcg_temp_free_i64(tmp64); - stride = nregs * interleave * 8; + /* Load store all elements -- handled already by decodetree */ + return 1; } else { size = (insn >> 10) & 3; if (size == 3) { -- 2.20.1 Convert the Neon "load single structure to all lanes" insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-13-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 5 +++ target/arm/translate-neon.inc.c | 73 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 55 +------------------------ 3 files changed, 80 insertions(+), 53 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ vd=%vd_dp + +# Neon load single element to all lanes + +VLD_all_lanes 1111 0100 1 . 1 0 rn:4 .... 11 n:2 size:2 t:1 a:1 rm:4 \ + vd=%vd_dp diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VLDST_multiple(DisasContext *s, arg_VLDST_multiple *a) gen_neon_ldst_base_update(s, a->rm, a->rn, nregs * interleave * 8); return true; } + +static bool trans_VLD_all_lanes(DisasContext *s, arg_VLD_all_lanes *a) +{ + /* Neon load single structure to all lanes */ + int reg, stride, vec_size; + int vd = a->vd; + int size = a->size; + int nregs = a->n + 1; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + + if (size == 3) { + if (nregs != 4 || a->a == 0) { + return false; + } + /* For VLD4 size == 3 a == 1 means 32 bits at 16 byte alignment */ + size = 2; + } + if (nregs == 1 && a->a == 1 && size == 0) { + return false; + } + if (nregs == 3 && a->a == 1) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + /* + * VLD1 to all lanes: T bit indicates how many Dregs to write. + * VLD2/3/4 to all lanes: T bit indicates register stride. + */ + stride = a->t ? 2 : 1; + vec_size = nregs == 1 ? stride * 8 : 8; + + tmp = tcg_temp_new_i32(); + addr = tcg_temp_new_i32(); + load_reg_var(s, addr, a->rn); + for (reg = 0; reg < nregs; reg++) { + gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), + s->be_data | size); + if ((vd & 1) && vec_size == 16) { + /* + * We cannot write 16 bytes at once because the + * destination is unaligned. + */ + tcg_gen_gvec_dup_i32(size, neon_reg_offset(vd, 0), + 8, 8, tmp); + tcg_gen_gvec_mov(0, neon_reg_offset(vd + 1, 0), + neon_reg_offset(vd, 0), 8, 8); + } else { + tcg_gen_gvec_dup_i32(size, neon_reg_offset(vd, 0), + vec_size, vec_size, tmp); + } + tcg_gen_addi_i32(addr, addr, 1 << size); + vd += stride; + } + tcg_temp_free_i32(tmp); + tcg_temp_free_i32(addr); + + gen_neon_ldst_base_update(s, a->rm, a->rn, (1 << size) * nregs); + + return true; +} 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_ls_insn(DisasContext *s, uint32_t insn) int size; int reg; int load; - int vec_size; TCGv_i32 addr; TCGv_i32 tmp; @@ -XXX,XX +XXX,XX @@ static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) } else { size = (insn >> 10) & 3; if (size == 3) { - /* Load single element to all lanes. */ - int a = (insn >> 4) & 1; - if (!load) { - return 1; - } - size = (insn >> 6) & 3; - nregs = ((insn >> 8) & 3) + 1; - - if (size == 3) { - if (nregs != 4 || a == 0) { - return 1; - } - /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */ - size = 2; - } - if (nregs == 1 && a == 1 && size == 0) { - return 1; - } - if (nregs == 3 && a == 1) { - return 1; - } - addr = tcg_temp_new_i32(); - load_reg_var(s, addr, rn); - - /* VLD1 to all lanes: bit 5 indicates how many Dregs to write. - * VLD2/3/4 to all lanes: bit 5 indicates register stride. - */ - stride = (insn & (1 << 5)) ? 2 : 1; - vec_size = nregs == 1 ? stride * 8 : 8; - - tmp = tcg_temp_new_i32(); - for (reg = 0; reg < nregs; reg++) { - gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - if ((rd & 1) && vec_size == 16) { - /* We cannot write 16 bytes at once because the - * destination is unaligned. - */ - tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), - 8, 8, tmp); - tcg_gen_gvec_mov(0, neon_reg_offset(rd + 1, 0), - neon_reg_offset(rd, 0), 8, 8); - } else { - tcg_gen_gvec_dup_i32(size, neon_reg_offset(rd, 0), - vec_size, vec_size, tmp); - } - tcg_gen_addi_i32(addr, addr, 1 << size); - rd += stride; - } - tcg_temp_free_i32(tmp); - tcg_temp_free_i32(addr); - stride = (1 << size) * nregs; + /* Load single element to all lanes -- handled by decodetree */ + return 1; } else { /* Single element. */ int idx = (insn >> 4) & 0xf; -- 2.20.1 Convert the Neon "load/store single structure to one lane" insns to decodetree. As this is the last set of insns in the neon load/store group, we can remove the whole disas_neon_ls_insn() function. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-14-peter.maydell@linaro.org --- target/arm/neon-ls.decode | 11 +++ target/arm/translate-neon.inc.c | 89 +++++++++++++++++++ target/arm/translate.c | 147 -------------------------------- 3 files changed, 100 insertions(+), 147 deletions(-) diff --git a/target/arm/neon-ls.decode b/target/arm/neon-ls.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-ls.decode +++ b/target/arm/neon-ls.decode @@ -XXX,XX +XXX,XX @@ VLDST_multiple 1111 0100 0 . l:1 0 rn:4 .... itype:4 size:2 align:2 rm:4 \ VLD_all_lanes 1111 0100 1 . 1 0 rn:4 .... 11 n:2 size:2 t:1 a:1 rm:4 \ vd=%vd_dp + +# Neon load/store single structure to one lane +%imm1_5_p1 5:1 !function=plus1 +%imm1_6_p1 6:1 !function=plus1 + +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 00 n:2 reg_idx:3 align:1 rm:4 \ + vd=%vd_dp size=0 stride=1 +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 01 n:2 reg_idx:2 align:2 rm:4 \ + vd=%vd_dp size=1 stride=%imm1_5_p1 +VLDST_single 1111 0100 1 . l:1 0 rn:4 .... 10 n:2 reg_idx:1 align:3 rm:4 \ + vd=%vd_dp size=2 stride=%imm1_6_p1 diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ * It might be possible to convert it to a standalone .c file eventually. */ +static inline int plus1(DisasContext *s, int x) +{ + return x + 1; +} + /* Include the generated Neon decoder */ #include "decode-neon-dp.inc.c" #include "decode-neon-ls.inc.c" @@ -XXX,XX +XXX,XX @@ static bool trans_VLD_all_lanes(DisasContext *s, arg_VLD_all_lanes *a) return true; } + +static bool trans_VLDST_single(DisasContext *s, arg_VLDST_single *a) +{ + /* Neon load/store single structure to one lane */ + int reg; + int nregs = a->n + 1; + int vd = a->vd; + TCGv_i32 addr, tmp; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist */ + if (!dc_isar_feature(aa32_simd_r32, s) && (a->vd & 0x10)) { + return false; + } + + /* Catch the UNDEF cases. This is unavoidably a bit messy. */ + switch (nregs) { + case 1: + if (((a->align & (1 << a->size)) != 0) || + (a->size == 2 && ((a->align & 3) == 1 || (a->align & 3) == 2))) { + return false; + } + break; + case 3: + if ((a->align & 1) != 0) { + return false; + } + /* fall through */ + case 2: + if (a->size == 2 && (a->align & 2) != 0) { + return false; + } + break; + case 4: + if ((a->size == 2) && ((a->align & 3) == 3)) { + return false; + } + break; + default: + abort(); + } + if ((vd + a->stride * (nregs - 1)) > 31) { + /* + * Attempts to write off the end of the register file are + * UNPREDICTABLE; we choose to UNDEF because otherwise we would + * access off the end of the array that holds the register data. + */ + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + tmp = tcg_temp_new_i32(); + addr = tcg_temp_new_i32(); + load_reg_var(s, addr, a->rn); + /* + * TODO: if we implemented alignment exceptions, we should check + * addr against the alignment encoded in a->align here. + */ + for (reg = 0; reg < nregs; reg++) { + if (a->l) { + gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), + s->be_data | a->size); + neon_store_element(vd, a->reg_idx, a->size, tmp); + } else { /* Store */ + neon_load_element(tmp, vd, a->reg_idx, a->size); + gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), + s->be_data | a->size); + } + vd += a->stride; + tcg_gen_addi_i32(addr, addr, 1 << a->size); + } + tcg_temp_free_i32(addr); + tcg_temp_free_i32(tmp); + + gen_neon_ldst_base_update(s, a->rm, a->rn, (1 << a->size) * nregs); + + return true; +} 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 void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1) tcg_temp_free_i32(rd); } - -/* Translate a NEON load/store element instruction. Return nonzero if the - instruction is invalid. */ -static int disas_neon_ls_insn(DisasContext *s, uint32_t insn) -{ - int rd, rn, rm; - int nregs; - int stride; - int size; - int reg; - int load; - TCGv_i32 addr; - TCGv_i32 tmp; - - if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { - return 1; - } - - /* FIXME: this access check should not take precedence over UNDEF - * for invalid encodings; we will generate incorrect syndrome information - * for attempts to execute invalid vfp/neon encodings with FP disabled. - */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_simd_access_trap(1, 0xe, false), s->fp_excp_el); - return 0; - } - - if (!s->vfp_enabled) - return 1; - VFP_DREG_D(rd, insn); - rn = (insn >> 16) & 0xf; - rm = insn & 0xf; - load = (insn & (1 << 21)) != 0; - if ((insn & (1 << 23)) == 0) { - /* Load store all elements -- handled already by decodetree */ - return 1; - } else { - size = (insn >> 10) & 3; - if (size == 3) { - /* Load single element to all lanes -- handled by decodetree */ - return 1; - } else { - /* Single element. */ - int idx = (insn >> 4) & 0xf; - int reg_idx; - switch (size) { - case 0: - reg_idx = (insn >> 5) & 7; - stride = 1; - break; - case 1: - reg_idx = (insn >> 6) & 3; - stride = (insn & (1 << 5)) ? 2 : 1; - break; - case 2: - reg_idx = (insn >> 7) & 1; - stride = (insn & (1 << 6)) ? 2 : 1; - break; - default: - abort(); - } - nregs = ((insn >> 8) & 3) + 1; - /* Catch the UNDEF cases. This is unavoidably a bit messy. */ - switch (nregs) { - case 1: - if (((idx & (1 << size)) != 0) || - (size == 2 && ((idx & 3) == 1 || (idx & 3) == 2))) { - return 1; - } - break; - case 3: - if ((idx & 1) != 0) { - return 1; - } - /* fall through */ - case 2: - if (size == 2 && (idx & 2) != 0) { - return 1; - } - break; - case 4: - if ((size == 2) && ((idx & 3) == 3)) { - return 1; - } - break; - default: - abort(); - } - if ((rd + stride * (nregs - 1)) > 31) { - /* Attempts to write off the end of the register file - * are UNPREDICTABLE; we choose to UNDEF because otherwise - * the neon_load_reg() would write off the end of the array. - */ - return 1; - } - tmp = tcg_temp_new_i32(); - addr = tcg_temp_new_i32(); - load_reg_var(s, addr, rn); - for (reg = 0; reg < nregs; reg++) { - if (load) { - gen_aa32_ld_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - neon_store_element(rd, reg_idx, size, tmp); - } else { /* Store */ - neon_load_element(tmp, rd, reg_idx, size); - gen_aa32_st_i32(s, tmp, addr, get_mem_index(s), - s->be_data | size); - } - rd += stride; - tcg_gen_addi_i32(addr, addr, 1 << size); - } - tcg_temp_free_i32(addr); - tcg_temp_free_i32(tmp); - stride = nregs * (1 << size); - } - } - if (rm != 15) { - TCGv_i32 base; - - base = load_reg(s, rn); - if (rm == 13) { - tcg_gen_addi_i32(base, base, stride); - } else { - TCGv_i32 index; - index = load_reg(s, rm); - tcg_gen_add_i32(base, base, index); - tcg_temp_free_i32(index); - } - store_reg(s, rn, base); - } - return 0; -} - static inline void gen_neon_narrow(int size, TCGv_i32 dest, TCGv_i64 src) { switch (size) { @@ -XXX,XX +XXX,XX @@ static void disas_arm_insn(DisasContext *s, unsigned int insn) } return; } - if ((insn & 0x0f100000) == 0x04000000) { - /* NEON load/store. */ - if (disas_neon_ls_insn(s, insn)) { - goto illegal_op; - } - return; - } if ((insn & 0x0e000f00) == 0x0c000100) { if (arm_dc_feature(s, ARM_FEATURE_IWMMXT)) { /* iWMMXt register transfer. */ @@ -XXX,XX +XXX,XX @@ static void disas_thumb2_insn(DisasContext *s, uint32_t insn) } break; case 12: - if ((insn & 0x01100000) == 0x01000000) { - if (disas_neon_ls_insn(s, insn)) { - goto illegal_op; - } - break; - } goto illegal_op; default: illegal_op: -- 2.20.1 Convert the Neon 3-reg-same VADD and VSUB insns to decodetree. Note that we don't need the neon_3r_sizes[op] check here because all size values are OK for VADD and VSUB; we'll add this when we convert the first insn that has size restrictions. For this we need one of the GVecGen*Fn typedefs currently in translate-a64.h; move them all to translate.h as a block so they are visible to the 32-bit decoder. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-15-peter.maydell@linaro.org --- target/arm/translate-a64.h | 9 -------- target/arm/translate.h | 9 ++++++++ target/arm/neon-dp.decode | 17 +++++++++++++++ target/arm/translate-neon.inc.c | 38 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 14 ++++-------- 5 files changed, 68 insertions(+), 19 deletions(-) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline int vec_full_reg_size(DisasContext *s) bool disas_sve(DisasContext *, uint32_t); -/* Note that the gvec expanders operate on offsets + sizes. */ -typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t); -typedef void GVecGen2iFn(unsigned, uint32_t, uint32_t, int64_t, - uint32_t, uint32_t); -typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, - uint32_t, uint32_t, uint32_t); -typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, - uint32_t, uint32_t, uint32_t); - #endif /* TARGET_ARM_TRANSLATE_A64_H */ diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ void gen_sshl_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); #define dc_isar_feature(name, ctx) \ ({ DisasContext *ctx_ = (ctx); isar_feature_##name(ctx_->isar); }) +/* Note that the gvec expanders operate on offsets + sizes. */ +typedef void GVecGen2Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t); +typedef void GVecGen2iFn(unsigned, uint32_t, uint32_t, int64_t, + uint32_t, uint32_t); +typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, + uint32_t, uint32_t, uint32_t); +typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, + uint32_t, uint32_t, uint32_t); + #endif /* TARGET_ARM_TRANSLATE_H */ diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ # # This file is processed by scripts/decodetree.py # +# VFP/Neon register fields; same as vfp.decode +%vm_dp 5:1 0:4 +%vn_dp 7:1 16:4 +%vd_dp 22:1 12:4 # Encodings for Neon data processing instructions where the T32 encoding # is a simple transformation of the A32 encoding. @@ -XXX,XX +XXX,XX @@ # 0b111p_1111_qqqq_qqqq_qqqq_qqqq_qqqq_qqqq # This file works on the A32 encoding only; calling code for T32 has to # transform the insn into the A32 version first. + +###################################################################### +# 3-reg-same grouping: +# 1111 001 U 0 D sz:2 Vn:4 Vd:4 opc:4 N Q M op Vm:4 +###################################################################### + +&3same vm vn vd q size + +@3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ + &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp + +VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same +VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool trans_VLDST_single(DisasContext *s, arg_VLDST_single *a) return true; } + +static bool do_3same(DisasContext *s, arg_3same *a, GVecGen3Fn fn) +{ + int vec_size = a->q ? 16 : 8; + int rd_ofs = neon_reg_offset(a->vd, 0); + int rn_ofs = neon_reg_offset(a->vn, 0); + int rm_ofs = neon_reg_offset(a->vm, 0); + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if ((a->vn | a->vm | a->vd) & a->q) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + fn(a->size, rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); + return true; +} + +#define DO_3SAME(INSN, FUNC) \ + static bool trans_##INSN##_3s(DisasContext *s, arg_3same *a) \ + { \ + return do_3same(s, a, FUNC); \ + } + +DO_3SAME(VADD, tcg_gen_gvec_add) +DO_3SAME(VSUB, tcg_gen_gvec_sub) 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) } return 0; - case NEON_3R_VADD_VSUB: - if (u) { - tcg_gen_gvec_sub(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_add(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VQADD: tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), rn_ofs, rm_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, u ? &ushl_op[size] : &sshl_op[size]); return 0; + + case NEON_3R_VADD_VSUB: + /* Already handled by decodetree */ + return 1; } if (size == 3) { -- 2.20.1 Convert the Neon logic ops in the 3-reg-same grouping to decodetree. Note that for the logic ops the 'size' field forms part of their decode and the actual operations are always bitwise. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-16-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 12 +++++++++++ target/arm/translate-neon.inc.c | 19 +++++++++++++++++ target/arm/translate.c | 38 +-------------------------------- 3 files changed, 32 insertions(+), 37 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ @3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp +@3same_logic .... ... . . . .. .... .... .... . q:1 .. .... \ + &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp size=0 + +VAND_3s 1111 001 0 0 . 00 .... .... 0001 ... 1 .... @3same_logic +VBIC_3s 1111 001 0 0 . 01 .... .... 0001 ... 1 .... @3same_logic +VORR_3s 1111 001 0 0 . 10 .... .... 0001 ... 1 .... @3same_logic +VORN_3s 1111 001 0 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VEOR_3s 1111 001 1 0 . 00 .... .... 0001 ... 1 .... @3same_logic +VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic +VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic +VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic + VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static bool do_3same(DisasContext *s, arg_3same *a, GVecGen3Fn fn) DO_3SAME(VADD, tcg_gen_gvec_add) DO_3SAME(VSUB, tcg_gen_gvec_sub) +DO_3SAME(VAND, tcg_gen_gvec_and) +DO_3SAME(VBIC, tcg_gen_gvec_andc) +DO_3SAME(VORR, tcg_gen_gvec_or) +DO_3SAME(VORN, tcg_gen_gvec_orc) +DO_3SAME(VEOR, tcg_gen_gvec_xor) + +/* These insns are all gvec_bitsel but with the inputs in various orders. */ +#define DO_3SAME_BITSEL(INSN, O1, O2, O3) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_bitsel(vece, rd_ofs, O1, O2, O3, oprsz, maxsz); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_BITSEL(VBSL, rd_ofs, rn_ofs, rm_ofs) +DO_3SAME_BITSEL(VBIT, rm_ofs, rn_ofs, rd_ofs) +DO_3SAME_BITSEL(VBIF, rm_ofs, rd_ofs, rn_ofs) 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) } return 1; - case NEON_3R_LOGIC: /* Logic ops. */ - switch ((u << 2) | size) { - case 0: /* VAND */ - tcg_gen_gvec_and(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 1: /* VBIC */ - tcg_gen_gvec_andc(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 2: /* VORR */ - tcg_gen_gvec_or(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 3: /* VORN */ - tcg_gen_gvec_orc(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 4: /* VEOR */ - tcg_gen_gvec_xor(0, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 5: /* VBSL */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - break; - case 6: /* VBIT */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rn_ofs, rd_ofs, - vec_size, vec_size); - break; - case 7: /* VBIF */ - tcg_gen_gvec_bitsel(MO_8, rd_ofs, rm_ofs, rd_ofs, rn_ofs, - vec_size, vec_size); - break; - } - return 0; - case NEON_3R_VQADD: tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), rn_ofs, rm_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) return 0; case NEON_3R_VADD_VSUB: + case NEON_3R_LOGIC: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon 3-reg-same VMAX and VMIN insns to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-17-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 5 +++++ target/arm/translate-neon.inc.c | 14 ++++++++++++++ target/arm/translate.c | 21 ++------------------- 3 files changed, 21 insertions(+), 19 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same +VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same +VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same +VMIN_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 1 .... @3same + VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME(VEOR, tcg_gen_gvec_xor) DO_3SAME_BITSEL(VBSL, rd_ofs, rn_ofs, rm_ofs) DO_3SAME_BITSEL(VBIT, rm_ofs, rn_ofs, rd_ofs) DO_3SAME_BITSEL(VBIF, rm_ofs, rd_ofs, rn_ofs) + +#define DO_3SAME_NO_SZ_3(INSN, FUNC) \ + static bool trans_##INSN##_3s(DisasContext *s, arg_3same *a) \ + { \ + if (a->size == 3) { \ + return false; \ + } \ + return do_3same(s, a, FUNC); \ + } + +DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) +DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) +DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) +DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) 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) rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); return 0; - case NEON_3R_VMAX: - if (u) { - tcg_gen_gvec_umax(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_smax(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VMIN: - if (u) { - tcg_gen_gvec_umin(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { - tcg_gen_gvec_smin(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - case NEON_3R_VSHL: /* Note the operation is vshl vd,vm,vn */ tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VADD_VSUB: case NEON_3R_LOGIC: + case NEON_3R_VMAX: + case NEON_3R_VMIN: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon comparison ops in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-18-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 8 ++++++++ target/arm/translate-neon.inc.c | 22 ++++++++++++++++++++++ target/arm/translate.c | 23 +++-------------------- 3 files changed, 33 insertions(+), 20 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VCGT_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 0 .... @3same +VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same +VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same +VCGE_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 1 .... @3same + VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same @@ -XXX,XX +XXX,XX @@ VMIN_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 1 .... @3same VADD_3s 1111 001 0 0 . .. .... .... 1000 . . . 0 .... @3same VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same + +VTST_3s 1111 001 0 0 . .. .... .... 1000 . . . 1 .... @3same +VCEQ_3s 1111 001 1 0 . .. .... .... 1000 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) + +#define DO_3SAME_CMP(INSN, COND) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_cmp(COND, vece, rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz); \ + } \ + DO_3SAME_NO_SZ_3(INSN, gen_##INSN##_3s) + +DO_3SAME_CMP(VCGT_S, TCG_COND_GT) +DO_3SAME_CMP(VCGT_U, TCG_COND_GTU) +DO_3SAME_CMP(VCGE_S, TCG_COND_GE) +DO_3SAME_CMP(VCGE_U, TCG_COND_GEU) +DO_3SAME_CMP(VCEQ, TCG_COND_EQ) + +static void gen_VTST_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, + uint32_t rm_ofs, uint32_t oprsz, uint32_t maxsz) +{ + tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, &cmtst_op[vece]); +} +DO_3SAME_NO_SZ_3(VTST, gen_VTST_3s) 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) u ? &mls_op[size] : &mla_op[size]); return 0; - case NEON_3R_VTST_VCEQ: - if (u) { /* VCEQ */ - tcg_gen_gvec_cmp(TCG_COND_EQ, size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } else { /* VTST */ - tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size, &cmtst_op[size]); - } - return 0; - - case NEON_3R_VCGT: - tcg_gen_gvec_cmp(u ? TCG_COND_GTU : TCG_COND_GT, size, - rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); - return 0; - - case NEON_3R_VCGE: - tcg_gen_gvec_cmp(u ? TCG_COND_GEU : TCG_COND_GE, size, - rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size); - return 0; - case NEON_3R_VSHL: /* Note the operation is vshl vd,vm,vn */ tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_LOGIC: case NEON_3R_VMAX: case NEON_3R_VMIN: + case NEON_3R_VTST_VCEQ: + case NEON_3R_VCGT: + case NEON_3R_VCGE: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon VQADD/VQSUB insns in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-19-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 6 ++++++ target/arm/translate-neon.inc.c | 15 +++++++++++++++ target/arm/translate.c | 14 ++------------ 3 files changed, 23 insertions(+), 12 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ @3same .... ... . . . size:2 .... .... .... . q:1 . . .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp +VQADD_S_3s 1111 001 0 0 . .. .... .... 0000 . . . 1 .... @3same +VQADD_U_3s 1111 001 1 0 . .. .... .... 0000 . . . 1 .... @3same + @3same_logic .... ... . . . .. .... .... .... . q:1 .. .... \ &3same vm=%vm_dp vn=%vn_dp vd=%vd_dp size=0 @@ -XXX,XX +XXX,XX @@ VBSL_3s 1111 001 1 0 . 01 .... .... 0001 ... 1 .... @3same_logic VBIT_3s 1111 001 1 0 . 10 .... .... 0001 ... 1 .... @3same_logic VBIF_3s 1111 001 1 0 . 11 .... .... 0001 ... 1 .... @3same_logic +VQSUB_S_3s 1111 001 0 0 . .. .... .... 0010 . . . 1 .... @3same +VQSUB_U_3s 1111 001 1 0 . .. .... .... 0010 . . . 1 .... @3same + VCGT_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 0 .... @3same VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ static void gen_VTST_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, &cmtst_op[vece]); } DO_3SAME_NO_SZ_3(VTST, gen_VTST_3s) + +#define DO_3SAME_GVEC4(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), \ + rn_ofs, rm_ofs, oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_GVEC4(VQADD_S, sqadd_op) +DO_3SAME_GVEC4(VQADD_U, uqadd_op) +DO_3SAME_GVEC4(VQSUB_S, sqsub_op) +DO_3SAME_GVEC4(VQSUB_U, uqsub_op) 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) } return 1; - case NEON_3R_VQADD: - tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), - rn_ofs, rm_ofs, vec_size, vec_size, - (u ? uqadd_op : sqadd_op) + size); - return 0; - - case NEON_3R_VQSUB: - tcg_gen_gvec_4(rd_ofs, offsetof(CPUARMState, vfp.qc), - rn_ofs, rm_ofs, vec_size, vec_size, - (u ? uqsub_op : sqsub_op) + size); - return 0; - case NEON_3R_VMUL: /* VMUL */ if (u) { /* Polynomial case allows only P8. */ @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VTST_VCEQ: case NEON_3R_VCGT: case NEON_3R_VCGE: + case NEON_3R_VQADD: + case NEON_3R_VQSUB: /* Already handled by decodetree */ return 1; } -- 2.20.1 Convert the Neon VMUL, VMLA, VMLS and VSHL insns in the 3-reg-same grouping to decodetree. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-20-peter.maydell@linaro.org --- target/arm/neon-dp.decode | 9 +++++++ target/arm/translate-neon.inc.c | 44 +++++++++++++++++++++++++++++++++ target/arm/translate.c | 28 +++------------------ 3 files changed, 56 insertions(+), 25 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -XXX,XX +XXX,XX @@ VCGT_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 0 .... @3same VCGE_S_3s 1111 001 0 0 . .. .... .... 0011 . . . 1 .... @3same VCGE_U_3s 1111 001 1 0 . .. .... .... 0011 . . . 1 .... @3same +VSHL_S_3s 1111 001 0 0 . .. .... .... 0100 . . . 0 .... @3same +VSHL_U_3s 1111 001 1 0 . .. .... .... 0100 . . . 0 .... @3same + VMAX_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 0 .... @3same VMAX_U_3s 1111 001 1 0 . .. .... .... 0110 . . . 0 .... @3same VMIN_S_3s 1111 001 0 0 . .. .... .... 0110 . . . 1 .... @3same @@ -XXX,XX +XXX,XX @@ VSUB_3s 1111 001 1 0 . .. .... .... 1000 . . . 0 .... @3same VTST_3s 1111 001 0 0 . .. .... .... 1000 . . . 1 .... @3same VCEQ_3s 1111 001 1 0 . .. .... .... 1000 . . . 1 .... @3same + +VMLA_3s 1111 001 0 0 . .. .... .... 1001 . . . 0 .... @3same +VMLS_3s 1111 001 1 0 . .. .... .... 1001 . . . 0 .... @3same + +VMUL_3s 1111 001 0 0 . .. .... .... 1001 . . . 1 .... @3same +VMUL_p_3s 1111 001 1 0 . .. .... .... 1001 . . . 1 .... @3same diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -XXX,XX +XXX,XX @@ DO_3SAME_NO_SZ_3(VMAX_S, tcg_gen_gvec_smax) DO_3SAME_NO_SZ_3(VMAX_U, tcg_gen_gvec_umax) DO_3SAME_NO_SZ_3(VMIN_S, tcg_gen_gvec_smin) DO_3SAME_NO_SZ_3(VMIN_U, tcg_gen_gvec_umin) +DO_3SAME_NO_SZ_3(VMUL, tcg_gen_gvec_mul) #define DO_3SAME_CMP(INSN, COND) \ static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ @@ -XXX,XX +XXX,XX @@ DO_3SAME_GVEC4(VQADD_S, sqadd_op) DO_3SAME_GVEC4(VQADD_U, uqadd_op) DO_3SAME_GVEC4(VQSUB_S, sqsub_op) DO_3SAME_GVEC4(VQSUB_U, uqsub_op) + +static void gen_VMUL_p_3s(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, + uint32_t rm_ofs, uint32_t oprsz, uint32_t maxsz) +{ + tcg_gen_gvec_3_ool(rd_ofs, rn_ofs, rm_ofs, oprsz, maxsz, + 0, gen_helper_gvec_pmul_b); +} + +static bool trans_VMUL_p_3s(DisasContext *s, arg_3same *a) +{ + if (a->size != 0) { + return false; + } + return do_3same(s, a, gen_VMUL_p_3s); +} + +#define DO_3SAME_GVEC3_NO_SZ_3(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, \ + oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME_NO_SZ_3(INSN, gen_##INSN##_3s) + + +DO_3SAME_GVEC3_NO_SZ_3(VMLA, mla_op) +DO_3SAME_GVEC3_NO_SZ_3(VMLS, mls_op) + +#define DO_3SAME_GVEC3_SHIFT(INSN, OPARRAY) \ + static void gen_##INSN##_3s(unsigned vece, uint32_t rd_ofs, \ + uint32_t rn_ofs, uint32_t rm_ofs, \ + uint32_t oprsz, uint32_t maxsz) \ + { \ + /* Note the operation is vshl vd,vm,vn */ \ + tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, \ + oprsz, maxsz, &OPARRAY[vece]); \ + } \ + DO_3SAME(INSN, gen_##INSN##_3s) + +DO_3SAME_GVEC3_SHIFT(VSHL_S, sshl_op) +DO_3SAME_GVEC3_SHIFT(VSHL_U, ushl_op) 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) } return 1; - case NEON_3R_VMUL: /* VMUL */ - if (u) { - /* Polynomial case allows only P8. */ - if (size != 0) { - return 1; - } - tcg_gen_gvec_3_ool(rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size, - 0, gen_helper_gvec_pmul_b); - } else { - tcg_gen_gvec_mul(size, rd_ofs, rn_ofs, rm_ofs, - vec_size, vec_size); - } - return 0; - - case NEON_3R_VML: /* VMLA, VMLS */ - tcg_gen_gvec_3(rd_ofs, rn_ofs, rm_ofs, vec_size, vec_size, - u ? &mls_op[size] : &mla_op[size]); - return 0; - - case NEON_3R_VSHL: - /* Note the operation is vshl vd,vm,vn */ - tcg_gen_gvec_3(rd_ofs, rm_ofs, rn_ofs, vec_size, vec_size, - u ? &ushl_op[size] : &sshl_op[size]); - return 0; - case NEON_3R_VADD_VSUB: case NEON_3R_LOGIC: case NEON_3R_VMAX: @@ -XXX,XX +XXX,XX @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) case NEON_3R_VCGE: case NEON_3R_VQADD: case NEON_3R_VQSUB: + case NEON_3R_VMUL: + case NEON_3R_VML: + case NEON_3R_VSHL: /* Already handled by decodetree */ return 1; } -- 2.20.1 We're going to want at least some of the NeonGen* typedefs for the refactored 32-bit Neon decoder, so move them all to translate.h since it makes more sense to keep them in one group. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20200430181003.21682-23-peter.maydell@linaro.org --- target/arm/translate.h | 17 +++++++++++++++++ target/arm/translate-a64.c | 17 ----------------- 2 files changed, 17 insertions(+), 17 deletions(-) diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ typedef void GVecGen3Fn(unsigned, uint32_t, uint32_t, typedef void GVecGen4Fn(unsigned, uint32_t, uint32_t, uint32_t, uint32_t, uint32_t, uint32_t); +/* Function prototype for gen_ functions for calling Neon helpers */ +typedef void NeonGenOneOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32); +typedef void NeonGenTwoOpFn(TCGv_i32, TCGv_i32, TCGv_i32); +typedef void NeonGenTwoOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32); +typedef void NeonGenTwo64OpFn(TCGv_i64, TCGv_i64, TCGv_i64); +typedef void NeonGenTwo64OpEnvFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i64); +typedef void NeonGenNarrowFn(TCGv_i32, TCGv_i64); +typedef void NeonGenNarrowEnvFn(TCGv_i32, TCGv_ptr, TCGv_i64); +typedef void NeonGenWidenFn(TCGv_i64, TCGv_i32); +typedef void NeonGenTwoSingleOPFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr); +typedef void NeonGenTwoDoubleOPFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_ptr); +typedef void NeonGenOneOpFn(TCGv_i64, TCGv_i64); +typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr); +typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32); +typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); +typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, MemOp); + #endif /* TARGET_ARM_TRANSLATE_H */ 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 @@ typedef struct AArch64DecodeTable { AArch64DecodeFn *disas_fn; } AArch64DecodeTable; -/* Function prototype for gen_ functions for calling Neon helpers */ -typedef void NeonGenOneOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32); -typedef void NeonGenTwoOpFn(TCGv_i32, TCGv_i32, TCGv_i32); -typedef void NeonGenTwoOpEnvFn(TCGv_i32, TCGv_ptr, TCGv_i32, TCGv_i32); -typedef void NeonGenTwo64OpFn(TCGv_i64, TCGv_i64, TCGv_i64); -typedef void NeonGenTwo64OpEnvFn(TCGv_i64, TCGv_ptr, TCGv_i64, TCGv_i64); -typedef void NeonGenNarrowFn(TCGv_i32, TCGv_i64); -typedef void NeonGenNarrowEnvFn(TCGv_i32, TCGv_ptr, TCGv_i64); -typedef void NeonGenWidenFn(TCGv_i64, TCGv_i32); -typedef void NeonGenTwoSingleOPFn(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_ptr); -typedef void NeonGenTwoDoubleOPFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_ptr); -typedef void NeonGenOneOpFn(TCGv_i64, TCGv_i64); -typedef void CryptoTwoOpFn(TCGv_ptr, TCGv_ptr); -typedef void CryptoThreeOpIntFn(TCGv_ptr, TCGv_ptr, TCGv_i32); -typedef void CryptoThreeOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); -typedef void AtomicThreeOpFn(TCGv_i64, TCGv_i64, TCGv_i64, TCGArg, MemOp); - /* initialize TCG globals. */ void a64_translate_init(void) { -- 2.20.1