The following changes since commit 53f306f316549d20c76886903181413d20842423: Merge remote-tracking branch 'remotes/ehabkost-gl/tags/x86-next-pull-request' into staging (2021-06-21 11:26:04 +0100) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20210621 for you to fetch changes up to a83f1d9263d281f938a3984cda7104d55affd43a: docs/system: arm: Add nRF boards description (2021-06-21 17:24:33 +0100) ---------------------------------------------------------------- target-arm queue: * Don't require 'virt' board to be compiled in for ACPI GHES code * docs: Document which architecture extensions we emulate * Fix bugs in M-profile FPCXT_NS accesses * First slice of MVE patches * Implement MTE3 * docs/system: arm: Add nRF boards description ---------------------------------------------------------------- Alexandre Iooss (1): docs/system: arm: Add nRF boards description Peter Collingbourne (1): target/arm: Implement MTE3 Peter Maydell (55): hw/acpi: Provide stub version of acpi_ghes_record_errors() hw/acpi: Provide function acpi_ghes_present() target/arm: Use acpi_ghes_present() to see if we report ACPI memory errors docs/system/arm: Document which architecture extensions we emulate target/arm/translate-vfp.c: Whitespace fixes target/arm: Handle FPU being disabled in FPCXT_NS accesses target/arm: Don't NOCP fault for FPCXT_NS accesses target/arm: Handle writeback in VLDR/VSTR sysreg with no memory access target/arm: Factor FP context update code out into helper function target/arm: Split vfp_access_check() into A and M versions target/arm: Handle FPU check for FPCXT_NS insns via vfp_access_check_m() target/arm: Implement MVE VLDR/VSTR (non-widening forms) target/arm: Implement widening/narrowing MVE VLDR/VSTR insns target/arm: Implement MVE VCLZ target/arm: Implement MVE VCLS target/arm: Implement MVE VREV16, VREV32, VREV64 target/arm: Implement MVE VMVN (register) target/arm: Implement MVE VABS target/arm: Implement MVE VNEG tcg: Make gen_dup_i32/i64() public as tcg_gen_dup_i32/i64 target/arm: Implement MVE VDUP target/arm: Implement MVE VAND, VBIC, VORR, VORN, VEOR target/arm: Implement MVE VADD, VSUB, VMUL target/arm: Implement MVE VMULH target/arm: Implement MVE VRMULH target/arm: Implement MVE VMAX, VMIN target/arm: Implement MVE VABD target/arm: Implement MVE VHADD, VHSUB target/arm: Implement MVE VMULL target/arm: Implement MVE VMLALDAV target/arm: Implement MVE VMLSLDAV target/arm: Implement MVE VRMLALDAVH, VRMLSLDAVH target/arm: Implement MVE VADD (scalar) target/arm: Implement MVE VSUB, VMUL (scalar) target/arm: Implement MVE VHADD, VHSUB (scalar) target/arm: Implement MVE VBRSR target/arm: Implement MVE VPST target/arm: Implement MVE VQADD and VQSUB target/arm: Implement MVE VQDMULH and VQRDMULH (scalar) target/arm: Implement MVE VQDMULL scalar target/arm: Implement MVE VQDMULH, VQRDMULH (vector) target/arm: Implement MVE VQADD, VQSUB (vector) target/arm: Implement MVE VQSHL (vector) target/arm: Implement MVE VQRSHL target/arm: Implement MVE VSHL insn target/arm: Implement MVE VRSHL target/arm: Implement MVE VQDMLADH and VQRDMLADH target/arm: Implement MVE VQDMLSDH and VQRDMLSDH target/arm: Implement MVE VQDMULL (vector) target/arm: Implement MVE VRHADD target/arm: Implement MVE VADC, VSBC target/arm: Implement MVE VCADD target/arm: Implement MVE VHCADD target/arm: Implement MVE VADDV target/arm: Make VMOV scalar <-> gpreg beatwise for MVE docs/system/arm/emulation.rst | 103 ++++ docs/system/arm/nrf.rst | 51 ++ docs/system/target-arm.rst | 7 + include/hw/acpi/ghes.h | 9 + include/tcg/tcg-op.h | 8 + include/tcg/tcg.h | 1 - target/arm/helper-mve.h | 357 +++++++++++++ target/arm/helper.h | 2 + target/arm/internals.h | 11 + target/arm/translate-a32.h | 3 + target/arm/translate.h | 10 + target/arm/m-nocp.decode | 24 + target/arm/mve.decode | 240 +++++++++ target/arm/vfp.decode | 14 - hw/acpi/ghes-stub.c | 22 + hw/acpi/ghes.c | 17 + target/arm/cpu64.c | 2 +- target/arm/kvm64.c | 6 +- target/arm/mte_helper.c | 82 +-- target/arm/mve_helper.c | 1160 +++++++++++++++++++++++++++++++++++++++++ target/arm/translate-m-nocp.c | 550 +++++++++++++++++++ target/arm/translate-mve.c | 759 +++++++++++++++++++++++++++ target/arm/translate-vfp.c | 741 +++++++------------------- tcg/tcg-op-gvec.c | 20 +- MAINTAINERS | 1 + hw/acpi/meson.build | 6 +- target/arm/meson.build | 1 + 27 files changed, 3578 insertions(+), 629 deletions(-) create mode 100644 docs/system/arm/emulation.rst create mode 100644 docs/system/arm/nrf.rst create mode 100644 target/arm/helper-mve.h create mode 100644 hw/acpi/ghes-stub.c create mode 100644 target/arm/mve_helper.c Generic code in target/arm wants to call acpi_ghes_record_errors(); provide a stub version so that we don't fail to link when CONFIG_ACPI_APEI is not set. This requires us to add a new ghes-stub.c file to contain it and the meson.build mechanics to use it when appropriate. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Dongjiu Geng <gengdongjiu1@gmail.com> Message-id: 20210603171259.27962-2-peter.maydell@linaro.org --- hw/acpi/ghes-stub.c | 17 +++++++++++++++++ hw/acpi/meson.build | 6 +++--- 2 files changed, 20 insertions(+), 3 deletions(-) create mode 100644 hw/acpi/ghes-stub.c diff --git a/hw/acpi/ghes-stub.c b/hw/acpi/ghes-stub.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/hw/acpi/ghes-stub.c @@ -XXX,XX +XXX,XX @@ +/* + * Support for generating APEI tables and recording CPER for Guests: + * stub functions. + * + * Copyright (c) 2021 Linaro, Ltd + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#include "qemu/osdep.h" +#include "hw/acpi/ghes.h" + +int acpi_ghes_record_errors(uint8_t source_id, uint64_t physical_address) +{ + return -1; +} diff --git a/hw/acpi/meson.build b/hw/acpi/meson.build index XXXXXXX..XXXXXXX 100644 --- a/hw/acpi/meson.build +++ b/hw/acpi/meson.build @@ -XXX,XX +XXX,XX @@ acpi_ss.add(when: 'CONFIG_ACPI_PCI', if_true: files('pci.c')) acpi_ss.add(when: 'CONFIG_ACPI_VMGENID', if_true: files('vmgenid.c')) acpi_ss.add(when: 'CONFIG_ACPI_HW_REDUCED', if_true: files('generic_event_device.c')) acpi_ss.add(when: 'CONFIG_ACPI_HMAT', if_true: files('hmat.c')) -acpi_ss.add(when: 'CONFIG_ACPI_APEI', if_true: files('ghes.c')) +acpi_ss.add(when: 'CONFIG_ACPI_APEI', if_true: files('ghes.c'), if_false: files('ghes-stub.c')) acpi_ss.add(when: 'CONFIG_ACPI_X86', if_true: files('core.c', 'piix4.c', 'pcihp.c'), if_false: files('acpi-stub.c')) acpi_ss.add(when: 'CONFIG_ACPI_X86_ICH', if_true: files('ich9.c', 'tco.c')) acpi_ss.add(when: 'CONFIG_IPMI', if_true: files('ipmi.c'), if_false: files('ipmi-stub.c')) acpi_ss.add(when: 'CONFIG_PC', if_false: files('acpi-x86-stub.c')) acpi_ss.add(when: 'CONFIG_TPM', if_true: files('tpm.c')) -softmmu_ss.add(when: 'CONFIG_ACPI', if_false: files('acpi-stub.c', 'aml-build-stub.c')) +softmmu_ss.add(when: 'CONFIG_ACPI', if_false: files('acpi-stub.c', 'aml-build-stub.c', 'ghes-stub.c')) softmmu_ss.add_all(when: 'CONFIG_ACPI', if_true: acpi_ss) softmmu_ss.add(when: 'CONFIG_ALL', if_true: files('acpi-stub.c', 'aml-build-stub.c', - 'acpi-x86-stub.c', 'ipmi-stub.c')) + 'acpi-x86-stub.c', 'ipmi-stub.c', 'ghes-stub.c')) -- 2.20.1 Allow code elsewhere in the system to check whether the ACPI GHES table is present, so it can determine whether it is OK to try to record an error by calling acpi_ghes_record_errors(). (We don't need to migrate the new 'present' field in AcpiGhesState, because it is set once at system initialization and doesn't change.) Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Dongjiu Geng <gengdongjiu1@gmail.com> Message-id: 20210603171259.27962-3-peter.maydell@linaro.org --- include/hw/acpi/ghes.h | 9 +++++++++ hw/acpi/ghes-stub.c | 5 +++++ hw/acpi/ghes.c | 17 +++++++++++++++++ 3 files changed, 31 insertions(+) diff --git a/include/hw/acpi/ghes.h b/include/hw/acpi/ghes.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/acpi/ghes.h +++ b/include/hw/acpi/ghes.h @@ -XXX,XX +XXX,XX @@ enum { typedef struct AcpiGhesState { uint64_t ghes_addr_le; + bool present; /* True if GHES is present at all on this board */ } AcpiGhesState; void build_ghes_error_table(GArray *hardware_errors, BIOSLinker *linker); @@ -XXX,XX +XXX,XX @@ void acpi_build_hest(GArray *table_data, BIOSLinker *linker, void acpi_ghes_add_fw_cfg(AcpiGhesState *vms, FWCfgState *s, GArray *hardware_errors); int acpi_ghes_record_errors(uint8_t notify, uint64_t error_physical_addr); + +/** + * acpi_ghes_present: Report whether ACPI GHES table is present + * + * Returns: true if the system has an ACPI GHES table and it is + * safe to call acpi_ghes_record_errors() to record a memory error. + */ +bool acpi_ghes_present(void); #endif diff --git a/hw/acpi/ghes-stub.c b/hw/acpi/ghes-stub.c index XXXXXXX..XXXXXXX 100644 --- a/hw/acpi/ghes-stub.c +++ b/hw/acpi/ghes-stub.c @@ -XXX,XX +XXX,XX @@ int acpi_ghes_record_errors(uint8_t source_id, uint64_t physical_address) { return -1; } + +bool acpi_ghes_present(void) +{ + return false; +} diff --git a/hw/acpi/ghes.c b/hw/acpi/ghes.c index XXXXXXX..XXXXXXX 100644 --- a/hw/acpi/ghes.c +++ b/hw/acpi/ghes.c @@ -XXX,XX +XXX,XX @@ void acpi_ghes_add_fw_cfg(AcpiGhesState *ags, FWCfgState *s, /* Create a read-write fw_cfg file for Address */ fw_cfg_add_file_callback(s, ACPI_GHES_DATA_ADDR_FW_CFG_FILE, NULL, NULL, NULL, &(ags->ghes_addr_le), sizeof(ags->ghes_addr_le), false); + + ags->present = true; } int acpi_ghes_record_errors(uint8_t source_id, uint64_t physical_address) @@ -XXX,XX +XXX,XX @@ int acpi_ghes_record_errors(uint8_t source_id, uint64_t physical_address) return ret; } + +bool acpi_ghes_present(void) +{ + AcpiGedState *acpi_ged_state; + AcpiGhesState *ags; + + acpi_ged_state = ACPI_GED(object_resolve_path_type("", TYPE_ACPI_GED, + NULL)); + + if (!acpi_ged_state) { + return false; + } + ags = &acpi_ged_state->ghes_state; + return ags->present; +} -- 2.20.1 The virt_is_acpi_enabled() function is specific to the virt board, as is the check for its 'ras' property. Use the new acpi_ghes_present() function to check whether we should report memory errors via acpi_ghes_record_errors(). This avoids a link error if QEMU was built without support for the virt board, and provides a mechanism that can be used by any future board models that want to add ACPI memory error reporting support (they only need to call acpi_ghes_add_fw_cfg()). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Dongjiu Geng <gengdongjiu1@gmail.com> Message-id: 20210603171259.27962-4-peter.maydell@linaro.org --- target/arm/kvm64.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/kvm64.c +++ b/target/arm/kvm64.c @@ -XXX,XX +XXX,XX @@ void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr) { ram_addr_t ram_addr; hwaddr paddr; - Object *obj = qdev_get_machine(); - VirtMachineState *vms = VIRT_MACHINE(obj); - bool acpi_enabled = virt_is_acpi_enabled(vms); assert(code == BUS_MCEERR_AR || code == BUS_MCEERR_AO); - if (acpi_enabled && addr && - object_property_get_bool(obj, "ras", NULL)) { + if (acpi_ghes_present() && addr) { ram_addr = qemu_ram_addr_from_host(addr); if (ram_addr != RAM_ADDR_INVALID && kvm_physical_memory_addr_from_host(c->kvm_state, addr, &paddr)) { -- 2.20.1 These days the Arm architecture has a wide range of fine-grained optional extra architectural features. We implement quite a lot of these but by no means all of them. Document what we do implement, so that users can find out without having to dig through back-issues of our Changelog on the wiki. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-id: 20210617140328.28622-1-peter.maydell@linaro.org Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- docs/system/arm/emulation.rst | 102 ++++++++++++++++++++++++++++++++++ docs/system/target-arm.rst | 6 ++ 2 files changed, 108 insertions(+) create mode 100644 docs/system/arm/emulation.rst diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/docs/system/arm/emulation.rst @@ -XXX,XX +XXX,XX @@ +A-profile CPU architecture support +================================== + +QEMU's TCG emulation includes support for the Armv5, Armv6, Armv7 and +Armv8 versions of the A-profile architecture. It also has support for +the following architecture extensions: + +- FEAT_AA32BF16 (AArch32 BFloat16 instructions) +- FEAT_AA32HPD (AArch32 hierarchical permission disables) +- FEAT_AA32I8MM (AArch32 Int8 matrix multiplication instructions) +- FEAT_AES (AESD and AESE instructions) +- FEAT_BF16 (AArch64 BFloat16 instructions) +- FEAT_BTI (Branch Target Identification) +- FEAT_DIT (Data Independent Timing instructions) +- FEAT_DPB (DC CVAP instruction) +- FEAT_DotProd (Advanced SIMD dot product instructions) +- FEAT_FCMA (Floating-point complex number instructions) +- FEAT_FHM (Floating-point half-precision multiplication instructions) +- FEAT_FP16 (Half-precision floating-point data processing) +- FEAT_FRINTTS (Floating-point to integer instructions) +- FEAT_FlagM (Flag manipulation instructions v2) +- FEAT_FlagM2 (Enhancements to flag manipulation instructions) +- FEAT_HPDS (Hierarchical permission disables) +- FEAT_I8MM (AArch64 Int8 matrix multiplication instructions) +- FEAT_JSCVT (JavaScript conversion instructions) +- FEAT_LOR (Limited ordering regions) +- FEAT_LRCPC (Load-acquire RCpc instructions) +- FEAT_LRCPC2 (Load-acquire RCpc instructions v2) +- FEAT_LSE (Large System Extensions) +- FEAT_MTE (Memory Tagging Extension) +- FEAT_MTE2 (Memory Tagging Extension) +- FEAT_PAN (Privileged access never) +- FEAT_PAN2 (AT S1E1R and AT S1E1W instruction variants affected by PSTATE.PAN) +- FEAT_PAuth (Pointer authentication) +- FEAT_PMULL (PMULL, PMULL2 instructions) +- FEAT_PMUv3p1 (PMU Extensions v3.1) +- FEAT_PMUv3p4 (PMU Extensions v3.4) +- FEAT_RDM (Advanced SIMD rounding double multiply accumulate instructions) +- FEAT_RNG (Random number generator) +- FEAT_SB (Speculation Barrier) +- FEAT_SEL2 (Secure EL2) +- FEAT_SHA1 (SHA1 instructions) +- FEAT_SHA256 (SHA256 instructions) +- FEAT_SHA3 (Advanced SIMD SHA3 instructions) +- FEAT_SHA512 (Advanced SIMD SHA512 instructions) +- FEAT_SM3 (Advanced SIMD SM3 instructions) +- FEAT_SM4 (Advanced SIMD SM4 instructions) +- FEAT_SPECRES (Speculation restriction instructions) +- FEAT_SSBS (Speculative Store Bypass Safe) +- FEAT_TLBIOS (TLB invalidate instructions in Outer Shareable domain) +- FEAT_TLBIRANGE (TLB invalidate range instructions) +- FEAT_TTCNP (Translation table Common not private translations) +- FEAT_TTST (Small translation tables) +- FEAT_UAO (Unprivileged Access Override control) +- FEAT_VHE (Virtualization Host Extensions) +- FEAT_VMID16 (16-bit VMID) +- FEAT_XNX (Translation table stage 2 Unprivileged Execute-never) +- SVE (The Scalable Vector Extension) +- SVE2 (The Scalable Vector Extension v2) + +For information on the specifics of these extensions, please refer +to the `Armv8-A Arm Architecture Reference Manual +<https://developer.arm.com/documentation/ddi0487/latest>`_. + +When a specific named CPU is being emulated, only those features which +are present in hardware for that CPU are emulated. (If a feature is +not in the list above then it is not supported, even if the real +hardware should have it.) The ``max`` CPU enables all features. + +R-profile CPU architecture support +================================== + +QEMU's TCG emulation support for R-profile CPUs is currently limited. +We emulate only the Cortex-R5 and Cortex-R5F CPUs. + +M-profile CPU architecture support +================================== + +QEMU's TCG emulation includes support for Armv6-M, Armv7-M, Armv8-M, and +Armv8.1-M versions of the M-profile architucture. It also has support +for the following architecture extensions: + +- FP (Floating-point Extension) +- FPCXT (FPCXT access instructions) +- HP (Half-precision floating-point instructions) +- LOB (Low Overhead loops and Branch future) +- M (Main Extension) +- MPU (Memory Protection Unit Extension) +- PXN (Privileged Execute Never) +- RAS (Reliability, Serviceability and Availability): "minimum RAS Extension" only +- S (Security Extension) +- ST (System Timer Extension) + +For information on the specifics of these extensions, please refer +to the `Armv8-M Arm Architecture Reference Manual +<https://developer.arm.com/documentation/ddi0553/latest>`_. + +When a specific named CPU is being emulated, only those features which +are present in hardware for that CPU are emulated. (If a feature is +not in the list above then it is not supported, even if the real +hardware should have it.) There is no equivalent of the ``max`` CPU for +M-profile. diff --git a/docs/system/target-arm.rst b/docs/system/target-arm.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/target-arm.rst +++ b/docs/system/target-arm.rst @@ -XXX,XX +XXX,XX @@ undocumented; you can get a complete list by running arm/virt arm/xlnx-versal-virt +Emulated CPU architecture support +================================= + +.. toctree:: + arm/emulation + Arm CPU features ================ -- 2.20.1 In the code for handling VFP system register accesses there is some stray whitespace after a unary '-' operator, and also some incorrect indent in a couple of function prototypes. We're about to move this code to another file, so fix the code style issues first so checkpatch doesn't complain about the code-movement patch. Cc: qemu-stable@nongnu.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-2-peter.maydell@linaro.org --- target/arm/translate-vfp.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static void gen_branch_fpInactive(DisasContext *s, TCGCond cond, } static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, - fp_sysreg_loadfn *loadfn, - void *opaque) + void *opaque) { /* Do a write to an M-profile floating point system register */ TCGv_i32 tmp; @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, } static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, - fp_sysreg_storefn *storefn, - void *opaque) + fp_sysreg_storefn *storefn, + void *opaque) { /* Do a read from an M-profile floating point system register */ TCGv_i32 tmp; @@ -XXX,XX +XXX,XX @@ static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) TCGv_i32 addr; if (!a->a) { - offset = - offset; + offset = -offset; } addr = load_reg(s, a->rn); @@ -XXX,XX +XXX,XX @@ static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque) TCGv_i32 value = tcg_temp_new_i32(); if (!a->a) { - offset = - offset; + offset = -offset; } addr = load_reg(s, a->rn); -- 2.20.1 If the guest makes an FPCXT_NS access when the FPU is disabled, one of two things happens: * if there is no active FP context, then the insn behaves the same way as if the FPU was enabled: writes ignored, reads same value as FPDSCR_NS * if there is an active FP context, then we take a NOCP exception Add code to the sysreg read/write functions which emits code to take the NOCP exception in the latter case. At the moment this will never be used, because the NOCP checks in m-nocp.decode happen first, and so the trans functions are never called when the FPU is disabled. The code will be needed when we move the sysreg access insns to before the NOCP patterns in the following commit. Cc: qemu-stable@nongnu.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-3-peter.maydell@linaro.org --- target/arm/translate-vfp.c | 32 ++++++++++++++++++++++++++++++-- 1 file changed, 30 insertions(+), 2 deletions(-) diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, lab_end = gen_new_label(); /* fpInactive case: write is a NOP, so branch to end */ gen_branch_fpInactive(s, TCG_COND_NE, lab_end); - /* !fpInactive: PreserveFPState(), and reads same as FPCXT_S */ + /* + * !fpInactive: if FPU disabled, take NOCP exception; + * otherwise PreserveFPState(), and then FPCXT_NS writes + * behave the same as FPCXT_S writes. + */ + if (s->fp_excp_el) { + gen_exception_insn(s, s->pc_curr, EXCP_NOCP, + syn_uncategorized(), s->fp_excp_el); + /* + * This was only a conditional exception, so override + * gen_exception_insn()'s default to DISAS_NORETURN + */ + s->base.is_jmp = DISAS_NEXT; + break; + } gen_preserve_fp_state(s); /* fall through */ case ARM_VFP_FPCXT_S: @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, tcg_gen_br(lab_end); gen_set_label(lab_active); - /* !fpInactive: Reads the same as FPCXT_S, but side effects differ */ + /* + * !fpInactive: if FPU disabled, take NOCP exception; + * otherwise PreserveFPState(), and then FPCXT_NS + * reads the same as FPCXT_S. + */ + if (s->fp_excp_el) { + gen_exception_insn(s, s->pc_curr, EXCP_NOCP, + syn_uncategorized(), s->fp_excp_el); + /* + * This was only a conditional exception, so override + * gen_exception_insn()'s default to DISAS_NORETURN + */ + s->base.is_jmp = DISAS_NEXT; + break; + } gen_preserve_fp_state(s); tmp = tcg_temp_new_i32(); sfpa = tcg_temp_new_i32(); -- 2.20.1 The M-profile architecture requires that accesses to FPCXT_NS when there is no active FP state must not take a NOCP fault even if the FPU is disabled. We were not implementing this correctly, because in our decode we catch the NOCP faults early in m-nocp.decode. Fix this bug by moving all the handling of M-profile FP system register accesses from vfp.decode into m-nocp.decode and putting it above the NOCP blocks. This provides the correct behaviour: * for accesses other than FPCXT_NS the trans functions call vfp_access_check(), which will check for FPU disabled and raise a NOCP exception if necessary * for FPCXT_NS we have the special case code that doesn't call vfp_access_check() * when these trans functions want to raise an UNDEF they return false, so the decoder will fall through into the NOCP blocks. This means that NOCP correctly takes precedence over UNDEF for these insns. (This is a difference from the other insns handled by m-nocp.decode, where UNDEF takes precedence and which we implement by having those trans functions call unallocated_encoding() in the appropriate places.) [Note for backport to stable: this commit has a semantic dependency on commit 9a486856e9173af, which was not marked as cc-stable because we didn't know we'd need it for a for-stable bugfix.] Cc: qemu-stable@nongnu.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-4-peter.maydell@linaro.org --- target/arm/translate-a32.h | 1 + target/arm/m-nocp.decode | 24 ++ target/arm/vfp.decode | 14 - target/arm/translate-m-nocp.c | 514 +++++++++++++++++++++++++++++++++ target/arm/translate-vfp.c | 517 +--------------------------------- 5 files changed, 542 insertions(+), 528 deletions(-) diff --git a/target/arm/translate-a32.h b/target/arm/translate-a32.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a32.h +++ b/target/arm/translate-a32.h @@ -XXX,XX +XXX,XX @@ bool disas_neon_shared(DisasContext *s, uint32_t insn); void load_reg_var(DisasContext *s, TCGv_i32 var, int reg); void arm_gen_condlabel(DisasContext *s); bool vfp_access_check(DisasContext *s); +void gen_preserve_fp_state(DisasContext *s); void read_neon_element32(TCGv_i32 dest, int reg, int ele, MemOp memop); void read_neon_element64(TCGv_i64 dest, int reg, int ele, MemOp memop); void write_neon_element32(TCGv_i32 src, int reg, int ele, MemOp memop); diff --git a/target/arm/m-nocp.decode b/target/arm/m-nocp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/m-nocp.decode +++ b/target/arm/m-nocp.decode @@ -XXX,XX +XXX,XX @@ &nocp cp +# M-profile VLDR/VSTR to sysreg +%vldr_sysreg 22:1 13:3 +%imm7_0x4 0:7 !function=times_4 + +&vldr_sysreg rn reg imm a w p +@vldr_sysreg .... ... . a:1 . . . rn:4 ... . ... .. ....... \ + reg=%vldr_sysreg imm=%imm7_0x4 &vldr_sysreg + { # Special cases which do not take an early NOCP: VLLDM and VLSTM VLLDM_VLSTM 1110 1100 001 l:1 rn:4 0000 1010 op:1 000 0000 @@ -XXX,XX +XXX,XX @@ VSCCLRM 1110 1100 1.01 1111 .... 1011 imm:7 0 vd=%vd_dp size=3 VSCCLRM 1110 1100 1.01 1111 .... 1010 imm:8 vd=%vd_sp size=2 + # FP system register accesses: these are a special case because accesses + # to FPCXT_NS succeed even if the FPU is disabled. We therefore need + # to handle them before the big NOCP blocks. Note that within these + # insns NOCP still has higher priority than UNDEFs; this is implemented + # by their returning 'false' for UNDEF so as to fall through into the + # NOCP check (in contrast to VLLDM etc, which call unallocated_encoding() + # for the UNDEFs there that must take precedence over NOCP.) + + VMSR_VMRS ---- 1110 111 l:1 reg:4 rt:4 1010 0001 0000 + + # P=0 W=0 is SEE "Related encodings", so split into two patterns + VLDR_sysreg ---- 110 1 . . w:1 1 .... ... 0 111 11 ....... @vldr_sysreg p=1 + VLDR_sysreg ---- 110 0 . . 1 1 .... ... 0 111 11 ....... @vldr_sysreg p=0 w=1 + VSTR_sysreg ---- 110 1 . . w:1 0 .... ... 0 111 11 ....... @vldr_sysreg p=1 + VSTR_sysreg ---- 110 0 . . 1 0 .... ... 0 111 11 ....... @vldr_sysreg p=0 w=1 + NOCP 111- 1110 ---- ---- ---- cp:4 ---- ---- &nocp NOCP 111- 110- ---- ---- ---- cp:4 ---- ---- &nocp # From v8.1M onwards this range will also NOCP: diff --git a/target/arm/vfp.decode b/target/arm/vfp.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vfp.decode +++ b/target/arm/vfp.decode @@ -XXX,XX +XXX,XX @@ VLDR_VSTR_hp ---- 1101 u:1 .0 l:1 rn:4 .... 1001 imm:8 vd=%vd_sp VLDR_VSTR_sp ---- 1101 u:1 .0 l:1 rn:4 .... 1010 imm:8 vd=%vd_sp VLDR_VSTR_dp ---- 1101 u:1 .0 l:1 rn:4 .... 1011 imm:8 vd=%vd_dp -# M-profile VLDR/VSTR to sysreg -%vldr_sysreg 22:1 13:3 -%imm7_0x4 0:7 !function=times_4 - -&vldr_sysreg rn reg imm a w p -@vldr_sysreg .... ... . a:1 . . . rn:4 ... . ... .. ....... \ - reg=%vldr_sysreg imm=%imm7_0x4 &vldr_sysreg - -# P=0 W=0 is SEE "Related encodings", so split into two patterns -VLDR_sysreg ---- 110 1 . . w:1 1 .... ... 0 111 11 ....... @vldr_sysreg p=1 -VLDR_sysreg ---- 110 0 . . 1 1 .... ... 0 111 11 ....... @vldr_sysreg p=0 w=1 -VSTR_sysreg ---- 110 1 . . w:1 0 .... ... 0 111 11 ....... @vldr_sysreg p=1 -VSTR_sysreg ---- 110 0 . . 1 0 .... ... 0 111 11 ....... @vldr_sysreg p=0 w=1 - # We split the load/store multiple up into two patterns to avoid # overlap with other insns in the "Advanced SIMD load/store and 64-bit move" # grouping: diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-m-nocp.c +++ b/target/arm/translate-m-nocp.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "tcg/tcg-op.h" +#include "tcg/tcg-op-gvec.h" #include "translate.h" #include "translate-a32.h" @@ -XXX,XX +XXX,XX @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a) return true; } +/* + * M-profile provides two different sets of instructions that can + * access floating point system registers: VMSR/VMRS (which move + * to/from a general purpose register) and VLDR/VSTR sysreg (which + * move directly to/from memory). In some cases there are also side + * effects which must happen after any write to memory (which could + * cause an exception). So we implement the common logic for the + * sysreg access in gen_M_fp_sysreg_write() and gen_M_fp_sysreg_read(), + * which take pointers to callback functions which will perform the + * actual "read/write general purpose register" and "read/write + * memory" operations. + */ + +/* + * Emit code to store the sysreg to its final destination; frees the + * TCG temp 'value' it is passed. + */ +typedef void fp_sysreg_storefn(DisasContext *s, void *opaque, TCGv_i32 value); +/* + * Emit code to load the value to be copied to the sysreg; returns + * a new TCG temporary + */ +typedef TCGv_i32 fp_sysreg_loadfn(DisasContext *s, void *opaque); + +/* Common decode/access checks for fp sysreg read/write */ +typedef enum FPSysRegCheckResult { + FPSysRegCheckFailed, /* caller should return false */ + FPSysRegCheckDone, /* caller should return true */ + FPSysRegCheckContinue, /* caller should continue generating code */ +} FPSysRegCheckResult; + +static FPSysRegCheckResult fp_sysreg_checks(DisasContext *s, int regno) +{ + if (!dc_isar_feature(aa32_fpsp_v2, s) && !dc_isar_feature(aa32_mve, s)) { + return FPSysRegCheckFailed; + } + + switch (regno) { + case ARM_VFP_FPSCR: + case QEMU_VFP_FPSCR_NZCV: + break; + case ARM_VFP_FPSCR_NZCVQC: + if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { + return FPSysRegCheckFailed; + } + break; + case ARM_VFP_FPCXT_S: + case ARM_VFP_FPCXT_NS: + if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { + return FPSysRegCheckFailed; + } + if (!s->v8m_secure) { + return FPSysRegCheckFailed; + } + break; + case ARM_VFP_VPR: + case ARM_VFP_P0: + if (!dc_isar_feature(aa32_mve, s)) { + return FPSysRegCheckFailed; + } + break; + default: + return FPSysRegCheckFailed; + } + + /* + * FPCXT_NS is a special case: it has specific handling for + * "current FP state is inactive", and must do the PreserveFPState() + * but not the usual full set of actions done by ExecuteFPCheck(). + * So we don't call vfp_access_check() and the callers must handle this. + */ + if (regno != ARM_VFP_FPCXT_NS && !vfp_access_check(s)) { + return FPSysRegCheckDone; + } + return FPSysRegCheckContinue; +} + +static void gen_branch_fpInactive(DisasContext *s, TCGCond cond, + TCGLabel *label) +{ + /* + * FPCXT_NS is a special case: it has specific handling for + * "current FP state is inactive", and must do the PreserveFPState() + * but not the usual full set of actions done by ExecuteFPCheck(). + * We don't have a TB flag that matches the fpInactive check, so we + * do it at runtime as we don't expect FPCXT_NS accesses to be frequent. + * + * Emit code that checks fpInactive and does a conditional + * branch to label based on it: + * if cond is TCG_COND_NE then branch if fpInactive != 0 (ie if inactive) + * if cond is TCG_COND_EQ then branch if fpInactive == 0 (ie if active) + */ + assert(cond == TCG_COND_EQ || cond == TCG_COND_NE); + + /* fpInactive = FPCCR_NS.ASPEN == 1 && CONTROL.FPCA == 0 */ + TCGv_i32 aspen, fpca; + aspen = load_cpu_field(v7m.fpccr[M_REG_NS]); + fpca = load_cpu_field(v7m.control[M_REG_S]); + tcg_gen_andi_i32(aspen, aspen, R_V7M_FPCCR_ASPEN_MASK); + tcg_gen_xori_i32(aspen, aspen, R_V7M_FPCCR_ASPEN_MASK); + tcg_gen_andi_i32(fpca, fpca, R_V7M_CONTROL_FPCA_MASK); + tcg_gen_or_i32(fpca, fpca, aspen); + tcg_gen_brcondi_i32(tcg_invert_cond(cond), fpca, 0, label); + tcg_temp_free_i32(aspen); + tcg_temp_free_i32(fpca); +} + +static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, + fp_sysreg_loadfn *loadfn, + void *opaque) +{ + /* Do a write to an M-profile floating point system register */ + TCGv_i32 tmp; + TCGLabel *lab_end = NULL; + + switch (fp_sysreg_checks(s, regno)) { + case FPSysRegCheckFailed: + return false; + case FPSysRegCheckDone: + return true; + case FPSysRegCheckContinue: + break; + } + + switch (regno) { + case ARM_VFP_FPSCR: + tmp = loadfn(s, opaque); + gen_helper_vfp_set_fpscr(cpu_env, tmp); + tcg_temp_free_i32(tmp); + gen_lookup_tb(s); + break; + case ARM_VFP_FPSCR_NZCVQC: + { + TCGv_i32 fpscr; + tmp = loadfn(s, opaque); + if (dc_isar_feature(aa32_mve, s)) { + /* QC is only present for MVE; otherwise RES0 */ + TCGv_i32 qc = tcg_temp_new_i32(); + tcg_gen_andi_i32(qc, tmp, FPCR_QC); + /* + * The 4 vfp.qc[] fields need only be "zero" vs "non-zero"; + * here writing the same value into all elements is simplest. + */ + tcg_gen_gvec_dup_i32(MO_32, offsetof(CPUARMState, vfp.qc), + 16, 16, qc); + } + tcg_gen_andi_i32(tmp, tmp, FPCR_NZCV_MASK); + fpscr = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); + tcg_gen_andi_i32(fpscr, fpscr, ~FPCR_NZCV_MASK); + tcg_gen_or_i32(fpscr, fpscr, tmp); + store_cpu_field(fpscr, vfp.xregs[ARM_VFP_FPSCR]); + tcg_temp_free_i32(tmp); + break; + } + case ARM_VFP_FPCXT_NS: + lab_end = gen_new_label(); + /* fpInactive case: write is a NOP, so branch to end */ + gen_branch_fpInactive(s, TCG_COND_NE, lab_end); + /* + * !fpInactive: if FPU disabled, take NOCP exception; + * otherwise PreserveFPState(), and then FPCXT_NS writes + * behave the same as FPCXT_S writes. + */ + if (s->fp_excp_el) { + gen_exception_insn(s, s->pc_curr, EXCP_NOCP, + syn_uncategorized(), s->fp_excp_el); + /* + * This was only a conditional exception, so override + * gen_exception_insn()'s default to DISAS_NORETURN + */ + s->base.is_jmp = DISAS_NEXT; + break; + } + gen_preserve_fp_state(s); + /* fall through */ + case ARM_VFP_FPCXT_S: + { + TCGv_i32 sfpa, control; + /* + * Set FPSCR and CONTROL.SFPA from value; the new FPSCR takes + * bits [27:0] from value and zeroes bits [31:28]. + */ + tmp = loadfn(s, opaque); + sfpa = tcg_temp_new_i32(); + tcg_gen_shri_i32(sfpa, tmp, 31); + control = load_cpu_field(v7m.control[M_REG_S]); + tcg_gen_deposit_i32(control, control, sfpa, + R_V7M_CONTROL_SFPA_SHIFT, 1); + store_cpu_field(control, v7m.control[M_REG_S]); + tcg_gen_andi_i32(tmp, tmp, ~FPCR_NZCV_MASK); + gen_helper_vfp_set_fpscr(cpu_env, tmp); + tcg_temp_free_i32(tmp); + tcg_temp_free_i32(sfpa); + break; + } + case ARM_VFP_VPR: + /* Behaves as NOP if not privileged */ + if (IS_USER(s)) { + break; + } + tmp = loadfn(s, opaque); + store_cpu_field(tmp, v7m.vpr); + break; + case ARM_VFP_P0: + { + TCGv_i32 vpr; + tmp = loadfn(s, opaque); + vpr = load_cpu_field(v7m.vpr); + tcg_gen_deposit_i32(vpr, vpr, tmp, + R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); + store_cpu_field(vpr, v7m.vpr); + tcg_temp_free_i32(tmp); + break; + } + default: + g_assert_not_reached(); + } + if (lab_end) { + gen_set_label(lab_end); + } + return true; +} + +static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, + fp_sysreg_storefn *storefn, + void *opaque) +{ + /* Do a read from an M-profile floating point system register */ + TCGv_i32 tmp; + TCGLabel *lab_end = NULL; + bool lookup_tb = false; + + switch (fp_sysreg_checks(s, regno)) { + case FPSysRegCheckFailed: + return false; + case FPSysRegCheckDone: + return true; + case FPSysRegCheckContinue: + break; + } + + if (regno == ARM_VFP_FPSCR_NZCVQC && !dc_isar_feature(aa32_mve, s)) { + /* QC is RES0 without MVE, so NZCVQC simplifies to NZCV */ + regno = QEMU_VFP_FPSCR_NZCV; + } + + switch (regno) { + case ARM_VFP_FPSCR: + tmp = tcg_temp_new_i32(); + gen_helper_vfp_get_fpscr(tmp, cpu_env); + storefn(s, opaque, tmp); + break; + case ARM_VFP_FPSCR_NZCVQC: + tmp = tcg_temp_new_i32(); + gen_helper_vfp_get_fpscr(tmp, cpu_env); + tcg_gen_andi_i32(tmp, tmp, FPCR_NZCVQC_MASK); + storefn(s, opaque, tmp); + break; + case QEMU_VFP_FPSCR_NZCV: + /* + * Read just NZCV; this is a special case to avoid the + * helper call for the "VMRS to CPSR.NZCV" insn. + */ + tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); + tcg_gen_andi_i32(tmp, tmp, FPCR_NZCV_MASK); + storefn(s, opaque, tmp); + break; + case ARM_VFP_FPCXT_S: + { + TCGv_i32 control, sfpa, fpscr; + /* Bits [27:0] from FPSCR, bit [31] from CONTROL.SFPA */ + tmp = tcg_temp_new_i32(); + sfpa = tcg_temp_new_i32(); + gen_helper_vfp_get_fpscr(tmp, cpu_env); + tcg_gen_andi_i32(tmp, tmp, ~FPCR_NZCV_MASK); + control = load_cpu_field(v7m.control[M_REG_S]); + tcg_gen_andi_i32(sfpa, control, R_V7M_CONTROL_SFPA_MASK); + tcg_gen_shli_i32(sfpa, sfpa, 31 - R_V7M_CONTROL_SFPA_SHIFT); + tcg_gen_or_i32(tmp, tmp, sfpa); + tcg_temp_free_i32(sfpa); + /* + * Store result before updating FPSCR etc, in case + * it is a memory write which causes an exception. + */ + storefn(s, opaque, tmp); + /* + * Now we must reset FPSCR from FPDSCR_NS, and clear + * CONTROL.SFPA; so we'll end the TB here. + */ + tcg_gen_andi_i32(control, control, ~R_V7M_CONTROL_SFPA_MASK); + store_cpu_field(control, v7m.control[M_REG_S]); + fpscr = load_cpu_field(v7m.fpdscr[M_REG_NS]); + gen_helper_vfp_set_fpscr(cpu_env, fpscr); + tcg_temp_free_i32(fpscr); + lookup_tb = true; + break; + } + case ARM_VFP_FPCXT_NS: + { + TCGv_i32 control, sfpa, fpscr, fpdscr, zero; + TCGLabel *lab_active = gen_new_label(); + + lookup_tb = true; + + gen_branch_fpInactive(s, TCG_COND_EQ, lab_active); + /* fpInactive case: reads as FPDSCR_NS */ + TCGv_i32 tmp = load_cpu_field(v7m.fpdscr[M_REG_NS]); + storefn(s, opaque, tmp); + lab_end = gen_new_label(); + tcg_gen_br(lab_end); + + gen_set_label(lab_active); + /* + * !fpInactive: if FPU disabled, take NOCP exception; + * otherwise PreserveFPState(), and then FPCXT_NS + * reads the same as FPCXT_S. + */ + if (s->fp_excp_el) { + gen_exception_insn(s, s->pc_curr, EXCP_NOCP, + syn_uncategorized(), s->fp_excp_el); + /* + * This was only a conditional exception, so override + * gen_exception_insn()'s default to DISAS_NORETURN + */ + s->base.is_jmp = DISAS_NEXT; + break; + } + gen_preserve_fp_state(s); + tmp = tcg_temp_new_i32(); + sfpa = tcg_temp_new_i32(); + fpscr = tcg_temp_new_i32(); + gen_helper_vfp_get_fpscr(fpscr, cpu_env); + tcg_gen_andi_i32(tmp, fpscr, ~FPCR_NZCV_MASK); + control = load_cpu_field(v7m.control[M_REG_S]); + tcg_gen_andi_i32(sfpa, control, R_V7M_CONTROL_SFPA_MASK); + tcg_gen_shli_i32(sfpa, sfpa, 31 - R_V7M_CONTROL_SFPA_SHIFT); + tcg_gen_or_i32(tmp, tmp, sfpa); + tcg_temp_free_i32(control); + /* Store result before updating FPSCR, in case it faults */ + storefn(s, opaque, tmp); + /* If SFPA is zero then set FPSCR from FPDSCR_NS */ + fpdscr = load_cpu_field(v7m.fpdscr[M_REG_NS]); + zero = tcg_const_i32(0); + tcg_gen_movcond_i32(TCG_COND_EQ, fpscr, sfpa, zero, fpdscr, fpscr); + gen_helper_vfp_set_fpscr(cpu_env, fpscr); + tcg_temp_free_i32(zero); + tcg_temp_free_i32(sfpa); + tcg_temp_free_i32(fpdscr); + tcg_temp_free_i32(fpscr); + break; + } + case ARM_VFP_VPR: + /* Behaves as NOP if not privileged */ + if (IS_USER(s)) { + break; + } + tmp = load_cpu_field(v7m.vpr); + storefn(s, opaque, tmp); + break; + case ARM_VFP_P0: + tmp = load_cpu_field(v7m.vpr); + tcg_gen_extract_i32(tmp, tmp, R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); + storefn(s, opaque, tmp); + break; + default: + g_assert_not_reached(); + } + + if (lab_end) { + gen_set_label(lab_end); + } + if (lookup_tb) { + gen_lookup_tb(s); + } + return true; +} + +static void fp_sysreg_to_gpr(DisasContext *s, void *opaque, TCGv_i32 value) +{ + arg_VMSR_VMRS *a = opaque; + + if (a->rt == 15) { + /* Set the 4 flag bits in the CPSR */ + gen_set_nzcv(value); + tcg_temp_free_i32(value); + } else { + store_reg(s, a->rt, value); + } +} + +static TCGv_i32 gpr_to_fp_sysreg(DisasContext *s, void *opaque) +{ + arg_VMSR_VMRS *a = opaque; + + return load_reg(s, a->rt); +} + +static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) +{ + /* + * Accesses to R15 are UNPREDICTABLE; we choose to undef. + * FPSCR -> r15 is a special case which writes to the PSR flags; + * set a->reg to a special value to tell gen_M_fp_sysreg_read() + * we only care about the top 4 bits of FPSCR there. + */ + if (a->rt == 15) { + if (a->l && a->reg == ARM_VFP_FPSCR) { + a->reg = QEMU_VFP_FPSCR_NZCV; + } else { + return false; + } + } + + if (a->l) { + /* VMRS, move FP system register to gp register */ + return gen_M_fp_sysreg_read(s, a->reg, fp_sysreg_to_gpr, a); + } else { + /* VMSR, move gp register to FP system register */ + return gen_M_fp_sysreg_write(s, a->reg, gpr_to_fp_sysreg, a); + } +} + +static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) +{ + arg_vldr_sysreg *a = opaque; + uint32_t offset = a->imm; + TCGv_i32 addr; + + if (!a->a) { + offset = -offset; + } + + addr = load_reg(s, a->rn); + if (a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + + if (s->v8m_stackcheck && a->rn == 13 && a->w) { + gen_helper_v8m_stackcheck(cpu_env, addr); + } + + gen_aa32_st_i32(s, value, addr, get_mem_index(s), + MO_UL | MO_ALIGN | s->be_data); + tcg_temp_free_i32(value); + + if (a->w) { + /* writeback */ + if (!a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + store_reg(s, a->rn, addr); + } else { + tcg_temp_free_i32(addr); + } +} + +static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque) +{ + arg_vldr_sysreg *a = opaque; + uint32_t offset = a->imm; + TCGv_i32 addr; + TCGv_i32 value = tcg_temp_new_i32(); + + if (!a->a) { + offset = -offset; + } + + addr = load_reg(s, a->rn); + if (a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + + if (s->v8m_stackcheck && a->rn == 13 && a->w) { + gen_helper_v8m_stackcheck(cpu_env, addr); + } + + gen_aa32_ld_i32(s, value, addr, get_mem_index(s), + MO_UL | MO_ALIGN | s->be_data); + + if (a->w) { + /* writeback */ + if (!a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + store_reg(s, a->rn, addr); + } else { + tcg_temp_free_i32(addr); + } + return value; +} + +static bool trans_VLDR_sysreg(DisasContext *s, arg_vldr_sysreg *a) +{ + if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { + return false; + } + if (a->rn == 15) { + return false; + } + return gen_M_fp_sysreg_write(s, a->reg, memory_to_fp_sysreg, a); +} + +static bool trans_VSTR_sysreg(DisasContext *s, arg_vldr_sysreg *a) +{ + if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { + return false; + } + if (a->rn == 15) { + return false; + } + return gen_M_fp_sysreg_read(s, a->reg, fp_sysreg_to_memory, a); +} + static bool trans_NOCP(DisasContext *s, arg_nocp *a) { /* diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static inline long vfp_f16_offset(unsigned reg, bool top) * Generate code for M-profile lazy FP state preservation if needed; * this corresponds to the pseudocode PreserveFPState() function. */ -static void gen_preserve_fp_state(DisasContext *s) +void gen_preserve_fp_state(DisasContext *s) { if (s->v7m_lspact) { /* @@ -XXX,XX +XXX,XX @@ static bool trans_VDUP(DisasContext *s, arg_VDUP *a) return true; } -/* - * M-profile provides two different sets of instructions that can - * access floating point system registers: VMSR/VMRS (which move - * to/from a general purpose register) and VLDR/VSTR sysreg (which - * move directly to/from memory). In some cases there are also side - * effects which must happen after any write to memory (which could - * cause an exception). So we implement the common logic for the - * sysreg access in gen_M_fp_sysreg_write() and gen_M_fp_sysreg_read(), - * which take pointers to callback functions which will perform the - * actual "read/write general purpose register" and "read/write - * memory" operations. - */ - -/* - * Emit code to store the sysreg to its final destination; frees the - * TCG temp 'value' it is passed. - */ -typedef void fp_sysreg_storefn(DisasContext *s, void *opaque, TCGv_i32 value); -/* - * Emit code to load the value to be copied to the sysreg; returns - * a new TCG temporary - */ -typedef TCGv_i32 fp_sysreg_loadfn(DisasContext *s, void *opaque); - -/* Common decode/access checks for fp sysreg read/write */ -typedef enum FPSysRegCheckResult { - FPSysRegCheckFailed, /* caller should return false */ - FPSysRegCheckDone, /* caller should return true */ - FPSysRegCheckContinue, /* caller should continue generating code */ -} FPSysRegCheckResult; - -static FPSysRegCheckResult fp_sysreg_checks(DisasContext *s, int regno) -{ - if (!dc_isar_feature(aa32_fpsp_v2, s) && !dc_isar_feature(aa32_mve, s)) { - return FPSysRegCheckFailed; - } - - switch (regno) { - case ARM_VFP_FPSCR: - case QEMU_VFP_FPSCR_NZCV: - break; - case ARM_VFP_FPSCR_NZCVQC: - if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { - return FPSysRegCheckFailed; - } - break; - case ARM_VFP_FPCXT_S: - case ARM_VFP_FPCXT_NS: - if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { - return FPSysRegCheckFailed; - } - if (!s->v8m_secure) { - return FPSysRegCheckFailed; - } - break; - case ARM_VFP_VPR: - case ARM_VFP_P0: - if (!dc_isar_feature(aa32_mve, s)) { - return FPSysRegCheckFailed; - } - break; - default: - return FPSysRegCheckFailed; - } - - /* - * FPCXT_NS is a special case: it has specific handling for - * "current FP state is inactive", and must do the PreserveFPState() - * but not the usual full set of actions done by ExecuteFPCheck(). - * So we don't call vfp_access_check() and the callers must handle this. - */ - if (regno != ARM_VFP_FPCXT_NS && !vfp_access_check(s)) { - return FPSysRegCheckDone; - } - return FPSysRegCheckContinue; -} - -static void gen_branch_fpInactive(DisasContext *s, TCGCond cond, - TCGLabel *label) -{ - /* - * FPCXT_NS is a special case: it has specific handling for - * "current FP state is inactive", and must do the PreserveFPState() - * but not the usual full set of actions done by ExecuteFPCheck(). - * We don't have a TB flag that matches the fpInactive check, so we - * do it at runtime as we don't expect FPCXT_NS accesses to be frequent. - * - * Emit code that checks fpInactive and does a conditional - * branch to label based on it: - * if cond is TCG_COND_NE then branch if fpInactive != 0 (ie if inactive) - * if cond is TCG_COND_EQ then branch if fpInactive == 0 (ie if active) - */ - assert(cond == TCG_COND_EQ || cond == TCG_COND_NE); - - /* fpInactive = FPCCR_NS.ASPEN == 1 && CONTROL.FPCA == 0 */ - TCGv_i32 aspen, fpca; - aspen = load_cpu_field(v7m.fpccr[M_REG_NS]); - fpca = load_cpu_field(v7m.control[M_REG_S]); - tcg_gen_andi_i32(aspen, aspen, R_V7M_FPCCR_ASPEN_MASK); - tcg_gen_xori_i32(aspen, aspen, R_V7M_FPCCR_ASPEN_MASK); - tcg_gen_andi_i32(fpca, fpca, R_V7M_CONTROL_FPCA_MASK); - tcg_gen_or_i32(fpca, fpca, aspen); - tcg_gen_brcondi_i32(tcg_invert_cond(cond), fpca, 0, label); - tcg_temp_free_i32(aspen); - tcg_temp_free_i32(fpca); -} - -static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, - fp_sysreg_loadfn *loadfn, - void *opaque) -{ - /* Do a write to an M-profile floating point system register */ - TCGv_i32 tmp; - TCGLabel *lab_end = NULL; - - switch (fp_sysreg_checks(s, regno)) { - case FPSysRegCheckFailed: - return false; - case FPSysRegCheckDone: - return true; - case FPSysRegCheckContinue: - break; - } - - switch (regno) { - case ARM_VFP_FPSCR: - tmp = loadfn(s, opaque); - gen_helper_vfp_set_fpscr(cpu_env, tmp); - tcg_temp_free_i32(tmp); - gen_lookup_tb(s); - break; - case ARM_VFP_FPSCR_NZCVQC: - { - TCGv_i32 fpscr; - tmp = loadfn(s, opaque); - if (dc_isar_feature(aa32_mve, s)) { - /* QC is only present for MVE; otherwise RES0 */ - TCGv_i32 qc = tcg_temp_new_i32(); - tcg_gen_andi_i32(qc, tmp, FPCR_QC); - /* - * The 4 vfp.qc[] fields need only be "zero" vs "non-zero"; - * here writing the same value into all elements is simplest. - */ - tcg_gen_gvec_dup_i32(MO_32, offsetof(CPUARMState, vfp.qc), - 16, 16, qc); - } - tcg_gen_andi_i32(tmp, tmp, FPCR_NZCV_MASK); - fpscr = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); - tcg_gen_andi_i32(fpscr, fpscr, ~FPCR_NZCV_MASK); - tcg_gen_or_i32(fpscr, fpscr, tmp); - store_cpu_field(fpscr, vfp.xregs[ARM_VFP_FPSCR]); - tcg_temp_free_i32(tmp); - break; - } - case ARM_VFP_FPCXT_NS: - lab_end = gen_new_label(); - /* fpInactive case: write is a NOP, so branch to end */ - gen_branch_fpInactive(s, TCG_COND_NE, lab_end); - /* - * !fpInactive: if FPU disabled, take NOCP exception; - * otherwise PreserveFPState(), and then FPCXT_NS writes - * behave the same as FPCXT_S writes. - */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, - syn_uncategorized(), s->fp_excp_el); - /* - * This was only a conditional exception, so override - * gen_exception_insn()'s default to DISAS_NORETURN - */ - s->base.is_jmp = DISAS_NEXT; - break; - } - gen_preserve_fp_state(s); - /* fall through */ - case ARM_VFP_FPCXT_S: - { - TCGv_i32 sfpa, control; - /* - * Set FPSCR and CONTROL.SFPA from value; the new FPSCR takes - * bits [27:0] from value and zeroes bits [31:28]. - */ - tmp = loadfn(s, opaque); - sfpa = tcg_temp_new_i32(); - tcg_gen_shri_i32(sfpa, tmp, 31); - control = load_cpu_field(v7m.control[M_REG_S]); - tcg_gen_deposit_i32(control, control, sfpa, - R_V7M_CONTROL_SFPA_SHIFT, 1); - store_cpu_field(control, v7m.control[M_REG_S]); - tcg_gen_andi_i32(tmp, tmp, ~FPCR_NZCV_MASK); - gen_helper_vfp_set_fpscr(cpu_env, tmp); - tcg_temp_free_i32(tmp); - tcg_temp_free_i32(sfpa); - break; - } - case ARM_VFP_VPR: - /* Behaves as NOP if not privileged */ - if (IS_USER(s)) { - break; - } - tmp = loadfn(s, opaque); - store_cpu_field(tmp, v7m.vpr); - break; - case ARM_VFP_P0: - { - TCGv_i32 vpr; - tmp = loadfn(s, opaque); - vpr = load_cpu_field(v7m.vpr); - tcg_gen_deposit_i32(vpr, vpr, tmp, - R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); - store_cpu_field(vpr, v7m.vpr); - tcg_temp_free_i32(tmp); - break; - } - default: - g_assert_not_reached(); - } - if (lab_end) { - gen_set_label(lab_end); - } - return true; -} - -static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, - fp_sysreg_storefn *storefn, - void *opaque) -{ - /* Do a read from an M-profile floating point system register */ - TCGv_i32 tmp; - TCGLabel *lab_end = NULL; - bool lookup_tb = false; - - switch (fp_sysreg_checks(s, regno)) { - case FPSysRegCheckFailed: - return false; - case FPSysRegCheckDone: - return true; - case FPSysRegCheckContinue: - break; - } - - if (regno == ARM_VFP_FPSCR_NZCVQC && !dc_isar_feature(aa32_mve, s)) { - /* QC is RES0 without MVE, so NZCVQC simplifies to NZCV */ - regno = QEMU_VFP_FPSCR_NZCV; - } - - switch (regno) { - case ARM_VFP_FPSCR: - tmp = tcg_temp_new_i32(); - gen_helper_vfp_get_fpscr(tmp, cpu_env); - storefn(s, opaque, tmp); - break; - case ARM_VFP_FPSCR_NZCVQC: - tmp = tcg_temp_new_i32(); - gen_helper_vfp_get_fpscr(tmp, cpu_env); - tcg_gen_andi_i32(tmp, tmp, FPCR_NZCVQC_MASK); - storefn(s, opaque, tmp); - break; - case QEMU_VFP_FPSCR_NZCV: - /* - * Read just NZCV; this is a special case to avoid the - * helper call for the "VMRS to CPSR.NZCV" insn. - */ - tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); - tcg_gen_andi_i32(tmp, tmp, FPCR_NZCV_MASK); - storefn(s, opaque, tmp); - break; - case ARM_VFP_FPCXT_S: - { - TCGv_i32 control, sfpa, fpscr; - /* Bits [27:0] from FPSCR, bit [31] from CONTROL.SFPA */ - tmp = tcg_temp_new_i32(); - sfpa = tcg_temp_new_i32(); - gen_helper_vfp_get_fpscr(tmp, cpu_env); - tcg_gen_andi_i32(tmp, tmp, ~FPCR_NZCV_MASK); - control = load_cpu_field(v7m.control[M_REG_S]); - tcg_gen_andi_i32(sfpa, control, R_V7M_CONTROL_SFPA_MASK); - tcg_gen_shli_i32(sfpa, sfpa, 31 - R_V7M_CONTROL_SFPA_SHIFT); - tcg_gen_or_i32(tmp, tmp, sfpa); - tcg_temp_free_i32(sfpa); - /* - * Store result before updating FPSCR etc, in case - * it is a memory write which causes an exception. - */ - storefn(s, opaque, tmp); - /* - * Now we must reset FPSCR from FPDSCR_NS, and clear - * CONTROL.SFPA; so we'll end the TB here. - */ - tcg_gen_andi_i32(control, control, ~R_V7M_CONTROL_SFPA_MASK); - store_cpu_field(control, v7m.control[M_REG_S]); - fpscr = load_cpu_field(v7m.fpdscr[M_REG_NS]); - gen_helper_vfp_set_fpscr(cpu_env, fpscr); - tcg_temp_free_i32(fpscr); - lookup_tb = true; - break; - } - case ARM_VFP_FPCXT_NS: - { - TCGv_i32 control, sfpa, fpscr, fpdscr, zero; - TCGLabel *lab_active = gen_new_label(); - - lookup_tb = true; - - gen_branch_fpInactive(s, TCG_COND_EQ, lab_active); - /* fpInactive case: reads as FPDSCR_NS */ - TCGv_i32 tmp = load_cpu_field(v7m.fpdscr[M_REG_NS]); - storefn(s, opaque, tmp); - lab_end = gen_new_label(); - tcg_gen_br(lab_end); - - gen_set_label(lab_active); - /* - * !fpInactive: if FPU disabled, take NOCP exception; - * otherwise PreserveFPState(), and then FPCXT_NS - * reads the same as FPCXT_S. - */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, - syn_uncategorized(), s->fp_excp_el); - /* - * This was only a conditional exception, so override - * gen_exception_insn()'s default to DISAS_NORETURN - */ - s->base.is_jmp = DISAS_NEXT; - break; - } - gen_preserve_fp_state(s); - tmp = tcg_temp_new_i32(); - sfpa = tcg_temp_new_i32(); - fpscr = tcg_temp_new_i32(); - gen_helper_vfp_get_fpscr(fpscr, cpu_env); - tcg_gen_andi_i32(tmp, fpscr, ~FPCR_NZCV_MASK); - control = load_cpu_field(v7m.control[M_REG_S]); - tcg_gen_andi_i32(sfpa, control, R_V7M_CONTROL_SFPA_MASK); - tcg_gen_shli_i32(sfpa, sfpa, 31 - R_V7M_CONTROL_SFPA_SHIFT); - tcg_gen_or_i32(tmp, tmp, sfpa); - tcg_temp_free_i32(control); - /* Store result before updating FPSCR, in case it faults */ - storefn(s, opaque, tmp); - /* If SFPA is zero then set FPSCR from FPDSCR_NS */ - fpdscr = load_cpu_field(v7m.fpdscr[M_REG_NS]); - zero = tcg_const_i32(0); - tcg_gen_movcond_i32(TCG_COND_EQ, fpscr, sfpa, zero, fpdscr, fpscr); - gen_helper_vfp_set_fpscr(cpu_env, fpscr); - tcg_temp_free_i32(zero); - tcg_temp_free_i32(sfpa); - tcg_temp_free_i32(fpdscr); - tcg_temp_free_i32(fpscr); - break; - } - case ARM_VFP_VPR: - /* Behaves as NOP if not privileged */ - if (IS_USER(s)) { - break; - } - tmp = load_cpu_field(v7m.vpr); - storefn(s, opaque, tmp); - break; - case ARM_VFP_P0: - tmp = load_cpu_field(v7m.vpr); - tcg_gen_extract_i32(tmp, tmp, R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); - storefn(s, opaque, tmp); - break; - default: - g_assert_not_reached(); - } - - if (lab_end) { - gen_set_label(lab_end); - } - if (lookup_tb) { - gen_lookup_tb(s); - } - return true; -} - -static void fp_sysreg_to_gpr(DisasContext *s, void *opaque, TCGv_i32 value) -{ - arg_VMSR_VMRS *a = opaque; - - if (a->rt == 15) { - /* Set the 4 flag bits in the CPSR */ - gen_set_nzcv(value); - tcg_temp_free_i32(value); - } else { - store_reg(s, a->rt, value); - } -} - -static TCGv_i32 gpr_to_fp_sysreg(DisasContext *s, void *opaque) -{ - arg_VMSR_VMRS *a = opaque; - - return load_reg(s, a->rt); -} - -static bool gen_M_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) -{ - /* - * Accesses to R15 are UNPREDICTABLE; we choose to undef. - * FPSCR -> r15 is a special case which writes to the PSR flags; - * set a->reg to a special value to tell gen_M_fp_sysreg_read() - * we only care about the top 4 bits of FPSCR there. - */ - if (a->rt == 15) { - if (a->l && a->reg == ARM_VFP_FPSCR) { - a->reg = QEMU_VFP_FPSCR_NZCV; - } else { - return false; - } - } - - if (a->l) { - /* VMRS, move FP system register to gp register */ - return gen_M_fp_sysreg_read(s, a->reg, fp_sysreg_to_gpr, a); - } else { - /* VMSR, move gp register to FP system register */ - return gen_M_fp_sysreg_write(s, a->reg, gpr_to_fp_sysreg, a); - } -} - static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) { TCGv_i32 tmp; bool ignore_vfp_enabled = false; if (arm_dc_feature(s, ARM_FEATURE_M)) { - return gen_M_VMSR_VMRS(s, a); + /* M profile version was already handled in m-nocp.decode */ + return false; } if (!dc_isar_feature(aa32_fpsp_v2, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) return true; } -static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) -{ - arg_vldr_sysreg *a = opaque; - uint32_t offset = a->imm; - TCGv_i32 addr; - - if (!a->a) { - offset = -offset; - } - - addr = load_reg(s, a->rn); - if (a->p) { - tcg_gen_addi_i32(addr, addr, offset); - } - - if (s->v8m_stackcheck && a->rn == 13 && a->w) { - gen_helper_v8m_stackcheck(cpu_env, addr); - } - - gen_aa32_st_i32(s, value, addr, get_mem_index(s), - MO_UL | MO_ALIGN | s->be_data); - tcg_temp_free_i32(value); - - if (a->w) { - /* writeback */ - if (!a->p) { - tcg_gen_addi_i32(addr, addr, offset); - } - store_reg(s, a->rn, addr); - } else { - tcg_temp_free_i32(addr); - } -} - -static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque) -{ - arg_vldr_sysreg *a = opaque; - uint32_t offset = a->imm; - TCGv_i32 addr; - TCGv_i32 value = tcg_temp_new_i32(); - - if (!a->a) { - offset = -offset; - } - - addr = load_reg(s, a->rn); - if (a->p) { - tcg_gen_addi_i32(addr, addr, offset); - } - - if (s->v8m_stackcheck && a->rn == 13 && a->w) { - gen_helper_v8m_stackcheck(cpu_env, addr); - } - - gen_aa32_ld_i32(s, value, addr, get_mem_index(s), - MO_UL | MO_ALIGN | s->be_data); - - if (a->w) { - /* writeback */ - if (!a->p) { - tcg_gen_addi_i32(addr, addr, offset); - } - store_reg(s, a->rn, addr); - } else { - tcg_temp_free_i32(addr); - } - return value; -} - -static bool trans_VLDR_sysreg(DisasContext *s, arg_vldr_sysreg *a) -{ - if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { - return false; - } - if (a->rn == 15) { - return false; - } - return gen_M_fp_sysreg_write(s, a->reg, memory_to_fp_sysreg, a); -} - -static bool trans_VSTR_sysreg(DisasContext *s, arg_vldr_sysreg *a) -{ - if (!arm_dc_feature(s, ARM_FEATURE_V8_1M)) { - return false; - } - if (a->rn == 15) { - return false; - } - return gen_M_fp_sysreg_read(s, a->reg, fp_sysreg_to_memory, a); -} static bool trans_VMOV_half(DisasContext *s, arg_VMOV_single *a) { -- 2.20.1 A few subcases of VLDR/VSTR sysreg succeed but do not perform a memory access: * VSTR of VPR when unprivileged * VLDR to VPR when unprivileged * VLDR to FPCXT_NS when fpInactive In these cases, even though we don't do the memory access we should still update the base register and perform the stack limit check if the insn's addressing mode specifies writeback. Our implementation failed to do this, because we handle these side-effects inside the memory_to_fp_sysreg() and fp_sysreg_to_memory() callback functions, which are only called if there's something to load or store. Fix this by adding an extra argument to the callbacks which is set to true to actually perform the access and false to only do side effects like writeback, and calling the callback with do_access = false for the three cases listed above. This produces slightly suboptimal code for the case of a write to FPCXT_NS when the FPU is inactive and the insn didn't have side effects (ie no writeback, or via VMSR), in which case we'll generate a conditional branch over an unconditional branch. But this doesn't seem to be important enough to merit requiring the callback to report back whether it generated any code or not. Cc: qemu-stable@nongnu.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-5-peter.maydell@linaro.org --- target/arm/translate-m-nocp.c | 102 ++++++++++++++++++++++++---------- 1 file changed, 72 insertions(+), 30 deletions(-) diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-m-nocp.c +++ b/target/arm/translate-m-nocp.c @@ -XXX,XX +XXX,XX @@ static bool trans_VSCCLRM(DisasContext *s, arg_VSCCLRM *a) /* * Emit code to store the sysreg to its final destination; frees the - * TCG temp 'value' it is passed. + * TCG temp 'value' it is passed. do_access is true to do the store, + * and false to skip it and only perform side-effects like base + * register writeback. */ -typedef void fp_sysreg_storefn(DisasContext *s, void *opaque, TCGv_i32 value); +typedef void fp_sysreg_storefn(DisasContext *s, void *opaque, TCGv_i32 value, + bool do_access); /* * Emit code to load the value to be copied to the sysreg; returns - * a new TCG temporary + * a new TCG temporary. do_access is true to do the store, + * and false to skip it and only perform side-effects like base + * register writeback. */ -typedef TCGv_i32 fp_sysreg_loadfn(DisasContext *s, void *opaque); +typedef TCGv_i32 fp_sysreg_loadfn(DisasContext *s, void *opaque, + bool do_access); /* Common decode/access checks for fp sysreg read/write */ typedef enum FPSysRegCheckResult { @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, switch (regno) { case ARM_VFP_FPSCR: - tmp = loadfn(s, opaque); + tmp = loadfn(s, opaque, true); gen_helper_vfp_set_fpscr(cpu_env, tmp); tcg_temp_free_i32(tmp); gen_lookup_tb(s); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, case ARM_VFP_FPSCR_NZCVQC: { TCGv_i32 fpscr; - tmp = loadfn(s, opaque); + tmp = loadfn(s, opaque, true); if (dc_isar_feature(aa32_mve, s)) { /* QC is only present for MVE; otherwise RES0 */ TCGv_i32 qc = tcg_temp_new_i32(); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, break; } case ARM_VFP_FPCXT_NS: + { + TCGLabel *lab_active = gen_new_label(); + lab_end = gen_new_label(); - /* fpInactive case: write is a NOP, so branch to end */ - gen_branch_fpInactive(s, TCG_COND_NE, lab_end); + gen_branch_fpInactive(s, TCG_COND_EQ, lab_active); + /* + * fpInactive case: write is a NOP, so only do side effects + * like register writeback before we branch to end + */ + loadfn(s, opaque, false); + tcg_gen_br(lab_end); + + gen_set_label(lab_active); /* * !fpInactive: if FPU disabled, take NOCP exception; * otherwise PreserveFPState(), and then FPCXT_NS writes @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, break; } gen_preserve_fp_state(s); - /* fall through */ + } + /* fall through */ case ARM_VFP_FPCXT_S: { TCGv_i32 sfpa, control; @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, * Set FPSCR and CONTROL.SFPA from value; the new FPSCR takes * bits [27:0] from value and zeroes bits [31:28]. */ - tmp = loadfn(s, opaque); + tmp = loadfn(s, opaque, true); sfpa = tcg_temp_new_i32(); tcg_gen_shri_i32(sfpa, tmp, 31); control = load_cpu_field(v7m.control[M_REG_S]); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, case ARM_VFP_VPR: /* Behaves as NOP if not privileged */ if (IS_USER(s)) { + loadfn(s, opaque, false); break; } - tmp = loadfn(s, opaque); + tmp = loadfn(s, opaque, true); store_cpu_field(tmp, v7m.vpr); break; case ARM_VFP_P0: { TCGv_i32 vpr; - tmp = loadfn(s, opaque); + tmp = loadfn(s, opaque, true); vpr = load_cpu_field(v7m.vpr); tcg_gen_deposit_i32(vpr, vpr, tmp, R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, case ARM_VFP_FPSCR: tmp = tcg_temp_new_i32(); gen_helper_vfp_get_fpscr(tmp, cpu_env); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); break; case ARM_VFP_FPSCR_NZCVQC: tmp = tcg_temp_new_i32(); gen_helper_vfp_get_fpscr(tmp, cpu_env); tcg_gen_andi_i32(tmp, tmp, FPCR_NZCVQC_MASK); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); break; case QEMU_VFP_FPSCR_NZCV: /* @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, */ tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); tcg_gen_andi_i32(tmp, tmp, FPCR_NZCV_MASK); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); break; case ARM_VFP_FPCXT_S: { @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, * Store result before updating FPSCR etc, in case * it is a memory write which causes an exception. */ - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); /* * Now we must reset FPSCR from FPDSCR_NS, and clear * CONTROL.SFPA; so we'll end the TB here. @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, gen_branch_fpInactive(s, TCG_COND_EQ, lab_active); /* fpInactive case: reads as FPDSCR_NS */ TCGv_i32 tmp = load_cpu_field(v7m.fpdscr[M_REG_NS]); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); lab_end = gen_new_label(); tcg_gen_br(lab_end); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, tcg_gen_or_i32(tmp, tmp, sfpa); tcg_temp_free_i32(control); /* Store result before updating FPSCR, in case it faults */ - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); /* If SFPA is zero then set FPSCR from FPDSCR_NS */ fpdscr = load_cpu_field(v7m.fpdscr[M_REG_NS]); zero = tcg_const_i32(0); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, case ARM_VFP_VPR: /* Behaves as NOP if not privileged */ if (IS_USER(s)) { + storefn(s, opaque, NULL, false); break; } tmp = load_cpu_field(v7m.vpr); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); break; case ARM_VFP_P0: tmp = load_cpu_field(v7m.vpr); tcg_gen_extract_i32(tmp, tmp, R_V7M_VPR_P0_SHIFT, R_V7M_VPR_P0_LENGTH); - storefn(s, opaque, tmp); + storefn(s, opaque, tmp, true); break; default: g_assert_not_reached(); @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, return true; } -static void fp_sysreg_to_gpr(DisasContext *s, void *opaque, TCGv_i32 value) +static void fp_sysreg_to_gpr(DisasContext *s, void *opaque, TCGv_i32 value, + bool do_access) { arg_VMSR_VMRS *a = opaque; + if (!do_access) { + return; + } + if (a->rt == 15) { /* Set the 4 flag bits in the CPSR */ gen_set_nzcv(value); @@ -XXX,XX +XXX,XX @@ static void fp_sysreg_to_gpr(DisasContext *s, void *opaque, TCGv_i32 value) } } -static TCGv_i32 gpr_to_fp_sysreg(DisasContext *s, void *opaque) +static TCGv_i32 gpr_to_fp_sysreg(DisasContext *s, void *opaque, bool do_access) { arg_VMSR_VMRS *a = opaque; + if (!do_access) { + return NULL; + } return load_reg(s, a->rt); } @@ -XXX,XX +XXX,XX @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) } } -static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) +static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value, + bool do_access) { arg_vldr_sysreg *a = opaque; uint32_t offset = a->imm; @@ -XXX,XX +XXX,XX @@ static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) offset = -offset; } + if (!do_access && !a->w) { + return; + } + addr = load_reg(s, a->rn); if (a->p) { tcg_gen_addi_i32(addr, addr, offset); @@ -XXX,XX +XXX,XX @@ static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) gen_helper_v8m_stackcheck(cpu_env, addr); } - gen_aa32_st_i32(s, value, addr, get_mem_index(s), - MO_UL | MO_ALIGN | s->be_data); - tcg_temp_free_i32(value); + if (do_access) { + gen_aa32_st_i32(s, value, addr, get_mem_index(s), + MO_UL | MO_ALIGN | s->be_data); + tcg_temp_free_i32(value); + } if (a->w) { /* writeback */ @@ -XXX,XX +XXX,XX @@ static void fp_sysreg_to_memory(DisasContext *s, void *opaque, TCGv_i32 value) } } -static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque) +static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque, + bool do_access) { arg_vldr_sysreg *a = opaque; uint32_t offset = a->imm; TCGv_i32 addr; - TCGv_i32 value = tcg_temp_new_i32(); + TCGv_i32 value = NULL; if (!a->a) { offset = -offset; } + if (!do_access && !a->w) { + return NULL; + } + addr = load_reg(s, a->rn); if (a->p) { tcg_gen_addi_i32(addr, addr, offset); @@ -XXX,XX +XXX,XX @@ static TCGv_i32 memory_to_fp_sysreg(DisasContext *s, void *opaque) gen_helper_v8m_stackcheck(cpu_env, addr); } - gen_aa32_ld_i32(s, value, addr, get_mem_index(s), - MO_UL | MO_ALIGN | s->be_data); + if (do_access) { + value = tcg_temp_new_i32(); + gen_aa32_ld_i32(s, value, addr, get_mem_index(s), + MO_UL | MO_ALIGN | s->be_data); + } if (a->w) { /* writeback */ -- 2.20.1 Factor the code in full_vfp_access_check() which updates the ownership of the FP context and creates a new FP context out into its own function. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-6-peter.maydell@linaro.org --- target/arm/translate-vfp.c | 104 +++++++++++++++++++++---------------- 1 file changed, 58 insertions(+), 46 deletions(-) diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ void gen_preserve_fp_state(DisasContext *s) } } +/* + * Generate code for M-profile FP context handling: update the + * ownership of the FP context, and create a new context if + * necessary. This corresponds to the parts of the pseudocode + * ExecuteFPCheck() after the inital PreserveFPState() call. + */ +static void gen_update_fp_context(DisasContext *s) +{ + /* Update ownership of FP context: set FPCCR.S to match current state */ + if (s->v8m_fpccr_s_wrong) { + TCGv_i32 tmp; + + tmp = load_cpu_field(v7m.fpccr[M_REG_S]); + if (s->v8m_secure) { + tcg_gen_ori_i32(tmp, tmp, R_V7M_FPCCR_S_MASK); + } else { + tcg_gen_andi_i32(tmp, tmp, ~R_V7M_FPCCR_S_MASK); + } + store_cpu_field(tmp, v7m.fpccr[M_REG_S]); + /* Don't need to do this for any further FP insns in this TB */ + s->v8m_fpccr_s_wrong = false; + } + + if (s->v7m_new_fp_ctxt_needed) { + /* + * Create new FP context by updating CONTROL.FPCA, CONTROL.SFPA, + * the FPSCR, and VPR. + */ + TCGv_i32 control, fpscr; + uint32_t bits = R_V7M_CONTROL_FPCA_MASK; + + fpscr = load_cpu_field(v7m.fpdscr[s->v8m_secure]); + gen_helper_vfp_set_fpscr(cpu_env, fpscr); + tcg_temp_free_i32(fpscr); + if (dc_isar_feature(aa32_mve, s)) { + TCGv_i32 z32 = tcg_const_i32(0); + store_cpu_field(z32, v7m.vpr); + } + + /* + * We don't need to arrange to end the TB, because the only + * parts of FPSCR which we cache in the TB flags are the VECLEN + * and VECSTRIDE, and those don't exist for M-profile. + */ + + if (s->v8m_secure) { + bits |= R_V7M_CONTROL_SFPA_MASK; + } + control = load_cpu_field(v7m.control[M_REG_S]); + tcg_gen_ori_i32(control, control, bits); + store_cpu_field(control, v7m.control[M_REG_S]); + /* Don't need to do this for any further FP insns in this TB */ + s->v7m_new_fp_ctxt_needed = false; + } +} + /* * Check that VFP access is enabled. If it is, do the necessary * M-profile lazy-FP handling and then return true. @@ -XXX,XX +XXX,XX @@ static bool full_vfp_access_check(DisasContext *s, bool ignore_vfp_enabled) /* Trigger lazy-state preservation if necessary */ gen_preserve_fp_state(s); - /* Update ownership of FP context: set FPCCR.S to match current state */ - if (s->v8m_fpccr_s_wrong) { - TCGv_i32 tmp; - - tmp = load_cpu_field(v7m.fpccr[M_REG_S]); - if (s->v8m_secure) { - tcg_gen_ori_i32(tmp, tmp, R_V7M_FPCCR_S_MASK); - } else { - tcg_gen_andi_i32(tmp, tmp, ~R_V7M_FPCCR_S_MASK); - } - store_cpu_field(tmp, v7m.fpccr[M_REG_S]); - /* Don't need to do this for any further FP insns in this TB */ - s->v8m_fpccr_s_wrong = false; - } - - if (s->v7m_new_fp_ctxt_needed) { - /* - * Create new FP context by updating CONTROL.FPCA, CONTROL.SFPA, - * the FPSCR, and VPR. - */ - TCGv_i32 control, fpscr; - uint32_t bits = R_V7M_CONTROL_FPCA_MASK; - - fpscr = load_cpu_field(v7m.fpdscr[s->v8m_secure]); - gen_helper_vfp_set_fpscr(cpu_env, fpscr); - tcg_temp_free_i32(fpscr); - if (dc_isar_feature(aa32_mve, s)) { - TCGv_i32 z32 = tcg_const_i32(0); - store_cpu_field(z32, v7m.vpr); - } - - /* - * We don't need to arrange to end the TB, because the only - * parts of FPSCR which we cache in the TB flags are the VECLEN - * and VECSTRIDE, and those don't exist for M-profile. - */ - - if (s->v8m_secure) { - bits |= R_V7M_CONTROL_SFPA_MASK; - } - control = load_cpu_field(v7m.control[M_REG_S]); - tcg_gen_ori_i32(control, control, bits); - store_cpu_field(control, v7m.control[M_REG_S]); - /* Don't need to do this for any further FP insns in this TB */ - s->v7m_new_fp_ctxt_needed = false; - } + /* Update ownership of FP context and create new FP context if needed */ + gen_update_fp_context(s); } return true; -- 2.20.1 vfp_access_check and its helper routine full_vfp_access_check() has gradually grown and is now an awkward mix of A-profile only and M-profile only pieces. Refactor it into an A-profile only and an M-profile only version, taking advantage of the fact that now the only direct call to full_vfp_access_check() is in A-profile-only code. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-7-peter.maydell@linaro.org --- target/arm/translate-vfp.c | 79 +++++++++++++++++++++++--------------- 1 file changed, 48 insertions(+), 31 deletions(-) diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static void gen_update_fp_context(DisasContext *s) } /* - * Check that VFP access is enabled. If it is, do the necessary - * M-profile lazy-FP handling and then return true. - * If not, emit code to generate an appropriate exception and - * return false. + * Check that VFP access is enabled, A-profile specific version. + * + * If VFP is enabled, return true. If not, emit code to generate an + * appropriate exception and return false. * The ignore_vfp_enabled argument specifies that we should ignore - * whether VFP is enabled via FPEXC[EN]: this should be true for FMXR/FMRX + * whether VFP is enabled via FPEXC.EN: this should be true for FMXR/FMRX * accesses to FPSID, FPEXC, MVFR0, MVFR1, MVFR2, and false for all other insns. */ -static bool full_vfp_access_check(DisasContext *s, bool ignore_vfp_enabled) +static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) { if (s->fp_excp_el) { - if (arm_dc_feature(s, ARM_FEATURE_M)) { - /* - * M-profile mostly catches the "FPU disabled" case early, in - * disas_m_nocp(), but a few insns (eg LCTP, WLSTP, DLSTP) - * which do coprocessor-checks are outside the large ranges of - * the encoding space handled by the patterns in m-nocp.decode, - * and for them we may need to raise NOCP here. - */ - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, - syn_uncategorized(), s->fp_excp_el); - } else { - gen_exception_insn(s, s->pc_curr, EXCP_UDEF, - syn_fp_access_trap(1, 0xe, false), - s->fp_excp_el); - } + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_fp_access_trap(1, 0xe, false), s->fp_excp_el); return false; } @@ -XXX,XX +XXX,XX @@ static bool full_vfp_access_check(DisasContext *s, bool ignore_vfp_enabled) unallocated_encoding(s); return false; } + return true; +} - if (arm_dc_feature(s, ARM_FEATURE_M)) { - /* Handle M-profile lazy FP state mechanics */ - - /* Trigger lazy-state preservation if necessary */ - gen_preserve_fp_state(s); - - /* Update ownership of FP context and create new FP context if needed */ - gen_update_fp_context(s); +/* + * Check that VFP access is enabled, M-profile specific version. + * + * If VFP is enabled, do the necessary M-profile lazy-FP handling and then + * return true. If not, emit code to generate an appropriate exception and + * return false. + */ +static bool vfp_access_check_m(DisasContext *s) +{ + if (s->fp_excp_el) { + /* + * M-profile mostly catches the "FPU disabled" case early, in + * disas_m_nocp(), but a few insns (eg LCTP, WLSTP, DLSTP) + * which do coprocessor-checks are outside the large ranges of + * the encoding space handled by the patterns in m-nocp.decode, + * and for them we may need to raise NOCP here. + */ + gen_exception_insn(s, s->pc_curr, EXCP_NOCP, + syn_uncategorized(), s->fp_excp_el); + return false; } + /* Handle M-profile lazy FP state mechanics */ + + /* Trigger lazy-state preservation if necessary */ + gen_preserve_fp_state(s); + + /* Update ownership of FP context and create new FP context if needed */ + gen_update_fp_context(s); + return true; } @@ -XXX,XX +XXX,XX @@ static bool full_vfp_access_check(DisasContext *s, bool ignore_vfp_enabled) */ bool vfp_access_check(DisasContext *s) { - return full_vfp_access_check(s, false); + if (arm_dc_feature(s, ARM_FEATURE_M)) { + return vfp_access_check_m(s); + } else { + return vfp_access_check_a(s, false); + } } static bool trans_VSEL(DisasContext *s, arg_VSEL *a) @@ -XXX,XX +XXX,XX @@ static bool trans_VMSR_VMRS(DisasContext *s, arg_VMSR_VMRS *a) return false; } - if (!full_vfp_access_check(s, ignore_vfp_enabled)) { + /* + * Call vfp_access_check_a() directly, because we need to tell + * it to ignore FPEXC.EN for some register accesses. + */ + if (!vfp_access_check_a(s, ignore_vfp_enabled)) { return true; } -- 2.20.1 Instead of open-coding the "take NOCP exception if FPU disabled, otherwise call gen_preserve_fp_state()" code in the accessors for FPCXT_NS, add an argument to vfp_access_check_m() which tells it to skip the gen_update_fp_context() call, so we can use it for the FPCXT_NS case. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210618141019.10671-8-peter.maydell@linaro.org --- target/arm/translate-a32.h | 2 +- target/arm/translate-m-nocp.c | 10 ++-------- target/arm/translate-vfp.c | 13 ++++++++----- 3 files changed, 11 insertions(+), 14 deletions(-) diff --git a/target/arm/translate-a32.h b/target/arm/translate-a32.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a32.h +++ b/target/arm/translate-a32.h @@ -XXX,XX +XXX,XX @@ bool disas_neon_shared(DisasContext *s, uint32_t insn); void load_reg_var(DisasContext *s, TCGv_i32 var, int reg); void arm_gen_condlabel(DisasContext *s); bool vfp_access_check(DisasContext *s); -void gen_preserve_fp_state(DisasContext *s); +bool vfp_access_check_m(DisasContext *s, bool skip_context_update); void read_neon_element32(TCGv_i32 dest, int reg, int ele, MemOp memop); void read_neon_element64(TCGv_i64 dest, int reg, int ele, MemOp memop); void write_neon_element32(TCGv_i32 src, int reg, int ele, MemOp memop); diff --git a/target/arm/translate-m-nocp.c b/target/arm/translate-m-nocp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-m-nocp.c +++ b/target/arm/translate-m-nocp.c @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, * otherwise PreserveFPState(), and then FPCXT_NS writes * behave the same as FPCXT_S writes. */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, - syn_uncategorized(), s->fp_excp_el); + if (!vfp_access_check_m(s, true)) { /* * This was only a conditional exception, so override * gen_exception_insn()'s default to DISAS_NORETURN @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_write(DisasContext *s, int regno, s->base.is_jmp = DISAS_NEXT; break; } - gen_preserve_fp_state(s); } /* fall through */ case ARM_VFP_FPCXT_S: @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, * otherwise PreserveFPState(), and then FPCXT_NS * reads the same as FPCXT_S. */ - if (s->fp_excp_el) { - gen_exception_insn(s, s->pc_curr, EXCP_NOCP, - syn_uncategorized(), s->fp_excp_el); + if (!vfp_access_check_m(s, true)) { /* * This was only a conditional exception, so override * gen_exception_insn()'s default to DISAS_NORETURN @@ -XXX,XX +XXX,XX @@ static bool gen_M_fp_sysreg_read(DisasContext *s, int regno, s->base.is_jmp = DISAS_NEXT; break; } - gen_preserve_fp_state(s); tmp = tcg_temp_new_i32(); sfpa = tcg_temp_new_i32(); fpscr = tcg_temp_new_i32(); diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static inline long vfp_f16_offset(unsigned reg, bool top) * Generate code for M-profile lazy FP state preservation if needed; * this corresponds to the pseudocode PreserveFPState() function. */ -void gen_preserve_fp_state(DisasContext *s) +static void gen_preserve_fp_state(DisasContext *s) { if (s->v7m_lspact) { /* @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) * If VFP is enabled, do the necessary M-profile lazy-FP handling and then * return true. If not, emit code to generate an appropriate exception and * return false. + * skip_context_update is true to skip the "update FP context" part of this. */ -static bool vfp_access_check_m(DisasContext *s) +bool vfp_access_check_m(DisasContext *s, bool skip_context_update) { if (s->fp_excp_el) { /* @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_m(DisasContext *s) /* Trigger lazy-state preservation if necessary */ gen_preserve_fp_state(s); - /* Update ownership of FP context and create new FP context if needed */ - gen_update_fp_context(s); + if (!skip_context_update) { + /* Update ownership of FP context and create new FP context if needed */ + gen_update_fp_context(s); + } return true; } @@ -XXX,XX +XXX,XX @@ static bool vfp_access_check_m(DisasContext *s) bool vfp_access_check(DisasContext *s) { if (arm_dc_feature(s, ARM_FEATURE_M)) { - return vfp_access_check_m(s); + return vfp_access_check_m(s, false); } else { return vfp_access_check_a(s, false); } -- 2.20.1 Implement the forms of the MVE VLDR and VSTR insns which perform non-widening loads of bytes, halfwords or words from memory into vector elements of the same width (encodings T5, T6, T7). (At the moment we know for MVE and M-profile in general that vfp_access_check() can never return false, but we include the conventional return-true-on-failure check for consistency with non-M-profile translation code.) Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-2-peter.maydell@linaro.org --- target/arm/{translate-mve.c => helper-mve.h} | 19 +- target/arm/helper.h | 2 + target/arm/internals.h | 11 ++ target/arm/mve.decode | 22 +++ target/arm/mve_helper.c | 172 +++++++++++++++++++ target/arm/translate-mve.c | 119 +++++++++++++ target/arm/meson.build | 1 + 7 files changed, 334 insertions(+), 12 deletions(-) copy target/arm/{translate-mve.c => helper-mve.h} (61%) create mode 100644 target/arm/mve_helper.c diff --git a/target/arm/translate-mve.c b/target/arm/helper-mve.h similarity index 61% copy from target/arm/translate-mve.c copy to target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ /* - * ARM translation: M-profile MVE instructions + * M-profile MVE specific helper definitions * * Copyright (c) 2021 Linaro, Ltd. * @@ -XXX,XX +XXX,XX @@ * 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/>. */ - -#include "qemu/osdep.h" -#include "tcg/tcg-op.h" -#include "tcg/tcg-op-gvec.h" -#include "exec/exec-all.h" -#include "exec/gen-icount.h" -#include "translate.h" -#include "translate-a32.h" - -/* Include the generated decoder */ -#include "decode-mve.c.inc" +DEF_HELPER_FLAGS_3(mve_vldrb, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrh, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrw, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrb, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrh, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrw, TCG_CALL_NO_WG, void, env, 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_6(gvec_bfmlal_idx, TCG_CALL_NO_RWG, #include "helper-a64.h" #include "helper-sve.h" #endif + +#include "helper-mve.h" diff --git a/target/arm/internals.h b/target/arm/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/internals.h +++ b/target/arm/internals.h @@ -XXX,XX +XXX,XX @@ static inline uint64_t useronly_maybe_clean_ptr(uint32_t desc, uint64_t ptr) return ptr; } +/* Values for M-profile PSR.ECI for MVE insns */ +enum MVEECIState { + ECI_NONE = 0, /* No completed beats */ + ECI_A0 = 1, /* Completed: A0 */ + ECI_A0A1 = 2, /* Completed: A0, A1 */ + /* 3 is reserved */ + ECI_A0A1A2 = 4, /* Completed: A0, A1, A2 */ + ECI_A0A1A2B0 = 5, /* Completed: A0, A1, A2, B0 */ + /* All other values reserved */ +}; + #endif diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ # # This file is processed by scripts/decodetree.py # + +%qd 22:1 13:3 + +&vldr_vstr rn qd imm p a w size l + +@vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd + +# Vector loads and stores + +# Non-widening loads/stores (P=0 W=0 is 'related encoding') +VLDR_VSTR 1110110 0 a:1 . 1 . .... ... 111100 ....... @vldr_vstr \ + size=0 p=0 w=1 +VLDR_VSTR 1110110 0 a:1 . 1 . .... ... 111101 ....... @vldr_vstr \ + size=1 p=0 w=1 +VLDR_VSTR 1110110 0 a:1 . 1 . .... ... 111110 ....... @vldr_vstr \ + size=2 p=0 w=1 +VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111100 ....... @vldr_vstr \ + size=0 p=1 +VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111101 ....... @vldr_vstr \ + size=1 p=1 +VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111110 ....... @vldr_vstr \ + size=2 p=1 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ +/* + * M-profile MVE Operations + * + * Copyright (c) 2021 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 2.1 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/>. + */ + +#include "qemu/osdep.h" +#include "cpu.h" +#include "internals.h" +#include "vec_internal.h" +#include "exec/helper-proto.h" +#include "exec/cpu_ldst.h" +#include "exec/exec-all.h" + +static uint16_t mve_element_mask(CPUARMState *env) +{ + /* + * Return the mask of which elements in the MVE vector should be + * updated. This is a combination of multiple things: + * (1) by default, we update every lane in the vector + * (2) VPT predication stores its state in the VPR register; + * (3) low-overhead-branch tail predication will mask out part + * the vector on the final iteration of the loop + * (4) if EPSR.ECI is set then we must execute only some beats + * of the insn + * We combine all these into a 16-bit result with the same semantics + * as VPR.P0: 0 to mask the lane, 1 if it is active. + * 8-bit vector ops will look at all bits of the result; + * 16-bit ops will look at bits 0, 2, 4, ...; + * 32-bit ops will look at bits 0, 4, 8 and 12. + * Compare pseudocode GetCurInstrBeat(), though that only returns + * the 4-bit slice of the mask corresponding to a single beat. + */ + uint16_t mask = FIELD_EX32(env->v7m.vpr, V7M_VPR, P0); + + if (!(env->v7m.vpr & R_V7M_VPR_MASK01_MASK)) { + mask |= 0xff; + } + if (!(env->v7m.vpr & R_V7M_VPR_MASK23_MASK)) { + mask |= 0xff00; + } + + if (env->v7m.ltpsize < 4 && + env->regs[14] <= (1 << (4 - env->v7m.ltpsize))) { + /* + * Tail predication active, and this is the last loop iteration. + * The element size is (1 << ltpsize), and we only want to process + * loopcount elements, so we want to retain the least significant + * (loopcount * esize) predicate bits and zero out bits above that. + */ + int masklen = env->regs[14] << env->v7m.ltpsize; + assert(masklen <= 16); + mask &= MAKE_64BIT_MASK(0, masklen); + } + + if ((env->condexec_bits & 0xf) == 0) { + /* + * ECI bits indicate which beats are already executed; + * we handle this by effectively predicating them out. + */ + int eci = env->condexec_bits >> 4; + switch (eci) { + case ECI_NONE: + break; + case ECI_A0: + mask &= 0xfff0; + break; + case ECI_A0A1: + mask &= 0xff00; + break; + case ECI_A0A1A2: + case ECI_A0A1A2B0: + mask &= 0xf000; + break; + default: + g_assert_not_reached(); + } + } + + return mask; +} + +static void mve_advance_vpt(CPUARMState *env) +{ + /* Advance the VPT and ECI state if necessary */ + uint32_t vpr = env->v7m.vpr; + unsigned mask01, mask23; + + if ((env->condexec_bits & 0xf) == 0) { + env->condexec_bits = (env->condexec_bits == (ECI_A0A1A2B0 << 4)) ? + (ECI_A0 << 4) : (ECI_NONE << 4); + } + + if (!(vpr & (R_V7M_VPR_MASK01_MASK | R_V7M_VPR_MASK23_MASK))) { + /* VPT not enabled, nothing to do */ + return; + } + + mask01 = FIELD_EX32(vpr, V7M_VPR, MASK01); + mask23 = FIELD_EX32(vpr, V7M_VPR, MASK23); + if (mask01 > 8) { + /* high bit set, but not 0b1000: invert the relevant half of P0 */ + vpr ^= 0xff; + } + if (mask23 > 8) { + /* high bit set, but not 0b1000: invert the relevant half of P0 */ + vpr ^= 0xff00; + } + vpr = FIELD_DP32(vpr, V7M_VPR, MASK01, mask01 << 1); + vpr = FIELD_DP32(vpr, V7M_VPR, MASK23, mask23 << 1); + env->v7m.vpr = vpr; +} + + +#define DO_VLDR(OP, MSIZE, LDTYPE, ESIZE, TYPE) \ + void HELPER(mve_##OP)(CPUARMState *env, void *vd, uint32_t addr) \ + { \ + TYPE *d = vd; \ + uint16_t mask = mve_element_mask(env); \ + unsigned b, e; \ + /* \ + * R_SXTM allows the dest reg to become UNKNOWN for abandoned \ + * beats so we don't care if we update part of the dest and \ + * then take an exception. \ + */ \ + for (b = 0, e = 0; b < 16; b += ESIZE, e++) { \ + if (mask & (1 << b)) { \ + d[H##ESIZE(e)] = cpu_##LDTYPE##_data_ra(env, addr, GETPC()); \ + } \ + addr += MSIZE; \ + } \ + mve_advance_vpt(env); \ + } + +#define DO_VSTR(OP, MSIZE, STTYPE, ESIZE, TYPE) \ + void HELPER(mve_##OP)(CPUARMState *env, void *vd, uint32_t addr) \ + { \ + TYPE *d = vd; \ + uint16_t mask = mve_element_mask(env); \ + unsigned b, e; \ + for (b = 0, e = 0; b < 16; b += ESIZE, e++) { \ + if (mask & (1 << b)) { \ + cpu_##STTYPE##_data_ra(env, addr, d[H##ESIZE(e)], GETPC()); \ + } \ + addr += MSIZE; \ + } \ + mve_advance_vpt(env); \ + } + +DO_VLDR(vldrb, 1, ldub, 1, uint8_t) +DO_VLDR(vldrh, 2, lduw, 2, uint16_t) +DO_VLDR(vldrw, 4, ldl, 4, uint32_t) + +DO_VSTR(vstrb, 1, stb, 1, uint8_t) +DO_VSTR(vstrh, 2, stw, 2, uint16_t) +DO_VSTR(vstrw, 4, stl, 4, uint32_t) + +#undef DO_VLDR +#undef DO_VSTR diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ /* Include the generated decoder */ #include "decode-mve.c.inc" + +typedef void MVEGenLdStFn(TCGv_ptr, TCGv_ptr, TCGv_i32); + +/* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ +static inline long mve_qreg_offset(unsigned reg) +{ + return offsetof(CPUARMState, vfp.zregs[reg].d[0]); +} + +static TCGv_ptr mve_qreg_ptr(unsigned reg) +{ + TCGv_ptr ret = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(ret, cpu_env, mve_qreg_offset(reg)); + return ret; +} + +static bool mve_check_qreg_bank(DisasContext *s, int qmask) +{ + /* + * Check whether Qregs are in range. For v8.1M only Q0..Q7 + * are supported, see VFPSmallRegisterBank(). + */ + return qmask < 8; +} + +static bool mve_eci_check(DisasContext *s) +{ + /* + * This is a beatwise insn: check that ECI is valid (not a + * reserved value) and note that we are handling it. + * Return true if OK, false if we generated an exception. + */ + s->eci_handled = true; + switch (s->eci) { + case ECI_NONE: + case ECI_A0: + case ECI_A0A1: + case ECI_A0A1A2: + case ECI_A0A1A2B0: + return true; + default: + /* Reserved value: INVSTATE UsageFault */ + gen_exception_insn(s, s->pc_curr, EXCP_INVSTATE, syn_uncategorized(), + default_exception_el(s)); + return false; + } +} + +static void mve_update_eci(DisasContext *s) +{ + /* + * The helper function will always update the CPUState field, + * so we only need to update the DisasContext field. + */ + if (s->eci) { + s->eci = (s->eci == ECI_A0A1A2B0) ? ECI_A0 : ECI_NONE; + } +} + +static bool do_ldst(DisasContext *s, arg_VLDR_VSTR *a, MVEGenLdStFn *fn) +{ + TCGv_i32 addr; + uint32_t offset; + TCGv_ptr qreg; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qd) || + !fn) { + return false; + } + + /* CONSTRAINED UNPREDICTABLE: we choose to UNDEF */ + if (a->rn == 15 || (a->rn == 13 && a->w)) { + return false; + } + + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + offset = a->imm << a->size; + if (!a->a) { + offset = -offset; + } + addr = load_reg(s, a->rn); + if (a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + + qreg = mve_qreg_ptr(a->qd); + fn(cpu_env, qreg, addr); + tcg_temp_free_ptr(qreg); + + /* + * Writeback always happens after the last beat of the insn, + * regardless of predication + */ + if (a->w) { + if (!a->p) { + tcg_gen_addi_i32(addr, addr, offset); + } + store_reg(s, a->rn, addr); + } else { + tcg_temp_free_i32(addr); + } + mve_update_eci(s); + return true; +} + +static bool trans_VLDR_VSTR(DisasContext *s, arg_VLDR_VSTR *a) +{ + static MVEGenLdStFn * const ldstfns[4][2] = { + { gen_helper_mve_vstrb, gen_helper_mve_vldrb }, + { gen_helper_mve_vstrh, gen_helper_mve_vldrh }, + { gen_helper_mve_vstrw, gen_helper_mve_vldrw }, + { NULL, NULL } + }; + return do_ldst(s, a, ldstfns[a->size][a->l]); +} diff --git a/target/arm/meson.build b/target/arm/meson.build index XXXXXXX..XXXXXXX 100644 --- a/target/arm/meson.build +++ b/target/arm/meson.build @@ -XXX,XX +XXX,XX @@ arm_ss.add(files( 'helper.c', 'iwmmxt_helper.c', 'm_helper.c', + 'mve_helper.c', 'neon_helper.c', 'op_helper.c', 'tlb_helper.c', -- 2.20.1 Implement the variants of MVE VLDR (encodings T1, T2) which perform "widening" loads where bytes or halfwords are loaded from memory and zero or sign-extended into halfword or word length vector elements, and the narrowing MVE VSTR (encodings T1, T2) where bytes or halfwords are stored from halfword or word elements. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-3-peter.maydell@linaro.org --- target/arm/helper-mve.h | 10 ++++++++++ target/arm/mve.decode | 25 +++++++++++++++++++++++-- target/arm/mve_helper.c | 11 +++++++++++ target/arm/translate-mve.c | 14 ++++++++++++++ 4 files changed, 58 insertions(+), 2 deletions(-) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vldrw, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrb, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrh, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrw, TCG_CALL_NO_WG, void, env, ptr, i32) + +DEF_HELPER_FLAGS_3(mve_vldrb_sh, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrb_sw, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrb_uh, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrb_uw, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrh_sw, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vldrh_uw, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrb_h, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrb_w, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vstrh_w, TCG_CALL_NO_WG, void, env, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ %qd 22:1 13:3 -&vldr_vstr rn qd imm p a w size l +&vldr_vstr rn qd imm p a w size l u -@vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd +@vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd u=0 +# Note that both Rn and Qd are 3 bits only (no D bit) +@vldst_wn ... u:1 ... . . . . l:1 . rn:3 qd:3 . ... .. imm:7 &vldr_vstr # Vector loads and stores +# Widening loads and narrowing stores: +# for these P=0 W=0 is 'related encoding'; sz=11 is 'related encoding' +# This means we need to expand out to multiple patterns for P, W, SZ. +# For stores the U bit must be 0 but we catch that in the trans_ function. +# The naming scheme here is "VLDSTB_H == in-memory byte load/store to/from +# signed halfword element in register", etc. +VLDSTB_H 111 . 110 0 a:1 0 1 . 0 ... ... 0 111 01 ....... @vldst_wn \ + p=0 w=1 size=1 +VLDSTB_H 111 . 110 1 a:1 0 w:1 . 0 ... ... 0 111 01 ....... @vldst_wn \ + p=1 size=1 +VLDSTB_W 111 . 110 0 a:1 0 1 . 0 ... ... 0 111 10 ....... @vldst_wn \ + p=0 w=1 size=2 +VLDSTB_W 111 . 110 1 a:1 0 w:1 . 0 ... ... 0 111 10 ....... @vldst_wn \ + p=1 size=2 +VLDSTH_W 111 . 110 0 a:1 0 1 . 1 ... ... 0 111 10 ....... @vldst_wn \ + p=0 w=1 size=2 +VLDSTH_W 111 . 110 1 a:1 0 w:1 . 1 ... ... 0 111 10 ....... @vldst_wn \ + p=1 size=2 + # Non-widening loads/stores (P=0 W=0 is 'related encoding') VLDR_VSTR 1110110 0 a:1 . 1 . .... ... 111100 ....... @vldr_vstr \ size=0 p=0 w=1 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_VSTR(vstrb, 1, stb, 1, uint8_t) DO_VSTR(vstrh, 2, stw, 2, uint16_t) DO_VSTR(vstrw, 4, stl, 4, uint32_t) +DO_VLDR(vldrb_sh, 1, ldsb, 2, int16_t) +DO_VLDR(vldrb_sw, 1, ldsb, 4, int32_t) +DO_VLDR(vldrb_uh, 1, ldub, 2, uint16_t) +DO_VLDR(vldrb_uw, 1, ldub, 4, uint32_t) +DO_VLDR(vldrh_sw, 2, ldsw, 4, int32_t) +DO_VLDR(vldrh_uw, 2, lduw, 4, uint32_t) + +DO_VSTR(vstrb_h, 1, stb, 2, int16_t) +DO_VSTR(vstrb_w, 1, stb, 4, int32_t) +DO_VSTR(vstrh_w, 2, stw, 4, int32_t) + #undef DO_VLDR #undef DO_VSTR diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool trans_VLDR_VSTR(DisasContext *s, arg_VLDR_VSTR *a) }; return do_ldst(s, a, ldstfns[a->size][a->l]); } + +#define DO_VLDST_WIDE_NARROW(OP, SLD, ULD, ST) \ + static bool trans_##OP(DisasContext *s, arg_VLDR_VSTR *a) \ + { \ + static MVEGenLdStFn * const ldstfns[2][2] = { \ + { gen_helper_mve_##ST, gen_helper_mve_##SLD }, \ + { NULL, gen_helper_mve_##ULD }, \ + }; \ + return do_ldst(s, a, ldstfns[a->u][a->l]); \ + } + +DO_VLDST_WIDE_NARROW(VLDSTB_H, vldrb_sh, vldrb_uh, vstrb_h) +DO_VLDST_WIDE_NARROW(VLDSTB_W, vldrb_sw, vldrb_uw, vstrb_w) +DO_VLDST_WIDE_NARROW(VLDSTH_W, vldrh_sw, vldrh_uw, vstrh_w) -- 2.20.1 Implement the MVE VCLZ insn (and the necessary machinery for MVE 1-input vector ops). Note that for non-load instructions predication is always performed at a byte level granularity regardless of element size (R_ZLSJ), and so the masking logic here differs from that used in the VLDR and VSTR helpers. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-4-peter.maydell@linaro.org --- target/arm/helper-mve.h | 4 ++ target/arm/mve.decode | 8 ++++ target/arm/mve_helper.c | 82 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 38 ++++++++++++++++++ 4 files changed, 132 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vldrh_uw, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrb_h, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrb_w, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrh_w, TCG_CALL_NO_WG, void, env, ptr, i32) + +DEF_HELPER_FLAGS_3(mve_vclzb, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vclzh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vclzw, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ # %qd 22:1 13:3 +%qm 5:1 1:3 &vldr_vstr rn qd imm p a w size l u +&1op qd qm size @vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd u=0 # Note that both Rn and Qd are 3 bits only (no D bit) @vldst_wn ... u:1 ... . . . . l:1 . rn:3 qd:3 . ... .. imm:7 &vldr_vstr +@1op .... .... .... size:2 .. .... .... .... .... &1op qd=%qd qm=%qm + # Vector loads and stores # Widening loads and narrowing stores: @@ -XXX,XX +XXX,XX @@ VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111101 ....... @vldr_vstr \ size=1 p=1 VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111110 ....... @vldr_vstr \ size=2 p=1 + +# Vector miscellaneous + +VCLZ 1111 1111 1 . 11 .. 00 ... 0 0100 11 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_VSTR(vstrh_w, 2, stw, 4, int32_t) #undef DO_VLDR #undef DO_VSTR + +/* + * The mergemask(D, R, M) macro performs the operation "*D = R" but + * storing only the bytes which correspond to 1 bits in M, + * leaving other bytes in *D unchanged. We use _Generic + * to select the correct implementation based on the type of D. + */ + +static void mergemask_ub(uint8_t *d, uint8_t r, uint16_t mask) +{ + if (mask & 1) { + *d = r; + } +} + +static void mergemask_sb(int8_t *d, int8_t r, uint16_t mask) +{ + mergemask_ub((uint8_t *)d, r, mask); +} + +static void mergemask_uh(uint16_t *d, uint16_t r, uint16_t mask) +{ + uint16_t bmask = expand_pred_b_data[mask & 3]; + *d = (*d & ~bmask) | (r & bmask); +} + +static void mergemask_sh(int16_t *d, int16_t r, uint16_t mask) +{ + mergemask_uh((uint16_t *)d, r, mask); +} + +static void mergemask_uw(uint32_t *d, uint32_t r, uint16_t mask) +{ + uint32_t bmask = expand_pred_b_data[mask & 0xf]; + *d = (*d & ~bmask) | (r & bmask); +} + +static void mergemask_sw(int32_t *d, int32_t r, uint16_t mask) +{ + mergemask_uw((uint32_t *)d, r, mask); +} + +static void mergemask_uq(uint64_t *d, uint64_t r, uint16_t mask) +{ + uint64_t bmask = expand_pred_b_data[mask & 0xff]; + *d = (*d & ~bmask) | (r & bmask); +} + +static void mergemask_sq(int64_t *d, int64_t r, uint16_t mask) +{ + mergemask_uq((uint64_t *)d, r, mask); +} + +#define mergemask(D, R, M) \ + _Generic(D, \ + uint8_t *: mergemask_ub, \ + int8_t *: mergemask_sb, \ + uint16_t *: mergemask_uh, \ + int16_t *: mergemask_sh, \ + uint32_t *: mergemask_uw, \ + int32_t *: mergemask_sw, \ + uint64_t *: mergemask_uq, \ + int64_t *: mergemask_sq)(D, R, M) + +#define DO_1OP(OP, ESIZE, TYPE, FN) \ + void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm) \ + { \ + TYPE *d = vd, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + mergemask(&d[H##ESIZE(e)], FN(m[H##ESIZE(e)]), mask); \ + } \ + mve_advance_vpt(env); \ + } + +#define DO_CLZ_B(N) (clz32(N) - 24) +#define DO_CLZ_H(N) (clz32(N) - 16) + +DO_1OP(vclzb, 1, uint8_t, DO_CLZ_B) +DO_1OP(vclzh, 2, uint16_t, DO_CLZ_H) +DO_1OP(vclzw, 4, uint32_t, clz32) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ #include "decode-mve.c.inc" typedef void MVEGenLdStFn(TCGv_ptr, TCGv_ptr, TCGv_i32); +typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ static inline long mve_qreg_offset(unsigned reg) @@ -XXX,XX +XXX,XX @@ static bool trans_VLDR_VSTR(DisasContext *s, arg_VLDR_VSTR *a) DO_VLDST_WIDE_NARROW(VLDSTB_H, vldrb_sh, vldrb_uh, vstrb_h) DO_VLDST_WIDE_NARROW(VLDSTB_W, vldrb_sw, vldrb_uw, vstrb_w) DO_VLDST_WIDE_NARROW(VLDSTH_W, vldrh_sw, vldrh_uw, vstrh_w) + +static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) +{ + TCGv_ptr qd, qm; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qd | a->qm) || + !fn) { + return false; + } + + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + qd = mve_qreg_ptr(a->qd); + qm = mve_qreg_ptr(a->qm); + fn(cpu_env, qd, qm); + tcg_temp_free_ptr(qd); + tcg_temp_free_ptr(qm); + mve_update_eci(s); + return true; +} + +#define DO_1OP(INSN, FN) \ + static bool trans_##INSN(DisasContext *s, arg_1op *a) \ + { \ + static MVEGenOneOpFn * const fns[] = { \ + gen_helper_mve_##FN##b, \ + gen_helper_mve_##FN##h, \ + gen_helper_mve_##FN##w, \ + NULL, \ + }; \ + return do_1op(s, a, fns[a->size]); \ + } + +DO_1OP(VCLZ, vclz) -- 2.20.1 Implement the MVE VCLS insn. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-5-peter.maydell@linaro.org --- target/arm/helper-mve.h | 4 ++++ target/arm/mve.decode | 1 + target/arm/mve_helper.c | 7 +++++++ target/arm/translate-mve.c | 1 + 4 files changed, 13 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vstrb_h, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrb_w, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrh_w, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vclsb, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vclsh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vclsw, TCG_CALL_NO_WG, void, env, ptr, ptr) + DEF_HELPER_FLAGS_3(mve_vclzb, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclzh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclzw, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111110 ....... @vldr_vstr \ # Vector miscellaneous +VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op VCLZ 1111 1111 1 . 11 .. 00 ... 0 0100 11 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ static void mergemask_sq(int64_t *d, int64_t r, uint16_t mask) mve_advance_vpt(env); \ } +#define DO_CLS_B(N) (clrsb32(N) - 24) +#define DO_CLS_H(N) (clrsb32(N) - 16) + +DO_1OP(vclsb, 1, int8_t, DO_CLS_B) +DO_1OP(vclsh, 2, int16_t, DO_CLS_H) +DO_1OP(vclsw, 4, int32_t, clrsb32) + #define DO_CLZ_B(N) (clz32(N) - 24) #define DO_CLZ_H(N) (clz32(N) - 16) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) } DO_1OP(VCLZ, vclz) +DO_1OP(VCLS, vcls) -- 2.20.1 Implement the MVE instructions VREV16, VREV32 and VREV64. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-6-peter.maydell@linaro.org --- target/arm/helper-mve.h | 7 +++++++ target/arm/mve.decode | 4 ++++ target/arm/mve_helper.c | 7 +++++++ target/arm/translate-mve.c | 33 +++++++++++++++++++++++++++++++++ 4 files changed, 51 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vclsw, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclzb, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclzh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclzw, TCG_CALL_NO_WG, void, env, ptr, ptr) + +DEF_HELPER_FLAGS_3(mve_vrev16b, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vrev32b, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vrev32h, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vrev64b, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vrev64h, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vrev64w, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111110 ....... @vldr_vstr \ VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op VCLZ 1111 1111 1 . 11 .. 00 ... 0 0100 11 . 0 ... 0 @1op + +VREV16 1111 1111 1 . 11 .. 00 ... 0 0001 01 . 0 ... 0 @1op +VREV32 1111 1111 1 . 11 .. 00 ... 0 0000 11 . 0 ... 0 @1op +VREV64 1111 1111 1 . 11 .. 00 ... 0 0000 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vclsw, 4, int32_t, clrsb32) DO_1OP(vclzb, 1, uint8_t, DO_CLZ_B) DO_1OP(vclzh, 2, uint16_t, DO_CLZ_H) DO_1OP(vclzw, 4, uint32_t, clz32) + +DO_1OP(vrev16b, 2, uint16_t, bswap16) +DO_1OP(vrev32b, 4, uint32_t, bswap32) +DO_1OP(vrev32h, 4, uint32_t, hswap32) +DO_1OP(vrev64b, 8, uint64_t, bswap64) +DO_1OP(vrev64h, 8, uint64_t, hswap64) +DO_1OP(vrev64w, 8, uint64_t, wswap64) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) DO_1OP(VCLZ, vclz) DO_1OP(VCLS, vcls) + +static bool trans_VREV16(DisasContext *s, arg_1op *a) +{ + static MVEGenOneOpFn * const fns[] = { + gen_helper_mve_vrev16b, + NULL, + NULL, + NULL, + }; + return do_1op(s, a, fns[a->size]); +} + +static bool trans_VREV32(DisasContext *s, arg_1op *a) +{ + static MVEGenOneOpFn * const fns[] = { + gen_helper_mve_vrev32b, + gen_helper_mve_vrev32h, + NULL, + NULL, + }; + return do_1op(s, a, fns[a->size]); +} + +static bool trans_VREV64(DisasContext *s, arg_1op *a) +{ + static MVEGenOneOpFn * const fns[] = { + gen_helper_mve_vrev64b, + gen_helper_mve_vrev64h, + gen_helper_mve_vrev64w, + NULL, + }; + return do_1op(s, a, fns[a->size]); +} -- 2.20.1 Implement the MVE VMVN(register) operation. Note that for predication this operation is byte-by-byte. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-7-peter.maydell@linaro.org --- target/arm/helper-mve.h | 2 ++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 4 ++++ target/arm/translate-mve.c | 5 +++++ 4 files changed, 14 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vrev32h, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vrev64b, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vrev64h, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vrev64w, TCG_CALL_NO_WG, void, env, ptr, ptr) + +DEF_HELPER_FLAGS_3(mve_vmvn, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ @vldst_wn ... u:1 ... . . . . l:1 . rn:3 qd:3 . ... .. imm:7 &vldr_vstr @1op .... .... .... size:2 .. .... .... .... .... &1op qd=%qd qm=%qm +@1op_nosz .... .... .... .... .... .... .... .... &1op qd=%qd qm=%qm size=0 # Vector loads and stores @@ -XXX,XX +XXX,XX @@ VCLZ 1111 1111 1 . 11 .. 00 ... 0 0100 11 . 0 ... 0 @1op VREV16 1111 1111 1 . 11 .. 00 ... 0 0001 01 . 0 ... 0 @1op VREV32 1111 1111 1 . 11 .. 00 ... 0 0000 11 . 0 ... 0 @1op VREV64 1111 1111 1 . 11 .. 00 ... 0 0000 01 . 0 ... 0 @1op + +VMVN 1111 1111 1 . 11 00 00 ... 0 0101 11 . 0 ... 0 @1op_nosz diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vrev32h, 4, uint32_t, hswap32) DO_1OP(vrev64b, 8, uint64_t, bswap64) DO_1OP(vrev64h, 8, uint64_t, hswap64) DO_1OP(vrev64w, 8, uint64_t, wswap64) + +#define DO_NOT(N) (~(N)) + +DO_1OP(vmvn, 8, uint64_t, DO_NOT) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool trans_VREV64(DisasContext *s, arg_1op *a) }; return do_1op(s, a, fns[a->size]); } + +static bool trans_VMVN(DisasContext *s, arg_1op *a) +{ + return do_1op(s, a, gen_helper_mve_vmvn); +} -- 2.20.1 Implement the MVE VABS functions (both integer and floating point). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-8-peter.maydell@linaro.org --- target/arm/helper-mve.h | 6 ++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 13 +++++++++++++ target/arm/translate-mve.c | 15 +++++++++++++++ 4 files changed, 37 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vrev64h, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vrev64w, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vmvn, TCG_CALL_NO_WG, void, env, ptr, ptr) + +DEF_HELPER_FLAGS_3(mve_vabsb, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vabsh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vabsw, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vfabsh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vfabss, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VREV32 1111 1111 1 . 11 .. 00 ... 0 0000 11 . 0 ... 0 @1op VREV64 1111 1111 1 . 11 .. 00 ... 0 0000 01 . 0 ... 0 @1op VMVN 1111 1111 1 . 11 00 00 ... 0 0101 11 . 0 ... 0 @1op_nosz + +VABS 1111 1111 1 . 11 .. 01 ... 0 0011 01 . 0 ... 0 @1op +VABS_fp 1111 1111 1 . 11 .. 01 ... 0 0111 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ #include "exec/helper-proto.h" #include "exec/cpu_ldst.h" #include "exec/exec-all.h" +#include "tcg/tcg.h" static uint16_t mve_element_mask(CPUARMState *env) { @@ -XXX,XX +XXX,XX @@ DO_1OP(vrev64w, 8, uint64_t, wswap64) #define DO_NOT(N) (~(N)) DO_1OP(vmvn, 8, uint64_t, DO_NOT) + +#define DO_ABS(N) ((N) < 0 ? -(N) : (N)) +#define DO_FABSH(N) ((N) & dup_const(MO_16, 0x7fff)) +#define DO_FABSS(N) ((N) & dup_const(MO_32, 0x7fffffff)) + +DO_1OP(vabsb, 1, int8_t, DO_ABS) +DO_1OP(vabsh, 2, int16_t, DO_ABS) +DO_1OP(vabsw, 4, int32_t, DO_ABS) + +/* We can do these 64 bits at a time */ +DO_1OP(vfabsh, 8, uint64_t, DO_FABSH) +DO_1OP(vfabss, 8, uint64_t, DO_FABSS) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) DO_1OP(VCLZ, vclz) DO_1OP(VCLS, vcls) +DO_1OP(VABS, vabs) static bool trans_VREV16(DisasContext *s, arg_1op *a) { @@ -XXX,XX +XXX,XX @@ static bool trans_VMVN(DisasContext *s, arg_1op *a) { return do_1op(s, a, gen_helper_mve_vmvn); } + +static bool trans_VABS_fp(DisasContext *s, arg_1op *a) +{ + static MVEGenOneOpFn * const fns[] = { + NULL, + gen_helper_mve_vfabsh, + gen_helper_mve_vfabss, + NULL, + }; + if (!dc_isar_feature(aa32_mve_fp, s)) { + return false; + } + return do_1op(s, a, fns[a->size]); +} -- 2.20.1 Implement the MVE VNEG insn (both integer and floating point forms). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-9-peter.maydell@linaro.org --- target/arm/helper-mve.h | 6 ++++++ target/arm/mve.decode | 2 ++ target/arm/mve_helper.c | 12 ++++++++++++ target/arm/translate-mve.c | 15 +++++++++++++++ 4 files changed, 35 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vabsh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vabsw, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vfabsh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vfabss, TCG_CALL_NO_WG, void, env, ptr, ptr) + +DEF_HELPER_FLAGS_3(mve_vnegb, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vnegh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vnegw, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vfnegh, TCG_CALL_NO_WG, void, env, ptr, ptr) +DEF_HELPER_FLAGS_3(mve_vfnegs, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMVN 1111 1111 1 . 11 00 00 ... 0 0101 11 . 0 ... 0 @1op_nosz VABS 1111 1111 1 . 11 .. 01 ... 0 0011 01 . 0 ... 0 @1op VABS_fp 1111 1111 1 . 11 .. 01 ... 0 0111 01 . 0 ... 0 @1op +VNEG 1111 1111 1 . 11 .. 01 ... 0 0011 11 . 0 ... 0 @1op +VNEG_fp 1111 1111 1 . 11 .. 01 ... 0 0111 11 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vabsw, 4, int32_t, DO_ABS) /* We can do these 64 bits at a time */ DO_1OP(vfabsh, 8, uint64_t, DO_FABSH) DO_1OP(vfabss, 8, uint64_t, DO_FABSS) + +#define DO_NEG(N) (-(N)) +#define DO_FNEGH(N) ((N) ^ dup_const(MO_16, 0x8000)) +#define DO_FNEGS(N) ((N) ^ dup_const(MO_32, 0x80000000)) + +DO_1OP(vnegb, 1, int8_t, DO_NEG) +DO_1OP(vnegh, 2, int16_t, DO_NEG) +DO_1OP(vnegw, 4, int32_t, DO_NEG) + +/* We can do these 64 bits at a time */ +DO_1OP(vfnegh, 8, uint64_t, DO_FNEGH) +DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) DO_1OP(VCLZ, vclz) DO_1OP(VCLS, vcls) DO_1OP(VABS, vabs) +DO_1OP(VNEG, vneg) static bool trans_VREV16(DisasContext *s, arg_1op *a) { @@ -XXX,XX +XXX,XX @@ static bool trans_VABS_fp(DisasContext *s, arg_1op *a) } return do_1op(s, a, fns[a->size]); } + +static bool trans_VNEG_fp(DisasContext *s, arg_1op *a) +{ + static MVEGenOneOpFn * const fns[] = { + NULL, + gen_helper_mve_vfnegh, + gen_helper_mve_vfnegs, + NULL, + }; + if (!dc_isar_feature(aa32_mve_fp, s)) { + return false; + } + return do_1op(s, a, fns[a->size]); +} -- 2.20.1 The Arm MVE VDUP implementation would like to be able to emit code to duplicate a byte or halfword value into an i32. We have code to do this already in tcg-op-gvec.c, so all we need to do is make the functions global. For consistency with other functions made available to the frontends: * we rename to tcg_gen_dup_* * we expose both the _i32 and _i64 forms * we provide the #define for a _tl form Suggested-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20210617121628.20116-10-peter.maydell@linaro.org --- include/tcg/tcg-op.h | 8 ++++++++ include/tcg/tcg.h | 1 - tcg/tcg-op-gvec.c | 20 ++++++++++---------- 3 files changed, 18 insertions(+), 11 deletions(-) diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op.h +++ b/include/tcg/tcg-op.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_umin_i32(TCGv_i32, TCGv_i32 arg1, TCGv_i32 arg2); void tcg_gen_umax_i32(TCGv_i32, TCGv_i32 arg1, TCGv_i32 arg2); void tcg_gen_abs_i32(TCGv_i32, TCGv_i32); +/* Replicate a value of size @vece from @in to all the lanes in @out */ +void tcg_gen_dup_i32(unsigned vece, TCGv_i32 out, TCGv_i32 in); + static inline void tcg_gen_discard_i32(TCGv_i32 arg) { tcg_gen_op1_i32(INDEX_op_discard, arg); @@ -XXX,XX +XXX,XX @@ void tcg_gen_umin_i64(TCGv_i64, TCGv_i64 arg1, TCGv_i64 arg2); void tcg_gen_umax_i64(TCGv_i64, TCGv_i64 arg1, TCGv_i64 arg2); void tcg_gen_abs_i64(TCGv_i64, TCGv_i64); +/* Replicate a value of size @vece from @in to all the lanes in @out */ +void tcg_gen_dup_i64(unsigned vece, TCGv_i64 out, TCGv_i64 in); + #if TCG_TARGET_REG_BITS == 64 static inline void tcg_gen_discard_i64(TCGv_i64 arg) { @@ -XXX,XX +XXX,XX @@ void tcg_gen_stl_vec(TCGv_vec r, TCGv_ptr base, TCGArg offset, TCGType t); #define tcg_gen_atomic_smax_fetch_tl tcg_gen_atomic_smax_fetch_i64 #define tcg_gen_atomic_umax_fetch_tl tcg_gen_atomic_umax_fetch_i64 #define tcg_gen_dup_tl_vec tcg_gen_dup_i64_vec +#define tcg_gen_dup_tl tcg_gen_dup_i64 #else #define tcg_gen_movi_tl tcg_gen_movi_i32 #define tcg_gen_mov_tl tcg_gen_mov_i32 @@ -XXX,XX +XXX,XX @@ void tcg_gen_stl_vec(TCGv_vec r, TCGv_ptr base, TCGArg offset, TCGType t); #define tcg_gen_atomic_smax_fetch_tl tcg_gen_atomic_smax_fetch_i32 #define tcg_gen_atomic_umax_fetch_tl tcg_gen_atomic_umax_fetch_i32 #define tcg_gen_dup_tl_vec tcg_gen_dup_i32_vec +#define tcg_gen_dup_tl tcg_gen_dup_i32 #endif #if UINTPTR_MAX == UINT32_MAX diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ uint64_t dup_const(unsigned vece, uint64_t c); : (qemu_build_not_reached_always(), 0)) \ : dup_const(VECE, C)) - /* * Memory helpers that will be used by TCG generated code. */ diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -XXX,XX +XXX,XX @@ uint64_t (dup_const)(unsigned vece, uint64_t c) } /* Duplicate IN into OUT as per VECE. */ -static void gen_dup_i32(unsigned vece, TCGv_i32 out, TCGv_i32 in) +void tcg_gen_dup_i32(unsigned vece, TCGv_i32 out, TCGv_i32 in) { switch (vece) { case MO_8: @@ -XXX,XX +XXX,XX @@ static void gen_dup_i32(unsigned vece, TCGv_i32 out, TCGv_i32 in) } } -static void gen_dup_i64(unsigned vece, TCGv_i64 out, TCGv_i64 in) +void tcg_gen_dup_i64(unsigned vece, TCGv_i64 out, TCGv_i64 in) { switch (vece) { case MO_8: @@ -XXX,XX +XXX,XX @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz, && (vece != MO_32 || !check_size_impl(oprsz, 4))) { t_64 = tcg_temp_new_i64(); tcg_gen_extu_i32_i64(t_64, in_32); - gen_dup_i64(vece, t_64, t_64); + tcg_gen_dup_i64(vece, t_64, t_64); } else { t_32 = tcg_temp_new_i32(); - gen_dup_i32(vece, t_32, in_32); + tcg_gen_dup_i32(vece, t_32, in_32); } } else if (in_64) { /* We are given a 64-bit variable input. */ t_64 = tcg_temp_new_i64(); - gen_dup_i64(vece, t_64, in_64); + tcg_gen_dup_i64(vece, t_64, in_64); } else { /* We are given a constant input. */ /* For 64-bit hosts, use 64-bit constants for "simple" constants @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz, } else if (g->fni8 && check_size_impl(oprsz, 8)) { TCGv_i64 t64 = tcg_temp_new_i64(); - gen_dup_i64(g->vece, t64, c); + tcg_gen_dup_i64(g->vece, t64, c); expand_2s_i64(dofs, aofs, oprsz, t64, g->scalar_first, g->fni8); tcg_temp_free_i64(t64); } else if (g->fni4 && check_size_impl(oprsz, 4)) { TCGv_i32 t32 = tcg_temp_new_i32(); tcg_gen_extrl_i64_i32(t32, c); - gen_dup_i32(g->vece, t32, t32); + tcg_gen_dup_i32(g->vece, t32, t32); expand_2s_i32(dofs, aofs, oprsz, t32, g->scalar_first, g->fni4); tcg_temp_free_i32(t32); } else { @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_ands(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i64 c, uint32_t oprsz, uint32_t maxsz) { TCGv_i64 tmp = tcg_temp_new_i64(); - gen_dup_i64(vece, tmp, c); + tcg_gen_dup_i64(vece, tmp, c); tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_ands); tcg_temp_free_i64(tmp); } @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_xors(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i64 c, uint32_t oprsz, uint32_t maxsz) { TCGv_i64 tmp = tcg_temp_new_i64(); - gen_dup_i64(vece, tmp, c); + tcg_gen_dup_i64(vece, tmp, c); tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_xors); tcg_temp_free_i64(tmp); } @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_ors(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i64 c, uint32_t oprsz, uint32_t maxsz) { TCGv_i64 tmp = tcg_temp_new_i64(); - gen_dup_i64(vece, tmp, c); + tcg_gen_dup_i64(vece, tmp, c); tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, tmp, &gop_ors); tcg_temp_free_i64(tmp); } -- 2.20.1 Implement the MVE VDUP insn, which duplicates a value from a general-purpose register into every lane of a vector register (subject to predication). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-11-peter.maydell@linaro.org --- target/arm/helper-mve.h | 2 ++ target/arm/mve.decode | 10 ++++++++++ target/arm/mve_helper.c | 16 ++++++++++++++++ target/arm/translate-mve.c | 27 +++++++++++++++++++++++++++ 4 files changed, 55 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vstrb_h, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrb_w, TCG_CALL_NO_WG, void, env, ptr, i32) DEF_HELPER_FLAGS_3(mve_vstrh_w, TCG_CALL_NO_WG, void, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vdup, TCG_CALL_NO_WG, void, env, ptr, i32) + DEF_HELPER_FLAGS_3(mve_vclsb, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclsh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vclsw, TCG_CALL_NO_WG, void, env, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ %qd 22:1 13:3 %qm 5:1 1:3 +%qn 7:1 17:3 &vldr_vstr rn qd imm p a w size l u &1op qd qm size @@ -XXX,XX +XXX,XX @@ VABS 1111 1111 1 . 11 .. 01 ... 0 0011 01 . 0 ... 0 @1op VABS_fp 1111 1111 1 . 11 .. 01 ... 0 0111 01 . 0 ... 0 @1op VNEG 1111 1111 1 . 11 .. 01 ... 0 0011 11 . 0 ... 0 @1op VNEG_fp 1111 1111 1 . 11 .. 01 ... 0 0111 11 . 0 ... 0 @1op + +&vdup qd rt size +# Qd is in the fields usually named Qn +@vdup .... .... . . .. ... . rt:4 .... . . . . .... qd=%qn &vdup + +# B and E bits encode size, which we decode here to the usual size values +VDUP 1110 1110 1 1 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=0 +VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 1 1 0000 @vdup size=1 +VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=2 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ static void mergemask_sq(int64_t *d, int64_t r, uint16_t mask) uint64_t *: mergemask_uq, \ int64_t *: mergemask_sq)(D, R, M) +void HELPER(mve_vdup)(CPUARMState *env, void *vd, uint32_t val) +{ + /* + * The generated code already replicated an 8 or 16 bit constant + * into the 32-bit value, so we only need to write the 32-bit + * value to all elements of the Qreg, allowing for predication. + */ + uint32_t *d = vd; + uint16_t mask = mve_element_mask(env); + unsigned e; + for (e = 0; e < 16 / 4; e++, mask >>= 4) { + mergemask(&d[H4(e)], val, mask); + } + mve_advance_vpt(env); +} + #define DO_1OP(OP, ESIZE, TYPE, FN) \ void HELPER(mve_##OP)(CPUARMState *env, void *vd, void *vm) \ { \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_VLDST_WIDE_NARROW(VLDSTB_H, vldrb_sh, vldrb_uh, vstrb_h) DO_VLDST_WIDE_NARROW(VLDSTB_W, vldrb_sw, vldrb_uw, vstrb_w) DO_VLDST_WIDE_NARROW(VLDSTH_W, vldrh_sw, vldrh_uw, vstrh_w) +static bool trans_VDUP(DisasContext *s, arg_VDUP *a) +{ + TCGv_ptr qd; + TCGv_i32 rt; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qd)) { + return false; + } + if (a->rt == 13 || a->rt == 15) { + /* UNPREDICTABLE; we choose to UNDEF */ + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + qd = mve_qreg_ptr(a->qd); + rt = load_reg(s, a->rt); + tcg_gen_dup_i32(a->size, rt, rt); + gen_helper_mve_vdup(cpu_env, qd, rt); + tcg_temp_free_ptr(qd); + tcg_temp_free_i32(rt); + mve_update_eci(s); + return true; +} + static bool do_1op(DisasContext *s, arg_1op *a, MVEGenOneOpFn fn) { TCGv_ptr qd, qm; -- 2.20.1 Implement the MVE vector logical operations operating on two registers. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-12-peter.maydell@linaro.org --- target/arm/helper-mve.h | 6 ++++++ target/arm/mve.decode | 9 +++++++++ target/arm/mve_helper.c | 26 ++++++++++++++++++++++++++ target/arm/translate-mve.c | 37 +++++++++++++++++++++++++++++++++++++ 4 files changed, 78 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(mve_vnegh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vnegw, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vfnegh, TCG_CALL_NO_WG, void, env, ptr, ptr) DEF_HELPER_FLAGS_3(mve_vfnegs, TCG_CALL_NO_WG, void, env, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vand, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vbic, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vorr, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vorn, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_veor, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ &vldr_vstr rn qd imm p a w size l u &1op qd qm size +&2op qd qm qn size @vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd u=0 # Note that both Rn and Qd are 3 bits only (no D bit) @@ -XXX,XX +XXX,XX @@ @1op .... .... .... size:2 .. .... .... .... .... &1op qd=%qd qm=%qm @1op_nosz .... .... .... .... .... .... .... .... &1op qd=%qd qm=%qm size=0 +@2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 # Vector loads and stores @@ -XXX,XX +XXX,XX @@ VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111101 ....... @vldr_vstr \ VLDR_VSTR 1110110 1 a:1 . w:1 . .... ... 111110 ....... @vldr_vstr \ size=2 p=1 +# Vector 2-op +VAND 1110 1111 0 . 00 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz +VBIC 1110 1111 0 . 01 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz +VORR 1110 1111 0 . 10 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz +VORN 1110 1111 0 . 11 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz +VEOR 1111 1111 0 . 00 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vnegw, 4, int32_t, DO_NEG) /* We can do these 64 bits at a time */ DO_1OP(vfnegh, 8, uint64_t, DO_FNEGH) DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) + +#define DO_2OP(OP, ESIZE, TYPE, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, \ + void *vd, void *vn, void *vm) \ + { \ + TYPE *d = vd, *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + mergemask(&d[H##ESIZE(e)], \ + FN(n[H##ESIZE(e)], m[H##ESIZE(e)]), mask); \ + } \ + mve_advance_vpt(env); \ + } + +#define DO_AND(N, M) ((N) & (M)) +#define DO_BIC(N, M) ((N) & ~(M)) +#define DO_ORR(N, M) ((N) | (M)) +#define DO_ORN(N, M) ((N) | ~(M)) +#define DO_EOR(N, M) ((N) ^ (M)) + +DO_2OP(vand, 8, uint64_t, DO_AND) +DO_2OP(vbic, 8, uint64_t, DO_BIC) +DO_2OP(vorr, 8, uint64_t, DO_ORR) +DO_2OP(vorn, 8, uint64_t, DO_ORN) +DO_2OP(veor, 8, uint64_t, DO_EOR) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ typedef void MVEGenLdStFn(TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); +typedef void MVEGenTwoOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ static inline long mve_qreg_offset(unsigned reg) @@ -XXX,XX +XXX,XX @@ static bool trans_VNEG_fp(DisasContext *s, arg_1op *a) } return do_1op(s, a, fns[a->size]); } + +static bool do_2op(DisasContext *s, arg_2op *a, MVEGenTwoOpFn fn) +{ + TCGv_ptr qd, qn, qm; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qd | a->qn | a->qm) || + !fn) { + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + qd = mve_qreg_ptr(a->qd); + qn = mve_qreg_ptr(a->qn); + qm = mve_qreg_ptr(a->qm); + fn(cpu_env, qd, qn, qm); + tcg_temp_free_ptr(qd); + tcg_temp_free_ptr(qn); + tcg_temp_free_ptr(qm); + mve_update_eci(s); + return true; +} + +#define DO_LOGIC(INSN, HELPER) \ + static bool trans_##INSN(DisasContext *s, arg_2op *a) \ + { \ + return do_2op(s, a, HELPER); \ + } + +DO_LOGIC(VAND, gen_helper_mve_vand) +DO_LOGIC(VBIC, gen_helper_mve_vbic) +DO_LOGIC(VORR, gen_helper_mve_vorr) +DO_LOGIC(VORN, gen_helper_mve_vorn) +DO_LOGIC(VEOR, gen_helper_mve_veor) -- 2.20.1 Implement the MVE VADD, VSUB and VMUL insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-13-peter.maydell@linaro.org --- target/arm/helper-mve.h | 12 ++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 14 ++++++++++++++ target/arm/translate-mve.c | 16 ++++++++++++++++ 4 files changed, 47 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vbic, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vorr, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vorn, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_veor, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vaddb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vaddh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vaddw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vsubb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vsubh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vsubw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmulb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ @1op .... .... .... size:2 .. .... .... .... .... &1op qd=%qd qm=%qm @1op_nosz .... .... .... .... .... .... .... .... &1op qd=%qd qm=%qm size=0 +@2op .... .... .. size:2 .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn @2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 # Vector loads and stores @@ -XXX,XX +XXX,XX @@ VORR 1110 1111 0 . 10 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz VORN 1110 1111 0 . 11 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz VEOR 1111 1111 0 . 00 ... 0 ... 0 0001 . 1 . 1 ... 0 @2op_nosz +VADD 1110 1111 0 . .. ... 0 ... 0 1000 . 1 . 0 ... 0 @2op +VSUB 1111 1111 0 . .. ... 0 ... 0 1000 . 1 . 0 ... 0 @2op +VMUL 1110 1111 0 . .. ... 0 ... 0 1001 . 1 . 1 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) mve_advance_vpt(env); \ } +/* provide unsigned 2-op helpers for all sizes */ +#define DO_2OP_U(OP, FN) \ + DO_2OP(OP##b, 1, uint8_t, FN) \ + DO_2OP(OP##h, 2, uint16_t, FN) \ + DO_2OP(OP##w, 4, uint32_t, FN) + #define DO_AND(N, M) ((N) & (M)) #define DO_BIC(N, M) ((N) & ~(M)) #define DO_ORR(N, M) ((N) | (M)) @@ -XXX,XX +XXX,XX @@ DO_2OP(vbic, 8, uint64_t, DO_BIC) DO_2OP(vorr, 8, uint64_t, DO_ORR) DO_2OP(vorn, 8, uint64_t, DO_ORN) DO_2OP(veor, 8, uint64_t, DO_EOR) + +#define DO_ADD(N, M) ((N) + (M)) +#define DO_SUB(N, M) ((N) - (M)) +#define DO_MUL(N, M) ((N) * (M)) + +DO_2OP_U(vadd, DO_ADD) +DO_2OP_U(vsub, DO_SUB) +DO_2OP_U(vmul, DO_MUL) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_LOGIC(VBIC, gen_helper_mve_vbic) DO_LOGIC(VORR, gen_helper_mve_vorr) DO_LOGIC(VORN, gen_helper_mve_vorn) DO_LOGIC(VEOR, gen_helper_mve_veor) + +#define DO_2OP(INSN, FN) \ + static bool trans_##INSN(DisasContext *s, arg_2op *a) \ + { \ + static MVEGenTwoOpFn * const fns[] = { \ + gen_helper_mve_##FN##b, \ + gen_helper_mve_##FN##h, \ + gen_helper_mve_##FN##w, \ + NULL, \ + }; \ + return do_2op(s, a, fns[a->size]); \ + } + +DO_2OP(VADD, vadd) +DO_2OP(VSUB, vsub) +DO_2OP(VMUL, vmul) -- 2.20.1 Implement the MVE VMULH insn, which performs a vector multiply and returns the high half of the result. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-14-peter.maydell@linaro.org --- target/arm/helper-mve.h | 7 +++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 26 ++++++++++++++++++++++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 38 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vsubw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmulhsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulhsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulhsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulhub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulhuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulhuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VADD 1110 1111 0 . .. ... 0 ... 0 1000 . 1 . 0 ... 0 @2op VSUB 1111 1111 0 . .. ... 0 ... 0 1000 . 1 . 0 ... 0 @2op VMUL 1110 1111 0 . .. ... 0 ... 0 1001 . 1 . 1 ... 0 @2op +VMULH_S 111 0 1110 0 . .. ...1 ... 0 1110 . 0 . 0 ... 1 @2op +VMULH_U 111 1 1110 0 . .. ...1 ... 0 1110 . 0 . 0 ... 1 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP(veor, 8, uint64_t, DO_EOR) DO_2OP_U(vadd, DO_ADD) DO_2OP_U(vsub, DO_SUB) DO_2OP_U(vmul, DO_MUL) + +/* + * Because the computation type is at least twice as large as required, + * these work for both signed and unsigned source types. + */ +static inline uint8_t do_mulh_b(int32_t n, int32_t m) +{ + return (n * m) >> 8; +} + +static inline uint16_t do_mulh_h(int32_t n, int32_t m) +{ + return (n * m) >> 16; +} + +static inline uint32_t do_mulh_w(int64_t n, int64_t m) +{ + return (n * m) >> 32; +} + +DO_2OP(vmulhsb, 1, int8_t, do_mulh_b) +DO_2OP(vmulhsh, 2, int16_t, do_mulh_h) +DO_2OP(vmulhsw, 4, int32_t, do_mulh_w) +DO_2OP(vmulhub, 1, uint8_t, do_mulh_b) +DO_2OP(vmulhuh, 2, uint16_t, do_mulh_h) +DO_2OP(vmulhuw, 4, uint32_t, do_mulh_w) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_LOGIC(VEOR, gen_helper_mve_veor) DO_2OP(VADD, vadd) DO_2OP(VSUB, vsub) DO_2OP(VMUL, vmul) +DO_2OP(VMULH_S, vmulhs) +DO_2OP(VMULH_U, vmulhu) -- 2.20.1 Implement the MVE VRMULH insn, which performs a rounding multiply and then returns the high half. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-15-peter.maydell@linaro.org --- target/arm/helper-mve.h | 7 +++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 22 ++++++++++++++++++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 34 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmulhsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulhub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulhuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulhuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vrmulhsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrmulhsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrmulhsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrmulhub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrmulhuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrmulhuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMUL 1110 1111 0 . .. ... 0 ... 0 1001 . 1 . 1 ... 0 @2op VMULH_S 111 0 1110 0 . .. ...1 ... 0 1110 . 0 . 0 ... 1 @2op VMULH_U 111 1 1110 0 . .. ...1 ... 0 1110 . 0 . 0 ... 1 @2op +VRMULH_S 111 0 1110 0 . .. ...1 ... 1 1110 . 0 . 0 ... 1 @2op +VRMULH_U 111 1 1110 0 . .. ...1 ... 1 1110 . 0 . 0 ... 1 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ static inline uint32_t do_mulh_w(int64_t n, int64_t m) return (n * m) >> 32; } +static inline uint8_t do_rmulh_b(int32_t n, int32_t m) +{ + return (n * m + (1U << 7)) >> 8; +} + +static inline uint16_t do_rmulh_h(int32_t n, int32_t m) +{ + return (n * m + (1U << 15)) >> 16; +} + +static inline uint32_t do_rmulh_w(int64_t n, int64_t m) +{ + return (n * m + (1U << 31)) >> 32; +} + DO_2OP(vmulhsb, 1, int8_t, do_mulh_b) DO_2OP(vmulhsh, 2, int16_t, do_mulh_h) DO_2OP(vmulhsw, 4, int32_t, do_mulh_w) DO_2OP(vmulhub, 1, uint8_t, do_mulh_b) DO_2OP(vmulhuh, 2, uint16_t, do_mulh_h) DO_2OP(vmulhuw, 4, uint32_t, do_mulh_w) + +DO_2OP(vrmulhsb, 1, int8_t, do_rmulh_b) +DO_2OP(vrmulhsh, 2, int16_t, do_rmulh_h) +DO_2OP(vrmulhsw, 4, int32_t, do_rmulh_w) +DO_2OP(vrmulhub, 1, uint8_t, do_rmulh_b) +DO_2OP(vrmulhuh, 2, uint16_t, do_rmulh_h) +DO_2OP(vrmulhuw, 4, uint32_t, do_rmulh_w) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VSUB, vsub) DO_2OP(VMUL, vmul) DO_2OP(VMULH_S, vmulhs) DO_2OP(VMULH_U, vmulhu) +DO_2OP(VRMULH_S, vrmulhs) +DO_2OP(VRMULH_U, vrmulhu) -- 2.20.1 Implement the MVE VMAX and VMIN insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-16-peter.maydell@linaro.org --- target/arm/helper-mve.h | 14 ++++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 14 ++++++++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 37 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vrmulhsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrmulhub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrmulhuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrmulhuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmaxsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmaxsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmaxsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmaxub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmaxuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmaxuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vminsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vminsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vminsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vminub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vminuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vminuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMULH_U 111 1 1110 0 . .. ...1 ... 0 1110 . 0 . 0 ... 1 @2op VRMULH_S 111 0 1110 0 . .. ...1 ... 1 1110 . 0 . 0 ... 1 @2op VRMULH_U 111 1 1110 0 . .. ...1 ... 1 1110 . 0 . 0 ... 1 @2op +VMAX_S 111 0 1111 0 . .. ... 0 ... 0 0110 . 1 . 0 ... 0 @2op +VMAX_U 111 1 1111 0 . .. ... 0 ... 0 0110 . 1 . 0 ... 0 @2op +VMIN_S 111 0 1111 0 . .. ... 0 ... 0 0110 . 1 . 1 ... 0 @2op +VMIN_U 111 1 1111 0 . .. ... 0 ... 0 0110 . 1 . 1 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) DO_2OP(OP##h, 2, uint16_t, FN) \ DO_2OP(OP##w, 4, uint32_t, FN) +/* provide signed 2-op helpers for all sizes */ +#define DO_2OP_S(OP, FN) \ + DO_2OP(OP##b, 1, int8_t, FN) \ + DO_2OP(OP##h, 2, int16_t, FN) \ + DO_2OP(OP##w, 4, int32_t, FN) + #define DO_AND(N, M) ((N) & (M)) #define DO_BIC(N, M) ((N) & ~(M)) #define DO_ORR(N, M) ((N) | (M)) @@ -XXX,XX +XXX,XX @@ DO_2OP(vrmulhsw, 4, int32_t, do_rmulh_w) DO_2OP(vrmulhub, 1, uint8_t, do_rmulh_b) DO_2OP(vrmulhuh, 2, uint16_t, do_rmulh_h) DO_2OP(vrmulhuw, 4, uint32_t, do_rmulh_w) + +#define DO_MAX(N, M) ((N) >= (M) ? (N) : (M)) +#define DO_MIN(N, M) ((N) >= (M) ? (M) : (N)) + +DO_2OP_S(vmaxs, DO_MAX) +DO_2OP_U(vmaxu, DO_MAX) +DO_2OP_S(vmins, DO_MIN) +DO_2OP_U(vminu, DO_MIN) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VMULH_S, vmulhs) DO_2OP(VMULH_U, vmulhu) DO_2OP(VRMULH_S, vrmulhs) DO_2OP(VRMULH_U, vrmulhu) +DO_2OP(VMAX_S, vmaxs) +DO_2OP(VMAX_U, vmaxu) +DO_2OP(VMIN_S, vmins) +DO_2OP(VMIN_U, vminu) -- 2.20.1 Implement the MVE VABD insn. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-17-peter.maydell@linaro.org --- target/arm/helper-mve.h | 7 +++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 5 +++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 17 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vminsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vminub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vminuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vminuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vabdsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vabdsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vabdsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vabdub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vabduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vabduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMAX_U 111 1 1111 0 . .. ... 0 ... 0 0110 . 1 . 0 ... 0 @2op VMIN_S 111 0 1111 0 . .. ... 0 ... 0 0110 . 1 . 1 ... 0 @2op VMIN_U 111 1 1111 0 . .. ... 0 ... 0 0110 . 1 . 1 ... 0 @2op +VABD_S 111 0 1111 0 . .. ... 0 ... 0 0111 . 1 . 0 ... 0 @2op +VABD_U 111 1 1111 0 . .. ... 0 ... 0 0111 . 1 . 0 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_S(vmaxs, DO_MAX) DO_2OP_U(vmaxu, DO_MAX) DO_2OP_S(vmins, DO_MIN) DO_2OP_U(vminu, DO_MIN) + +#define DO_ABD(N, M) ((N) >= (M) ? (N) - (M) : (M) - (N)) + +DO_2OP_S(vabds, DO_ABD) +DO_2OP_U(vabdu, DO_ABD) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VMAX_S, vmaxs) DO_2OP(VMAX_U, vmaxu) DO_2OP(VMIN_S, vmins) DO_2OP(VMIN_U, vminu) +DO_2OP(VABD_S, vabds) +DO_2OP(VABD_U, vabdu) -- 2.20.1 Implement MVE VHADD and VHSUB insns, which perform an addition or subtraction and then halve the result. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-18-peter.maydell@linaro.org --- target/arm/helper-mve.h | 14 ++++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 25 +++++++++++++++++++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 48 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vabdsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vabdub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vabduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vabduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vhaddsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhaddsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhaddsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhaddub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhadduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhadduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vhsubsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhsubsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhsubsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhsubub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhsubuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhsubuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMIN_U 111 1 1111 0 . .. ... 0 ... 0 0110 . 1 . 1 ... 0 @2op VABD_S 111 0 1111 0 . .. ... 0 ... 0 0111 . 1 . 0 ... 0 @2op VABD_U 111 1 1111 0 . .. ... 0 ... 0 0111 . 1 . 0 ... 0 @2op +VHADD_S 111 0 1111 0 . .. ... 0 ... 0 0000 . 1 . 0 ... 0 @2op +VHADD_U 111 1 1111 0 . .. ... 0 ... 0 0000 . 1 . 0 ... 0 @2op +VHSUB_S 111 0 1111 0 . .. ... 0 ... 0 0010 . 1 . 0 ... 0 @2op +VHSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 0 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vminu, DO_MIN) DO_2OP_S(vabds, DO_ABD) DO_2OP_U(vabdu, DO_ABD) + +static inline uint32_t do_vhadd_u(uint32_t n, uint32_t m) +{ + return ((uint64_t)n + m) >> 1; +} + +static inline int32_t do_vhadd_s(int32_t n, int32_t m) +{ + return ((int64_t)n + m) >> 1; +} + +static inline uint32_t do_vhsub_u(uint32_t n, uint32_t m) +{ + return ((uint64_t)n - m) >> 1; +} + +static inline int32_t do_vhsub_s(int32_t n, int32_t m) +{ + return ((int64_t)n - m) >> 1; +} + +DO_2OP_S(vhadds, do_vhadd_s) +DO_2OP_U(vhaddu, do_vhadd_u) +DO_2OP_S(vhsubs, do_vhsub_s) +DO_2OP_U(vhsubu, do_vhsub_u) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VMIN_S, vmins) DO_2OP(VMIN_U, vminu) DO_2OP(VABD_S, vabds) DO_2OP(VABD_U, vabdu) +DO_2OP(VHADD_S, vhadds) +DO_2OP(VHADD_U, vhaddu) +DO_2OP(VHSUB_S, vhsubs) +DO_2OP(VHSUB_U, vhsubu) -- 2.20.1 Implement the MVE VMULL insn, which multiplies two single width integer elements to produce a double width result. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-19-peter.maydell@linaro.org --- target/arm/helper-mve.h | 14 ++++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 34 ++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 57 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vhsubsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vhsubub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vhsubuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vhsubuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmullbsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmullbsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmullbsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmullbub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmullbuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmullbuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmulltsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulltsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulltsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulltub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulltuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vmulltuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VHADD_U 111 1 1111 0 . .. ... 0 ... 0 0000 . 1 . 0 ... 0 @2op VHSUB_S 111 0 1111 0 . .. ... 0 ... 0 0010 . 1 . 0 ... 0 @2op VHSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 0 ... 0 @2op +VMULL_BS 111 0 1110 0 . .. ... 1 ... 0 1110 . 0 . 0 ... 0 @2op +VMULL_BU 111 1 1110 0 . .. ... 1 ... 0 1110 . 0 . 0 ... 0 @2op +VMULL_TS 111 0 1110 0 . .. ... 1 ... 1 1110 . 0 . 0 ... 0 @2op +VMULL_TU 111 1 1110 0 . .. ... 1 ... 1 1110 . 0 . 0 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) DO_2OP(OP##h, 2, int16_t, FN) \ DO_2OP(OP##w, 4, int32_t, FN) +/* + * "Long" operations where two half-sized inputs (taken from either the + * top or the bottom of the input vector) produce a double-width result. + * Here ESIZE, TYPE are for the input, and LESIZE, LTYPE for the output. + */ +#define DO_2OP_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \ + { \ + LTYPE *d = vd; \ + TYPE *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned le; \ + for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) { \ + LTYPE r = FN((LTYPE)n[H##ESIZE(le * 2 + TOP)], \ + m[H##ESIZE(le * 2 + TOP)]); \ + mergemask(&d[H##LESIZE(le)], r, mask); \ + } \ + mve_advance_vpt(env); \ + } + #define DO_AND(N, M) ((N) & (M)) #define DO_BIC(N, M) ((N) & ~(M)) #define DO_ORR(N, M) ((N) | (M)) @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vadd, DO_ADD) DO_2OP_U(vsub, DO_SUB) DO_2OP_U(vmul, DO_MUL) +DO_2OP_L(vmullbsb, 0, 1, int8_t, 2, int16_t, DO_MUL) +DO_2OP_L(vmullbsh, 0, 2, int16_t, 4, int32_t, DO_MUL) +DO_2OP_L(vmullbsw, 0, 4, int32_t, 8, int64_t, DO_MUL) +DO_2OP_L(vmullbub, 0, 1, uint8_t, 2, uint16_t, DO_MUL) +DO_2OP_L(vmullbuh, 0, 2, uint16_t, 4, uint32_t, DO_MUL) +DO_2OP_L(vmullbuw, 0, 4, uint32_t, 8, uint64_t, DO_MUL) + +DO_2OP_L(vmulltsb, 1, 1, int8_t, 2, int16_t, DO_MUL) +DO_2OP_L(vmulltsh, 1, 2, int16_t, 4, int32_t, DO_MUL) +DO_2OP_L(vmulltsw, 1, 4, int32_t, 8, int64_t, DO_MUL) +DO_2OP_L(vmulltub, 1, 1, uint8_t, 2, uint16_t, DO_MUL) +DO_2OP_L(vmulltuh, 1, 2, uint16_t, 4, uint32_t, DO_MUL) +DO_2OP_L(vmulltuw, 1, 4, uint32_t, 8, uint64_t, DO_MUL) + /* * Because the computation type is at least twice as large as required, * these work for both signed and unsigned source types. diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VHADD_S, vhadds) DO_2OP(VHADD_U, vhaddu) DO_2OP(VHSUB_S, vhsubs) DO_2OP(VHSUB_U, vhsubu) +DO_2OP(VMULL_BS, vmullbs) +DO_2OP(VMULL_BU, vmullbu) +DO_2OP(VMULL_TS, vmullts) +DO_2OP(VMULL_TU, vmulltu) -- 2.20.1 Implement the MVE VMLALDAV insn, which multiplies pairs of integer elements, accumulating them into a 64-bit result in a pair of general-purpose registers. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-20-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++ target/arm/translate.h | 10 ++++ target/arm/mve.decode | 15 ++++++ target/arm/mve_helper.c | 34 ++++++++++++++ target/arm/translate-mve.c | 96 ++++++++++++++++++++++++++++++++++++++ 5 files changed, 163 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmulltsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlaldavxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_4(mve_vmlaldavuh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlaldavuw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) 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 @@ static inline int negate(DisasContext *s, int x) return -x; } +static inline int plus_1(DisasContext *s, int x) +{ + return x + 1; +} + static inline int plus_2(DisasContext *s, int x) { return x + 2; @@ -XXX,XX +XXX,XX @@ static inline int times_4(DisasContext *s, int x) return x * 4; } +static inline int times_2_plus_1(DisasContext *s, int x) +{ + return x * 2 + 1; +} + static inline int arm_dc_feature(DisasContext *dc, int feature) { return (dc->features & (1ULL << feature)) != 0; diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VNEG_fp 1111 1111 1 . 11 .. 01 ... 0 0111 11 . 0 ... 0 @1op VDUP 1110 1110 1 1 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=0 VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 1 1 0000 @vdup size=1 VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=2 + +# multiply-add long dual accumulate +# rdahi: bits [3:1] from insn, bit 0 is 1 +# rdalo: bits [3:1] from insn, bit 0 is 0 +%rdahi 20:3 !function=times_2_plus_1 +%rdalo 13:3 !function=times_2 +# size bit is 0 for 16 bit, 1 for 32 bit +%size_16 16:1 !function=plus_1 + +&vmlaldav rdahi rdalo size qn qm x a + +@vmlaldav .... .... . ... ... . ... . .... .... qm:3 . \ + qn=%qn rdahi=%rdahi rdalo=%rdalo size=%size_16 &vmlaldav +VMLALDAV_S 1110 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav +VMLALDAV_U 1111 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_S(vhadds, do_vhadd_s) DO_2OP_U(vhaddu, do_vhadd_u) DO_2OP_S(vhsubs, do_vhsub_s) DO_2OP_U(vhsubu, do_vhsub_u) + + +/* + * Multiply add long dual accumulate ops. + */ +#define DO_LDAV(OP, ESIZE, TYPE, XCHG, EVENACC, ODDACC) \ + uint64_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn, \ + void *vm, uint64_t a) \ + { \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + TYPE *n = vn, *m = vm; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + if (mask & 1) { \ + if (e & 1) { \ + a ODDACC \ + (int64_t)n[H##ESIZE(e - 1 * XCHG)] * m[H##ESIZE(e)]; \ + } else { \ + a EVENACC \ + (int64_t)n[H##ESIZE(e + 1 * XCHG)] * m[H##ESIZE(e)]; \ + } \ + } \ + } \ + mve_advance_vpt(env); \ + return a; \ + } + +DO_LDAV(vmlaldavsh, 2, int16_t, false, +=, +=) +DO_LDAV(vmlaldavxsh, 2, int16_t, true, +=, +=) +DO_LDAV(vmlaldavsw, 4, int32_t, false, +=, +=) +DO_LDAV(vmlaldavxsw, 4, int32_t, true, +=, +=) + +DO_LDAV(vmlaldavuh, 2, uint16_t, false, +=, +=) +DO_LDAV(vmlaldavuw, 4, uint32_t, false, +=, +=) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ typedef void MVEGenLdStFn(TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr); +typedef void MVEGenDualAccOpFn(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i64); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ static inline long mve_qreg_offset(unsigned reg) @@ -XXX,XX +XXX,XX @@ static void mve_update_eci(DisasContext *s) } } +static bool mve_skip_first_beat(DisasContext *s) +{ + /* Return true if PSR.ECI says we must skip the first beat of this insn */ + switch (s->eci) { + case ECI_NONE: + return false; + case ECI_A0: + case ECI_A0A1: + case ECI_A0A1A2: + case ECI_A0A1A2B0: + return true; + default: + g_assert_not_reached(); + } +} + static bool do_ldst(DisasContext *s, arg_VLDR_VSTR *a, MVEGenLdStFn *fn) { TCGv_i32 addr; @@ -XXX,XX +XXX,XX @@ DO_2OP(VMULL_BS, vmullbs) DO_2OP(VMULL_BU, vmullbu) DO_2OP(VMULL_TS, vmullts) DO_2OP(VMULL_TU, vmulltu) + +static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, + MVEGenDualAccOpFn *fn) +{ + TCGv_ptr qn, qm; + TCGv_i64 rda; + TCGv_i32 rdalo, rdahi; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qn | a->qm) || + !fn) { + return false; + } + /* + * rdahi == 13 is UNPREDICTABLE; rdahi == 15 is a related + * encoding; rdalo always has bit 0 clear so cannot be 13 or 15. + */ + if (a->rdahi == 13 || a->rdahi == 15) { + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + qn = mve_qreg_ptr(a->qn); + qm = mve_qreg_ptr(a->qm); + + /* + * This insn is subject to beat-wise execution. Partial execution + * of an A=0 (no-accumulate) insn which does not execute the first + * beat must start with the current rda value, not 0. + */ + if (a->a || mve_skip_first_beat(s)) { + rda = tcg_temp_new_i64(); + rdalo = load_reg(s, a->rdalo); + rdahi = load_reg(s, a->rdahi); + tcg_gen_concat_i32_i64(rda, rdalo, rdahi); + tcg_temp_free_i32(rdalo); + tcg_temp_free_i32(rdahi); + } else { + rda = tcg_const_i64(0); + } + + fn(rda, cpu_env, qn, qm, rda); + tcg_temp_free_ptr(qn); + tcg_temp_free_ptr(qm); + + rdalo = tcg_temp_new_i32(); + rdahi = tcg_temp_new_i32(); + tcg_gen_extrl_i64_i32(rdalo, rda); + tcg_gen_extrh_i64_i32(rdahi, rda); + store_reg(s, a->rdalo, rdalo); + store_reg(s, a->rdahi, rdahi); + tcg_temp_free_i64(rda); + mve_update_eci(s); + return true; +} + +static bool trans_VMLALDAV_S(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[4][2] = { + { NULL, NULL }, + { gen_helper_mve_vmlaldavsh, gen_helper_mve_vmlaldavxsh }, + { gen_helper_mve_vmlaldavsw, gen_helper_mve_vmlaldavxsw }, + { NULL, NULL }, + }; + return do_long_dual_acc(s, a, fns[a->size][a->x]); +} + +static bool trans_VMLALDAV_U(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[4][2] = { + { NULL, NULL }, + { gen_helper_mve_vmlaldavuh, NULL }, + { gen_helper_mve_vmlaldavuw, NULL }, + { NULL, NULL }, + }; + return do_long_dual_acc(s, a, fns[a->size][a->x]); +} -- 2.20.1 Implement the MVE insn VMLSLDAV, which multiplies source elements, alternately adding and subtracting them, and accumulates into a 64-bit result in a pair of general purpose registers. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-21-peter.maydell@linaro.org --- target/arm/helper-mve.h | 5 +++++ target/arm/mve.decode | 2 ++ target/arm/mve_helper.c | 5 +++++ target/arm/translate-mve.c | 11 +++++++++++ 4 files changed, 23 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmlaldavxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavuh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavuw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_4(mve_vmlsldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlsldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlsldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vmlsldavxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=2 qn=%qn rdahi=%rdahi rdalo=%rdalo size=%size_16 &vmlaldav VMLALDAV_S 1110 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav VMLALDAV_U 1111 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav + +VMLSLDAV 1110 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_LDAV(vmlaldavxsw, 4, int32_t, true, +=, +=) DO_LDAV(vmlaldavuh, 2, uint16_t, false, +=, +=) DO_LDAV(vmlaldavuw, 4, uint32_t, false, +=, +=) + +DO_LDAV(vmlsldavsh, 2, int16_t, false, +=, -=) +DO_LDAV(vmlsldavxsh, 2, int16_t, true, +=, -=) +DO_LDAV(vmlsldavsw, 4, int32_t, false, +=, -=) +DO_LDAV(vmlsldavxsw, 4, int32_t, true, +=, -=) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool trans_VMLALDAV_U(DisasContext *s, arg_vmlaldav *a) }; return do_long_dual_acc(s, a, fns[a->size][a->x]); } + +static bool trans_VMLSLDAV(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[4][2] = { + { NULL, NULL }, + { gen_helper_mve_vmlsldavsh, gen_helper_mve_vmlsldavxsh }, + { gen_helper_mve_vmlsldavsw, gen_helper_mve_vmlsldavxsw }, + { NULL, NULL }, + }; + return do_long_dual_acc(s, a, fns[a->size][a->x]); +} -- 2.20.1 Implement the MVE VRMLALDAVH and VRMLSLDAVH insns, which accumulate the results of a rounded multiply of pairs of elements into a 72-bit accumulator, returning the top 64 bits in a pair of general purpose registers. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-22-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 7 +++++++ target/arm/mve_helper.c | 37 +++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 24 ++++++++++++++++++++++++ 4 files changed, 76 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmlsldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlsldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlsldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlsldavxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_4(mve_vrmlaldavhsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vrmlaldavhxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_4(mve_vrmlaldavhuw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_4(mve_vrmlsldavhsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) +DEF_HELPER_FLAGS_4(mve_vrmlsldavhxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VDUP 1110 1110 1 0 10 ... 0 .... 1011 . 0 0 1 0000 @vdup size=2 @vmlaldav .... .... . ... ... . ... . .... .... qm:3 . \ qn=%qn rdahi=%rdahi rdalo=%rdalo size=%size_16 &vmlaldav +@vmlaldav_nosz .... .... . ... ... . ... . .... .... qm:3 . \ + qn=%qn rdahi=%rdahi rdalo=%rdalo size=0 &vmlaldav VMLALDAV_S 1110 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav VMLALDAV_U 1111 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 0 @vmlaldav VMLSLDAV 1110 1110 1 ... ... . ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav + +VRMLALDAVH_S 1110 1110 1 ... ... 0 ... x:1 1111 . 0 a:1 0 ... 0 @vmlaldav_nosz +VRMLALDAVH_U 1111 1110 1 ... ... 0 ... x:1 1111 . 0 a:1 0 ... 0 @vmlaldav_nosz + +VRMLSLDAVH 1111 1110 1 ... ... 0 ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav_nosz diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "qemu/int128.h" #include "cpu.h" #include "internals.h" #include "vec_internal.h" @@ -XXX,XX +XXX,XX @@ DO_LDAV(vmlsldavsh, 2, int16_t, false, +=, -=) DO_LDAV(vmlsldavxsh, 2, int16_t, true, +=, -=) DO_LDAV(vmlsldavsw, 4, int32_t, false, +=, -=) DO_LDAV(vmlsldavxsw, 4, int32_t, true, +=, -=) + +/* + * Rounding multiply add long dual accumulate high: we must keep + * a 72-bit internal accumulator value and return the top 64 bits. + */ +#define DO_LDAVH(OP, ESIZE, TYPE, XCHG, EVENACC, ODDACC, TO128) \ + uint64_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vn, \ + void *vm, uint64_t a) \ + { \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + TYPE *n = vn, *m = vm; \ + Int128 acc = int128_lshift(TO128(a), 8); \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + if (mask & 1) { \ + if (e & 1) { \ + acc = ODDACC(acc, TO128(n[H##ESIZE(e - 1 * XCHG)] * \ + m[H##ESIZE(e)])); \ + } else { \ + acc = EVENACC(acc, TO128(n[H##ESIZE(e + 1 * XCHG)] * \ + m[H##ESIZE(e)])); \ + } \ + acc = int128_add(acc, 1 << 7); \ + } \ + } \ + mve_advance_vpt(env); \ + return int128_getlo(int128_rshift(acc, 8)); \ + } + +DO_LDAVH(vrmlaldavhsw, 4, int32_t, false, int128_add, int128_add, int128_makes64) +DO_LDAVH(vrmlaldavhxsw, 4, int32_t, true, int128_add, int128_add, int128_makes64) + +DO_LDAVH(vrmlaldavhuw, 4, uint32_t, false, int128_add, int128_add, int128_make64) + +DO_LDAVH(vrmlsldavhsw, 4, int32_t, false, int128_add, int128_sub, int128_makes64) +DO_LDAVH(vrmlsldavhxsw, 4, int32_t, true, int128_add, int128_sub, int128_makes64) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool trans_VMLSLDAV(DisasContext *s, arg_vmlaldav *a) }; return do_long_dual_acc(s, a, fns[a->size][a->x]); } + +static bool trans_VRMLALDAVH_S(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[] = { + gen_helper_mve_vrmlaldavhsw, gen_helper_mve_vrmlaldavhxsw, + }; + return do_long_dual_acc(s, a, fns[a->x]); +} + +static bool trans_VRMLALDAVH_U(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[] = { + gen_helper_mve_vrmlaldavhuw, NULL, + }; + return do_long_dual_acc(s, a, fns[a->x]); +} + +static bool trans_VRMLSLDAVH(DisasContext *s, arg_vmlaldav *a) +{ + static MVEGenDualAccOpFn * const fns[] = { + gen_helper_mve_vrmlsldavhsw, gen_helper_mve_vrmlsldavhxsw, + }; + return do_long_dual_acc(s, a, fns[a->x]); +} -- 2.20.1 Implement the scalar form of the MVE VADD insn. This takes the scalar operand from a general purpose register. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-23-peter.maydell@linaro.org --- target/arm/helper-mve.h | 4 ++++ target/arm/mve.decode | 7 ++++++ target/arm/mve_helper.c | 22 +++++++++++++++++++ target/arm/translate-mve.c | 45 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 78 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmulltub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ &vldr_vstr rn qd imm p a w size l u &1op qd qm size &2op qd qm qn size +&2scalar qd qn rm size @vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd u=0 # Note that both Rn and Qd are 3 bits only (no D bit) @@ -XXX,XX +XXX,XX @@ @2op .... .... .. size:2 .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn @2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 +@2scalar .... .... .. size:2 .... .... .... .... rm:4 &2scalar qd=%qd qn=%qn + # Vector loads and stores # Widening loads and narrowing stores: @@ -XXX,XX +XXX,XX @@ VRMLALDAVH_S 1110 1110 1 ... ... 0 ... x:1 1111 . 0 a:1 0 ... 0 @vmlaldav_no VRMLALDAVH_U 1111 1110 1 ... ... 0 ... x:1 1111 . 0 a:1 0 ... 0 @vmlaldav_nosz VRMLSLDAVH 1111 1110 1 ... ... 0 ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav_nosz + +# Scalar operations + +VADD_scalar 1110 1110 0 . .. ... 1 ... 0 1111 . 100 .... @2scalar diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_S(vhsubs, do_vhsub_s) DO_2OP_U(vhsubu, do_vhsub_u) +#define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ + uint32_t rm) \ + { \ + TYPE *d = vd, *n = vn; \ + TYPE m = rm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + mergemask(&d[H##ESIZE(e)], FN(n[H##ESIZE(e)], m), mask); \ + } \ + mve_advance_vpt(env); \ + } + +/* provide unsigned 2-op scalar helpers for all sizes */ +#define DO_2OP_SCALAR_U(OP, FN) \ + DO_2OP_SCALAR(OP##b, 1, uint8_t, FN) \ + DO_2OP_SCALAR(OP##h, 2, uint16_t, FN) \ + DO_2OP_SCALAR(OP##w, 4, uint32_t, FN) + +DO_2OP_SCALAR_U(vadd_scalar, DO_ADD) + /* * Multiply add long dual accumulate ops. */ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ typedef void MVEGenLdStFn(TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr); +typedef void MVEGenTwoOpScalarFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenDualAccOpFn(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i64); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ @@ -XXX,XX +XXX,XX @@ DO_2OP(VMULL_BU, vmullbu) DO_2OP(VMULL_TS, vmullts) DO_2OP(VMULL_TU, vmulltu) +static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, + MVEGenTwoOpScalarFn fn) +{ + TCGv_ptr qd, qn; + TCGv_i32 rm; + + if (!dc_isar_feature(aa32_mve, s) || + !mve_check_qreg_bank(s, a->qd | a->qn) || + !fn) { + return false; + } + if (a->rm == 13 || a->rm == 15) { + /* UNPREDICTABLE */ + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + qd = mve_qreg_ptr(a->qd); + qn = mve_qreg_ptr(a->qn); + rm = load_reg(s, a->rm); + fn(cpu_env, qd, qn, rm); + tcg_temp_free_i32(rm); + tcg_temp_free_ptr(qd); + tcg_temp_free_ptr(qn); + mve_update_eci(s); + return true; +} + +#define DO_2OP_SCALAR(INSN, FN) \ + static bool trans_##INSN(DisasContext *s, arg_2scalar *a) \ + { \ + static MVEGenTwoOpScalarFn * const fns[] = { \ + gen_helper_mve_##FN##b, \ + gen_helper_mve_##FN##h, \ + gen_helper_mve_##FN##w, \ + NULL, \ + }; \ + return do_2op_scalar(s, a, fns[a->size]); \ + } + +DO_2OP_SCALAR(VADD_scalar, vadd_scalar) + static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, MVEGenDualAccOpFn *fn) { -- 2.20.1 Implement the scalar forms of the MVE VSUB and VMUL insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-24-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 2 ++ target/arm/mve_helper.c | 2 ++ target/arm/translate-mve.c | 2 ++ 4 files changed, 14 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vsub_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vsub_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vsub_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vmul_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vmul_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vmul_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VRMLSLDAVH 1111 1110 1 ... ... 0 ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav_no # Scalar operations VADD_scalar 1110 1110 0 . .. ... 1 ... 0 1111 . 100 .... @2scalar +VSUB_scalar 1110 1110 0 . .. ... 1 ... 1 1111 . 100 .... @2scalar +VMUL_scalar 1110 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhsubu, do_vhsub_u) DO_2OP_SCALAR(OP##w, 4, uint32_t, FN) DO_2OP_SCALAR_U(vadd_scalar, DO_ADD) +DO_2OP_SCALAR_U(vsub_scalar, DO_SUB) +DO_2OP_SCALAR_U(vmul_scalar, DO_MUL) /* * Multiply add long dual accumulate ops. diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, } DO_2OP_SCALAR(VADD_scalar, vadd_scalar) +DO_2OP_SCALAR(VSUB_scalar, vsub_scalar) +DO_2OP_SCALAR(VMUL_scalar, vmul_scalar) static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, MVEGenDualAccOpFn *fn) -- 2.20.1 Implement the scalar variants of the MVE VHADD and VHSUB insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-25-peter.maydell@linaro.org --- target/arm/helper-mve.h | 16 ++++++++++++++++ target/arm/mve.decode | 4 ++++ target/arm/mve_helper.c | 8 ++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 32 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmul_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vmul_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vmul_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhadds_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhadds_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhadds_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vhaddu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhaddu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhaddu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vhsubs_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhsubs_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhsubs_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vhsubu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhsubu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vhsubu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VRMLSLDAVH 1111 1110 1 ... ... 0 ... x:1 1110 . 0 a:1 0 ... 1 @vmlaldav_no VADD_scalar 1110 1110 0 . .. ... 1 ... 0 1111 . 100 .... @2scalar VSUB_scalar 1110 1110 0 . .. ... 1 ... 1 1111 . 100 .... @2scalar VMUL_scalar 1110 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar +VHADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar +VHADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar +VHSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar +VHSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhsubu, do_vhsub_u) DO_2OP_SCALAR(OP##b, 1, uint8_t, FN) \ DO_2OP_SCALAR(OP##h, 2, uint16_t, FN) \ DO_2OP_SCALAR(OP##w, 4, uint32_t, FN) +#define DO_2OP_SCALAR_S(OP, FN) \ + DO_2OP_SCALAR(OP##b, 1, int8_t, FN) \ + DO_2OP_SCALAR(OP##h, 2, int16_t, FN) \ + DO_2OP_SCALAR(OP##w, 4, int32_t, FN) DO_2OP_SCALAR_U(vadd_scalar, DO_ADD) DO_2OP_SCALAR_U(vsub_scalar, DO_SUB) DO_2OP_SCALAR_U(vmul_scalar, DO_MUL) +DO_2OP_SCALAR_S(vhadds_scalar, do_vhadd_s) +DO_2OP_SCALAR_U(vhaddu_scalar, do_vhadd_u) +DO_2OP_SCALAR_S(vhsubs_scalar, do_vhsub_s) +DO_2OP_SCALAR_U(vhsubu_scalar, do_vhsub_u) /* * Multiply add long dual accumulate ops. diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, DO_2OP_SCALAR(VADD_scalar, vadd_scalar) DO_2OP_SCALAR(VSUB_scalar, vsub_scalar) DO_2OP_SCALAR(VMUL_scalar, vmul_scalar) +DO_2OP_SCALAR(VHADD_S_scalar, vhadds_scalar) +DO_2OP_SCALAR(VHADD_U_scalar, vhaddu_scalar) +DO_2OP_SCALAR(VHSUB_S_scalar, vhsubs_scalar) +DO_2OP_SCALAR(VHSUB_U_scalar, vhsubu_scalar) static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, MVEGenDualAccOpFn *fn) -- 2.20.1 Implement the MVE VBRSR insn, which reverses a specified number of bits in each element, setting the rest to zero. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-26-peter.maydell@linaro.org --- target/arm/helper-mve.h | 4 ++++ target/arm/mve.decode | 1 + target/arm/mve_helper.c | 43 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 1 + 4 files changed, 49 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vhsubu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vhsubu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vhsubu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vbrsrb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vbrsrh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vbrsrw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VHADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar VHSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar +VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR_U(vhaddu_scalar, do_vhadd_u) DO_2OP_SCALAR_S(vhsubs_scalar, do_vhsub_s) DO_2OP_SCALAR_U(vhsubu_scalar, do_vhsub_u) +static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) +{ + m &= 0xff; + if (m == 0) { + return 0; + } + n = revbit8(n); + if (m < 8) { + n >>= 8 - m; + } + return n; +} + +static inline uint32_t do_vbrsrh(uint32_t n, uint32_t m) +{ + m &= 0xff; + if (m == 0) { + return 0; + } + n = revbit16(n); + if (m < 16) { + n >>= 16 - m; + } + return n; +} + +static inline uint32_t do_vbrsrw(uint32_t n, uint32_t m) +{ + m &= 0xff; + if (m == 0) { + return 0; + } + n = revbit32(n); + if (m < 32) { + n >>= 32 - m; + } + return n; +} + +DO_2OP_SCALAR(vbrsrb, 1, uint8_t, do_vbrsrb) +DO_2OP_SCALAR(vbrsrh, 2, uint16_t, do_vbrsrh) +DO_2OP_SCALAR(vbrsrw, 4, uint32_t, do_vbrsrw) + /* * Multiply add long dual accumulate ops. */ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR(VHADD_S_scalar, vhadds_scalar) DO_2OP_SCALAR(VHADD_U_scalar, vhaddu_scalar) DO_2OP_SCALAR(VHSUB_S_scalar, vhsubs_scalar) DO_2OP_SCALAR(VHSUB_U_scalar, vhsubu_scalar) +DO_2OP_SCALAR(VBRSR, vbrsr) static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, MVEGenDualAccOpFn *fn) -- 2.20.1 Implement the MVE VPST insn, which sets the predicate mask fields in the VPR to the immediate value encoded in the insn. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-27-peter.maydell@linaro.org --- target/arm/mve.decode | 4 +++ target/arm/translate-mve.c | 59 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 63 insertions(+) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VHADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar VHSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar + +# Predicate operations +%mask_22_13 22:1 13:3 +VPST 1111 1110 0 . 11 000 1 ... 0 1111 0100 1101 mask=%mask_22_13 diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static void mve_update_eci(DisasContext *s) } } +static void mve_update_and_store_eci(DisasContext *s) +{ + /* + * For insns which don't call a helper function that will call + * mve_advance_vpt(), this version updates s->eci and also stores + * it out to the CPUState field. + */ + if (s->eci) { + mve_update_eci(s); + store_cpu_field(tcg_constant_i32(s->eci << 4), condexec_bits); + } +} + static bool mve_skip_first_beat(DisasContext *s) { /* Return true if PSR.ECI says we must skip the first beat of this insn */ @@ -XXX,XX +XXX,XX @@ static bool trans_VRMLSLDAVH(DisasContext *s, arg_vmlaldav *a) }; return do_long_dual_acc(s, a, fns[a->x]); } + +static bool trans_VPST(DisasContext *s, arg_VPST *a) +{ + TCGv_i32 vpr; + + /* mask == 0 is a "related encoding" */ + if (!dc_isar_feature(aa32_mve, s) || !a->mask) { + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + /* + * Set the VPR mask fields. We take advantage of MASK01 and MASK23 + * being adjacent fields in the register. + * + * This insn is not predicated, but it is subject to beat-wise + * execution, and the mask is updated on the odd-numbered beats. + * So if PSR.ECI says we should skip beat 1, we mustn't update the + * 01 mask field. + */ + vpr = load_cpu_field(v7m.vpr); + switch (s->eci) { + case ECI_NONE: + case ECI_A0: + /* Update both 01 and 23 fields */ + tcg_gen_deposit_i32(vpr, vpr, + tcg_constant_i32(a->mask | (a->mask << 4)), + R_V7M_VPR_MASK01_SHIFT, + R_V7M_VPR_MASK01_LENGTH + R_V7M_VPR_MASK23_LENGTH); + break; + case ECI_A0A1: + case ECI_A0A1A2: + case ECI_A0A1A2B0: + /* Update only the 23 mask field */ + tcg_gen_deposit_i32(vpr, vpr, + tcg_constant_i32(a->mask), + R_V7M_VPR_MASK23_SHIFT, R_V7M_VPR_MASK23_LENGTH); + break; + default: + g_assert_not_reached(); + } + store_cpu_field(vpr, v7m.vpr); + mve_update_and_store_eci(s); + return true; +} -- 2.20.1 Implement the MVE VQADD and VQSUB insns, which perform saturating addition of a scalar to each element. Note that individual bytes of each result element are used or discarded according to the predicate mask, but FPSCR.QC is only set if the predicate mask for the lowest byte of the element is set. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-28-peter.maydell@linaro.org --- target/arm/helper-mve.h | 16 ++++++++++ target/arm/mve.decode | 5 +++ target/arm/mve_helper.c | 62 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 4 +++ 4 files changed, 87 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vhsubu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vhsubu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vhsubu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqadds_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqadds_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqadds_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vqaddu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqaddu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqaddu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vqsubs_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqsubs_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqsubs_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vqsubu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqsubu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqsubu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vbrsrb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VHADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar VHSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar + +VQADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar +VQADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar +VQSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar +VQSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar # Predicate operations diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhaddu, do_vhadd_u) DO_2OP_S(vhsubs, do_vhsub_s) DO_2OP_U(vhsubu, do_vhsub_u) +static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) +{ + if (val > max) { + *s = true; + return max; + } else if (val < min) { + *s = true; + return min; + } + return val; +} + +#define DO_SQADD_B(n, m, s) do_sat_bhw((int64_t)n + m, INT8_MIN, INT8_MAX, s) +#define DO_SQADD_H(n, m, s) do_sat_bhw((int64_t)n + m, INT16_MIN, INT16_MAX, s) +#define DO_SQADD_W(n, m, s) do_sat_bhw((int64_t)n + m, INT32_MIN, INT32_MAX, s) + +#define DO_UQADD_B(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT8_MAX, s) +#define DO_UQADD_H(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT16_MAX, s) +#define DO_UQADD_W(n, m, s) do_sat_bhw((int64_t)n + m, 0, UINT32_MAX, s) + +#define DO_SQSUB_B(n, m, s) do_sat_bhw((int64_t)n - m, INT8_MIN, INT8_MAX, s) +#define DO_SQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, INT16_MIN, INT16_MAX, s) +#define DO_SQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, INT32_MIN, INT32_MAX, s) + +#define DO_UQSUB_B(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT8_MAX, s) +#define DO_UQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT16_MAX, s) +#define DO_UQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT32_MAX, s) #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhsubu, do_vhsub_u) mve_advance_vpt(env); \ } +#define DO_2OP_SAT_SCALAR(OP, ESIZE, TYPE, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ + uint32_t rm) \ + { \ + TYPE *d = vd, *n = vn; \ + TYPE m = rm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + bool qc = false; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + bool sat = false; \ + mergemask(&d[H##ESIZE(e)], FN(n[H##ESIZE(e)], m, &sat), \ + mask); \ + qc |= sat & mask & 1; \ + } \ + if (qc) { \ + env->vfp.qc[0] = qc; \ + } \ + mve_advance_vpt(env); \ + } + /* provide unsigned 2-op scalar helpers for all sizes */ #define DO_2OP_SCALAR_U(OP, FN) \ DO_2OP_SCALAR(OP##b, 1, uint8_t, FN) \ @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR_U(vhaddu_scalar, do_vhadd_u) DO_2OP_SCALAR_S(vhsubs_scalar, do_vhsub_s) DO_2OP_SCALAR_U(vhsubu_scalar, do_vhsub_u) +DO_2OP_SAT_SCALAR(vqaddu_scalarb, 1, uint8_t, DO_UQADD_B) +DO_2OP_SAT_SCALAR(vqaddu_scalarh, 2, uint16_t, DO_UQADD_H) +DO_2OP_SAT_SCALAR(vqaddu_scalarw, 4, uint32_t, DO_UQADD_W) +DO_2OP_SAT_SCALAR(vqadds_scalarb, 1, int8_t, DO_SQADD_B) +DO_2OP_SAT_SCALAR(vqadds_scalarh, 2, int16_t, DO_SQADD_H) +DO_2OP_SAT_SCALAR(vqadds_scalarw, 4, int32_t, DO_SQADD_W) + +DO_2OP_SAT_SCALAR(vqsubu_scalarb, 1, uint8_t, DO_UQSUB_B) +DO_2OP_SAT_SCALAR(vqsubu_scalarh, 2, uint16_t, DO_UQSUB_H) +DO_2OP_SAT_SCALAR(vqsubu_scalarw, 4, uint32_t, DO_UQSUB_W) +DO_2OP_SAT_SCALAR(vqsubs_scalarb, 1, int8_t, DO_SQSUB_B) +DO_2OP_SAT_SCALAR(vqsubs_scalarh, 2, int16_t, DO_SQSUB_H) +DO_2OP_SAT_SCALAR(vqsubs_scalarw, 4, int32_t, DO_SQSUB_W) + static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) { m &= 0xff; diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR(VHADD_S_scalar, vhadds_scalar) DO_2OP_SCALAR(VHADD_U_scalar, vhaddu_scalar) DO_2OP_SCALAR(VHSUB_S_scalar, vhsubs_scalar) DO_2OP_SCALAR(VHSUB_U_scalar, vhsubu_scalar) +DO_2OP_SCALAR(VQADD_S_scalar, vqadds_scalar) +DO_2OP_SCALAR(VQADD_U_scalar, vqaddu_scalar) +DO_2OP_SCALAR(VQSUB_S_scalar, vqsubs_scalar) +DO_2OP_SCALAR(VQSUB_U_scalar, vqsubu_scalar) DO_2OP_SCALAR(VBRSR, vbrsr) static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, -- 2.20.1 Implement the MVE VQDMULH and VQRDMULH scalar insns, which multiply elements by the scalar, double, possibly round, take the high half and saturate. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-29-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 25 +++++++++++++++++++++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 38 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqsubu_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vqsubu_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vqsubu_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmulh_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmulh_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmulh_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(mve_vqrdmulh_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqrdmulh_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqrdmulh_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vbrsrb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar VQSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar +VQDMULH_scalar 1110 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar +VQRDMULH_scalar 1111 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar + # Predicate operations %mask_22_13 22:1 13:3 VPST 1111 1110 0 . 11 000 1 ... 0 1111 0100 1101 mask=%mask_22_13 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) #define DO_UQSUB_H(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT16_MAX, s) #define DO_UQSUB_W(n, m, s) do_sat_bhw((int64_t)n - m, 0, UINT32_MAX, s) +/* + * For QDMULH and QRDMULH we simplify "double and shift by esize" into + * "shift by esize-1", adjusting the QRDMULH rounding constant to match. + */ +#define DO_QDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m) >> 7, \ + INT8_MIN, INT8_MAX, s) +#define DO_QDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m) >> 15, \ + INT16_MIN, INT16_MAX, s) +#define DO_QDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m) >> 31, \ + INT32_MIN, INT32_MAX, s) + +#define DO_QRDMULH_B(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 6)) >> 7, \ + INT8_MIN, INT8_MAX, s) +#define DO_QRDMULH_H(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 14)) >> 15, \ + INT16_MIN, INT16_MAX, s) +#define DO_QRDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 30)) >> 31, \ + INT32_MIN, INT32_MAX, s) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT_SCALAR(vqsubs_scalarb, 1, int8_t, DO_SQSUB_B) DO_2OP_SAT_SCALAR(vqsubs_scalarh, 2, int16_t, DO_SQSUB_H) DO_2OP_SAT_SCALAR(vqsubs_scalarw, 4, int32_t, DO_SQSUB_W) +DO_2OP_SAT_SCALAR(vqdmulh_scalarb, 1, int8_t, DO_QDMULH_B) +DO_2OP_SAT_SCALAR(vqdmulh_scalarh, 2, int16_t, DO_QDMULH_H) +DO_2OP_SAT_SCALAR(vqdmulh_scalarw, 4, int32_t, DO_QDMULH_W) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarb, 1, int8_t, DO_QRDMULH_B) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarh, 2, int16_t, DO_QRDMULH_H) +DO_2OP_SAT_SCALAR(vqrdmulh_scalarw, 4, int32_t, DO_QRDMULH_W) + static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) { m &= 0xff; diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR(VQADD_S_scalar, vqadds_scalar) DO_2OP_SCALAR(VQADD_U_scalar, vqaddu_scalar) DO_2OP_SCALAR(VQSUB_S_scalar, vqsubs_scalar) DO_2OP_SCALAR(VQSUB_U_scalar, vqsubu_scalar) +DO_2OP_SCALAR(VQDMULH_scalar, vqdmulh_scalar) +DO_2OP_SCALAR(VQRDMULH_scalar, vqrdmulh_scalar) DO_2OP_SCALAR(VBRSR, vbrsr) static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, -- 2.20.1 Implement the MVE VQDMULL scalar insn. This multiplies the top or bottom half of each element by the scalar, doubles and saturates to a double-width result. Note that this encoding overlaps with VQADD and VQSUB; it uses what in VQADD and VQSUB would be the 'size=0b11' encoding. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-30-peter.maydell@linaro.org --- target/arm/helper-mve.h | 5 +++ target/arm/mve.decode | 23 +++++++++++--- target/arm/mve_helper.c | 65 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 30 ++++++++++++++++++ 4 files changed, 119 insertions(+), 4 deletions(-) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vbrsrb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vbrsrw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmullb_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmullb_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmullt_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(mve_vqdmullt_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vmlaldavsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vmlaldavxsh, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ %qm 5:1 1:3 %qn 7:1 17:3 +# VQDMULL has size in bit 28: 0 for 16 bit, 1 for 32 bit +%size_28 28:1 !function=plus_1 + &vldr_vstr rn qd imm p a w size l u &1op qd qm size &2op qd qm qn size @@ -XXX,XX +XXX,XX @@ @2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 @2scalar .... .... .. size:2 .... .... .... .... rm:4 &2scalar qd=%qd qn=%qn +@2scalar_nosz .... .... .... .... .... .... .... rm:4 &2scalar qd=%qd qn=%qn # Vector loads and stores @@ -XXX,XX +XXX,XX @@ VHADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 100 .... @2scalar VHSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar VHSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 100 .... @2scalar -VQADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar -VQADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar -VQSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar -VQSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar +{ + VQADD_S_scalar 1110 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar + VQADD_U_scalar 1111 1110 0 . .. ... 0 ... 0 1111 . 110 .... @2scalar + VQDMULLB_scalar 111 . 1110 0 . 11 ... 0 ... 0 1111 . 110 .... @2scalar_nosz \ + size=%size_28 +} + +{ + VQSUB_S_scalar 1110 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar + VQSUB_U_scalar 1111 1110 0 . .. ... 0 ... 1 1111 . 110 .... @2scalar + VQDMULLT_scalar 111 . 1110 0 . 11 ... 0 ... 1 1111 . 110 .... @2scalar_nosz \ + size=%size_28 +} + VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar VQDMULH_scalar 1110 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar VQRDMULH_scalar 1111 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar + # Predicate operations %mask_22_13 22:1 13:3 VPST 1111 1110 0 . 11 000 1 ... 0 1111 0100 1101 mask=%mask_22_13 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT_SCALAR(vqrdmulh_scalarb, 1, int8_t, DO_QRDMULH_B) DO_2OP_SAT_SCALAR(vqrdmulh_scalarh, 2, int16_t, DO_QRDMULH_H) DO_2OP_SAT_SCALAR(vqrdmulh_scalarw, 4, int32_t, DO_QRDMULH_W) +/* + * Long saturating scalar ops. As with DO_2OP_L, TYPE and H are for the + * input (smaller) type and LESIZE, LTYPE, LH for the output (long) type. + * SATMASK specifies which bits of the predicate mask matter for determining + * whether to propagate a saturation indication into FPSCR.QC -- for + * the 16x16->32 case we must check only the bit corresponding to the T or B + * half that we used, but for the 32x32->64 case we propagate if the mask + * bit is set for either half. + */ +#define DO_2OP_SAT_SCALAR_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN, SATMASK) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ + uint32_t rm) \ + { \ + LTYPE *d = vd; \ + TYPE *n = vn; \ + TYPE m = rm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned le; \ + bool qc = false; \ + for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) { \ + bool sat = false; \ + LTYPE r = FN((LTYPE)n[H##ESIZE(le * 2 + TOP)], m, &sat); \ + mergemask(&d[H##LESIZE(le)], r, mask); \ + qc |= sat && (mask & SATMASK); \ + } \ + if (qc) { \ + env->vfp.qc[0] = qc; \ + } \ + mve_advance_vpt(env); \ + } + +static inline int32_t do_qdmullh(int16_t n, int16_t m, bool *sat) +{ + int64_t r = ((int64_t)n * m) * 2; + return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat); +} + +static inline int64_t do_qdmullw(int32_t n, int32_t m, bool *sat) +{ + /* The multiply can't overflow, but the doubling might */ + int64_t r = (int64_t)n * m; + if (r > INT64_MAX / 2) { + *sat = true; + return INT64_MAX; + } else if (r < INT64_MIN / 2) { + *sat = true; + return INT64_MIN; + } else { + return r * 2; + } +} + +#define SATMASK16B 1 +#define SATMASK16T (1 << 2) +#define SATMASK32 ((1 << 4) | 1) + +DO_2OP_SAT_SCALAR_L(vqdmullb_scalarh, 0, 2, int16_t, 4, int32_t, \ + do_qdmullh, SATMASK16B) +DO_2OP_SAT_SCALAR_L(vqdmullb_scalarw, 0, 4, int32_t, 8, int64_t, \ + do_qdmullw, SATMASK32) +DO_2OP_SAT_SCALAR_L(vqdmullt_scalarh, 1, 2, int16_t, 4, int32_t, \ + do_qdmullh, SATMASK16T) +DO_2OP_SAT_SCALAR_L(vqdmullt_scalarw, 1, 4, int32_t, 8, int64_t, \ + do_qdmullw, SATMASK32) + static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) { m &= 0xff; diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SCALAR(VQDMULH_scalar, vqdmulh_scalar) DO_2OP_SCALAR(VQRDMULH_scalar, vqrdmulh_scalar) DO_2OP_SCALAR(VBRSR, vbrsr) +static bool trans_VQDMULLB_scalar(DisasContext *s, arg_2scalar *a) +{ + static MVEGenTwoOpScalarFn * const fns[] = { + NULL, + gen_helper_mve_vqdmullb_scalarh, + gen_helper_mve_vqdmullb_scalarw, + NULL, + }; + if (a->qd == a->qn && a->size == MO_32) { + /* UNPREDICTABLE; we choose to undef */ + return false; + } + return do_2op_scalar(s, a, fns[a->size]); +} + +static bool trans_VQDMULLT_scalar(DisasContext *s, arg_2scalar *a) +{ + static MVEGenTwoOpScalarFn * const fns[] = { + NULL, + gen_helper_mve_vqdmullt_scalarh, + gen_helper_mve_vqdmullt_scalarw, + NULL, + }; + if (a->qd == a->qn && a->size == MO_32) { + /* UNPREDICTABLE; we choose to undef */ + return false; + } + return do_2op_scalar(s, a, fns[a->size]); +} + static bool do_long_dual_acc(DisasContext *s, arg_vmlaldav *a, MVEGenDualAccOpFn *fn) { -- 2.20.1 Implement the vector forms of the MVE VQDMULH and VQRDMULH insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-31-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 27 +++++++++++++++++++++++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 40 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vmulltub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vmulltuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmulhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmulhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmulhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrdmulhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmulhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmulhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMULL_BU 111 1 1110 0 . .. ... 1 ... 0 1110 . 0 . 0 ... 0 @2op VMULL_TS 111 0 1110 0 . .. ... 1 ... 1 1110 . 0 . 0 ... 0 @2op VMULL_TU 111 1 1110 0 . .. ... 1 ... 1 1110 . 0 . 0 ... 0 @2op +VQDMULH 1110 1111 0 . .. ... 0 ... 0 1011 . 1 . 0 ... 0 @2op +VQRDMULH 1111 1111 0 . .. ... 0 ... 0 1011 . 1 . 0 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) mve_advance_vpt(env); \ } +#define DO_2OP_SAT(OP, ESIZE, TYPE, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \ + { \ + TYPE *d = vd, *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + bool qc = false; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + bool sat = false; \ + TYPE r = FN(n[H##ESIZE(e)], m[H##ESIZE(e)], &sat); \ + mergemask(&d[H##ESIZE(e)], r, mask); \ + qc |= sat & mask & 1; \ + } \ + if (qc) { \ + env->vfp.qc[0] = qc; \ + } \ + mve_advance_vpt(env); \ + } + #define DO_AND(N, M) ((N) & (M)) #define DO_BIC(N, M) ((N) & ~(M)) #define DO_ORR(N, M) ((N) | (M)) @@ -XXX,XX +XXX,XX @@ static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) #define DO_QRDMULH_W(n, m, s) do_sat_bhw(((int64_t)n * m + (1 << 30)) >> 31, \ INT32_MIN, INT32_MAX, s) +DO_2OP_SAT(vqdmulhb, 1, int8_t, DO_QDMULH_B) +DO_2OP_SAT(vqdmulhh, 2, int16_t, DO_QDMULH_H) +DO_2OP_SAT(vqdmulhw, 4, int32_t, DO_QDMULH_W) + +DO_2OP_SAT(vqrdmulhb, 1, int8_t, DO_QRDMULH_B) +DO_2OP_SAT(vqrdmulhh, 2, int16_t, DO_QRDMULH_H) +DO_2OP_SAT(vqrdmulhw, 4, int32_t, DO_QRDMULH_W) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VMULL_BS, vmullbs) DO_2OP(VMULL_BU, vmullbu) DO_2OP(VMULL_TS, vmullts) DO_2OP(VMULL_TU, vmulltu) +DO_2OP(VQDMULH, vqdmulh) +DO_2OP(VQRDMULH, vqrdmulh) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the vector forms of the MVE VQADD and VQSUB insns. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-32-peter.maydell@linaro.org --- target/arm/helper-mve.h | 16 ++++++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 14 ++++++++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 39 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqrdmulhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmulhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmulhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqaddsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqaddsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqaddsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqaddub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqadduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqadduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqsubsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqsubsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqsubsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqsubub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqsubuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqsubuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VMULL_TU 111 1 1110 0 . .. ... 1 ... 1 1110 . 0 . 0 ... 0 @2op VQDMULH 1110 1111 0 . .. ... 0 ... 0 1011 . 1 . 0 ... 0 @2op VQRDMULH 1111 1111 0 . .. ... 0 ... 0 1011 . 1 . 0 ... 0 @2op +VQADD_S 111 0 1111 0 . .. ... 0 ... 0 0000 . 1 . 1 ... 0 @2op +VQADD_U 111 1 1111 0 . .. ... 0 ... 0 0000 . 1 . 1 ... 0 @2op +VQSUB_S 111 0 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op +VQSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT(vqrdmulhb, 1, int8_t, DO_QRDMULH_B) DO_2OP_SAT(vqrdmulhh, 2, int16_t, DO_QRDMULH_H) DO_2OP_SAT(vqrdmulhw, 4, int32_t, DO_QRDMULH_W) +DO_2OP_SAT(vqaddub, 1, uint8_t, DO_UQADD_B) +DO_2OP_SAT(vqadduh, 2, uint16_t, DO_UQADD_H) +DO_2OP_SAT(vqadduw, 4, uint32_t, DO_UQADD_W) +DO_2OP_SAT(vqaddsb, 1, int8_t, DO_SQADD_B) +DO_2OP_SAT(vqaddsh, 2, int16_t, DO_SQADD_H) +DO_2OP_SAT(vqaddsw, 4, int32_t, DO_SQADD_W) + +DO_2OP_SAT(vqsubub, 1, uint8_t, DO_UQSUB_B) +DO_2OP_SAT(vqsubuh, 2, uint16_t, DO_UQSUB_H) +DO_2OP_SAT(vqsubuw, 4, uint32_t, DO_UQSUB_W) +DO_2OP_SAT(vqsubsb, 1, int8_t, DO_SQSUB_B) +DO_2OP_SAT(vqsubsh, 2, int16_t, DO_SQSUB_H) +DO_2OP_SAT(vqsubsw, 4, int32_t, DO_SQSUB_W) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VMULL_TS, vmullts) DO_2OP(VMULL_TU, vmulltu) DO_2OP(VQDMULH, vqdmulh) DO_2OP(VQRDMULH, vqrdmulh) +DO_2OP(VQADD_S, vqadds) +DO_2OP(VQADD_U, vqaddu) +DO_2OP(VQSUB_S, vqsubs) +DO_2OP(VQSUB_U, vqsubu) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the MVE VQSHL insn (encoding T4, which is the vector-shift-by-vector version). The DO_SQSHL_OP and DO_UQSHL_OP macros here are derived from the neon_helper.c code for qshl_u{8,16,32} and qshl_s{8,16,32}. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-33-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 12 ++++++++++++ target/arm/mve_helper.c | 34 ++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 56 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqsubub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqsubuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqsubuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ @2op .... .... .. size:2 .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn @2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 +# The _rev suffix indicates that Vn and Vm are reversed. This is +# the case for shifts. In the Arm ARM these insns are documented +# with the Vm and Vn fields in their usual places, but in the +# assembly the operands are listed "backwards", ie in the order +# Qd, Qm, Qn where other insns use Qd, Qn, Qm. For QEMU we choose +# to consider Vm and Vn as being in different fields in the insn. +# This gives us consistency with A64 and Neon. +@2op_rev .... .... .. size:2 .... .... .... .... .... &2op qd=%qd qm=%qn qn=%qm + @2scalar .... .... .. size:2 .... .... .... .... rm:4 &2scalar qd=%qd qn=%qn @2scalar_nosz .... .... .... .... .... .... .... rm:4 &2scalar qd=%qd qn=%qn @@ -XXX,XX +XXX,XX @@ VQADD_U 111 1 1111 0 . .. ... 0 ... 0 0000 . 1 . 1 ... 0 @2op VQSUB_S 111 0 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op VQSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op +VQSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev +VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_1OP(vfnegs, 8, uint64_t, DO_FNEGS) mve_advance_vpt(env); \ } +/* provide unsigned 2-op helpers for all sizes */ +#define DO_2OP_SAT_U(OP, FN) \ + DO_2OP_SAT(OP##b, 1, uint8_t, FN) \ + DO_2OP_SAT(OP##h, 2, uint16_t, FN) \ + DO_2OP_SAT(OP##w, 4, uint32_t, FN) + +/* provide signed 2-op helpers for all sizes */ +#define DO_2OP_SAT_S(OP, FN) \ + DO_2OP_SAT(OP##b, 1, int8_t, FN) \ + DO_2OP_SAT(OP##h, 2, int16_t, FN) \ + DO_2OP_SAT(OP##w, 4, int32_t, FN) + #define DO_AND(N, M) ((N) & (M)) #define DO_BIC(N, M) ((N) & ~(M)) #define DO_ORR(N, M) ((N) | (M)) @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT(vqsubsb, 1, int8_t, DO_SQSUB_B) DO_2OP_SAT(vqsubsh, 2, int16_t, DO_SQSUB_H) DO_2OP_SAT(vqsubsw, 4, int32_t, DO_SQSUB_W) +/* + * This wrapper fixes up the impedance mismatch between do_sqrshl_bhs() + * and friends wanting a uint32_t* sat and our needing a bool*. + */ +#define WRAP_QRSHL_HELPER(FN, N, M, ROUND, satp) \ + ({ \ + uint32_t su32 = 0; \ + typeof(N) r = FN(N, (int8_t)(M), sizeof(N) * 8, ROUND, &su32); \ + if (su32) { \ + *satp = true; \ + } \ + r; \ + }) + +#define DO_SQSHL_OP(N, M, satp) \ + WRAP_QRSHL_HELPER(do_sqrshl_bhs, N, M, false, satp) +#define DO_UQSHL_OP(N, M, satp) \ + WRAP_QRSHL_HELPER(do_uqrshl_bhs, N, M, false, satp) + +DO_2OP_SAT_S(vqshls, DO_SQSHL_OP) +DO_2OP_SAT_U(vqshlu, DO_UQSHL_OP) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQADD_S, vqadds) DO_2OP(VQADD_U, vqaddu) DO_2OP(VQSUB_S, vqsubs) DO_2OP(VQSUB_U, vqsubu) +DO_2OP(VQSHL_S, vqshls) +DO_2OP(VQSHL_U, vqshlu) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the MV VQRSHL (vector) insn. Again, the code to perform the actual shifts is borrowed from neon_helper.c. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-34-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 6 ++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 19 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op VQSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev +VQRSHL_S 111 0 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev +VQRSHL_U 111 1 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT(vqsubsw, 4, int32_t, DO_SQSUB_W) WRAP_QRSHL_HELPER(do_sqrshl_bhs, N, M, false, satp) #define DO_UQSHL_OP(N, M, satp) \ WRAP_QRSHL_HELPER(do_uqrshl_bhs, N, M, false, satp) +#define DO_SQRSHL_OP(N, M, satp) \ + WRAP_QRSHL_HELPER(do_sqrshl_bhs, N, M, true, satp) +#define DO_UQRSHL_OP(N, M, satp) \ + WRAP_QRSHL_HELPER(do_uqrshl_bhs, N, M, true, satp) DO_2OP_SAT_S(vqshls, DO_SQSHL_OP) DO_2OP_SAT_U(vqshlu, DO_UQSHL_OP) +DO_2OP_SAT_S(vqrshls, DO_SQRSHL_OP) +DO_2OP_SAT_U(vqrshlu, DO_UQRSHL_OP) #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQSUB_S, vqsubs) DO_2OP(VQSUB_U, vqsubu) DO_2OP(VQSHL_S, vqshls) DO_2OP(VQSHL_U, vqshlu) +DO_2OP(VQRSHL_S, vqrshls) +DO_2OP(VQRSHL_U, vqrshlu) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the MVE VSHL insn (vector form). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-35-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 6 ++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 19 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqsubub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqsubuh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqsubuw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vqshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQADD_U 111 1 1111 0 . .. ... 0 ... 0 0000 . 1 . 1 ... 0 @2op VQSUB_S 111 0 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op VQSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op +VSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 0 ... 0 @2op_rev +VSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 0 ... 0 @2op_rev + VQSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhaddu, do_vhadd_u) DO_2OP_S(vhsubs, do_vhsub_s) DO_2OP_U(vhsubu, do_vhsub_u) +#define DO_VSHLS(N, M) do_sqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL) +#define DO_VSHLU(N, M) do_uqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL) + +DO_2OP_S(vshls, DO_VSHLS) +DO_2OP_U(vshlu, DO_VSHLU) + static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { if (val > max) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQADD_S, vqadds) DO_2OP(VQADD_U, vqaddu) DO_2OP(VQSUB_S, vqsubs) DO_2OP(VQSUB_U, vqsubu) +DO_2OP(VSHL_S, vshls) +DO_2OP(VSHL_U, vshlu) DO_2OP(VQSHL_S, vqshls) DO_2OP(VQSHL_U, vqshlu) DO_2OP(VQRSHL_S, vqrshls) -- 2.20.1 Implement the MVE VRSHL insn (vector form). Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-36-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 4 ++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 17 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vrshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vqshlsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshlsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqshlsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQSUB_U 111 1 1111 0 . .. ... 0 ... 0 0010 . 1 . 1 ... 0 @2op VSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 0 ... 0 @2op_rev VSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 0 ... 0 @2op_rev +VRSHL_S 111 0 1111 0 . .. ... 0 ... 0 0101 . 1 . 0 ... 0 @2op_rev +VRSHL_U 111 1 1111 0 . .. ... 0 ... 0 0101 . 1 . 0 ... 0 @2op_rev + VQSHL_S 111 0 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vhsubu, do_vhsub_u) #define DO_VSHLS(N, M) do_sqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL) #define DO_VSHLU(N, M) do_uqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, false, NULL) +#define DO_VRSHLS(N, M) do_sqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, true, NULL) +#define DO_VRSHLU(N, M) do_uqrshl_bhs(N, (int8_t)(M), sizeof(N) * 8, true, NULL) DO_2OP_S(vshls, DO_VSHLS) DO_2OP_U(vshlu, DO_VSHLU) +DO_2OP_S(vrshls, DO_VRSHLS) +DO_2OP_U(vrshlu, DO_VRSHLU) static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQSUB_S, vqsubs) DO_2OP(VQSUB_U, vqsubu) DO_2OP(VSHL_S, vshls) DO_2OP(VSHL_U, vshlu) +DO_2OP(VRSHL_S, vrshls) +DO_2OP(VRSHL_U, vrshlu) DO_2OP(VQSHL_S, vqshls) DO_2OP(VQSHL_U, vqshlu) DO_2OP(VQRSHL_S, vqrshls) -- 2.20.1 Implement the MVE VQDMLADH and VQRDMLADH insns. These multiply elements, and then add pairs of products, double, possibly round, saturate and return the high half of the result. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-37-peter.maydell@linaro.org --- target/arm/helper-mve.h | 16 +++++++ target/arm/mve.decode | 5 +++ target/arm/mve_helper.c | 89 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 4 ++ 4 files changed, 114 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqrshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmladhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmladhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmladhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrdmladhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmladhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmladhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev VQRSHL_S 111 0 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev VQRSHL_U 111 1 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev +VQDMLADH 1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 0 @2op +VQDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op +VQRDMLADH 1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op +VQRDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT_U(vqshlu, DO_UQSHL_OP) DO_2OP_SAT_S(vqrshls, DO_SQRSHL_OP) DO_2OP_SAT_U(vqrshlu, DO_UQRSHL_OP) +/* + * Multiply add dual returning high half + * The 'FN' here takes four inputs A, B, C, D, a 0/1 indicator of + * whether to add the rounding constant, and the pointer to the + * saturation flag, and should do "(A * B + C * D) * 2 + rounding constant", + * saturate to twice the input size and return the high half; or + * (A * B - C * D) etc for VQDMLSDH. + */ +#define DO_VQDMLADH_OP(OP, ESIZE, TYPE, XCHG, ROUND, FN) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ + void *vm) \ + { \ + TYPE *d = vd, *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + bool qc = false; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + bool sat = false; \ + if ((e & 1) == XCHG) { \ + TYPE r = FN(n[H##ESIZE(e)], \ + m[H##ESIZE(e - XCHG)], \ + n[H##ESIZE(e + (1 - 2 * XCHG))], \ + m[H##ESIZE(e + (1 - XCHG))], \ + ROUND, &sat); \ + mergemask(&d[H##ESIZE(e)], r, mask); \ + qc |= sat & mask & 1; \ + } \ + } \ + if (qc) { \ + env->vfp.qc[0] = qc; \ + } \ + mve_advance_vpt(env); \ + } + +static int8_t do_vqdmladh_b(int8_t a, int8_t b, int8_t c, int8_t d, + int round, bool *sat) +{ + int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 7); + return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8; +} + +static int16_t do_vqdmladh_h(int16_t a, int16_t b, int16_t c, int16_t d, + int round, bool *sat) +{ + int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 15); + return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16; +} + +static int32_t do_vqdmladh_w(int32_t a, int32_t b, int32_t c, int32_t d, + int round, bool *sat) +{ + int64_t m1 = (int64_t)a * b; + int64_t m2 = (int64_t)c * d; + int64_t r; + /* + * Architecturally we should do the entire add, double, round + * and then check for saturation. We do three saturating adds, + * but we need to be careful about the order. If the first + * m1 + m2 saturates then it's impossible for the *2+rc to + * bring it back into the non-saturated range. However, if + * m1 + m2 is negative then it's possible that doing the doubling + * would take the intermediate result below INT64_MAX and the + * addition of the rounding constant then brings it back in range. + * So we add half the rounding constant before doubling rather + * than adding the rounding constant after the doubling. + */ + if (sadd64_overflow(m1, m2, &r) || + sadd64_overflow(r, (round << 30), &r) || + sadd64_overflow(r, r, &r)) { + *sat = true; + return r < 0 ? INT32_MAX : INT32_MIN; + } + return r >> 32; +} + +DO_VQDMLADH_OP(vqdmladhb, 1, int8_t, 0, 0, do_vqdmladh_b) +DO_VQDMLADH_OP(vqdmladhh, 2, int16_t, 0, 0, do_vqdmladh_h) +DO_VQDMLADH_OP(vqdmladhw, 4, int32_t, 0, 0, do_vqdmladh_w) +DO_VQDMLADH_OP(vqdmladhxb, 1, int8_t, 1, 0, do_vqdmladh_b) +DO_VQDMLADH_OP(vqdmladhxh, 2, int16_t, 1, 0, do_vqdmladh_h) +DO_VQDMLADH_OP(vqdmladhxw, 4, int32_t, 1, 0, do_vqdmladh_w) + +DO_VQDMLADH_OP(vqrdmladhb, 1, int8_t, 0, 1, do_vqdmladh_b) +DO_VQDMLADH_OP(vqrdmladhh, 2, int16_t, 0, 1, do_vqdmladh_h) +DO_VQDMLADH_OP(vqrdmladhw, 4, int32_t, 0, 1, do_vqdmladh_w) +DO_VQDMLADH_OP(vqrdmladhxb, 1, int8_t, 1, 1, do_vqdmladh_b) +DO_VQDMLADH_OP(vqrdmladhxh, 2, int16_t, 1, 1, do_vqdmladh_h) +DO_VQDMLADH_OP(vqrdmladhxw, 4, int32_t, 1, 1, do_vqdmladh_w) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQSHL_S, vqshls) DO_2OP(VQSHL_U, vqshlu) DO_2OP(VQRSHL_S, vqrshls) DO_2OP(VQRSHL_U, vqrshlu) +DO_2OP(VQDMLADH, vqdmladh) +DO_2OP(VQDMLADHX, vqdmladhx) +DO_2OP(VQRDMLADH, vqrdmladh) +DO_2OP(VQRDMLADHX, vqrdmladhx) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the MVE VQDMLSDH and VQRDMLSDH insns, which are like VQDMLADH and VQRDMLADH except that products are subtracted rather than added. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-38-peter.maydell@linaro.org --- target/arm/helper-mve.h | 16 ++++++++++++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 44 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 4 ++++ 4 files changed, 69 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqrdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmlsdhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmlsdhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmlsdhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqdmlsdhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmlsdhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmlsdhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op VQRDMLADH 1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op VQRDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op +VQDMLSDH 1111 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 0 @2op +VQDMLSDHX 1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op +VQRDMLSDH 1111 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op +VQRDMLSDHX 1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ static int32_t do_vqdmladh_w(int32_t a, int32_t b, int32_t c, int32_t d, return r >> 32; } +static int8_t do_vqdmlsdh_b(int8_t a, int8_t b, int8_t c, int8_t d, + int round, bool *sat) +{ + int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 7); + return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8; +} + +static int16_t do_vqdmlsdh_h(int16_t a, int16_t b, int16_t c, int16_t d, + int round, bool *sat) +{ + int64_t r = ((int64_t)a * b - (int64_t)c * d) * 2 + (round << 15); + return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16; +} + +static int32_t do_vqdmlsdh_w(int32_t a, int32_t b, int32_t c, int32_t d, + int round, bool *sat) +{ + int64_t m1 = (int64_t)a * b; + int64_t m2 = (int64_t)c * d; + int64_t r; + /* The same ordering issue as in do_vqdmladh_w applies here too */ + if (ssub64_overflow(m1, m2, &r) || + sadd64_overflow(r, (round << 30), &r) || + sadd64_overflow(r, r, &r)) { + *sat = true; + return r < 0 ? INT32_MAX : INT32_MIN; + } + return r >> 32; +} + DO_VQDMLADH_OP(vqdmladhb, 1, int8_t, 0, 0, do_vqdmladh_b) DO_VQDMLADH_OP(vqdmladhh, 2, int16_t, 0, 0, do_vqdmladh_h) DO_VQDMLADH_OP(vqdmladhw, 4, int32_t, 0, 0, do_vqdmladh_w) @@ -XXX,XX +XXX,XX @@ DO_VQDMLADH_OP(vqrdmladhxb, 1, int8_t, 1, 1, do_vqdmladh_b) DO_VQDMLADH_OP(vqrdmladhxh, 2, int16_t, 1, 1, do_vqdmladh_h) DO_VQDMLADH_OP(vqrdmladhxw, 4, int32_t, 1, 1, do_vqdmladh_w) +DO_VQDMLADH_OP(vqdmlsdhb, 1, int8_t, 0, 0, do_vqdmlsdh_b) +DO_VQDMLADH_OP(vqdmlsdhh, 2, int16_t, 0, 0, do_vqdmlsdh_h) +DO_VQDMLADH_OP(vqdmlsdhw, 4, int32_t, 0, 0, do_vqdmlsdh_w) +DO_VQDMLADH_OP(vqdmlsdhxb, 1, int8_t, 1, 0, do_vqdmlsdh_b) +DO_VQDMLADH_OP(vqdmlsdhxh, 2, int16_t, 1, 0, do_vqdmlsdh_h) +DO_VQDMLADH_OP(vqdmlsdhxw, 4, int32_t, 1, 0, do_vqdmlsdh_w) + +DO_VQDMLADH_OP(vqrdmlsdhb, 1, int8_t, 0, 1, do_vqdmlsdh_b) +DO_VQDMLADH_OP(vqrdmlsdhh, 2, int16_t, 0, 1, do_vqdmlsdh_h) +DO_VQDMLADH_OP(vqrdmlsdhw, 4, int32_t, 0, 1, do_vqdmlsdh_w) +DO_VQDMLADH_OP(vqrdmlsdhxb, 1, int8_t, 1, 1, do_vqdmlsdh_b) +DO_VQDMLADH_OP(vqrdmlsdhxh, 2, int16_t, 1, 1, do_vqdmlsdh_h) +DO_VQDMLADH_OP(vqrdmlsdhxw, 4, int32_t, 1, 1, do_vqdmlsdh_w) + #define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ uint32_t rm) \ diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQDMLADH, vqdmladh) DO_2OP(VQDMLADHX, vqdmladhx) DO_2OP(VQRDMLADH, vqrdmladh) DO_2OP(VQRDMLADHX, vqrdmladhx) +DO_2OP(VQDMLSDH, vqdmlsdh) +DO_2OP(VQDMLSDHX, vqdmlsdhx) +DO_2OP(VQRDMLSDH, vqrdmlsdh) +DO_2OP(VQRDMLSDHX, vqrdmlsdhx) static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) -- 2.20.1 Implement the vector form of the MVE VQDMULL insn. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-39-peter.maydell@linaro.org --- target/arm/helper-mve.h | 5 +++++ target/arm/mve.decode | 5 +++++ target/arm/mve_helper.c | 30 ++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 30 ++++++++++++++++++++++++++++++ 4 files changed, 70 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqrdmlsdhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmullbh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmullbw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmullth, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vqdmulltw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ @1op_nosz .... .... .... .... .... .... .... .... &1op qd=%qd qm=%qm size=0 @2op .... .... .. size:2 .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn @2op_nosz .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn size=0 +@2op_sz28 .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn \ + size=%size_28 # The _rev suffix indicates that Vn and Vm are reversed. This is # the case for shifts. In the Arm ARM these insns are documented @@ -XXX,XX +XXX,XX @@ VQDMLSDHX 1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op VQRDMLSDH 1111 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op VQRDMLSDHX 1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op +VQDMULLB 111 . 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 1 @2op_sz28 +VQDMULLT 111 . 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 1 @2op_sz28 + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_SAT_SCALAR_L(vqdmullt_scalarh, 1, 2, int16_t, 4, int32_t, \ DO_2OP_SAT_SCALAR_L(vqdmullt_scalarw, 1, 4, int32_t, 8, int64_t, \ do_qdmullw, SATMASK32) +/* + * Long saturating ops + */ +#define DO_2OP_SAT_L(OP, TOP, ESIZE, TYPE, LESIZE, LTYPE, FN, SATMASK) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \ + void *vm) \ + { \ + LTYPE *d = vd; \ + TYPE *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned le; \ + bool qc = false; \ + for (le = 0; le < 16 / LESIZE; le++, mask >>= LESIZE) { \ + bool sat = false; \ + LTYPE op1 = n[H##ESIZE(le * 2 + TOP)]; \ + LTYPE op2 = m[H##ESIZE(le * 2 + TOP)]; \ + mergemask(&d[H##LESIZE(le)], FN(op1, op2, &sat), mask); \ + qc |= sat && (mask & SATMASK); \ + } \ + if (qc) { \ + env->vfp.qc[0] = qc; \ + } \ + mve_advance_vpt(env); \ + } + +DO_2OP_SAT_L(vqdmullbh, 0, 2, int16_t, 4, int32_t, do_qdmullh, SATMASK16B) +DO_2OP_SAT_L(vqdmullbw, 0, 4, int32_t, 8, int64_t, do_qdmullw, SATMASK32) +DO_2OP_SAT_L(vqdmullth, 1, 2, int16_t, 4, int32_t, do_qdmullh, SATMASK16T) +DO_2OP_SAT_L(vqdmulltw, 1, 4, int32_t, 8, int64_t, do_qdmullw, SATMASK32) + static inline uint32_t do_vbrsrb(uint32_t n, uint32_t m) { m &= 0xff; diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQDMLSDHX, vqdmlsdhx) DO_2OP(VQRDMLSDH, vqrdmlsdh) DO_2OP(VQRDMLSDHX, vqrdmlsdhx) +static bool trans_VQDMULLB(DisasContext *s, arg_2op *a) +{ + static MVEGenTwoOpFn * const fns[] = { + NULL, + gen_helper_mve_vqdmullbh, + gen_helper_mve_vqdmullbw, + NULL, + }; + if (a->size == MO_32 && (a->qd == a->qm || a->qd == a->qn)) { + /* UNPREDICTABLE; we choose to undef */ + return false; + } + return do_2op(s, a, fns[a->size]); +} + +static bool trans_VQDMULLT(DisasContext *s, arg_2op *a) +{ + static MVEGenTwoOpFn * const fns[] = { + NULL, + gen_helper_mve_vqdmullth, + gen_helper_mve_vqdmulltw, + NULL, + }; + if (a->size == MO_32 && (a->qd == a->qm || a->qd == a->qn)) { + /* UNPREDICTABLE; we choose to undef */ + return false; + } + return do_2op(s, a, fns[a->size]); +} + static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) { -- 2.20.1 Implement the MVE VRHADD insn, which performs a rounded halving addition. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-40-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 3 +++ target/arm/mve_helper.c | 6 ++++++ target/arm/translate-mve.c | 2 ++ 4 files changed, 19 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vqdmullbw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqdmullth, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vqdmulltw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrhaddsb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrhaddsh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrhaddsw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vrhaddub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrhadduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vrhadduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQRDMLSDHX 1111 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op VQDMULLB 111 . 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 1 @2op_sz28 VQDMULLT 111 . 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 1 @2op_sz28 +VRHADD_S 111 0 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op +VRHADD_U 111 1 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vshlu, DO_VSHLU) DO_2OP_S(vrshls, DO_VRSHLS) DO_2OP_U(vrshlu, DO_VRSHLU) +#define DO_RHADD_S(N, M) (((int64_t)(N) + (M) + 1) >> 1) +#define DO_RHADD_U(N, M) (((uint64_t)(N) + (M) + 1) >> 1) + +DO_2OP_S(vrhadds, DO_RHADD_S) +DO_2OP_U(vrhaddu, DO_RHADD_U) + static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { if (val > max) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQDMLSDH, vqdmlsdh) DO_2OP(VQDMLSDHX, vqdmlsdhx) DO_2OP(VQRDMLSDH, vqrdmlsdh) DO_2OP(VQRDMLSDHX, vqrdmlsdhx) +DO_2OP(VRHADD_S, vrhadds) +DO_2OP(VRHADD_U, vrhaddu) static bool trans_VQDMULLB(DisasContext *s, arg_2op *a) { -- 2.20.1 Implement the MVE VADC and VSBC insns. These perform an add-with-carry or subtract-with-carry of the 32-bit elements in each lane of the input vectors, where the carry-out of each add is the carry-in of the next. The initial carry input is either 1 or is from FPSCR.C; the carry out at the end is written back to FPSCR.C. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-41-peter.maydell@linaro.org --- target/arm/helper-mve.h | 5 ++++ target/arm/mve.decode | 5 ++++ target/arm/mve_helper.c | 52 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-mve.c | 37 +++++++++++++++++++++++++++ 4 files changed, 99 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vrhaddub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrhadduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrhadduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vadc, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vadci, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vsbc, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vsbci, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQDMULLT 111 . 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 1 @2op_sz28 VRHADD_S 111 0 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op VRHADD_U 111 1 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op +VADC 1110 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz +VSBC 1111 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz +VADCI 1110 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz +VSBCI 1111 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_2OP_U(vrshlu, DO_VRSHLU) DO_2OP_S(vrhadds, DO_RHADD_S) DO_2OP_U(vrhaddu, DO_RHADD_U) +static void do_vadc(CPUARMState *env, uint32_t *d, uint32_t *n, uint32_t *m, + uint32_t inv, uint32_t carry_in, bool update_flags) +{ + uint16_t mask = mve_element_mask(env); + unsigned e; + + /* If any additions trigger, we will update flags. */ + if (mask & 0x1111) { + update_flags = true; + } + + for (e = 0; e < 16 / 4; e++, mask >>= 4) { + uint64_t r = carry_in; + r += n[H4(e)]; + r += m[H4(e)] ^ inv; + if (mask & 1) { + carry_in = r >> 32; + } + mergemask(&d[H4(e)], r, mask); + } + + if (update_flags) { + /* Store C, clear NZV. */ + env->vfp.xregs[ARM_VFP_FPSCR] &= ~FPCR_NZCV_MASK; + env->vfp.xregs[ARM_VFP_FPSCR] |= carry_in * FPCR_C; + } + mve_advance_vpt(env); +} + +void HELPER(mve_vadc)(CPUARMState *env, void *vd, void *vn, void *vm) +{ + bool carry_in = env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_C; + do_vadc(env, vd, vn, vm, 0, carry_in, false); +} + +void HELPER(mve_vsbc)(CPUARMState *env, void *vd, void *vn, void *vm) +{ + bool carry_in = env->vfp.xregs[ARM_VFP_FPSCR] & FPCR_C; + do_vadc(env, vd, vn, vm, -1, carry_in, false); +} + + +void HELPER(mve_vadci)(CPUARMState *env, void *vd, void *vn, void *vm) +{ + do_vadc(env, vd, vn, vm, 0, 0, true); +} + +void HELPER(mve_vsbci)(CPUARMState *env, void *vd, void *vn, void *vm) +{ + do_vadc(env, vd, vn, vm, -1, 1, true); +} + static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { if (val > max) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool trans_VQDMULLT(DisasContext *s, arg_2op *a) return do_2op(s, a, fns[a->size]); } +/* + * VADC and VSBC: these perform an add-with-carry or subtract-with-carry + * of the 32-bit elements in each lane of the input vectors, where the + * carry-out of each add is the carry-in of the next. The initial carry + * input is either fixed (0 for VADCI, 1 for VSBCI) or is from FPSCR.C + * (for VADC and VSBC); the carry out at the end is written back to FPSCR.C. + * These insns are subject to beat-wise execution. Partial execution + * of an I=1 (initial carry input fixed) insn which does not + * execute the first beat must start with the current FPSCR.NZCV + * value, not the fixed constant input. + */ +static bool trans_VADC(DisasContext *s, arg_2op *a) +{ + return do_2op(s, a, gen_helper_mve_vadc); +} + +static bool trans_VADCI(DisasContext *s, arg_2op *a) +{ + if (mve_skip_first_beat(s)) { + return trans_VADC(s, a); + } + return do_2op(s, a, gen_helper_mve_vadci); +} + +static bool trans_VSBC(DisasContext *s, arg_2op *a) +{ + return do_2op(s, a, gen_helper_mve_vsbc); +} + +static bool trans_VSBCI(DisasContext *s, arg_2op *a) +{ + if (mve_skip_first_beat(s)) { + return trans_VSBC(s, a); + } + return do_2op(s, a, gen_helper_mve_vsbci); +} + static bool do_2op_scalar(DisasContext *s, arg_2scalar *a, MVEGenTwoOpScalarFn fn) { -- 2.20.1 Implement the MVE VCADD insn, which performs a complex add with rotate. Note that the size=0b11 encoding is VSBC. The architecture grants some leeway for the "destination and Vm source overlap" case for the size MO_32 case, but we choose not to make use of it, instead always calculating all 16 bytes worth of results before setting the destination register. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-42-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 9 +++++++-- target/arm/mve_helper.c | 29 +++++++++++++++++++++++++++++ target/arm/translate-mve.c | 7 +++++++ 4 files changed, 51 insertions(+), 2 deletions(-) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vadci, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vsbc, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vsbci, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vcadd90b, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vcadd90h, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vcadd90w, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vcadd270b, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vcadd270h, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vcadd270w, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VRHADD_S 111 0 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op VRHADD_U 111 1 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op VADC 1110 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz -VSBC 1111 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz VADCI 1110 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz -VSBCI 1111 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz + +{ + VSBC 1111 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz + VSBCI 1111 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz + VCADD90 1111 1110 0 . .. ... 0 ... 0 1111 . 0 . 0 ... 0 @2op + VCADD270 1111 1110 0 . .. ... 0 ... 1 1111 . 0 . 0 ... 0 @2op +} # Vector miscellaneous diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(mve_vsbci)(CPUARMState *env, void *vd, void *vn, void *vm) do_vadc(env, vd, vn, vm, -1, 1, true); } +#define DO_VCADD(OP, ESIZE, TYPE, FN0, FN1) \ + void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, void *vm) \ + { \ + TYPE *d = vd, *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + TYPE r[16 / ESIZE]; \ + /* Calculate all results first to avoid overwriting inputs */ \ + for (e = 0; e < 16 / ESIZE; e++) { \ + if (!(e & 1)) { \ + r[e] = FN0(n[H##ESIZE(e)], m[H##ESIZE(e + 1)]); \ + } else { \ + r[e] = FN1(n[H##ESIZE(e)], m[H##ESIZE(e - 1)]); \ + } \ + } \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + mergemask(&d[H##ESIZE(e)], r[e], mask); \ + } \ + mve_advance_vpt(env); \ + } + +#define DO_VCADD_ALL(OP, FN0, FN1) \ + DO_VCADD(OP##b, 1, int8_t, FN0, FN1) \ + DO_VCADD(OP##h, 2, int16_t, FN0, FN1) \ + DO_VCADD(OP##w, 4, int32_t, FN0, FN1) + +DO_VCADD_ALL(vcadd90, DO_SUB, DO_ADD) +DO_VCADD_ALL(vcadd270, DO_ADD, DO_SUB) + static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { if (val > max) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VQRDMLSDH, vqrdmlsdh) DO_2OP(VQRDMLSDHX, vqrdmlsdhx) DO_2OP(VRHADD_S, vrhadds) DO_2OP(VRHADD_U, vrhaddu) +/* + * VCADD Qd == Qm at size MO_32 is UNPREDICTABLE; we choose not to diagnose + * so we can reuse the DO_2OP macro. (Our implementation calculates the + * "expected" results in this case.) + */ +DO_2OP(VCADD90, vcadd90) +DO_2OP(VCADD270, vcadd270) static bool trans_VQDMULLB(DisasContext *s, arg_2op *a) { -- 2.20.1 Implement the MVE VHCADD insn, which is similar to VCADD but performs a halving step. This one overlaps with VADC. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-43-peter.maydell@linaro.org --- target/arm/helper-mve.h | 8 ++++++++ target/arm/mve.decode | 8 ++++++-- target/arm/mve_helper.c | 2 ++ target/arm/translate-mve.c | 4 +++- 4 files changed, 19 insertions(+), 3 deletions(-) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vcadd270b, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vcadd270h, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vcadd270w, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhcadd90b, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhcadd90h, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhcadd90w, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + +DEF_HELPER_FLAGS_4(mve_vhcadd270b, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhcadd270h, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_4(mve_vhcadd270w, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VQDMULLT 111 . 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 1 @2op_sz28 VRHADD_S 111 0 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op VRHADD_U 111 1 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op -VADC 1110 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz -VADCI 1110 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz +{ + VADC 1110 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz + VADCI 1110 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 0 @2op_nosz + VHCADD90 1110 1110 0 . .. ... 0 ... 0 1111 . 0 . 0 ... 0 @2op + VHCADD270 1110 1110 0 . .. ... 0 ... 1 1111 . 0 . 0 ... 0 @2op +} { VSBC 1111 1110 0 . 11 ... 0 ... 0 1111 . 0 . 0 ... 0 @2op_nosz diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(mve_vsbci)(CPUARMState *env, void *vd, void *vn, void *vm) DO_VCADD_ALL(vcadd90, DO_SUB, DO_ADD) DO_VCADD_ALL(vcadd270, DO_ADD, DO_SUB) +DO_VCADD_ALL(vhcadd90, do_vhsub_s, do_vhadd_s) +DO_VCADD_ALL(vhcadd270, do_vhadd_s, do_vhsub_s) static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ DO_2OP(VRHADD_U, vrhaddu) /* * VCADD Qd == Qm at size MO_32 is UNPREDICTABLE; we choose not to diagnose * so we can reuse the DO_2OP macro. (Our implementation calculates the - * "expected" results in this case.) + * "expected" results in this case.) Similarly for VHCADD. */ DO_2OP(VCADD90, vcadd90) DO_2OP(VCADD270, vcadd270) +DO_2OP(VHCADD90, vhcadd90) +DO_2OP(VHCADD270, vhcadd270) static bool trans_VQDMULLB(DisasContext *s, arg_2op *a) { -- 2.20.1 Implement the MVE VADDV insn, which performs an addition across vector lanes. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-44-peter.maydell@linaro.org --- target/arm/helper-mve.h | 7 +++++++ target/arm/mve.decode | 2 ++ target/arm/mve_helper.c | 24 +++++++++++++++++++++ target/arm/translate-mve.c | 43 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 76 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(mve_vrmlaldavhuw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vrmlsldavhsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) DEF_HELPER_FLAGS_4(mve_vrmlsldavhxsw, TCG_CALL_NO_WG, i64, env, ptr, ptr, i64) + +DEF_HELPER_FLAGS_3(mve_vaddvsb, TCG_CALL_NO_WG, i32, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vaddvub, TCG_CALL_NO_WG, i32, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vaddvsh, TCG_CALL_NO_WG, i32, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vaddvuh, TCG_CALL_NO_WG, i32, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vaddvsw, TCG_CALL_NO_WG, i32, env, ptr, i32) +DEF_HELPER_FLAGS_3(mve_vaddvuw, TCG_CALL_NO_WG, i32, env, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -XXX,XX +XXX,XX @@ VBRSR 1111 1110 0 . .. ... 1 ... 1 1110 . 110 .... @2scalar VQDMULH_scalar 1110 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar VQRDMULH_scalar 1111 1110 0 . .. ... 1 ... 0 1110 . 110 .... @2scalar +# Vector add across vector +VADDV 111 u:1 1110 1111 size:2 01 ... 0 1111 0 0 a:1 0 qm:3 0 rda=%rdalo # Predicate operations %mask_22_13 22:1 13:3 diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -XXX,XX +XXX,XX @@ DO_LDAVH(vrmlaldavhuw, 4, uint32_t, false, int128_add, int128_add, int128_make64 DO_LDAVH(vrmlsldavhsw, 4, int32_t, false, int128_add, int128_sub, int128_makes64) DO_LDAVH(vrmlsldavhxsw, 4, int32_t, true, int128_add, int128_sub, int128_makes64) + +/* Vector add across vector */ +#define DO_VADDV(OP, ESIZE, TYPE) \ + uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vm, \ + uint32_t ra) \ + { \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + TYPE *m = vm; \ + for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \ + if (mask & 1) { \ + ra += m[H##ESIZE(e)]; \ + } \ + } \ + mve_advance_vpt(env); \ + return ra; \ + } \ + +DO_VADDV(vaddvsb, 1, uint8_t) +DO_VADDV(vaddvsh, 2, uint16_t) +DO_VADDV(vaddvsw, 4, uint32_t) +DO_VADDV(vaddvub, 1, uint8_t) +DO_VADDV(vaddvuh, 2, uint16_t) +DO_VADDV(vaddvuw, 4, uint32_t) diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpScalarFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenDualAccOpFn(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i64); +typedef void MVEGenVADDVFn(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_i32); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ static inline long mve_qreg_offset(unsigned reg) @@ -XXX,XX +XXX,XX @@ static bool trans_VPST(DisasContext *s, arg_VPST *a) mve_update_and_store_eci(s); return true; } + +static bool trans_VADDV(DisasContext *s, arg_VADDV *a) +{ + /* VADDV: vector add across vector */ + static MVEGenVADDVFn * const fns[4][2] = { + { gen_helper_mve_vaddvsb, gen_helper_mve_vaddvub }, + { gen_helper_mve_vaddvsh, gen_helper_mve_vaddvuh }, + { gen_helper_mve_vaddvsw, gen_helper_mve_vaddvuw }, + { NULL, NULL } + }; + TCGv_ptr qm; + TCGv_i32 rda; + + if (!dc_isar_feature(aa32_mve, s) || + a->size == 3) { + return false; + } + if (!mve_eci_check(s) || !vfp_access_check(s)) { + return true; + } + + /* + * This insn is subject to beat-wise execution. Partial execution + * of an A=0 (no-accumulate) insn which does not execute the first + * beat must start with the current value of Rda, not zero. + */ + if (a->a || mve_skip_first_beat(s)) { + /* Accumulate input from Rda */ + rda = load_reg(s, a->rda); + } else { + /* Accumulate starting at zero */ + rda = tcg_const_i32(0); + } + + qm = mve_qreg_ptr(a->qm); + fns[a->size][a->u](rda, cpu_env, qm, rda); + store_reg(s, a->rda, rda); + tcg_temp_free_ptr(qm); + + mve_update_eci(s); + return true; +} -- 2.20.1 In a CPU with MVE, the VMOV (vector lane to general-purpose register) and VMOV (general-purpose register to vector lane) insns are not predicated, but they are subject to beatwise execution if they are not in an IT block. Since our implementation always executes all 4 beats in one tick, this means only that we need to handle PSR.ECI: * we must do the usual check for bad ECI state * we must advance ECI state if the insn succeeds * if ECI says we should not be executing the beat corresponding to the lane of the vector register being accessed then we should skip performing the move Note that if PSR.ECI is non-zero then we cannot be in an IT block. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210617121628.20116-45-peter.maydell@linaro.org --- target/arm/translate-a32.h | 2 + target/arm/translate-mve.c | 4 +- target/arm/translate-vfp.c | 77 +++++++++++++++++++++++++++++++++++--- 3 files changed, 75 insertions(+), 8 deletions(-) diff --git a/target/arm/translate-a32.h b/target/arm/translate-a32.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a32.h +++ b/target/arm/translate-a32.h @@ -XXX,XX +XXX,XX @@ long neon_full_reg_offset(unsigned reg); long neon_element_offset(int reg, int element, MemOp memop); void gen_rev16(TCGv_i32 dest, TCGv_i32 var); void clear_eci_state(DisasContext *s); +bool mve_eci_check(DisasContext *s); +void mve_update_and_store_eci(DisasContext *s); static inline TCGv_i32 load_cpu_offset(int offset) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -XXX,XX +XXX,XX @@ static bool mve_check_qreg_bank(DisasContext *s, int qmask) return qmask < 8; } -static bool mve_eci_check(DisasContext *s) +bool mve_eci_check(DisasContext *s) { /* * This is a beatwise insn: check that ECI is valid (not a @@ -XXX,XX +XXX,XX @@ static void mve_update_eci(DisasContext *s) } } -static void mve_update_and_store_eci(DisasContext *s) +void mve_update_and_store_eci(DisasContext *s) { /* * For insns which don't call a helper function that will call diff --git a/target/arm/translate-vfp.c b/target/arm/translate-vfp.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-vfp.c +++ b/target/arm/translate-vfp.c @@ -XXX,XX +XXX,XX @@ static bool trans_VCVT(DisasContext *s, arg_VCVT *a) return true; } +static bool mve_skip_vmov(DisasContext *s, int vn, int index, int size) +{ + /* + * In a CPU with MVE, the VMOV (vector lane to general-purpose register) + * and VMOV (general-purpose register to vector lane) insns are not + * predicated, but they are subject to beatwise execution if they are + * not in an IT block. + * + * Since our implementation always executes all 4 beats in one tick, + * this means only that if PSR.ECI says we should not be executing + * the beat corresponding to the lane of the vector register being + * accessed then we should skip performing the move, and that we need + * to do the usual check for bad ECI state and advance of ECI state. + * + * Note that if PSR.ECI is non-zero then we cannot be in an IT block. + * + * Return true if this VMOV scalar <-> gpreg should be skipped because + * the MVE PSR.ECI state says we skip the beat where the store happens. + */ + + /* Calculate the byte offset into Qn which we're going to access */ + int ofs = (index << size) + ((vn & 1) * 8); + + if (!dc_isar_feature(aa32_mve, s)) { + return false; + } + + switch (s->eci) { + case ECI_NONE: + return false; + case ECI_A0: + return ofs < 4; + case ECI_A0A1: + return ofs < 8; + case ECI_A0A1A2: + case ECI_A0A1A2B0: + return ofs < 12; + default: + g_assert_not_reached(); + } +} + static bool trans_VMOV_to_gp(DisasContext *s, arg_VMOV_to_gp *a) { /* VMOV scalar to general purpose register */ @@ -XXX,XX +XXX,XX @@ static bool trans_VMOV_to_gp(DisasContext *s, arg_VMOV_to_gp *a) return false; } + if (dc_isar_feature(aa32_mve, s)) { + if (!mve_eci_check(s)) { + return true; + } + } + if (!vfp_access_check(s)) { return true; } - tmp = tcg_temp_new_i32(); - read_neon_element32(tmp, a->vn, a->index, a->size | (a->u ? 0 : MO_SIGN)); - store_reg(s, a->rt, tmp); + if (!mve_skip_vmov(s, a->vn, a->index, a->size)) { + tmp = tcg_temp_new_i32(); + read_neon_element32(tmp, a->vn, a->index, + a->size | (a->u ? 0 : MO_SIGN)); + store_reg(s, a->rt, tmp); + } + if (dc_isar_feature(aa32_mve, s)) { + mve_update_and_store_eci(s); + } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_VMOV_from_gp(DisasContext *s, arg_VMOV_from_gp *a) return false; } + if (dc_isar_feature(aa32_mve, s)) { + if (!mve_eci_check(s)) { + return true; + } + } + if (!vfp_access_check(s)) { return true; } - tmp = load_reg(s, a->rt); - write_neon_element32(tmp, a->vn, a->index, a->size); - tcg_temp_free_i32(tmp); + if (!mve_skip_vmov(s, a->vn, a->index, a->size)) { + tmp = load_reg(s, a->rt); + write_neon_element32(tmp, a->vn, a->index, a->size); + tcg_temp_free_i32(tmp); + } + if (dc_isar_feature(aa32_mve, s)) { + mve_update_and_store_eci(s); + } return true; } -- 2.20.1 From: Peter Collingbourne <pcc@google.com> MTE3 introduces an asymmetric tag checking mode, in which loads are checked synchronously and stores are checked asynchronously. Add support for it. Signed-off-by: Peter Collingbourne <pcc@google.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20210616195614.11785-1-pcc@google.com [PMM: Add line to emulation.rst] Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- docs/system/arm/emulation.rst | 1 + target/arm/cpu64.c | 2 +- target/arm/mte_helper.c | 82 ++++++++++++++++++++++------------- 3 files changed, 53 insertions(+), 32 deletions(-) diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/arm/emulation.rst +++ b/docs/system/arm/emulation.rst @@ -XXX,XX +XXX,XX @@ the following architecture extensions: - FEAT_LSE (Large System Extensions) - FEAT_MTE (Memory Tagging Extension) - FEAT_MTE2 (Memory Tagging Extension) +- FEAT_MTE3 (MTE Asymmetric Fault Handling) - FEAT_PAN (Privileged access never) - FEAT_PAN2 (AT S1E1R and AT S1E1W instruction variants affected by PSTATE.PAN) - FEAT_PAuth (Pointer authentication) 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) * during realize if the board provides no tag memory, much like * we do for EL2 with the virtualization=on property. */ - t = FIELD_DP64(t, ID_AA64PFR1, MTE, 2); + t = FIELD_DP64(t, ID_AA64PFR1, MTE, 3); cpu->isar.id_aa64pfr1 = t; t = cpu->isar.id_aa64mmfr0; diff --git a/target/arm/mte_helper.c b/target/arm/mte_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/mte_helper.c +++ b/target/arm/mte_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(stzgm_tags)(CPUARMState *env, uint64_t ptr, uint64_t val) } } +static void mte_sync_check_fail(CPUARMState *env, uint32_t desc, + uint64_t dirty_ptr, uintptr_t ra) +{ + int is_write, syn; + + env->exception.vaddress = dirty_ptr; + + is_write = FIELD_EX32(desc, MTEDESC, WRITE); + syn = syn_data_abort_no_iss(arm_current_el(env) != 0, 0, 0, 0, 0, is_write, + 0x11); + raise_exception_ra(env, EXCP_DATA_ABORT, syn, exception_target_el(env), ra); + g_assert_not_reached(); +} + +static void mte_async_check_fail(CPUARMState *env, uint64_t dirty_ptr, + uintptr_t ra, ARMMMUIdx arm_mmu_idx, int el) +{ + int select; + + if (regime_has_2_ranges(arm_mmu_idx)) { + select = extract64(dirty_ptr, 55, 1); + } else { + select = 0; + } + env->cp15.tfsr_el[el] |= 1 << select; +#ifdef CONFIG_USER_ONLY + /* + * Stand in for a timer irq, setting _TIF_MTE_ASYNC_FAULT, + * which then sends a SIGSEGV when the thread is next scheduled. + * This cpu will return to the main loop at the end of the TB, + * which is rather sooner than "normal". But the alternative + * is waiting until the next syscall. + */ + qemu_cpu_kick(env_cpu(env)); +#endif +} + /* Record a tag check failure. */ static void mte_check_fail(CPUARMState *env, uint32_t desc, uint64_t dirty_ptr, uintptr_t ra) { int mmu_idx = FIELD_EX32(desc, MTEDESC, MIDX); ARMMMUIdx arm_mmu_idx = core_to_aa64_mmu_idx(mmu_idx); - int el, reg_el, tcf, select, is_write, syn; + int el, reg_el, tcf; uint64_t sctlr; reg_el = regime_el(env, arm_mmu_idx); @@ -XXX,XX +XXX,XX @@ static void mte_check_fail(CPUARMState *env, uint32_t desc, switch (tcf) { case 1: /* Tag check fail causes a synchronous exception. */ - env->exception.vaddress = dirty_ptr; - - is_write = FIELD_EX32(desc, MTEDESC, WRITE); - syn = syn_data_abort_no_iss(arm_current_el(env) != 0, 0, 0, 0, 0, - is_write, 0x11); - raise_exception_ra(env, EXCP_DATA_ABORT, syn, - exception_target_el(env), ra); - /* noreturn, but fall through to the assert anyway */ + mte_sync_check_fail(env, desc, dirty_ptr, ra); + break; case 0: /* @@ -XXX,XX +XXX,XX @@ static void mte_check_fail(CPUARMState *env, uint32_t desc, case 2: /* Tag check fail causes asynchronous flag set. */ - if (regime_has_2_ranges(arm_mmu_idx)) { - select = extract64(dirty_ptr, 55, 1); - } else { - select = 0; - } - env->cp15.tfsr_el[el] |= 1 << select; -#ifdef CONFIG_USER_ONLY - /* - * Stand in for a timer irq, setting _TIF_MTE_ASYNC_FAULT, - * which then sends a SIGSEGV when the thread is next scheduled. - * This cpu will return to the main loop at the end of the TB, - * which is rather sooner than "normal". But the alternative - * is waiting until the next syscall. - */ - qemu_cpu_kick(env_cpu(env)); -#endif + mte_async_check_fail(env, dirty_ptr, ra, arm_mmu_idx, el); break; - default: - /* Case 3: Reserved. */ - qemu_log_mask(LOG_GUEST_ERROR, - "Tag check failure with SCTLR_EL%d.TCF%s " - "set to reserved value %d\n", - reg_el, el ? "" : "0", tcf); + case 3: + /* + * Tag check fail causes asynchronous flag set for stores, or + * a synchronous exception for loads. + */ + if (FIELD_EX32(desc, MTEDESC, WRITE)) { + mte_async_check_fail(env, dirty_ptr, ra, arm_mmu_idx, el); + } else { + mte_sync_check_fail(env, desc, dirty_ptr, ra); + } break; } } -- 2.20.1 From: Alexandre Iooss <erdnaxe@crans.org> This adds the target guide for BBC Micro:bit. Information is taken from https://wiki.qemu.org/Features/MicroBit and from hw/arm/nrf51_soc.c. Signed-off-by: Alexandre Iooss <erdnaxe@crans.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Joel Stanley <joel@jms.id.au> Message-id: 20210621075625.540471-1-erdnaxe@crans.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- docs/system/arm/nrf.rst | 51 ++++++++++++++++++++++++++++++++++++++ docs/system/target-arm.rst | 1 + MAINTAINERS | 1 + 3 files changed, 53 insertions(+) create mode 100644 docs/system/arm/nrf.rst diff --git a/docs/system/arm/nrf.rst b/docs/system/arm/nrf.rst new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/docs/system/arm/nrf.rst @@ -XXX,XX +XXX,XX @@ +Nordic nRF boards (``microbit``) +================================ + +The `Nordic nRF`_ chips are a family of ARM-based System-on-Chip that +are designed to be used for low-power and short-range wireless solutions. + +.. _Nordic nRF: https://www.nordicsemi.com/Products + +The nRF51 series is the first series for short range wireless applications. +It is superseded by the nRF52 series. +The following machines are based on this chip : + +- ``microbit`` BBC micro:bit board with nRF51822 SoC + +There are other series such as nRF52, nRF53 and nRF91 which are currently not +supported by QEMU. + +Supported devices +----------------- + + * ARM Cortex-M0 (ARMv6-M) + * Serial ports (UART) + * Clock controller + * Timers + * Random Number Generator (RNG) + * GPIO controller + * NVMC + * SWI + +Missing devices +--------------- + + * Watchdog + * Real-Time Clock (RTC) controller + * TWI (i2c) + * SPI controller + * Analog to Digital Converter (ADC) + * Quadrature decoder + * Radio + +Boot options +------------ + +The Micro:bit machine can be started using the ``-device`` option to load a +firmware in `ihex format`_. Example: + +.. _ihex format: https://en.wikipedia.org/wiki/Intel_HEX + +.. code-block:: bash + + $ qemu-system-arm -M microbit -device loader,file=test.hex diff --git a/docs/system/target-arm.rst b/docs/system/target-arm.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/target-arm.rst +++ b/docs/system/target-arm.rst @@ -XXX,XX +XXX,XX @@ undocumented; you can get a complete list by running arm/digic arm/musicpal arm/gumstix + arm/nrf arm/nseries arm/nuvoton arm/orangepi diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: hw/*/microbit*.c F: include/hw/*/nrf51*.h F: include/hw/*/microbit*.h F: tests/qtest/microbit-test.c +F: docs/system/arm/nrf.rst AVR Machines ------------- -- 2.20.1 v3->v4: Windows headers define an INT type which clashed with an enum value name in arm_gicv3_its.c... The following changes since commit eae587e8e3694b1aceab23239493fb4c7e1a80f5: Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2021-09-13' into staging (2021-09-13 11:00:30 +0100) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20210913-3 for you to fetch changes up to 28e987a7e7edaa3ca7feeac65edca26145df8814: hw/arm/mps2.c: Mark internal-only I2C buses as 'full' (2021-09-13 21:01:08 +0100) ---------------------------------------------------------------- target-arm queue: * mark MPS2/MPS3 board-internal i2c buses as 'full' so that command line user-created devices are not plugged into them * Take an exception if PSTATE.IL is set * Support an emulated ITS in the virt board * Add support for kudo-bmc board * Probe for KVM_CAP_ARM_VM_IPA_SIZE when creating scratch VM * cadence_uart: Fix clock handling issues that prevented u-boot from running ---------------------------------------------------------------- Bin Meng (6): hw/misc: zynq_slcr: Correctly compute output clocks in the reset exit phase hw/char: cadence_uart: Disable transmit when input clock is disabled hw/char: cadence_uart: Move clock/reset check to uart_can_receive() hw/char: cadence_uart: Convert to memop_with_attrs() ops hw/char: cadence_uart: Ignore access when unclocked or in reset for uart_{read, write}() hw/char: cadence_uart: Log a guest error when device is unclocked or in reset Chris Rauer (1): hw/arm: Add support for kudo-bmc board. Marc Zyngier (1): hw/arm/virt: KVM: Probe for KVM_CAP_ARM_VM_IPA_SIZE when creating scratch VM Peter Maydell (5): target/arm: Take an exception if PSTATE.IL is set qdev: Support marking individual buses as 'full' hw/arm/mps2-tz.c: Add extra data parameter to MakeDevFn hw/arm/mps2-tz.c: Mark internal-only I2C buses as 'full' hw/arm/mps2.c: Mark internal-only I2C buses as 'full' Richard Henderson (1): target/arm: Merge disas_a64_insn into aarch64_tr_translate_insn Shashi Mallela (9): hw/intc: GICv3 ITS initial framework hw/intc: GICv3 ITS register definitions added hw/intc: GICv3 ITS command queue framework hw/intc: GICv3 ITS Command processing hw/intc: GICv3 ITS Feature enablement hw/intc: GICv3 redistributor ITS processing tests/data/acpi/virt: Add IORT files for ITS hw/arm/virt: add ITS support in virt GIC tests/data/acpi/virt: Update IORT files for ITS docs/system/arm/nuvoton.rst | 1 + hw/intc/gicv3_internal.h | 188 ++++- include/hw/arm/virt.h | 2 + include/hw/intc/arm_gicv3_common.h | 13 + include/hw/intc/arm_gicv3_its_common.h | 32 +- include/hw/qdev-core.h | 24 + target/arm/cpu.h | 1 + target/arm/kvm_arm.h | 4 +- target/arm/syndrome.h | 5 + target/arm/translate.h | 2 + hw/arm/mps2-tz.c | 92 ++- hw/arm/mps2.c | 12 +- hw/arm/npcm7xx_boards.c | 34 + hw/arm/virt.c | 29 +- hw/char/cadence_uart.c | 61 +- hw/intc/arm_gicv3.c | 14 + hw/intc/arm_gicv3_common.c | 13 + hw/intc/arm_gicv3_cpuif.c | 7 +- hw/intc/arm_gicv3_dist.c | 5 +- hw/intc/arm_gicv3_its.c | 1322 ++++++++++++++++++++++++++++++++ hw/intc/arm_gicv3_its_common.c | 7 +- hw/intc/arm_gicv3_its_kvm.c | 2 +- hw/intc/arm_gicv3_redist.c | 153 +++- hw/misc/zynq_slcr.c | 31 +- softmmu/qdev-monitor.c | 7 +- target/arm/helper-a64.c | 1 + target/arm/helper.c | 8 + target/arm/kvm.c | 7 +- target/arm/translate-a64.c | 255 +++--- target/arm/translate.c | 21 + hw/intc/meson.build | 1 + tests/data/acpi/virt/IORT | Bin 0 -> 124 bytes tests/data/acpi/virt/IORT.memhp | Bin 0 -> 124 bytes tests/data/acpi/virt/IORT.numamem | Bin 0 -> 124 bytes tests/data/acpi/virt/IORT.pxb | Bin 0 -> 124 bytes 35 files changed, 2144 insertions(+), 210 deletions(-) create mode 100644 hw/intc/arm_gicv3_its.c create mode 100644 tests/data/acpi/virt/IORT create mode 100644 tests/data/acpi/virt/IORT.memhp create mode 100644 tests/data/acpi/virt/IORT.numamem create mode 100644 tests/data/acpi/virt/IORT.pxb