Arm queue -- mostly the first slice of my Musca patches. thanks -- PMM The following changes since commit fc3dbb90f2eb069801bfb4cfe9cbc83cf9c5f4a9: Merge remote-tracking branch 'remotes/jnsnow/tags/bitmaps-pull-request' into staging (2019-02-21 13:09:33 +0000) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20190221 for you to fetch changes up to 3733f80308d2a7f23f5e39b039e0547aba6c07f1: hw/arm/armsse: Make 0x5... alias region work for per-CPU devices (2019-02-21 18:17:48 +0000) ---------------------------------------------------------------- target-arm queue: * Model the Arm "Musca" development boards: "musca-a" and "musca-b1" * Implement the ARMv8.3-JSConv extension * v8M MPU should use background region as default, not always * Stop unintentional sign extension in pmu_init ---------------------------------------------------------------- Aaron Lindsay OS (1): target/arm: Stop unintentional sign extension in pmu_init Peter Maydell (16): hw/arm/armsse: Fix memory leak in error-exit path target/arm: v8M MPU should use background region as default, not always hw/misc/tz-ppc: Support having unused ports in the middle of the range hw/timer/pl031: Allow use as an embedded-struct device hw/timer/pl031: Convert to using trace events hw/char/pl011: Allow use as an embedded-struct device hw/char/pl011: Support all interrupt lines hw/char/pl011: Use '0x' prefix when logging hex numbers hw/arm/armsse: Document SRAM_ADDR_WIDTH property in header comment hw/arm/armsse: Allow boards to specify init-svtor hw/arm/musca.c: Implement models of the Musca-A and -B1 boards hw/arm/musca: Add PPCs hw/arm/musca: Add MPCs hw/arm/musca: Wire up PL031 RTC hw/arm/musca: Wire up PL011 UARTs hw/arm/armsse: Make 0x5... alias region work for per-CPU devices Richard Henderson (4): target/arm: Restructure disas_fp_int_conv target/arm: Split out vfp_helper.c target/arm: Rearrange Floating-point data-processing (2 regs) target/arm: Implement ARMv8.3-JSConv hw/arm/Makefile.objs | 1 + target/arm/Makefile.objs | 2 +- include/hw/arm/armsse.h | 7 +- include/hw/char/pl011.h | 34 ++ include/hw/misc/tz-ppc.h | 8 +- include/hw/timer/pl031.h | 44 ++ target/arm/cpu.h | 10 + target/arm/helper.h | 3 + hw/arm/armsse.c | 44 +- hw/arm/musca.c | 669 ++++++++++++++++++++++ hw/char/pl011.c | 81 +-- hw/misc/tz-ppc.c | 32 ++ hw/timer/pl031.c | 80 ++- target/arm/cpu.c | 1 + target/arm/cpu64.c | 2 + target/arm/helper.c | 1072 +---------------------------------- target/arm/translate-a64.c | 120 ++-- target/arm/translate.c | 237 ++++---- target/arm/vfp_helper.c | 1176 +++++++++++++++++++++++++++++++++++++++ MAINTAINERS | 7 + default-configs/arm-softmmu.mak | 1 + hw/timer/trace-events | 6 + 22 files changed, 2307 insertions(+), 1330 deletions(-) create mode 100644 include/hw/timer/pl031.h create mode 100644 hw/arm/musca.c create mode 100644 target/arm/vfp_helper.c Coverity points out (CID 1398632, CID 1398650) that we leak a couple of allocated strings in the error-exit code path for setting up the MHUs in the ARMSSE. Fix this bug by moving the allocate-and-free of each string to be closer to the use, so we do the free before doing the error-exit check. Fixes: f8574705f62b38a ("hw/arm/armsse: Add unimplemented-device stubs for MHUs") Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Message-id: 20190215113707.24553-1-peter.maydell@linaro.org --- hw/arm/armsse.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/hw/arm/armsse.c b/hw/arm/armsse.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/armsse.c +++ b/hw/arm/armsse.c @@ -XXX,XX +XXX,XX @@ static void armsse_realize(DeviceState *dev, Error **errp) if (info->has_mhus) { for (i = 0; i < ARRAY_SIZE(s->mhu); i++) { - char *name = g_strdup_printf("MHU%d", i); - char *port = g_strdup_printf("port[%d]", i + 3); + char *name; + char *port; + name = g_strdup_printf("MHU%d", i); qdev_prop_set_string(DEVICE(&s->mhu[i]), "name", name); qdev_prop_set_uint64(DEVICE(&s->mhu[i]), "size", 0x1000); object_property_set_bool(OBJECT(&s->mhu[i]), true, "realized", &err); + g_free(name); if (err) { error_propagate(errp, err); return; } + port = g_strdup_printf("port[%d]", i + 3); mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(&s->mhu[i]), 0); object_property_set_link(OBJECT(&s->apb_ppc0), OBJECT(mr), port, &err); + g_free(port); if (err) { error_propagate(errp, err); return; } - g_free(name); - g_free(port); } } -- 2.20.1 The "background region" for a v8M MPU is a default which will be used (if enabled, and if the access is privileged) if the access does not match any specific MPU region. We were incorrectly using it always (by putting the condition at the wrong nesting level). This meant that we would always return the default background permissions rather than the correct permissions for a specific region, and also that we would not return the right information in response to a TT instruction. Move the check for the background region to the same place in the logic as the equivalent v8M MPUCheck() pseudocode puts it. This in turn means we must adjust the condition we use to detect matches in multiple regions to avoid false-positives. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20190214113408.10214-1-peter.maydell@linaro.org --- target/arm/helper.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ static bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address, hit = true; } else if (m_is_ppb_region(env, address)) { hit = true; - } else if (pmsav7_use_background_region(cpu, mmu_idx, is_user)) { - hit = true; } else { + if (pmsav7_use_background_region(cpu, mmu_idx, is_user)) { + hit = true; + } + for (n = (int)cpu->pmsav7_dregion - 1; n >= 0; n--) { /* region search */ /* Note that the base address is bits [31:5] from the register @@ -XXX,XX +XXX,XX @@ static bool pmsav8_mpu_lookup(CPUARMState *env, uint32_t address, *is_subpage = true; } - if (hit) { + if (matchregion != -1) { /* Multiple regions match -- always a failure (unlike * PMSAv7 where highest-numbered-region wins) */ -- 2.20.1 From: Aaron Lindsay OS <aaron@os.amperecomputing.com> This was introduced by commit bf8d09694ccc07487cd73d7562081fdaec3370c8 target/arm: Don't clear supported PMU events when initializing PMCEID1 and identified by Coverity (CID 1398645). Signed-off-by: Aaron Lindsay <aaron@os.amperecomputing.com> Reported-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20190219144621.450-1-aaron@os.amperecomputing.com Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ void pmu_init(ARMCPU *cpu) if (cnt->supported(&cpu->env)) { supported_event_map[cnt->number] = i; - uint64_t event_mask = 1 << (cnt->number & 0x1f); + uint64_t event_mask = 1ULL << (cnt->number & 0x1f); if (cnt->number & 0x20) { cpu->pmceid1 |= event_mask; } else { -- 2.20.1 From: Richard Henderson <richard.henderson@linaro.org> For opcodes 0-5, move some if conditions into the structure of a switch statement. For opcodes 6 & 7, decode everything at once with a second switch. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20190215192302.27855-2-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 94 ++++++++++++++++++++------------------ 1 file changed, 49 insertions(+), 45 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void disas_fp_int_conv(DisasContext *s, uint32_t insn) int type = extract32(insn, 22, 2); bool sbit = extract32(insn, 29, 1); bool sf = extract32(insn, 31, 1); + bool itof = false; if (sbit) { - unallocated_encoding(s); - return; + goto do_unallocated; } - if (opcode > 5) { - /* FMOV */ - bool itof = opcode & 1; - - if (rmode >= 2) { - unallocated_encoding(s); - return; - } - - switch (sf << 3 | type << 1 | rmode) { - case 0x0: /* 32 bit */ - case 0xa: /* 64 bit */ - case 0xd: /* 64 bit to top half of quad */ - break; - case 0x6: /* 16-bit float, 32-bit int */ - case 0xe: /* 16-bit float, 64-bit int */ - if (dc_isar_feature(aa64_fp16, s)) { - break; - } - /* fallthru */ - default: - /* all other sf/type/rmode combinations are invalid */ - unallocated_encoding(s); - return; - } - - if (!fp_access_check(s)) { - return; - } - handle_fmov(s, rd, rn, type, itof); - } else { - /* actual FP conversions */ - bool itof = extract32(opcode, 1, 1); - - if (rmode != 0 && opcode > 1) { - unallocated_encoding(s); - return; + switch (opcode) { + case 2: /* SCVTF */ + case 3: /* UCVTF */ + itof = true; + /* fallthru */ + case 4: /* FCVTAS */ + case 5: /* FCVTAU */ + if (rmode != 0) { + goto do_unallocated; } + /* fallthru */ + case 0: /* FCVT[NPMZ]S */ + case 1: /* FCVT[NPMZ]U */ switch (type) { case 0: /* float32 */ case 1: /* float64 */ break; case 3: /* float16 */ - if (dc_isar_feature(aa64_fp16, s)) { - break; + if (!dc_isar_feature(aa64_fp16, s)) { + goto do_unallocated; } - /* fallthru */ + break; default: - unallocated_encoding(s); - return; + goto do_unallocated; } - if (!fp_access_check(s)) { return; } handle_fpfpcvt(s, rd, rn, opcode, itof, rmode, 64, sf, type); + break; + + default: + switch (sf << 7 | type << 5 | rmode << 3 | opcode) { + case 0b01100110: /* FMOV half <-> 32-bit int */ + case 0b01100111: + case 0b11100110: /* FMOV half <-> 64-bit int */ + case 0b11100111: + if (!dc_isar_feature(aa64_fp16, s)) { + goto do_unallocated; + } + /* fallthru */ + case 0b00000110: /* FMOV 32-bit */ + case 0b00000111: + case 0b10100110: /* FMOV 64-bit */ + case 0b10100111: + case 0b11001110: /* FMOV top half of 128-bit */ + case 0b11001111: + if (!fp_access_check(s)) { + return; + } + itof = opcode & 1; + handle_fmov(s, rd, rn, type, itof); + break; + + default: + do_unallocated: + unallocated_encoding(s); + return; + } + break; } } -- 2.20.1 From: Richard Henderson <richard.henderson@linaro.org> Move all of the fp helpers out of helper.c into a new file. This is code movement only. Since helper.c has no copyright header, take the one from cpu.h for the new file. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20190215192302.27855-3-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/Makefile.objs | 2 +- target/arm/helper.c | 1062 ------------------------------------- target/arm/vfp_helper.c | 1088 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 1089 insertions(+), 1063 deletions(-) create mode 100644 target/arm/vfp_helper.c diff --git a/target/arm/Makefile.objs b/target/arm/Makefile.objs index XXXXXXX..XXXXXXX 100644 --- a/target/arm/Makefile.objs +++ b/target/arm/Makefile.objs @@ -XXX,XX +XXX,XX @@ obj-$(call land,$(CONFIG_KVM),$(call lnot,$(TARGET_AARCH64))) += kvm32.o obj-$(call land,$(CONFIG_KVM),$(TARGET_AARCH64)) += kvm64.o obj-$(call lnot,$(CONFIG_KVM)) += kvm-stub.o obj-y += translate.o op_helper.o helper.o cpu.o -obj-y += neon_helper.o iwmmxt_helper.o vec_helper.o +obj-y += neon_helper.o iwmmxt_helper.o vec_helper.o vfp_helper.o obj-y += gdbstub.o obj-$(TARGET_AARCH64) += cpu64.o translate-a64.o helper-a64.o gdbstub64.o obj-$(TARGET_AARCH64) += pauth_helper.o diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(sel_flags)(uint32_t flags, uint32_t a, uint32_t b) return (a & mask) | (b & ~mask); } -/* VFP support. We follow the convention used for VFP instructions: - Single precision routines have a "s" suffix, double precision a - "d" suffix. */ - -/* Convert host exception flags to vfp form. */ -static inline int vfp_exceptbits_from_host(int host_bits) -{ - int target_bits = 0; - - if (host_bits & float_flag_invalid) - target_bits |= 1; - if (host_bits & float_flag_divbyzero) - target_bits |= 2; - if (host_bits & float_flag_overflow) - target_bits |= 4; - if (host_bits & (float_flag_underflow | float_flag_output_denormal)) - target_bits |= 8; - if (host_bits & float_flag_inexact) - target_bits |= 0x10; - if (host_bits & float_flag_input_denormal) - target_bits |= 0x80; - return target_bits; -} - -uint32_t HELPER(vfp_get_fpscr)(CPUARMState *env) -{ - uint32_t i, fpscr; - - fpscr = env->vfp.xregs[ARM_VFP_FPSCR] - | (env->vfp.vec_len << 16) - | (env->vfp.vec_stride << 20); - - i = get_float_exception_flags(&env->vfp.fp_status); - i |= get_float_exception_flags(&env->vfp.standard_fp_status); - /* FZ16 does not generate an input denormal exception. */ - i |= (get_float_exception_flags(&env->vfp.fp_status_f16) - & ~float_flag_input_denormal); - fpscr |= vfp_exceptbits_from_host(i); - - i = env->vfp.qc[0] | env->vfp.qc[1] | env->vfp.qc[2] | env->vfp.qc[3]; - fpscr |= i ? FPCR_QC : 0; - - return fpscr; -} - -uint32_t vfp_get_fpscr(CPUARMState *env) -{ - return HELPER(vfp_get_fpscr)(env); -} - -/* Convert vfp exception flags to target form. */ -static inline int vfp_exceptbits_to_host(int target_bits) -{ - int host_bits = 0; - - if (target_bits & 1) - host_bits |= float_flag_invalid; - if (target_bits & 2) - host_bits |= float_flag_divbyzero; - if (target_bits & 4) - host_bits |= float_flag_overflow; - if (target_bits & 8) - host_bits |= float_flag_underflow; - if (target_bits & 0x10) - host_bits |= float_flag_inexact; - if (target_bits & 0x80) - host_bits |= float_flag_input_denormal; - return host_bits; -} - -void HELPER(vfp_set_fpscr)(CPUARMState *env, uint32_t val) -{ - int i; - uint32_t changed = env->vfp.xregs[ARM_VFP_FPSCR]; - - /* When ARMv8.2-FP16 is not supported, FZ16 is RES0. */ - if (!cpu_isar_feature(aa64_fp16, arm_env_get_cpu(env))) { - val &= ~FPCR_FZ16; - } - - /* - * We don't implement trapped exception handling, so the - * trap enable bits, IDE|IXE|UFE|OFE|DZE|IOE are all RAZ/WI (not RES0!) - * - * If we exclude the exception flags, IOC|DZC|OFC|UFC|IXC|IDC - * (which are stored in fp_status), and the other RES0 bits - * in between, then we clear all of the low 16 bits. - */ - env->vfp.xregs[ARM_VFP_FPSCR] = val & 0xf7c80000; - env->vfp.vec_len = (val >> 16) & 7; - env->vfp.vec_stride = (val >> 20) & 3; - - /* - * The bit we set within fpscr_q is arbitrary; the register as a - * whole being zero/non-zero is what counts. - */ - env->vfp.qc[0] = val & FPCR_QC; - env->vfp.qc[1] = 0; - env->vfp.qc[2] = 0; - env->vfp.qc[3] = 0; - - changed ^= val; - if (changed & (3 << 22)) { - i = (val >> 22) & 3; - switch (i) { - case FPROUNDING_TIEEVEN: - i = float_round_nearest_even; - break; - case FPROUNDING_POSINF: - i = float_round_up; - break; - case FPROUNDING_NEGINF: - i = float_round_down; - break; - case FPROUNDING_ZERO: - i = float_round_to_zero; - break; - } - set_float_rounding_mode(i, &env->vfp.fp_status); - set_float_rounding_mode(i, &env->vfp.fp_status_f16); - } - if (changed & FPCR_FZ16) { - bool ftz_enabled = val & FPCR_FZ16; - set_flush_to_zero(ftz_enabled, &env->vfp.fp_status_f16); - set_flush_inputs_to_zero(ftz_enabled, &env->vfp.fp_status_f16); - } - if (changed & FPCR_FZ) { - bool ftz_enabled = val & FPCR_FZ; - set_flush_to_zero(ftz_enabled, &env->vfp.fp_status); - set_flush_inputs_to_zero(ftz_enabled, &env->vfp.fp_status); - } - if (changed & FPCR_DN) { - bool dnan_enabled = val & FPCR_DN; - set_default_nan_mode(dnan_enabled, &env->vfp.fp_status); - set_default_nan_mode(dnan_enabled, &env->vfp.fp_status_f16); - } - - /* The exception flags are ORed together when we read fpscr so we - * only need to preserve the current state in one of our - * float_status values. - */ - i = vfp_exceptbits_to_host(val); - set_float_exception_flags(i, &env->vfp.fp_status); - set_float_exception_flags(0, &env->vfp.fp_status_f16); - set_float_exception_flags(0, &env->vfp.standard_fp_status); -} - -void vfp_set_fpscr(CPUARMState *env, uint32_t val) -{ - HELPER(vfp_set_fpscr)(env, val); -} - -#define VFP_HELPER(name, p) HELPER(glue(glue(vfp_,name),p)) - -#define VFP_BINOP(name) \ -float32 VFP_HELPER(name, s)(float32 a, float32 b, void *fpstp) \ -{ \ - float_status *fpst = fpstp; \ - return float32_ ## name(a, b, fpst); \ -} \ -float64 VFP_HELPER(name, d)(float64 a, float64 b, void *fpstp) \ -{ \ - float_status *fpst = fpstp; \ - return float64_ ## name(a, b, fpst); \ -} -VFP_BINOP(add) -VFP_BINOP(sub) -VFP_BINOP(mul) -VFP_BINOP(div) -VFP_BINOP(min) -VFP_BINOP(max) -VFP_BINOP(minnum) -VFP_BINOP(maxnum) -#undef VFP_BINOP - -float32 VFP_HELPER(neg, s)(float32 a) -{ - return float32_chs(a); -} - -float64 VFP_HELPER(neg, d)(float64 a) -{ - return float64_chs(a); -} - -float32 VFP_HELPER(abs, s)(float32 a) -{ - return float32_abs(a); -} - -float64 VFP_HELPER(abs, d)(float64 a) -{ - return float64_abs(a); -} - -float32 VFP_HELPER(sqrt, s)(float32 a, CPUARMState *env) -{ - return float32_sqrt(a, &env->vfp.fp_status); -} - -float64 VFP_HELPER(sqrt, d)(float64 a, CPUARMState *env) -{ - return float64_sqrt(a, &env->vfp.fp_status); -} - -static void softfloat_to_vfp_compare(CPUARMState *env, int cmp) -{ - uint32_t flags; - switch (cmp) { - case float_relation_equal: - flags = 0x6; - break; - case float_relation_less: - flags = 0x8; - break; - case float_relation_greater: - flags = 0x2; - break; - case float_relation_unordered: - flags = 0x3; - break; - default: - g_assert_not_reached(); - } - env->vfp.xregs[ARM_VFP_FPSCR] = - deposit32(env->vfp.xregs[ARM_VFP_FPSCR], 28, 4, flags); -} - -/* XXX: check quiet/signaling case */ -#define DO_VFP_cmp(p, type) \ -void VFP_HELPER(cmp, p)(type a, type b, CPUARMState *env) \ -{ \ - softfloat_to_vfp_compare(env, \ - type ## _compare_quiet(a, b, &env->vfp.fp_status)); \ -} \ -void VFP_HELPER(cmpe, p)(type a, type b, CPUARMState *env) \ -{ \ - softfloat_to_vfp_compare(env, \ - type ## _compare(a, b, &env->vfp.fp_status)); \ -} -DO_VFP_cmp(s, float32) -DO_VFP_cmp(d, float64) -#undef DO_VFP_cmp - -/* Integer to float and float to integer conversions */ - -#define CONV_ITOF(name, ftype, fsz, sign) \ -ftype HELPER(name)(uint32_t x, void *fpstp) \ -{ \ - float_status *fpst = fpstp; \ - return sign##int32_to_##float##fsz((sign##int32_t)x, fpst); \ -} - -#define CONV_FTOI(name, ftype, fsz, sign, round) \ -sign##int32_t HELPER(name)(ftype x, void *fpstp) \ -{ \ - float_status *fpst = fpstp; \ - if (float##fsz##_is_any_nan(x)) { \ - float_raise(float_flag_invalid, fpst); \ - return 0; \ - } \ - return float##fsz##_to_##sign##int32##round(x, fpst); \ -} - -#define FLOAT_CONVS(name, p, ftype, fsz, sign) \ - CONV_ITOF(vfp_##name##to##p, ftype, fsz, sign) \ - CONV_FTOI(vfp_to##name##p, ftype, fsz, sign, ) \ - CONV_FTOI(vfp_to##name##z##p, ftype, fsz, sign, _round_to_zero) - -FLOAT_CONVS(si, h, uint32_t, 16, ) -FLOAT_CONVS(si, s, float32, 32, ) -FLOAT_CONVS(si, d, float64, 64, ) -FLOAT_CONVS(ui, h, uint32_t, 16, u) -FLOAT_CONVS(ui, s, float32, 32, u) -FLOAT_CONVS(ui, d, float64, 64, u) - -#undef CONV_ITOF -#undef CONV_FTOI -#undef FLOAT_CONVS - -/* floating point conversion */ -float64 VFP_HELPER(fcvtd, s)(float32 x, CPUARMState *env) -{ - return float32_to_float64(x, &env->vfp.fp_status); -} - -float32 VFP_HELPER(fcvts, d)(float64 x, CPUARMState *env) -{ - return float64_to_float32(x, &env->vfp.fp_status); -} - -/* VFP3 fixed point conversion. */ -#define VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ -float##fsz HELPER(vfp_##name##to##p)(uint##isz##_t x, uint32_t shift, \ - void *fpstp) \ -{ return itype##_to_##float##fsz##_scalbn(x, -shift, fpstp); } - -#define VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, ROUND, suff) \ -uint##isz##_t HELPER(vfp_to##name##p##suff)(float##fsz x, uint32_t shift, \ - void *fpst) \ -{ \ - if (unlikely(float##fsz##_is_any_nan(x))) { \ - float_raise(float_flag_invalid, fpst); \ - return 0; \ - } \ - return float##fsz##_to_##itype##_scalbn(x, ROUND, shift, fpst); \ -} - -#define VFP_CONV_FIX(name, p, fsz, isz, itype) \ -VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ -VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ - float_round_to_zero, _round_to_zero) \ -VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ - get_float_rounding_mode(fpst), ) - -#define VFP_CONV_FIX_A64(name, p, fsz, isz, itype) \ -VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ -VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ - get_float_rounding_mode(fpst), ) - -VFP_CONV_FIX(sh, d, 64, 64, int16) -VFP_CONV_FIX(sl, d, 64, 64, int32) -VFP_CONV_FIX_A64(sq, d, 64, 64, int64) -VFP_CONV_FIX(uh, d, 64, 64, uint16) -VFP_CONV_FIX(ul, d, 64, 64, uint32) -VFP_CONV_FIX_A64(uq, d, 64, 64, uint64) -VFP_CONV_FIX(sh, s, 32, 32, int16) -VFP_CONV_FIX(sl, s, 32, 32, int32) -VFP_CONV_FIX_A64(sq, s, 32, 64, int64) -VFP_CONV_FIX(uh, s, 32, 32, uint16) -VFP_CONV_FIX(ul, s, 32, 32, uint32) -VFP_CONV_FIX_A64(uq, s, 32, 64, uint64) - -#undef VFP_CONV_FIX -#undef VFP_CONV_FIX_FLOAT -#undef VFP_CONV_FLOAT_FIX_ROUND -#undef VFP_CONV_FIX_A64 - -uint32_t HELPER(vfp_sltoh)(uint32_t x, uint32_t shift, void *fpst) -{ - return int32_to_float16_scalbn(x, -shift, fpst); -} - -uint32_t HELPER(vfp_ultoh)(uint32_t x, uint32_t shift, void *fpst) -{ - return uint32_to_float16_scalbn(x, -shift, fpst); -} - -uint32_t HELPER(vfp_sqtoh)(uint64_t x, uint32_t shift, void *fpst) -{ - return int64_to_float16_scalbn(x, -shift, fpst); -} - -uint32_t HELPER(vfp_uqtoh)(uint64_t x, uint32_t shift, void *fpst) -{ - return uint64_to_float16_scalbn(x, -shift, fpst); -} - -uint32_t HELPER(vfp_toshh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_int16_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -uint32_t HELPER(vfp_touhh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_uint16_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -uint32_t HELPER(vfp_toslh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_int32_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -uint32_t HELPER(vfp_toulh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_uint32_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -uint64_t HELPER(vfp_tosqh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_int64_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -uint64_t HELPER(vfp_touqh)(uint32_t x, uint32_t shift, void *fpst) -{ - if (unlikely(float16_is_any_nan(x))) { - float_raise(float_flag_invalid, fpst); - return 0; - } - return float16_to_uint64_scalbn(x, get_float_rounding_mode(fpst), - shift, fpst); -} - -/* Set the current fp rounding mode and return the old one. - * The argument is a softfloat float_round_ value. - */ -uint32_t HELPER(set_rmode)(uint32_t rmode, void *fpstp) -{ - float_status *fp_status = fpstp; - - uint32_t prev_rmode = get_float_rounding_mode(fp_status); - set_float_rounding_mode(rmode, fp_status); - - return prev_rmode; -} - -/* Set the current fp rounding mode in the standard fp status and return - * the old one. This is for NEON instructions that need to change the - * rounding mode but wish to use the standard FPSCR values for everything - * else. Always set the rounding mode back to the correct value after - * modifying it. - * The argument is a softfloat float_round_ value. - */ -uint32_t HELPER(set_neon_rmode)(uint32_t rmode, CPUARMState *env) -{ - float_status *fp_status = &env->vfp.standard_fp_status; - - uint32_t prev_rmode = get_float_rounding_mode(fp_status); - set_float_rounding_mode(rmode, fp_status); - - return prev_rmode; -} - -/* Half precision conversions. */ -float32 HELPER(vfp_fcvt_f16_to_f32)(uint32_t a, void *fpstp, uint32_t ahp_mode) -{ - /* Squash FZ16 to 0 for the duration of conversion. In this case, - * it would affect flushing input denormals. - */ - float_status *fpst = fpstp; - flag save = get_flush_inputs_to_zero(fpst); - set_flush_inputs_to_zero(false, fpst); - float32 r = float16_to_float32(a, !ahp_mode, fpst); - set_flush_inputs_to_zero(save, fpst); - return r; -} - -uint32_t HELPER(vfp_fcvt_f32_to_f16)(float32 a, void *fpstp, uint32_t ahp_mode) -{ - /* Squash FZ16 to 0 for the duration of conversion. In this case, - * it would affect flushing output denormals. - */ - float_status *fpst = fpstp; - flag save = get_flush_to_zero(fpst); - set_flush_to_zero(false, fpst); - float16 r = float32_to_float16(a, !ahp_mode, fpst); - set_flush_to_zero(save, fpst); - return r; -} - -float64 HELPER(vfp_fcvt_f16_to_f64)(uint32_t a, void *fpstp, uint32_t ahp_mode) -{ - /* Squash FZ16 to 0 for the duration of conversion. In this case, - * it would affect flushing input denormals. - */ - float_status *fpst = fpstp; - flag save = get_flush_inputs_to_zero(fpst); - set_flush_inputs_to_zero(false, fpst); - float64 r = float16_to_float64(a, !ahp_mode, fpst); - set_flush_inputs_to_zero(save, fpst); - return r; -} - -uint32_t HELPER(vfp_fcvt_f64_to_f16)(float64 a, void *fpstp, uint32_t ahp_mode) -{ - /* Squash FZ16 to 0 for the duration of conversion. In this case, - * it would affect flushing output denormals. - */ - float_status *fpst = fpstp; - flag save = get_flush_to_zero(fpst); - set_flush_to_zero(false, fpst); - float16 r = float64_to_float16(a, !ahp_mode, fpst); - set_flush_to_zero(save, fpst); - return r; -} - -#define float32_two make_float32(0x40000000) -#define float32_three make_float32(0x40400000) -#define float32_one_point_five make_float32(0x3fc00000) - -float32 HELPER(recps_f32)(float32 a, float32 b, CPUARMState *env) -{ - float_status *s = &env->vfp.standard_fp_status; - if ((float32_is_infinity(a) && float32_is_zero_or_denormal(b)) || - (float32_is_infinity(b) && float32_is_zero_or_denormal(a))) { - if (!(float32_is_zero(a) || float32_is_zero(b))) { - float_raise(float_flag_input_denormal, s); - } - return float32_two; - } - return float32_sub(float32_two, float32_mul(a, b, s), s); -} - -float32 HELPER(rsqrts_f32)(float32 a, float32 b, CPUARMState *env) -{ - float_status *s = &env->vfp.standard_fp_status; - float32 product; - if ((float32_is_infinity(a) && float32_is_zero_or_denormal(b)) || - (float32_is_infinity(b) && float32_is_zero_or_denormal(a))) { - if (!(float32_is_zero(a) || float32_is_zero(b))) { - float_raise(float_flag_input_denormal, s); - } - return float32_one_point_five; - } - product = float32_mul(a, b, s); - return float32_div(float32_sub(float32_three, product, s), float32_two, s); -} - -/* NEON helpers. */ - -/* Constants 256 and 512 are used in some helpers; we avoid relying on - * int->float conversions at run-time. */ -#define float64_256 make_float64(0x4070000000000000LL) -#define float64_512 make_float64(0x4080000000000000LL) -#define float16_maxnorm make_float16(0x7bff) -#define float32_maxnorm make_float32(0x7f7fffff) -#define float64_maxnorm make_float64(0x7fefffffffffffffLL) - -/* Reciprocal functions - * - * The algorithm that must be used to calculate the estimate - * is specified by the ARM ARM, see FPRecipEstimate()/RecipEstimate - */ - -/* See RecipEstimate() - * - * input is a 9 bit fixed point number - * input range 256 .. 511 for a number from 0.5 <= x < 1.0. - * result range 256 .. 511 for a number from 1.0 to 511/256. - */ - -static int recip_estimate(int input) -{ - int a, b, r; - assert(256 <= input && input < 512); - a = (input * 2) + 1; - b = (1 << 19) / a; - r = (b + 1) >> 1; - assert(256 <= r && r < 512); - return r; -} - -/* - * Common wrapper to call recip_estimate - * - * The parameters are exponent and 64 bit fraction (without implicit - * bit) where the binary point is nominally at bit 52. Returns a - * float64 which can then be rounded to the appropriate size by the - * callee. - */ - -static uint64_t call_recip_estimate(int *exp, int exp_off, uint64_t frac) -{ - uint32_t scaled, estimate; - uint64_t result_frac; - int result_exp; - - /* Handle sub-normals */ - if (*exp == 0) { - if (extract64(frac, 51, 1) == 0) { - *exp = -1; - frac <<= 2; - } else { - frac <<= 1; - } - } - - /* scaled = UInt('1':fraction<51:44>) */ - scaled = deposit32(1 << 8, 0, 8, extract64(frac, 44, 8)); - estimate = recip_estimate(scaled); - - result_exp = exp_off - *exp; - result_frac = deposit64(0, 44, 8, estimate); - if (result_exp == 0) { - result_frac = deposit64(result_frac >> 1, 51, 1, 1); - } else if (result_exp == -1) { - result_frac = deposit64(result_frac >> 2, 50, 2, 1); - result_exp = 0; - } - - *exp = result_exp; - - return result_frac; -} - -static bool round_to_inf(float_status *fpst, bool sign_bit) -{ - switch (fpst->float_rounding_mode) { - case float_round_nearest_even: /* Round to Nearest */ - return true; - case float_round_up: /* Round to +Inf */ - return !sign_bit; - case float_round_down: /* Round to -Inf */ - return sign_bit; - case float_round_to_zero: /* Round to Zero */ - return false; - } - - g_assert_not_reached(); -} - -uint32_t HELPER(recpe_f16)(uint32_t input, void *fpstp) -{ - float_status *fpst = fpstp; - float16 f16 = float16_squash_input_denormal(input, fpst); - uint32_t f16_val = float16_val(f16); - uint32_t f16_sign = float16_is_neg(f16); - int f16_exp = extract32(f16_val, 10, 5); - uint32_t f16_frac = extract32(f16_val, 0, 10); - uint64_t f64_frac; - - if (float16_is_any_nan(f16)) { - float16 nan = f16; - if (float16_is_signaling_nan(f16, fpst)) { - float_raise(float_flag_invalid, fpst); - nan = float16_silence_nan(f16, fpst); - } - if (fpst->default_nan_mode) { - nan = float16_default_nan(fpst); - } - return nan; - } else if (float16_is_infinity(f16)) { - return float16_set_sign(float16_zero, float16_is_neg(f16)); - } else if (float16_is_zero(f16)) { - float_raise(float_flag_divbyzero, fpst); - return float16_set_sign(float16_infinity, float16_is_neg(f16)); - } else if (float16_abs(f16) < (1 << 8)) { - /* Abs(value) < 2.0^-16 */ - float_raise(float_flag_overflow | float_flag_inexact, fpst); - if (round_to_inf(fpst, f16_sign)) { - return float16_set_sign(float16_infinity, f16_sign); - } else { - return float16_set_sign(float16_maxnorm, f16_sign); - } - } else if (f16_exp >= 29 && fpst->flush_to_zero) { - float_raise(float_flag_underflow, fpst); - return float16_set_sign(float16_zero, float16_is_neg(f16)); - } - - f64_frac = call_recip_estimate(&f16_exp, 29, - ((uint64_t) f16_frac) << (52 - 10)); - - /* result = sign : result_exp<4:0> : fraction<51:42> */ - f16_val = deposit32(0, 15, 1, f16_sign); - f16_val = deposit32(f16_val, 10, 5, f16_exp); - f16_val = deposit32(f16_val, 0, 10, extract64(f64_frac, 52 - 10, 10)); - return make_float16(f16_val); -} - -float32 HELPER(recpe_f32)(float32 input, void *fpstp) -{ - float_status *fpst = fpstp; - float32 f32 = float32_squash_input_denormal(input, fpst); - uint32_t f32_val = float32_val(f32); - bool f32_sign = float32_is_neg(f32); - int f32_exp = extract32(f32_val, 23, 8); - uint32_t f32_frac = extract32(f32_val, 0, 23); - uint64_t f64_frac; - - if (float32_is_any_nan(f32)) { - float32 nan = f32; - if (float32_is_signaling_nan(f32, fpst)) { - float_raise(float_flag_invalid, fpst); - nan = float32_silence_nan(f32, fpst); - } - if (fpst->default_nan_mode) { - nan = float32_default_nan(fpst); - } - return nan; - } else if (float32_is_infinity(f32)) { - return float32_set_sign(float32_zero, float32_is_neg(f32)); - } else if (float32_is_zero(f32)) { - float_raise(float_flag_divbyzero, fpst); - return float32_set_sign(float32_infinity, float32_is_neg(f32)); - } else if (float32_abs(f32) < (1ULL << 21)) { - /* Abs(value) < 2.0^-128 */ - float_raise(float_flag_overflow | float_flag_inexact, fpst); - if (round_to_inf(fpst, f32_sign)) { - return float32_set_sign(float32_infinity, f32_sign); - } else { - return float32_set_sign(float32_maxnorm, f32_sign); - } - } else if (f32_exp >= 253 && fpst->flush_to_zero) { - float_raise(float_flag_underflow, fpst); - return float32_set_sign(float32_zero, float32_is_neg(f32)); - } - - f64_frac = call_recip_estimate(&f32_exp, 253, - ((uint64_t) f32_frac) << (52 - 23)); - - /* result = sign : result_exp<7:0> : fraction<51:29> */ - f32_val = deposit32(0, 31, 1, f32_sign); - f32_val = deposit32(f32_val, 23, 8, f32_exp); - f32_val = deposit32(f32_val, 0, 23, extract64(f64_frac, 52 - 23, 23)); - return make_float32(f32_val); -} - -float64 HELPER(recpe_f64)(float64 input, void *fpstp) -{ - float_status *fpst = fpstp; - float64 f64 = float64_squash_input_denormal(input, fpst); - uint64_t f64_val = float64_val(f64); - bool f64_sign = float64_is_neg(f64); - int f64_exp = extract64(f64_val, 52, 11); - uint64_t f64_frac = extract64(f64_val, 0, 52); - - /* Deal with any special cases */ - if (float64_is_any_nan(f64)) { - float64 nan = f64; - if (float64_is_signaling_nan(f64, fpst)) { - float_raise(float_flag_invalid, fpst); - nan = float64_silence_nan(f64, fpst); - } - if (fpst->default_nan_mode) { - nan = float64_default_nan(fpst); - } - return nan; - } else if (float64_is_infinity(f64)) { - return float64_set_sign(float64_zero, float64_is_neg(f64)); - } else if (float64_is_zero(f64)) { - float_raise(float_flag_divbyzero, fpst); - return float64_set_sign(float64_infinity, float64_is_neg(f64)); - } else if ((f64_val & ~(1ULL << 63)) < (1ULL << 50)) { - /* Abs(value) < 2.0^-1024 */ - float_raise(float_flag_overflow | float_flag_inexact, fpst); - if (round_to_inf(fpst, f64_sign)) { - return float64_set_sign(float64_infinity, f64_sign); - } else { - return float64_set_sign(float64_maxnorm, f64_sign); - } - } else if (f64_exp >= 2045 && fpst->flush_to_zero) { - float_raise(float_flag_underflow, fpst); - return float64_set_sign(float64_zero, float64_is_neg(f64)); - } - - f64_frac = call_recip_estimate(&f64_exp, 2045, f64_frac); - - /* result = sign : result_exp<10:0> : fraction<51:0>; */ - f64_val = deposit64(0, 63, 1, f64_sign); - f64_val = deposit64(f64_val, 52, 11, f64_exp); - f64_val = deposit64(f64_val, 0, 52, f64_frac); - return make_float64(f64_val); -} - -/* The algorithm that must be used to calculate the estimate - * is specified by the ARM ARM. - */ - -static int do_recip_sqrt_estimate(int a) -{ - int b, estimate; - - assert(128 <= a && a < 512); - if (a < 256) { - a = a * 2 + 1; - } else { - a = (a >> 1) << 1; - a = (a + 1) * 2; - } - b = 512; - while (a * (b + 1) * (b + 1) < (1 << 28)) { - b += 1; - } - estimate = (b + 1) / 2; - assert(256 <= estimate && estimate < 512); - - return estimate; -} - - -static uint64_t recip_sqrt_estimate(int *exp , int exp_off, uint64_t frac) -{ - int estimate; - uint32_t scaled; - - if (*exp == 0) { - while (extract64(frac, 51, 1) == 0) { - frac = frac << 1; - *exp -= 1; - } - frac = extract64(frac, 0, 51) << 1; - } - - if (*exp & 1) { - /* scaled = UInt('01':fraction<51:45>) */ - scaled = deposit32(1 << 7, 0, 7, extract64(frac, 45, 7)); - } else { - /* scaled = UInt('1':fraction<51:44>) */ - scaled = deposit32(1 << 8, 0, 8, extract64(frac, 44, 8)); - } - estimate = do_recip_sqrt_estimate(scaled); - - *exp = (exp_off - *exp) / 2; - return extract64(estimate, 0, 8) << 44; -} - -uint32_t HELPER(rsqrte_f16)(uint32_t input, void *fpstp) -{ - float_status *s = fpstp; - float16 f16 = float16_squash_input_denormal(input, s); - uint16_t val = float16_val(f16); - bool f16_sign = float16_is_neg(f16); - int f16_exp = extract32(val, 10, 5); - uint16_t f16_frac = extract32(val, 0, 10); - uint64_t f64_frac; - - if (float16_is_any_nan(f16)) { - float16 nan = f16; - if (float16_is_signaling_nan(f16, s)) { - float_raise(float_flag_invalid, s); - nan = float16_silence_nan(f16, s); - } - if (s->default_nan_mode) { - nan = float16_default_nan(s); - } - return nan; - } else if (float16_is_zero(f16)) { - float_raise(float_flag_divbyzero, s); - return float16_set_sign(float16_infinity, f16_sign); - } else if (f16_sign) { - float_raise(float_flag_invalid, s); - return float16_default_nan(s); - } else if (float16_is_infinity(f16)) { - return float16_zero; - } - - /* Scale and normalize to a double-precision value between 0.25 and 1.0, - * preserving the parity of the exponent. */ - - f64_frac = ((uint64_t) f16_frac) << (52 - 10); - - f64_frac = recip_sqrt_estimate(&f16_exp, 44, f64_frac); - - /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(2) */ - val = deposit32(0, 15, 1, f16_sign); - val = deposit32(val, 10, 5, f16_exp); - val = deposit32(val, 2, 8, extract64(f64_frac, 52 - 8, 8)); - return make_float16(val); -} - -float32 HELPER(rsqrte_f32)(float32 input, void *fpstp) -{ - float_status *s = fpstp; - float32 f32 = float32_squash_input_denormal(input, s); - uint32_t val = float32_val(f32); - uint32_t f32_sign = float32_is_neg(f32); - int f32_exp = extract32(val, 23, 8); - uint32_t f32_frac = extract32(val, 0, 23); - uint64_t f64_frac; - - if (float32_is_any_nan(f32)) { - float32 nan = f32; - if (float32_is_signaling_nan(f32, s)) { - float_raise(float_flag_invalid, s); - nan = float32_silence_nan(f32, s); - } - if (s->default_nan_mode) { - nan = float32_default_nan(s); - } - return nan; - } else if (float32_is_zero(f32)) { - float_raise(float_flag_divbyzero, s); - return float32_set_sign(float32_infinity, float32_is_neg(f32)); - } else if (float32_is_neg(f32)) { - float_raise(float_flag_invalid, s); - return float32_default_nan(s); - } else if (float32_is_infinity(f32)) { - return float32_zero; - } - - /* Scale and normalize to a double-precision value between 0.25 and 1.0, - * preserving the parity of the exponent. */ - - f64_frac = ((uint64_t) f32_frac) << 29; - - f64_frac = recip_sqrt_estimate(&f32_exp, 380, f64_frac); - - /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(15) */ - val = deposit32(0, 31, 1, f32_sign); - val = deposit32(val, 23, 8, f32_exp); - val = deposit32(val, 15, 8, extract64(f64_frac, 52 - 8, 8)); - return make_float32(val); -} - -float64 HELPER(rsqrte_f64)(float64 input, void *fpstp) -{ - float_status *s = fpstp; - float64 f64 = float64_squash_input_denormal(input, s); - uint64_t val = float64_val(f64); - bool f64_sign = float64_is_neg(f64); - int f64_exp = extract64(val, 52, 11); - uint64_t f64_frac = extract64(val, 0, 52); - - if (float64_is_any_nan(f64)) { - float64 nan = f64; - if (float64_is_signaling_nan(f64, s)) { - float_raise(float_flag_invalid, s); - nan = float64_silence_nan(f64, s); - } - if (s->default_nan_mode) { - nan = float64_default_nan(s); - } - return nan; - } else if (float64_is_zero(f64)) { - float_raise(float_flag_divbyzero, s); - return float64_set_sign(float64_infinity, float64_is_neg(f64)); - } else if (float64_is_neg(f64)) { - float_raise(float_flag_invalid, s); - return float64_default_nan(s); - } else if (float64_is_infinity(f64)) { - return float64_zero; - } - - f64_frac = recip_sqrt_estimate(&f64_exp, 3068, f64_frac); - - /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(44) */ - val = deposit64(0, 61, 1, f64_sign); - val = deposit64(val, 52, 11, f64_exp); - val = deposit64(val, 44, 8, extract64(f64_frac, 52 - 8, 8)); - return make_float64(val); -} - -uint32_t HELPER(recpe_u32)(uint32_t a, void *fpstp) -{ - /* float_status *s = fpstp; */ - int input, estimate; - - if ((a & 0x80000000) == 0) { - return 0xffffffff; - } - - input = extract32(a, 23, 9); - estimate = recip_estimate(input); - - return deposit32(0, (32 - 9), 9, estimate); -} - -uint32_t HELPER(rsqrte_u32)(uint32_t a, void *fpstp) -{ - int estimate; - - if ((a & 0xc0000000) == 0) { - return 0xffffffff; - } - - estimate = do_recip_sqrt_estimate(extract32(a, 23, 9)); - - return deposit32(0, 23, 9, estimate); -} - -/* VFPv4 fused multiply-accumulate */ -float32 VFP_HELPER(muladd, s)(float32 a, float32 b, float32 c, void *fpstp) -{ - float_status *fpst = fpstp; - return float32_muladd(a, b, c, 0, fpst); -} - -float64 VFP_HELPER(muladd, d)(float64 a, float64 b, float64 c, void *fpstp) -{ - float_status *fpst = fpstp; - return float64_muladd(a, b, c, 0, fpst); -} - -/* ARMv8 round to integral */ -float32 HELPER(rints_exact)(float32 x, void *fp_status) -{ - return float32_round_to_int(x, fp_status); -} - -float64 HELPER(rintd_exact)(float64 x, void *fp_status) -{ - return float64_round_to_int(x, fp_status); -} - -float32 HELPER(rints)(float32 x, void *fp_status) -{ - int old_flags = get_float_exception_flags(fp_status), new_flags; - float32 ret; - - ret = float32_round_to_int(x, fp_status); - - /* Suppress any inexact exceptions the conversion produced */ - if (!(old_flags & float_flag_inexact)) { - new_flags = get_float_exception_flags(fp_status); - set_float_exception_flags(new_flags & ~float_flag_inexact, fp_status); - } - - return ret; -} - -float64 HELPER(rintd)(float64 x, void *fp_status) -{ - int old_flags = get_float_exception_flags(fp_status), new_flags; - float64 ret; - - ret = float64_round_to_int(x, fp_status); - - new_flags = get_float_exception_flags(fp_status); - - /* Suppress any inexact exceptions the conversion produced */ - if (!(old_flags & float_flag_inexact)) { - new_flags = get_float_exception_flags(fp_status); - set_float_exception_flags(new_flags & ~float_flag_inexact, fp_status); - } - - return ret; -} - -/* Convert ARM rounding mode to softfloat */ -int arm_rmode_to_sf(int rmode) -{ - switch (rmode) { - case FPROUNDING_TIEAWAY: - rmode = float_round_ties_away; - break; - case FPROUNDING_ODD: - /* FIXME: add support for TIEAWAY and ODD */ - qemu_log_mask(LOG_UNIMP, "arm: unimplemented rounding mode: %d\n", - rmode); - /* fall through for now */ - case FPROUNDING_TIEEVEN: - default: - rmode = float_round_nearest_even; - break; - case FPROUNDING_POSINF: - rmode = float_round_up; - break; - case FPROUNDING_NEGINF: - rmode = float_round_down; - break; - case FPROUNDING_ZERO: - rmode = float_round_to_zero; - break; - } - return rmode; -} - /* CRC helpers. * The upper bytes of val (above the number specified by 'bytes') must have * been zeroed out by the caller. diff --git a/target/arm/vfp_helper.c b/target/arm/vfp_helper.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/vfp_helper.c @@ -XXX,XX +XXX,XX @@ +/* + * ARM VFP floating-point operations + * + * Copyright (c) 2003 Fabrice Bellard + * + * 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 "qemu/log.h" +#include "cpu.h" +#include "exec/helper-proto.h" +#include "fpu/softfloat.h" +#include "internals.h" + + +/* VFP support. We follow the convention used for VFP instructions: + Single precision routines have a "s" suffix, double precision a + "d" suffix. */ + +/* Convert host exception flags to vfp form. */ +static inline int vfp_exceptbits_from_host(int host_bits) +{ + int target_bits = 0; + + if (host_bits & float_flag_invalid) + target_bits |= 1; + if (host_bits & float_flag_divbyzero) + target_bits |= 2; + if (host_bits & float_flag_overflow) + target_bits |= 4; + if (host_bits & (float_flag_underflow | float_flag_output_denormal)) + target_bits |= 8; + if (host_bits & float_flag_inexact) + target_bits |= 0x10; + if (host_bits & float_flag_input_denormal) + target_bits |= 0x80; + return target_bits; +} + +uint32_t HELPER(vfp_get_fpscr)(CPUARMState *env) +{ + uint32_t i, fpscr; + + fpscr = env->vfp.xregs[ARM_VFP_FPSCR] + | (env->vfp.vec_len << 16) + | (env->vfp.vec_stride << 20); + + i = get_float_exception_flags(&env->vfp.fp_status); + i |= get_float_exception_flags(&env->vfp.standard_fp_status); + /* FZ16 does not generate an input denormal exception. */ + i |= (get_float_exception_flags(&env->vfp.fp_status_f16) + & ~float_flag_input_denormal); + fpscr |= vfp_exceptbits_from_host(i); + + i = env->vfp.qc[0] | env->vfp.qc[1] | env->vfp.qc[2] | env->vfp.qc[3]; + fpscr |= i ? FPCR_QC : 0; + + return fpscr; +} + +uint32_t vfp_get_fpscr(CPUARMState *env) +{ + return HELPER(vfp_get_fpscr)(env); +} + +/* Convert vfp exception flags to target form. */ +static inline int vfp_exceptbits_to_host(int target_bits) +{ + int host_bits = 0; + + if (target_bits & 1) + host_bits |= float_flag_invalid; + if (target_bits & 2) + host_bits |= float_flag_divbyzero; + if (target_bits & 4) + host_bits |= float_flag_overflow; + if (target_bits & 8) + host_bits |= float_flag_underflow; + if (target_bits & 0x10) + host_bits |= float_flag_inexact; + if (target_bits & 0x80) + host_bits |= float_flag_input_denormal; + return host_bits; +} + +void HELPER(vfp_set_fpscr)(CPUARMState *env, uint32_t val) +{ + int i; + uint32_t changed = env->vfp.xregs[ARM_VFP_FPSCR]; + + /* When ARMv8.2-FP16 is not supported, FZ16 is RES0. */ + if (!cpu_isar_feature(aa64_fp16, arm_env_get_cpu(env))) { + val &= ~FPCR_FZ16; + } + + /* + * We don't implement trapped exception handling, so the + * trap enable bits, IDE|IXE|UFE|OFE|DZE|IOE are all RAZ/WI (not RES0!) + * + * If we exclude the exception flags, IOC|DZC|OFC|UFC|IXC|IDC + * (which are stored in fp_status), and the other RES0 bits + * in between, then we clear all of the low 16 bits. + */ + env->vfp.xregs[ARM_VFP_FPSCR] = val & 0xf7c80000; + env->vfp.vec_len = (val >> 16) & 7; + env->vfp.vec_stride = (val >> 20) & 3; + + /* + * The bit we set within fpscr_q is arbitrary; the register as a + * whole being zero/non-zero is what counts. + */ + env->vfp.qc[0] = val & FPCR_QC; + env->vfp.qc[1] = 0; + env->vfp.qc[2] = 0; + env->vfp.qc[3] = 0; + + changed ^= val; + if (changed & (3 << 22)) { + i = (val >> 22) & 3; + switch (i) { + case FPROUNDING_TIEEVEN: + i = float_round_nearest_even; + break; + case FPROUNDING_POSINF: + i = float_round_up; + break; + case FPROUNDING_NEGINF: + i = float_round_down; + break; + case FPROUNDING_ZERO: + i = float_round_to_zero; + break; + } + set_float_rounding_mode(i, &env->vfp.fp_status); + set_float_rounding_mode(i, &env->vfp.fp_status_f16); + } + if (changed & FPCR_FZ16) { + bool ftz_enabled = val & FPCR_FZ16; + set_flush_to_zero(ftz_enabled, &env->vfp.fp_status_f16); + set_flush_inputs_to_zero(ftz_enabled, &env->vfp.fp_status_f16); + } + if (changed & FPCR_FZ) { + bool ftz_enabled = val & FPCR_FZ; + set_flush_to_zero(ftz_enabled, &env->vfp.fp_status); + set_flush_inputs_to_zero(ftz_enabled, &env->vfp.fp_status); + } + if (changed & FPCR_DN) { + bool dnan_enabled = val & FPCR_DN; + set_default_nan_mode(dnan_enabled, &env->vfp.fp_status); + set_default_nan_mode(dnan_enabled, &env->vfp.fp_status_f16); + } + + /* The exception flags are ORed together when we read fpscr so we + * only need to preserve the current state in one of our + * float_status values. + */ + i = vfp_exceptbits_to_host(val); + set_float_exception_flags(i, &env->vfp.fp_status); + set_float_exception_flags(0, &env->vfp.fp_status_f16); + set_float_exception_flags(0, &env->vfp.standard_fp_status); +} + +void vfp_set_fpscr(CPUARMState *env, uint32_t val) +{ + HELPER(vfp_set_fpscr)(env, val); +} + +#define VFP_HELPER(name, p) HELPER(glue(glue(vfp_,name),p)) + +#define VFP_BINOP(name) \ +float32 VFP_HELPER(name, s)(float32 a, float32 b, void *fpstp) \ +{ \ + float_status *fpst = fpstp; \ + return float32_ ## name(a, b, fpst); \ +} \ +float64 VFP_HELPER(name, d)(float64 a, float64 b, void *fpstp) \ +{ \ + float_status *fpst = fpstp; \ + return float64_ ## name(a, b, fpst); \ +} +VFP_BINOP(add) +VFP_BINOP(sub) +VFP_BINOP(mul) +VFP_BINOP(div) +VFP_BINOP(min) +VFP_BINOP(max) +VFP_BINOP(minnum) +VFP_BINOP(maxnum) +#undef VFP_BINOP + +float32 VFP_HELPER(neg, s)(float32 a) +{ + return float32_chs(a); +} + +float64 VFP_HELPER(neg, d)(float64 a) +{ + return float64_chs(a); +} + +float32 VFP_HELPER(abs, s)(float32 a) +{ + return float32_abs(a); +} + +float64 VFP_HELPER(abs, d)(float64 a) +{ + return float64_abs(a); +} + +float32 VFP_HELPER(sqrt, s)(float32 a, CPUARMState *env) +{ + return float32_sqrt(a, &env->vfp.fp_status); +} + +float64 VFP_HELPER(sqrt, d)(float64 a, CPUARMState *env) +{ + return float64_sqrt(a, &env->vfp.fp_status); +} + +static void softfloat_to_vfp_compare(CPUARMState *env, int cmp) +{ + uint32_t flags; + switch (cmp) { + case float_relation_equal: + flags = 0x6; + break; + case float_relation_less: + flags = 0x8; + break; + case float_relation_greater: + flags = 0x2; + break; + case float_relation_unordered: + flags = 0x3; + break; + default: + g_assert_not_reached(); + } + env->vfp.xregs[ARM_VFP_FPSCR] = + deposit32(env->vfp.xregs[ARM_VFP_FPSCR], 28, 4, flags); +} + +/* XXX: check quiet/signaling case */ +#define DO_VFP_cmp(p, type) \ +void VFP_HELPER(cmp, p)(type a, type b, CPUARMState *env) \ +{ \ + softfloat_to_vfp_compare(env, \ + type ## _compare_quiet(a, b, &env->vfp.fp_status)); \ +} \ +void VFP_HELPER(cmpe, p)(type a, type b, CPUARMState *env) \ +{ \ + softfloat_to_vfp_compare(env, \ + type ## _compare(a, b, &env->vfp.fp_status)); \ +} +DO_VFP_cmp(s, float32) +DO_VFP_cmp(d, float64) +#undef DO_VFP_cmp + +/* Integer to float and float to integer conversions */ + +#define CONV_ITOF(name, ftype, fsz, sign) \ +ftype HELPER(name)(uint32_t x, void *fpstp) \ +{ \ + float_status *fpst = fpstp; \ + return sign##int32_to_##float##fsz((sign##int32_t)x, fpst); \ +} + +#define CONV_FTOI(name, ftype, fsz, sign, round) \ +sign##int32_t HELPER(name)(ftype x, void *fpstp) \ +{ \ + float_status *fpst = fpstp; \ + if (float##fsz##_is_any_nan(x)) { \ + float_raise(float_flag_invalid, fpst); \ + return 0; \ + } \ + return float##fsz##_to_##sign##int32##round(x, fpst); \ +} + +#define FLOAT_CONVS(name, p, ftype, fsz, sign) \ + CONV_ITOF(vfp_##name##to##p, ftype, fsz, sign) \ + CONV_FTOI(vfp_to##name##p, ftype, fsz, sign, ) \ + CONV_FTOI(vfp_to##name##z##p, ftype, fsz, sign, _round_to_zero) + +FLOAT_CONVS(si, h, uint32_t, 16, ) +FLOAT_CONVS(si, s, float32, 32, ) +FLOAT_CONVS(si, d, float64, 64, ) +FLOAT_CONVS(ui, h, uint32_t, 16, u) +FLOAT_CONVS(ui, s, float32, 32, u) +FLOAT_CONVS(ui, d, float64, 64, u) + +#undef CONV_ITOF +#undef CONV_FTOI +#undef FLOAT_CONVS + +/* floating point conversion */ +float64 VFP_HELPER(fcvtd, s)(float32 x, CPUARMState *env) +{ + return float32_to_float64(x, &env->vfp.fp_status); +} + +float32 VFP_HELPER(fcvts, d)(float64 x, CPUARMState *env) +{ + return float64_to_float32(x, &env->vfp.fp_status); +} + +/* VFP3 fixed point conversion. */ +#define VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ +float##fsz HELPER(vfp_##name##to##p)(uint##isz##_t x, uint32_t shift, \ + void *fpstp) \ +{ return itype##_to_##float##fsz##_scalbn(x, -shift, fpstp); } + +#define VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, ROUND, suff) \ +uint##isz##_t HELPER(vfp_to##name##p##suff)(float##fsz x, uint32_t shift, \ + void *fpst) \ +{ \ + if (unlikely(float##fsz##_is_any_nan(x))) { \ + float_raise(float_flag_invalid, fpst); \ + return 0; \ + } \ + return float##fsz##_to_##itype##_scalbn(x, ROUND, shift, fpst); \ +} + +#define VFP_CONV_FIX(name, p, fsz, isz, itype) \ +VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ +VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ + float_round_to_zero, _round_to_zero) \ +VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ + get_float_rounding_mode(fpst), ) + +#define VFP_CONV_FIX_A64(name, p, fsz, isz, itype) \ +VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \ +VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, \ + get_float_rounding_mode(fpst), ) + +VFP_CONV_FIX(sh, d, 64, 64, int16) +VFP_CONV_FIX(sl, d, 64, 64, int32) +VFP_CONV_FIX_A64(sq, d, 64, 64, int64) +VFP_CONV_FIX(uh, d, 64, 64, uint16) +VFP_CONV_FIX(ul, d, 64, 64, uint32) +VFP_CONV_FIX_A64(uq, d, 64, 64, uint64) +VFP_CONV_FIX(sh, s, 32, 32, int16) +VFP_CONV_FIX(sl, s, 32, 32, int32) +VFP_CONV_FIX_A64(sq, s, 32, 64, int64) +VFP_CONV_FIX(uh, s, 32, 32, uint16) +VFP_CONV_FIX(ul, s, 32, 32, uint32) +VFP_CONV_FIX_A64(uq, s, 32, 64, uint64) + +#undef VFP_CONV_FIX +#undef VFP_CONV_FIX_FLOAT +#undef VFP_CONV_FLOAT_FIX_ROUND +#undef VFP_CONV_FIX_A64 + +uint32_t HELPER(vfp_sltoh)(uint32_t x, uint32_t shift, void *fpst) +{ + return int32_to_float16_scalbn(x, -shift, fpst); +} + +uint32_t HELPER(vfp_ultoh)(uint32_t x, uint32_t shift, void *fpst) +{ + return uint32_to_float16_scalbn(x, -shift, fpst); +} + +uint32_t HELPER(vfp_sqtoh)(uint64_t x, uint32_t shift, void *fpst) +{ + return int64_to_float16_scalbn(x, -shift, fpst); +} + +uint32_t HELPER(vfp_uqtoh)(uint64_t x, uint32_t shift, void *fpst) +{ + return uint64_to_float16_scalbn(x, -shift, fpst); +} + +uint32_t HELPER(vfp_toshh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_int16_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +uint32_t HELPER(vfp_touhh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_uint16_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +uint32_t HELPER(vfp_toslh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_int32_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +uint32_t HELPER(vfp_toulh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_uint32_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +uint64_t HELPER(vfp_tosqh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_int64_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +uint64_t HELPER(vfp_touqh)(uint32_t x, uint32_t shift, void *fpst) +{ + if (unlikely(float16_is_any_nan(x))) { + float_raise(float_flag_invalid, fpst); + return 0; + } + return float16_to_uint64_scalbn(x, get_float_rounding_mode(fpst), + shift, fpst); +} + +/* Set the current fp rounding mode and return the old one. + * The argument is a softfloat float_round_ value. + */ +uint32_t HELPER(set_rmode)(uint32_t rmode, void *fpstp) +{ + float_status *fp_status = fpstp; + + uint32_t prev_rmode = get_float_rounding_mode(fp_status); + set_float_rounding_mode(rmode, fp_status); + + return prev_rmode; +} + +/* Set the current fp rounding mode in the standard fp status and return + * the old one. This is for NEON instructions that need to change the + * rounding mode but wish to use the standard FPSCR values for everything + * else. Always set the rounding mode back to the correct value after + * modifying it. + * The argument is a softfloat float_round_ value. + */ +uint32_t HELPER(set_neon_rmode)(uint32_t rmode, CPUARMState *env) +{ + float_status *fp_status = &env->vfp.standard_fp_status; + + uint32_t prev_rmode = get_float_rounding_mode(fp_status); + set_float_rounding_mode(rmode, fp_status); + + return prev_rmode; +} + +/* Half precision conversions. */ +float32 HELPER(vfp_fcvt_f16_to_f32)(uint32_t a, void *fpstp, uint32_t ahp_mode) +{ + /* Squash FZ16 to 0 for the duration of conversion. In this case, + * it would affect flushing input denormals. + */ + float_status *fpst = fpstp; + flag save = get_flush_inputs_to_zero(fpst); + set_flush_inputs_to_zero(false, fpst); + float32 r = float16_to_float32(a, !ahp_mode, fpst); + set_flush_inputs_to_zero(save, fpst); + return r; +} + +uint32_t HELPER(vfp_fcvt_f32_to_f16)(float32 a, void *fpstp, uint32_t ahp_mode) +{ + /* Squash FZ16 to 0 for the duration of conversion. In this case, + * it would affect flushing output denormals. + */ + float_status *fpst = fpstp; + flag save = get_flush_to_zero(fpst); + set_flush_to_zero(false, fpst); + float16 r = float32_to_float16(a, !ahp_mode, fpst); + set_flush_to_zero(save, fpst); + return r; +} + +float64 HELPER(vfp_fcvt_f16_to_f64)(uint32_t a, void *fpstp, uint32_t ahp_mode) +{ + /* Squash FZ16 to 0 for the duration of conversion. In this case, + * it would affect flushing input denormals. + */ + float_status *fpst = fpstp; + flag save = get_flush_inputs_to_zero(fpst); + set_flush_inputs_to_zero(false, fpst); + float64 r = float16_to_float64(a, !ahp_mode, fpst); + set_flush_inputs_to_zero(save, fpst); + return r; +} + +uint32_t HELPER(vfp_fcvt_f64_to_f16)(float64 a, void *fpstp, uint32_t ahp_mode) +{ + /* Squash FZ16 to 0 for the duration of conversion. In this case, + * it would affect flushing output denormals. + */ + float_status *fpst = fpstp; + flag save = get_flush_to_zero(fpst); + set_flush_to_zero(false, fpst); + float16 r = float64_to_float16(a, !ahp_mode, fpst); + set_flush_to_zero(save, fpst); + return r; +} + +#define float32_two make_float32(0x40000000) +#define float32_three make_float32(0x40400000) +#define float32_one_point_five make_float32(0x3fc00000) + +float32 HELPER(recps_f32)(float32 a, float32 b, CPUARMState *env) +{ + float_status *s = &env->vfp.standard_fp_status; + if ((float32_is_infinity(a) && float32_is_zero_or_denormal(b)) || + (float32_is_infinity(b) && float32_is_zero_or_denormal(a))) { + if (!(float32_is_zero(a) || float32_is_zero(b))) { + float_raise(float_flag_input_denormal, s); + } + return float32_two; + } + return float32_sub(float32_two, float32_mul(a, b, s), s); +} + +float32 HELPER(rsqrts_f32)(float32 a, float32 b, CPUARMState *env) +{ + float_status *s = &env->vfp.standard_fp_status; + float32 product; + if ((float32_is_infinity(a) && float32_is_zero_or_denormal(b)) || + (float32_is_infinity(b) && float32_is_zero_or_denormal(a))) { + if (!(float32_is_zero(a) || float32_is_zero(b))) { + float_raise(float_flag_input_denormal, s); + } + return float32_one_point_five; + } + product = float32_mul(a, b, s); + return float32_div(float32_sub(float32_three, product, s), float32_two, s); +} + +/* NEON helpers. */ + +/* Constants 256 and 512 are used in some helpers; we avoid relying on + * int->float conversions at run-time. */ +#define float64_256 make_float64(0x4070000000000000LL) +#define float64_512 make_float64(0x4080000000000000LL) +#define float16_maxnorm make_float16(0x7bff) +#define float32_maxnorm make_float32(0x7f7fffff) +#define float64_maxnorm make_float64(0x7fefffffffffffffLL) + +/* Reciprocal functions + * + * The algorithm that must be used to calculate the estimate + * is specified by the ARM ARM, see FPRecipEstimate()/RecipEstimate + */ + +/* See RecipEstimate() + * + * input is a 9 bit fixed point number + * input range 256 .. 511 for a number from 0.5 <= x < 1.0. + * result range 256 .. 511 for a number from 1.0 to 511/256. + */ + +static int recip_estimate(int input) +{ + int a, b, r; + assert(256 <= input && input < 512); + a = (input * 2) + 1; + b = (1 << 19) / a; + r = (b + 1) >> 1; + assert(256 <= r && r < 512); + return r; +} + +/* + * Common wrapper to call recip_estimate + * + * The parameters are exponent and 64 bit fraction (without implicit + * bit) where the binary point is nominally at bit 52. Returns a + * float64 which can then be rounded to the appropriate size by the + * callee. + */ + +static uint64_t call_recip_estimate(int *exp, int exp_off, uint64_t frac) +{ + uint32_t scaled, estimate; + uint64_t result_frac; + int result_exp; + + /* Handle sub-normals */ + if (*exp == 0) { + if (extract64(frac, 51, 1) == 0) { + *exp = -1; + frac <<= 2; + } else { + frac <<= 1; + } + } + + /* scaled = UInt('1':fraction<51:44>) */ + scaled = deposit32(1 << 8, 0, 8, extract64(frac, 44, 8)); + estimate = recip_estimate(scaled); + + result_exp = exp_off - *exp; + result_frac = deposit64(0, 44, 8, estimate); + if (result_exp == 0) { + result_frac = deposit64(result_frac >> 1, 51, 1, 1); + } else if (result_exp == -1) { + result_frac = deposit64(result_frac >> 2, 50, 2, 1); + result_exp = 0; + } + + *exp = result_exp; + + return result_frac; +} + +static bool round_to_inf(float_status *fpst, bool sign_bit) +{ + switch (fpst->float_rounding_mode) { + case float_round_nearest_even: /* Round to Nearest */ + return true; + case float_round_up: /* Round to +Inf */ + return !sign_bit; + case float_round_down: /* Round to -Inf */ + return sign_bit; + case float_round_to_zero: /* Round to Zero */ + return false; + } + + g_assert_not_reached(); +} + +uint32_t HELPER(recpe_f16)(uint32_t input, void *fpstp) +{ + float_status *fpst = fpstp; + float16 f16 = float16_squash_input_denormal(input, fpst); + uint32_t f16_val = float16_val(f16); + uint32_t f16_sign = float16_is_neg(f16); + int f16_exp = extract32(f16_val, 10, 5); + uint32_t f16_frac = extract32(f16_val, 0, 10); + uint64_t f64_frac; + + if (float16_is_any_nan(f16)) { + float16 nan = f16; + if (float16_is_signaling_nan(f16, fpst)) { + float_raise(float_flag_invalid, fpst); + nan = float16_silence_nan(f16, fpst); + } + if (fpst->default_nan_mode) { + nan = float16_default_nan(fpst); + } + return nan; + } else if (float16_is_infinity(f16)) { + return float16_set_sign(float16_zero, float16_is_neg(f16)); + } else if (float16_is_zero(f16)) { + float_raise(float_flag_divbyzero, fpst); + return float16_set_sign(float16_infinity, float16_is_neg(f16)); + } else if (float16_abs(f16) < (1 << 8)) { + /* Abs(value) < 2.0^-16 */ + float_raise(float_flag_overflow | float_flag_inexact, fpst); + if (round_to_inf(fpst, f16_sign)) { + return float16_set_sign(float16_infinity, f16_sign); + } else { + return float16_set_sign(float16_maxnorm, f16_sign); + } + } else if (f16_exp >= 29 && fpst->flush_to_zero) { + float_raise(float_flag_underflow, fpst); + return float16_set_sign(float16_zero, float16_is_neg(f16)); + } + + f64_frac = call_recip_estimate(&f16_exp, 29, + ((uint64_t) f16_frac) << (52 - 10)); + + /* result = sign : result_exp<4:0> : fraction<51:42> */ + f16_val = deposit32(0, 15, 1, f16_sign); + f16_val = deposit32(f16_val, 10, 5, f16_exp); + f16_val = deposit32(f16_val, 0, 10, extract64(f64_frac, 52 - 10, 10)); + return make_float16(f16_val); +} + +float32 HELPER(recpe_f32)(float32 input, void *fpstp) +{ + float_status *fpst = fpstp; + float32 f32 = float32_squash_input_denormal(input, fpst); + uint32_t f32_val = float32_val(f32); + bool f32_sign = float32_is_neg(f32); + int f32_exp = extract32(f32_val, 23, 8); + uint32_t f32_frac = extract32(f32_val, 0, 23); + uint64_t f64_frac; + + if (float32_is_any_nan(f32)) { + float32 nan = f32; + if (float32_is_signaling_nan(f32, fpst)) { + float_raise(float_flag_invalid, fpst); + nan = float32_silence_nan(f32, fpst); + } + if (fpst->default_nan_mode) { + nan = float32_default_nan(fpst); + } + return nan; + } else if (float32_is_infinity(f32)) { + return float32_set_sign(float32_zero, float32_is_neg(f32)); + } else if (float32_is_zero(f32)) { + float_raise(float_flag_divbyzero, fpst); + return float32_set_sign(float32_infinity, float32_is_neg(f32)); + } else if (float32_abs(f32) < (1ULL << 21)) { + /* Abs(value) < 2.0^-128 */ + float_raise(float_flag_overflow | float_flag_inexact, fpst); + if (round_to_inf(fpst, f32_sign)) { + return float32_set_sign(float32_infinity, f32_sign); + } else { + return float32_set_sign(float32_maxnorm, f32_sign); + } + } else if (f32_exp >= 253 && fpst->flush_to_zero) { + float_raise(float_flag_underflow, fpst); + return float32_set_sign(float32_zero, float32_is_neg(f32)); + } + + f64_frac = call_recip_estimate(&f32_exp, 253, + ((uint64_t) f32_frac) << (52 - 23)); + + /* result = sign : result_exp<7:0> : fraction<51:29> */ + f32_val = deposit32(0, 31, 1, f32_sign); + f32_val = deposit32(f32_val, 23, 8, f32_exp); + f32_val = deposit32(f32_val, 0, 23, extract64(f64_frac, 52 - 23, 23)); + return make_float32(f32_val); +} + +float64 HELPER(recpe_f64)(float64 input, void *fpstp) +{ + float_status *fpst = fpstp; + float64 f64 = float64_squash_input_denormal(input, fpst); + uint64_t f64_val = float64_val(f64); + bool f64_sign = float64_is_neg(f64); + int f64_exp = extract64(f64_val, 52, 11); + uint64_t f64_frac = extract64(f64_val, 0, 52); + + /* Deal with any special cases */ + if (float64_is_any_nan(f64)) { + float64 nan = f64; + if (float64_is_signaling_nan(f64, fpst)) { + float_raise(float_flag_invalid, fpst); + nan = float64_silence_nan(f64, fpst); + } + if (fpst->default_nan_mode) { + nan = float64_default_nan(fpst); + } + return nan; + } else if (float64_is_infinity(f64)) { + return float64_set_sign(float64_zero, float64_is_neg(f64)); + } else if (float64_is_zero(f64)) { + float_raise(float_flag_divbyzero, fpst); + return float64_set_sign(float64_infinity, float64_is_neg(f64)); + } else if ((f64_val & ~(1ULL << 63)) < (1ULL << 50)) { + /* Abs(value) < 2.0^-1024 */ + float_raise(float_flag_overflow | float_flag_inexact, fpst); + if (round_to_inf(fpst, f64_sign)) { + return float64_set_sign(float64_infinity, f64_sign); + } else { + return float64_set_sign(float64_maxnorm, f64_sign); + } + } else if (f64_exp >= 2045 && fpst->flush_to_zero) { + float_raise(float_flag_underflow, fpst); + return float64_set_sign(float64_zero, float64_is_neg(f64)); + } + + f64_frac = call_recip_estimate(&f64_exp, 2045, f64_frac); + + /* result = sign : result_exp<10:0> : fraction<51:0>; */ + f64_val = deposit64(0, 63, 1, f64_sign); + f64_val = deposit64(f64_val, 52, 11, f64_exp); + f64_val = deposit64(f64_val, 0, 52, f64_frac); + return make_float64(f64_val); +} + +/* The algorithm that must be used to calculate the estimate + * is specified by the ARM ARM. + */ + +static int do_recip_sqrt_estimate(int a) +{ + int b, estimate; + + assert(128 <= a && a < 512); + if (a < 256) { + a = a * 2 + 1; + } else { + a = (a >> 1) << 1; + a = (a + 1) * 2; + } + b = 512; + while (a * (b + 1) * (b + 1) < (1 << 28)) { + b += 1; + } + estimate = (b + 1) / 2; + assert(256 <= estimate && estimate < 512); + + return estimate; +} + + +static uint64_t recip_sqrt_estimate(int *exp , int exp_off, uint64_t frac) +{ + int estimate; + uint32_t scaled; + + if (*exp == 0) { + while (extract64(frac, 51, 1) == 0) { + frac = frac << 1; + *exp -= 1; + } + frac = extract64(frac, 0, 51) << 1; + } + + if (*exp & 1) { + /* scaled = UInt('01':fraction<51:45>) */ + scaled = deposit32(1 << 7, 0, 7, extract64(frac, 45, 7)); + } else { + /* scaled = UInt('1':fraction<51:44>) */ + scaled = deposit32(1 << 8, 0, 8, extract64(frac, 44, 8)); + } + estimate = do_recip_sqrt_estimate(scaled); + + *exp = (exp_off - *exp) / 2; + return extract64(estimate, 0, 8) << 44; +} + +uint32_t HELPER(rsqrte_f16)(uint32_t input, void *fpstp) +{ + float_status *s = fpstp; + float16 f16 = float16_squash_input_denormal(input, s); + uint16_t val = float16_val(f16); + bool f16_sign = float16_is_neg(f16); + int f16_exp = extract32(val, 10, 5); + uint16_t f16_frac = extract32(val, 0, 10); + uint64_t f64_frac; + + if (float16_is_any_nan(f16)) { + float16 nan = f16; + if (float16_is_signaling_nan(f16, s)) { + float_raise(float_flag_invalid, s); + nan = float16_silence_nan(f16, s); + } + if (s->default_nan_mode) { + nan = float16_default_nan(s); + } + return nan; + } else if (float16_is_zero(f16)) { + float_raise(float_flag_divbyzero, s); + return float16_set_sign(float16_infinity, f16_sign); + } else if (f16_sign) { + float_raise(float_flag_invalid, s); + return float16_default_nan(s); + } else if (float16_is_infinity(f16)) { + return float16_zero; + } + + /* Scale and normalize to a double-precision value between 0.25 and 1.0, + * preserving the parity of the exponent. */ + + f64_frac = ((uint64_t) f16_frac) << (52 - 10); + + f64_frac = recip_sqrt_estimate(&f16_exp, 44, f64_frac); + + /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(2) */ + val = deposit32(0, 15, 1, f16_sign); + val = deposit32(val, 10, 5, f16_exp); + val = deposit32(val, 2, 8, extract64(f64_frac, 52 - 8, 8)); + return make_float16(val); +} + +float32 HELPER(rsqrte_f32)(float32 input, void *fpstp) +{ + float_status *s = fpstp; + float32 f32 = float32_squash_input_denormal(input, s); + uint32_t val = float32_val(f32); + uint32_t f32_sign = float32_is_neg(f32); + int f32_exp = extract32(val, 23, 8); + uint32_t f32_frac = extract32(val, 0, 23); + uint64_t f64_frac; + + if (float32_is_any_nan(f32)) { + float32 nan = f32; + if (float32_is_signaling_nan(f32, s)) { + float_raise(float_flag_invalid, s); + nan = float32_silence_nan(f32, s); + } + if (s->default_nan_mode) { + nan = float32_default_nan(s); + } + return nan; + } else if (float32_is_zero(f32)) { + float_raise(float_flag_divbyzero, s); + return float32_set_sign(float32_infinity, float32_is_neg(f32)); + } else if (float32_is_neg(f32)) { + float_raise(float_flag_invalid, s); + return float32_default_nan(s); + } else if (float32_is_infinity(f32)) { + return float32_zero; + } + + /* Scale and normalize to a double-precision value between 0.25 and 1.0, + * preserving the parity of the exponent. */ + + f64_frac = ((uint64_t) f32_frac) << 29; + + f64_frac = recip_sqrt_estimate(&f32_exp, 380, f64_frac); + + /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(15) */ + val = deposit32(0, 31, 1, f32_sign); + val = deposit32(val, 23, 8, f32_exp); + val = deposit32(val, 15, 8, extract64(f64_frac, 52 - 8, 8)); + return make_float32(val); +} + +float64 HELPER(rsqrte_f64)(float64 input, void *fpstp) +{ + float_status *s = fpstp; + float64 f64 = float64_squash_input_denormal(input, s); + uint64_t val = float64_val(f64); + bool f64_sign = float64_is_neg(f64); + int f64_exp = extract64(val, 52, 11); + uint64_t f64_frac = extract64(val, 0, 52); + + if (float64_is_any_nan(f64)) { + float64 nan = f64; + if (float64_is_signaling_nan(f64, s)) { + float_raise(float_flag_invalid, s); + nan = float64_silence_nan(f64, s); + } + if (s->default_nan_mode) { + nan = float64_default_nan(s); + } + return nan; + } else if (float64_is_zero(f64)) { + float_raise(float_flag_divbyzero, s); + return float64_set_sign(float64_infinity, float64_is_neg(f64)); + } else if (float64_is_neg(f64)) { + float_raise(float_flag_invalid, s); + return float64_default_nan(s); + } else if (float64_is_infinity(f64)) { + return float64_zero; + } + + f64_frac = recip_sqrt_estimate(&f64_exp, 3068, f64_frac); + + /* result = sign : result_exp<4:0> : estimate<7:0> : Zeros(44) */ + val = deposit64(0, 61, 1, f64_sign); + val = deposit64(val, 52, 11, f64_exp); + val = deposit64(val, 44, 8, extract64(f64_frac, 52 - 8, 8)); + return make_float64(val); +} + +uint32_t HELPER(recpe_u32)(uint32_t a, void *fpstp) +{ + /* float_status *s = fpstp; */ + int input, estimate; + + if ((a & 0x80000000) == 0) { + return 0xffffffff; + } + + input = extract32(a, 23, 9); + estimate = recip_estimate(input); + + return deposit32(0, (32 - 9), 9, estimate); +} + +uint32_t HELPER(rsqrte_u32)(uint32_t a, void *fpstp) +{ + int estimate; + + if ((a & 0xc0000000) == 0) { + return 0xffffffff; + } + + estimate = do_recip_sqrt_estimate(extract32(a, 23, 9)); + + return deposit32(0, 23, 9, estimate); +} + +/* VFPv4 fused multiply-accumulate */ +float32 VFP_HELPER(muladd, s)(float32 a, float32 b, float32 c, void *fpstp) +{ + float_status *fpst = fpstp; + return float32_muladd(a, b, c, 0, fpst); +} + +float64 VFP_HELPER(muladd, d)(float64 a, float64 b, float64 c, void *fpstp) +{ + float_status *fpst = fpstp; + return float64_muladd(a, b, c, 0, fpst); +} + +/* ARMv8 round to integral */ +float32 HELPER(rints_exact)(float32 x, void *fp_status) +{ + return float32_round_to_int(x, fp_status); +} + +float64 HELPER(rintd_exact)(float64 x, void *fp_status) +{ + return float64_round_to_int(x, fp_status); +} + +float32 HELPER(rints)(float32 x, void *fp_status) +{ + int old_flags = get_float_exception_flags(fp_status), new_flags; + float32 ret; + + ret = float32_round_to_int(x, fp_status); + + /* Suppress any inexact exceptions the conversion produced */ + if (!(old_flags & float_flag_inexact)) { + new_flags = get_float_exception_flags(fp_status); + set_float_exception_flags(new_flags & ~float_flag_inexact, fp_status); + } + + return ret; +} + +float64 HELPER(rintd)(float64 x, void *fp_status) +{ + int old_flags = get_float_exception_flags(fp_status), new_flags; + float64 ret; + + ret = float64_round_to_int(x, fp_status); + + new_flags = get_float_exception_flags(fp_status); + + /* Suppress any inexact exceptions the conversion produced */ + if (!(old_flags & float_flag_inexact)) { + new_flags = get_float_exception_flags(fp_status); + set_float_exception_flags(new_flags & ~float_flag_inexact, fp_status); + } + + return ret; +} + +/* Convert ARM rounding mode to softfloat */ +int arm_rmode_to_sf(int rmode) +{ + switch (rmode) { + case FPROUNDING_TIEAWAY: + rmode = float_round_ties_away; + break; + case FPROUNDING_ODD: + /* FIXME: add support for TIEAWAY and ODD */ + qemu_log_mask(LOG_UNIMP, "arm: unimplemented rounding mode: %d\n", + rmode); + /* fall through for now */ + case FPROUNDING_TIEEVEN: + default: + rmode = float_round_nearest_even; + break; + case FPROUNDING_POSINF: + rmode = float_round_up; + break; + case FPROUNDING_NEGINF: + rmode = float_round_down; + break; + case FPROUNDING_ZERO: + rmode = float_round_to_zero; + break; + } + return rmode; +} -- 2.20.1 From: Richard Henderson <richard.henderson@linaro.org> There are lots of special cases within these insns. Split the major argument decode/loading/saving into no_output (compares), rd_is_dp, and rm_is_dp. We still need to special case argument load for compare (rd as input, rm as zero) and vcvt fixed (rd as input+output), but lots of special cases do disappear. Now that we have a full switch at the beginning, hoist the ISA checks from the code generation. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20190215192302.27855-4-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.c | 227 ++++++++++++++++++++--------------------- 1 file changed, 111 insertions(+), 116 deletions(-) diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) } } else { /* data processing */ + bool rd_is_dp = dp; + bool rm_is_dp = dp; + bool no_output = false; + /* The opcode is in bits 23, 21, 20 and 6. */ op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1); - if (dp) { - if (op == 15) { - /* rn is opcode */ - rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1); - } else { - /* rn is register number */ - VFP_DREG_N(rn, insn); - } + rn = VFP_SREG_N(insn); - if (op == 15 && (rn == 15 || ((rn & 0x1c) == 0x18) || - ((rn & 0x1e) == 0x6))) { - /* Integer or single/half precision destination. */ - rd = VFP_SREG_D(insn); - } else { - VFP_DREG_D(rd, insn); - } - if (op == 15 && - (((rn & 0x1c) == 0x10) || ((rn & 0x14) == 0x14) || - ((rn & 0x1e) == 0x4))) { - /* VCVT from int or half precision is always from S reg - * regardless of dp bit. VCVT with immediate frac_bits - * has same format as SREG_M. + if (op == 15) { + /* rn is opcode, encoded as per VFP_SREG_N. */ + switch (rn) { + case 0x00: /* vmov */ + case 0x01: /* vabs */ + case 0x02: /* vneg */ + case 0x03: /* vsqrt */ + break; + + case 0x04: /* vcvtb.f64.f16, vcvtb.f32.f16 */ + case 0x05: /* vcvtt.f64.f16, vcvtt.f32.f16 */ + /* + * VCVTB, VCVTT: only present with the halfprec extension + * UNPREDICTABLE if bit 8 is set prior to ARMv8 + * (we choose to UNDEF) */ - rm = VFP_SREG_M(insn); - } else { - VFP_DREG_M(rm, insn); + if ((dp && !arm_dc_feature(s, ARM_FEATURE_V8)) || + !arm_dc_feature(s, ARM_FEATURE_VFP_FP16)) { + return 1; + } + rm_is_dp = false; + break; + case 0x06: /* vcvtb.f16.f32, vcvtb.f16.f64 */ + case 0x07: /* vcvtt.f16.f32, vcvtt.f16.f64 */ + if ((dp && !arm_dc_feature(s, ARM_FEATURE_V8)) || + !arm_dc_feature(s, ARM_FEATURE_VFP_FP16)) { + return 1; + } + rd_is_dp = false; + break; + + case 0x08: case 0x0a: /* vcmp, vcmpz */ + case 0x09: case 0x0b: /* vcmpe, vcmpez */ + no_output = true; + break; + + case 0x0c: /* vrintr */ + case 0x0d: /* vrintz */ + case 0x0e: /* vrintx */ + break; + + case 0x0f: /* vcvt double<->single */ + rd_is_dp = !dp; + break; + + case 0x10: /* vcvt.fxx.u32 */ + case 0x11: /* vcvt.fxx.s32 */ + rm_is_dp = false; + break; + case 0x18: /* vcvtr.u32.fxx */ + case 0x19: /* vcvtz.u32.fxx */ + case 0x1a: /* vcvtr.s32.fxx */ + case 0x1b: /* vcvtz.s32.fxx */ + rd_is_dp = false; + break; + + case 0x14: /* vcvt fp <-> fixed */ + case 0x15: + case 0x16: + case 0x17: + case 0x1c: + case 0x1d: + case 0x1e: + case 0x1f: + if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { + return 1; + } + /* Immediate frac_bits has same format as SREG_M. */ + rm_is_dp = false; + break; + + default: + return 1; } + } else if (dp) { + /* rn is register number */ + VFP_DREG_N(rn, insn); + } + + if (rd_is_dp) { + VFP_DREG_D(rd, insn); + } else { + rd = VFP_SREG_D(insn); + } + if (rm_is_dp) { + VFP_DREG_M(rm, insn); } else { - rn = VFP_SREG_N(insn); - if (op == 15 && rn == 15) { - /* Double precision destination. */ - VFP_DREG_D(rd, insn); - } else { - rd = VFP_SREG_D(insn); - } - /* NB that we implicitly rely on the encoding for the frac_bits - * in VCVT of fixed to float being the same as that of an SREG_M - */ rm = VFP_SREG_M(insn); } veclen = s->vec_len; - if (op == 15 && rn > 3) + if (op == 15 && rn > 3) { veclen = 0; + } /* Shut up compiler warnings. */ delta_m = 0; @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) /* Load the initial operands. */ if (op == 15) { switch (rn) { - case 16: - case 17: - /* Integer source */ - gen_mov_F0_vreg(0, rm); - break; - case 8: - case 9: - /* Compare */ + case 0x08: case 0x09: /* Compare */ gen_mov_F0_vreg(dp, rd); gen_mov_F1_vreg(dp, rm); break; - case 10: - case 11: - /* Compare with zero */ + case 0x0a: case 0x0b: /* Compare with zero */ gen_mov_F0_vreg(dp, rd); gen_vfp_F1_ld0(dp); break; - case 20: - case 21: - case 22: - case 23: - case 28: - case 29: - case 30: - case 31: + case 0x14: /* vcvt fp <-> fixed */ + case 0x15: + case 0x16: + case 0x17: + case 0x1c: + case 0x1d: + case 0x1e: + case 0x1f: /* Source and destination the same. */ gen_mov_F0_vreg(dp, rd); break; - case 4: - case 5: - case 6: - case 7: - /* VCVTB, VCVTT: only present with the halfprec extension - * UNPREDICTABLE if bit 8 is set prior to ARMv8 - * (we choose to UNDEF) - */ - if ((dp && !arm_dc_feature(s, ARM_FEATURE_V8)) || - !arm_dc_feature(s, ARM_FEATURE_VFP_FP16)) { - return 1; - } - if (!extract32(rn, 1, 1)) { - /* Half precision source. */ - gen_mov_F0_vreg(0, rm); - break; - } - /* Otherwise fall through */ default: /* One source operand. */ - gen_mov_F0_vreg(dp, rm); + gen_mov_F0_vreg(rm_is_dp, rm); break; } } else { @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) break; } case 15: /* single<->double conversion */ - if (dp) + if (dp) { gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env); - else + } else { gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env); + } break; case 16: /* fuito */ gen_vfp_uito(dp, 0); @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) gen_vfp_sito(dp, 0); break; case 20: /* fshto */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_shto(dp, 16 - rm, 0); break; case 21: /* fslto */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_slto(dp, 32 - rm, 0); break; case 22: /* fuhto */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_uhto(dp, 16 - rm, 0); break; case 23: /* fulto */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_ulto(dp, 32 - rm, 0); break; case 24: /* ftoui */ @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) gen_vfp_tosiz(dp, 0); break; case 28: /* ftosh */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_tosh(dp, 16 - rm, 0); break; case 29: /* ftosl */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_tosl(dp, 32 - rm, 0); break; case 30: /* ftouh */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_touh(dp, 16 - rm, 0); break; case 31: /* ftoul */ - if (!arm_dc_feature(s, ARM_FEATURE_VFP3)) { - return 1; - } gen_vfp_toul(dp, 32 - rm, 0); break; default: /* undefined */ - return 1; + g_assert_not_reached(); } break; default: /* undefined */ return 1; } - /* Write back the result. */ - if (op == 15 && (rn >= 8 && rn <= 11)) { - /* Comparison, do nothing. */ - } else if (op == 15 && dp && ((rn & 0x1c) == 0x18 || - (rn & 0x1e) == 0x6)) { - /* VCVT double to int: always integer result. - * VCVT double to half precision is always a single - * precision result. - */ - gen_mov_vreg_F0(0, rd); - } else if (op == 15 && rn == 15) { - /* conversion */ - gen_mov_vreg_F0(!dp, rd); - } else { - gen_mov_vreg_F0(dp, rd); + /* Write back the result, if any. */ + if (!no_output) { + gen_mov_vreg_F0(rd_is_dp, rd); } /* break out of the loop if we have finished */ - if (veclen == 0) + if (veclen == 0) { break; + } if (op == 15 && delta_m == 0) { /* single source one-many */ -- 2.20.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20190215192302.27855-5-richard.henderson@linaro.org Reviewed-by: Peter Maydell <peter.maydell@linaro.org> [PMM: fixed a couple of comment typos] Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 10 +++++ target/arm/helper.h | 3 ++ target/arm/cpu.c | 1 + target/arm/cpu64.c | 2 + target/arm/translate-a64.c | 26 +++++++++++ target/arm/translate.c | 10 +++++ target/arm/vfp_helper.c | 88 ++++++++++++++++++++++++++++++++++++++ 7 files changed, 140 insertions(+) diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_aa32_vcma(const ARMISARegisters *id) return FIELD_EX32(id->id_isar5, ID_ISAR5, VCMA) != 0; } +static inline bool isar_feature_aa32_jscvt(const ARMISARegisters *id) +{ + return FIELD_EX32(id->id_isar6, ID_ISAR6, JSCVT) != 0; +} + static inline bool isar_feature_aa32_dp(const ARMISARegisters *id) { return FIELD_EX32(id->id_isar6, ID_ISAR6, DP) != 0; @@ -XXX,XX +XXX,XX @@ static inline bool isar_feature_aa64_dp(const ARMISARegisters *id) return FIELD_EX64(id->id_aa64isar0, ID_AA64ISAR0, DP) != 0; } +static inline bool isar_feature_aa64_jscvt(const ARMISARegisters *id) +{ + return FIELD_EX64(id->id_aa64isar1, ID_AA64ISAR1, JSCVT) != 0; +} + static inline bool isar_feature_aa64_fcma(const ARMISARegisters *id) { return FIELD_EX64(id->id_aa64isar1, ID_AA64ISAR1, FCMA) != 0; 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_2(rintd_exact, TCG_CALL_NO_RWG, f64, f64, ptr) DEF_HELPER_FLAGS_2(rints, TCG_CALL_NO_RWG, f32, f32, ptr) DEF_HELPER_FLAGS_2(rintd, TCG_CALL_NO_RWG, f64, f64, ptr) +DEF_HELPER_FLAGS_2(vjcvt, TCG_CALL_NO_RWG, i32, f64, env) +DEF_HELPER_FLAGS_2(fjcvtzs, TCG_CALL_NO_RWG, i64, f64, ptr) + /* neon_helper.c */ DEF_HELPER_FLAGS_3(neon_qadd_u8, TCG_CALL_NO_RWG, i32, env, i32, i32) DEF_HELPER_FLAGS_3(neon_qadd_s8, TCG_CALL_NO_RWG, i32, env, i32, i32) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_max_initfn(Object *obj) cpu->isar.id_isar5 = t; t = cpu->isar.id_isar6; + t = FIELD_DP32(t, ID_ISAR6, JSCVT, 1); t = FIELD_DP32(t, ID_ISAR6, DP, 1); cpu->isar.id_isar6 = t; 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) cpu->isar.id_aa64isar0 = t; t = cpu->isar.id_aa64isar1; + t = FIELD_DP64(t, ID_AA64ISAR1, JSCVT, 1); t = FIELD_DP64(t, ID_AA64ISAR1, FCMA, 1); t = FIELD_DP64(t, ID_AA64ISAR1, APA, 1); /* PAuth, architected only */ t = FIELD_DP64(t, ID_AA64ISAR1, API, 0); @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) cpu->isar.id_isar5 = u; u = cpu->isar.id_isar6; + u = FIELD_DP32(u, ID_ISAR6, JSCVT, 1); u = FIELD_DP32(u, ID_ISAR6, DP, 1); cpu->isar.id_isar6 = u; diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void handle_fmov(DisasContext *s, int rd, int rn, int type, bool itof) } } +static void handle_fjcvtzs(DisasContext *s, int rd, int rn) +{ + TCGv_i64 t = read_fp_dreg(s, rn); + TCGv_ptr fpstatus = get_fpstatus_ptr(false); + + gen_helper_fjcvtzs(t, t, fpstatus); + + tcg_temp_free_ptr(fpstatus); + + tcg_gen_ext32u_i64(cpu_reg(s, rd), t); + tcg_gen_extrh_i64_i32(cpu_ZF, t); + tcg_gen_movi_i32(cpu_CF, 0); + tcg_gen_movi_i32(cpu_NF, 0); + tcg_gen_movi_i32(cpu_VF, 0); + + tcg_temp_free_i64(t); +} + /* Floating point <-> integer conversions * 31 30 29 28 24 23 22 21 20 19 18 16 15 10 9 5 4 0 * +----+---+---+-----------+------+---+-------+-----+-------------+----+----+ @@ -XXX,XX +XXX,XX @@ static void disas_fp_int_conv(DisasContext *s, uint32_t insn) handle_fmov(s, rd, rn, type, itof); break; + case 0b00111110: /* FJCVTZS */ + if (!dc_isar_feature(aa64_jscvt, s)) { + goto do_unallocated; + } else if (fp_access_check(s)) { + handle_fjcvtzs(s, rd, rn); + } + break; + default: do_unallocated: unallocated_encoding(s); diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) rm_is_dp = false; break; + case 0x13: /* vjcvt */ + if (!dp || !dc_isar_feature(aa32_jscvt, s)) { + return 1; + } + rd_is_dp = false; + break; + default: return 1; } @@ -XXX,XX +XXX,XX @@ static int disas_vfp_insn(DisasContext *s, uint32_t insn) case 17: /* fsito */ gen_vfp_sito(dp, 0); break; + case 19: /* vjcvt */ + gen_helper_vjcvt(cpu_F0s, cpu_F0d, cpu_env); + break; case 20: /* fshto */ gen_vfp_shto(dp, 16 - rm, 0); break; diff --git a/target/arm/vfp_helper.c b/target/arm/vfp_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vfp_helper.c +++ b/target/arm/vfp_helper.c @@ -XXX,XX +XXX,XX @@ int arm_rmode_to_sf(int rmode) } return rmode; } + +/* + * Implement float64 to int32_t conversion without saturation; + * the result is supplied modulo 2^32. + */ +uint64_t HELPER(fjcvtzs)(float64 value, void *vstatus) +{ + float_status *status = vstatus; + uint32_t exp, sign; + uint64_t frac; + uint32_t inexact = 1; /* !Z */ + + sign = extract64(value, 63, 1); + exp = extract64(value, 52, 11); + frac = extract64(value, 0, 52); + + if (exp == 0) { + /* While not inexact for IEEE FP, -0.0 is inexact for JavaScript. */ + inexact = sign; + if (frac != 0) { + if (status->flush_inputs_to_zero) { + float_raise(float_flag_input_denormal, status); + } else { + float_raise(float_flag_inexact, status); + inexact = 1; + } + } + frac = 0; + } else if (exp == 0x7ff) { + /* This operation raises Invalid for both NaN and overflow (Inf). */ + float_raise(float_flag_invalid, status); + frac = 0; + } else { + int true_exp = exp - 1023; + int shift = true_exp - 52; + + /* Restore implicit bit. */ + frac |= 1ull << 52; + + /* Shift the fraction into place. */ + if (shift >= 0) { + /* The number is so large we must shift the fraction left. */ + if (shift >= 64) { + /* The fraction is shifted out entirely. */ + frac = 0; + } else { + frac <<= shift; + } + } else if (shift > -64) { + /* Normal case -- shift right and notice if bits shift out. */ + inexact = (frac << (64 + shift)) != 0; + frac >>= -shift; + } else { + /* The fraction is shifted out entirely. */ + frac = 0; + } + + /* Notice overflow or inexact exceptions. */ + if (true_exp > 31 || frac > (sign ? 0x80000000ull : 0x7fffffff)) { + /* Overflow, for which this operation raises invalid. */ + float_raise(float_flag_invalid, status); + inexact = 1; + } else if (inexact) { + float_raise(float_flag_inexact, status); + } + + /* Honor the sign. */ + if (sign) { + frac = -frac; + } + } + + /* Pack the result and the env->ZF representation of Z together. */ + return deposit64(frac, 32, 32, inexact); +} + +uint32_t HELPER(vjcvt)(float64 value, CPUARMState *env) +{ + uint64_t pair = HELPER(fjcvtzs)(value, &env->vfp.fp_status); + uint32_t result = pair; + uint32_t z = (pair >> 32) == 0; + + /* Store Z, clear NCV, in FPSCR.NZCV. */ + env->vfp.xregs[ARM_VFP_FPSCR] + = (env->vfp.xregs[ARM_VFP_FPSCR] & ~CPSR_NZCV) | (z * CPSR_Z); + + return result; +} -- 2.20.1 The Peripheral Protection Controller's handling of unused ports is that if there is nothing connected to the port's downstream then it does not create the sysbus MMIO region for the upstream end of the port. This results in odd behaviour when there is an unused port in the middle of the range: since sysbus MMIO regions are implicitly consecutively allocated, any used ports above the unused ones end up with sysbus MMIO region numbers that don't match the port number. Avoid this numbering mismatch by creating dummy MMIO regions for the unused ports. This doesn't change anything for our existing boards, which don't have any gaps in the middle of the port ranges they use; but it will be needed for the Musca board. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/misc/tz-ppc.h | 8 +++++++- hw/misc/tz-ppc.c | 32 ++++++++++++++++++++++++++++++++ 2 files changed, 39 insertions(+), 1 deletion(-) diff --git a/include/hw/misc/tz-ppc.h b/include/hw/misc/tz-ppc.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/misc/tz-ppc.h +++ b/include/hw/misc/tz-ppc.h @@ -XXX,XX +XXX,XX @@ * * QEMU interface: * + sysbus MMIO regions 0..15: MemoryRegions defining the upstream end - * of each of the 16 ports of the PPC + * of each of the 16 ports of the PPC. When a port is unused (i.e. no + * downstream MemoryRegion is connected to it) at the end of the 0..15 + * range then no sysbus MMIO region is created for its upstream. When an + * unused port lies in the middle of the range with other used ports at + * higher port numbers, a dummy MMIO region is created to ensure that + * port N's upstream is always sysbus MMIO region N. Dummy regions should + * not be mapped, and will assert if any access is made to them. * + Property "port[0..15]": MemoryRegion defining the downstream device(s) * for each of the 16 ports of the PPC * + Named GPIO inputs "cfg_nonsec[0..15]": set to 1 if the port should be diff --git a/hw/misc/tz-ppc.c b/hw/misc/tz-ppc.c index XXXXXXX..XXXXXXX 100644 --- a/hw/misc/tz-ppc.c +++ b/hw/misc/tz-ppc.c @@ -XXX,XX +XXX,XX @@ static const MemoryRegionOps tz_ppc_ops = { .endianness = DEVICE_LITTLE_ENDIAN, }; +static bool tz_ppc_dummy_accepts(void *opaque, hwaddr addr, + unsigned size, bool is_write, + MemTxAttrs attrs) +{ + /* + * Board code should never map the upstream end of an unused port, + * so we should never try to make a memory access to it. + */ + g_assert_not_reached(); +} + +static const MemoryRegionOps tz_ppc_dummy_ops = { + .valid.accepts = tz_ppc_dummy_accepts, +}; + static void tz_ppc_reset(DeviceState *dev) { TZPPC *s = TZ_PPC(dev); @@ -XXX,XX +XXX,XX @@ static void tz_ppc_realize(DeviceState *dev, Error **errp) SysBusDevice *sbd = SYS_BUS_DEVICE(dev); TZPPC *s = TZ_PPC(dev); int i; + int max_port = 0; /* We can't create the upstream end of the port until realize, * as we don't know the size of the MR used as the downstream until then. */ for (i = 0; i < TZ_NUM_PORTS; i++) { + if (s->port[i].downstream) { + max_port = i; + } + } + + for (i = 0; i <= max_port; i++) { TZPPCPort *port = &s->port[i]; char *name; uint64_t size; if (!port->downstream) { + /* + * Create dummy sysbus MMIO region so the sysbus region + * numbering doesn't get out of sync with the port numbers. + * The size is entirely arbitrary. + */ + name = g_strdup_printf("tz-ppc-dummy-port[%d]", i); + memory_region_init_io(&port->upstream, obj, &tz_ppc_dummy_ops, + port, name, 0x10000); + sysbus_init_mmio(sbd, &port->upstream); + g_free(name); continue; } -- 2.20.1 Create a new include file for the pl031's device struct, type macros, etc, so that it can be instantiated using the "embedded struct" coding style. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/timer/pl031.h | 44 ++++++++++++++++++++++++++++++++++++++++ hw/timer/pl031.c | 25 +---------------------- MAINTAINERS | 1 + 3 files changed, 46 insertions(+), 24 deletions(-) create mode 100644 include/hw/timer/pl031.h diff --git a/include/hw/timer/pl031.h b/include/hw/timer/pl031.h new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/include/hw/timer/pl031.h @@ -XXX,XX +XXX,XX @@ +/* + * ARM AMBA PrimeCell PL031 RTC + * + * Copyright (c) 2007 CodeSourcery + * + * This file is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + * + * Contributions after 2012-01-13 are licensed under the terms of the + * GNU GPL, version 2 or (at your option) any later version. + */ + +#ifndef HW_TIMER_PL031 +#define HW_TIMER_PL031 + +#include "hw/sysbus.h" + +#define TYPE_PL031 "pl031" +#define PL031(obj) OBJECT_CHECK(PL031State, (obj), TYPE_PL031) + +typedef struct PL031State { + SysBusDevice parent_obj; + + MemoryRegion iomem; + QEMUTimer *timer; + qemu_irq irq; + + /* + * Needed to preserve the tick_count across migration, even if the + * absolute value of the rtc_clock is different on the source and + * destination. + */ + uint32_t tick_offset_vmstate; + uint32_t tick_offset; + + uint32_t mr; + uint32_t lr; + uint32_t cr; + uint32_t im; + uint32_t is; +} PL031State; + +#endif diff --git a/hw/timer/pl031.c b/hw/timer/pl031.c index XXXXXXX..XXXXXXX 100644 --- a/hw/timer/pl031.c +++ b/hw/timer/pl031.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "hw/timer/pl031.h" #include "hw/sysbus.h" #include "qemu/timer.h" #include "sysemu/sysemu.h" @@ -XXX,XX +XXX,XX @@ do { printf("pl031: " fmt , ## __VA_ARGS__); } while (0) #define RTC_MIS 0x18 /* Masked interrupt status register */ #define RTC_ICR 0x1c /* Interrupt clear register */ -#define TYPE_PL031 "pl031" -#define PL031(obj) OBJECT_CHECK(PL031State, (obj), TYPE_PL031) - -typedef struct PL031State { - SysBusDevice parent_obj; - - MemoryRegion iomem; - QEMUTimer *timer; - qemu_irq irq; - - /* Needed to preserve the tick_count across migration, even if the - * absolute value of the rtc_clock is different on the source and - * destination. - */ - uint32_t tick_offset_vmstate; - uint32_t tick_offset; - - uint32_t mr; - uint32_t lr; - uint32_t cr; - uint32_t im; - uint32_t is; -} PL031State; - static const unsigned char pl031_id[] = { 0x31, 0x10, 0x14, 0x00, /* Device ID */ 0x0d, 0xf0, 0x05, 0xb1 /* Cell ID */ diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: hw/sd/pl181.c F: hw/ssi/pl022.c F: include/hw/ssi/pl022.h F: hw/timer/pl031.c +F: include/hw/timer/pl031.h F: include/hw/arm/primecell.h F: hw/timer/cmsdk-apb-timer.c F: include/hw/timer/cmsdk-apb-timer.h -- 2.20.1 Convert the debug printing in the PL031 device to use trace events, and augment it to cover the interesting parts of device operation. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/timer/pl031.c | 55 +++++++++++++++++++++++-------------------- hw/timer/trace-events | 6 +++++ 2 files changed, 36 insertions(+), 25 deletions(-) diff --git a/hw/timer/pl031.c b/hw/timer/pl031.c index XXXXXXX..XXXXXXX 100644 --- a/hw/timer/pl031.c +++ b/hw/timer/pl031.c @@ -XXX,XX +XXX,XX @@ #include "sysemu/sysemu.h" #include "qemu/cutils.h" #include "qemu/log.h" - -//#define DEBUG_PL031 - -#ifdef DEBUG_PL031 -#define DPRINTF(fmt, ...) \ -do { printf("pl031: " fmt , ## __VA_ARGS__); } while (0) -#else -#define DPRINTF(fmt, ...) do {} while(0) -#endif +#include "trace.h" #define RTC_DR 0x00 /* Data read register */ #define RTC_MR 0x04 /* Match register */ @@ -XXX,XX +XXX,XX @@ static const unsigned char pl031_id[] = { static void pl031_update(PL031State *s) { - qemu_set_irq(s->irq, s->is & s->im); + uint32_t flags = s->is & s->im; + + trace_pl031_irq_state(flags); + qemu_set_irq(s->irq, flags); } static void pl031_interrupt(void * opaque) @@ -XXX,XX +XXX,XX @@ static void pl031_interrupt(void * opaque) PL031State *s = (PL031State *)opaque; s->is = 1; - DPRINTF("Alarm raised\n"); + trace_pl031_alarm_raised(); pl031_update(s); } @@ -XXX,XX +XXX,XX @@ static void pl031_set_alarm(PL031State *s) /* The timer wraps around. This subtraction also wraps in the same way, and gives correct results when alarm < now_ticks. */ ticks = s->mr - pl031_get_count(s); - DPRINTF("Alarm set in %ud ticks\n", ticks); + trace_pl031_set_alarm(ticks); if (ticks == 0) { timer_del(s->timer); pl031_interrupt(s); @@ -XXX,XX +XXX,XX @@ static uint64_t pl031_read(void *opaque, hwaddr offset, unsigned size) { PL031State *s = (PL031State *)opaque; - - if (offset >= 0xfe0 && offset < 0x1000) - return pl031_id[(offset - 0xfe0) >> 2]; + uint64_t r; switch (offset) { case RTC_DR: - return pl031_get_count(s); + r = pl031_get_count(s); + break; case RTC_MR: - return s->mr; + r = s->mr; + break; case RTC_IMSC: - return s->im; + r = s->im; + break; case RTC_RIS: - return s->is; + r = s->is; + break; case RTC_LR: - return s->lr; + r = s->lr; + break; case RTC_CR: /* RTC is permanently enabled. */ - return 1; + r = 1; + break; case RTC_MIS: - return s->is & s->im; + r = s->is & s->im; + break; + case 0xfe0 ... 0xfff: + r = pl031_id[(offset - 0xfe0) >> 2]; + break; case RTC_ICR: qemu_log_mask(LOG_GUEST_ERROR, "pl031: read of write-only register at offset 0x%x\n", (int)offset); + r = 0; break; default: qemu_log_mask(LOG_GUEST_ERROR, "pl031_read: Bad offset 0x%x\n", (int)offset); + r = 0; break; } - return 0; + trace_pl031_read(offset, r); + return r; } static void pl031_write(void * opaque, hwaddr offset, @@ -XXX,XX +XXX,XX @@ static void pl031_write(void * opaque, hwaddr offset, { PL031State *s = (PL031State *)opaque; + trace_pl031_write(offset, value); switch (offset) { case RTC_LR: @@ -XXX,XX +XXX,XX @@ static void pl031_write(void * opaque, hwaddr offset, break; case RTC_IMSC: s->im = value & 1; - DPRINTF("Interrupt mask %d\n", s->im); pl031_update(s); break; case RTC_ICR: @@ -XXX,XX +XXX,XX @@ static void pl031_write(void * opaque, hwaddr offset, cleared when bit 0 of the written value is set. However the arm926e documentation (DDI0287B) states that the interrupt is cleared when any value is written. */ - DPRINTF("Interrupt cleared"); s->is = 0; pl031_update(s); break; diff --git a/hw/timer/trace-events b/hw/timer/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/timer/trace-events +++ b/hw/timer/trace-events @@ -XXX,XX +XXX,XX @@ xlnx_zynqmp_rtc_gettime(int year, int month, int day, int hour, int min, int sec nrf51_timer_read(uint64_t addr, uint32_t value, unsigned size) "read addr 0x%" PRIx64 " data 0x%" PRIx32 " size %u" nrf51_timer_write(uint64_t addr, uint32_t value, unsigned size) "write addr 0x%" PRIx64 " data 0x%" PRIx32 " size %u" +# hw/timer/pl031.c +pl031_irq_state(int level) "irq state %d" +pl031_read(uint32_t addr, uint32_t value) "addr 0x%08x value 0x%08x" +pl031_write(uint32_t addr, uint32_t value) "addr 0x%08x value 0x%08x" +pl031_alarm_raised(void) "alarm raised" +pl031_set_alarm(uint32_t ticks) "alarm set for %u ticks" -- 2.20.1 Create a new include file for the pl011's device struct, type macros, etc, so that it can be instantiated using the "embedded struct" coding style. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/char/pl011.h | 34 ++++++++++++++++++++++++++++++++++ hw/char/pl011.c | 31 ++----------------------------- 2 files changed, 36 insertions(+), 29 deletions(-) diff --git a/include/hw/char/pl011.h b/include/hw/char/pl011.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/char/pl011.h +++ b/include/hw/char/pl011.h @@ -XXX,XX +XXX,XX @@ #ifndef HW_PL011_H #define HW_PL011_H +#include "hw/sysbus.h" +#include "chardev/char-fe.h" + +#define TYPE_PL011 "pl011" +#define PL011(obj) OBJECT_CHECK(PL011State, (obj), TYPE_PL011) + +/* This shares the same struct (and cast macro) as the base pl011 device */ +#define TYPE_PL011_LUMINARY "pl011_luminary" + +typedef struct PL011State { + SysBusDevice parent_obj; + + MemoryRegion iomem; + uint32_t readbuff; + uint32_t flags; + uint32_t lcr; + uint32_t rsr; + uint32_t cr; + uint32_t dmacr; + uint32_t int_enabled; + uint32_t int_level; + uint32_t read_fifo[16]; + uint32_t ilpr; + uint32_t ibrd; + uint32_t fbrd; + uint32_t ifl; + int read_pos; + int read_count; + int read_trigger; + CharBackend chr; + qemu_irq irq; + const unsigned char *id; +} PL011State; + static inline DeviceState *pl011_create(hwaddr addr, qemu_irq irq, Chardev *chr) diff --git a/hw/char/pl011.c b/hw/char/pl011.c index XXXXXXX..XXXXXXX 100644 --- a/hw/char/pl011.c +++ b/hw/char/pl011.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "hw/char/pl011.h" #include "hw/sysbus.h" #include "chardev/char-fe.h" #include "qemu/log.h" #include "trace.h" -#define TYPE_PL011 "pl011" -#define PL011(obj) OBJECT_CHECK(PL011State, (obj), TYPE_PL011) - -typedef struct PL011State { - SysBusDevice parent_obj; - - MemoryRegion iomem; - uint32_t readbuff; - uint32_t flags; - uint32_t lcr; - uint32_t rsr; - uint32_t cr; - uint32_t dmacr; - uint32_t int_enabled; - uint32_t int_level; - uint32_t read_fifo[16]; - uint32_t ilpr; - uint32_t ibrd; - uint32_t fbrd; - uint32_t ifl; - int read_pos; - int read_count; - int read_trigger; - CharBackend chr; - qemu_irq irq; - const unsigned char *id; -} PL011State; - #define PL011_INT_TX 0x20 #define PL011_INT_RX 0x10 @@ -XXX,XX +XXX,XX @@ static void pl011_luminary_init(Object *obj) } static const TypeInfo pl011_luminary_info = { - .name = "pl011_luminary", + .name = TYPE_PL011_LUMINARY, .parent = TYPE_PL011, .instance_init = pl011_luminary_init, }; -- 2.20.1 The PL011 UART has six interrupt lines: * RX (receive data) * TX (transmit data) * RT (receive timeout) * MS (modem status) * E (errors) * combined (logical OR of all the above) So far we have only emulated the combined interrupt line; add support for the others, so that boards that wire them up to different interrupt controller inputs can do so. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/char/pl011.h | 2 +- hw/char/pl011.c | 46 +++++++++++++++++++++++++++++++++++++++-- 2 files changed, 45 insertions(+), 3 deletions(-) diff --git a/include/hw/char/pl011.h b/include/hw/char/pl011.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/char/pl011.h +++ b/include/hw/char/pl011.h @@ -XXX,XX +XXX,XX @@ typedef struct PL011State { int read_count; int read_trigger; CharBackend chr; - qemu_irq irq; + qemu_irq irq[6]; const unsigned char *id; } PL011State; diff --git a/hw/char/pl011.c b/hw/char/pl011.c index XXXXXXX..XXXXXXX 100644 --- a/hw/char/pl011.c +++ b/hw/char/pl011.c @@ -XXX,XX +XXX,XX @@ * This code is licensed under the GPL. */ +/* + * QEMU interface: + * + sysbus MMIO region 0: device registers + * + sysbus IRQ 0: UARTINTR (combined interrupt line) + * + sysbus IRQ 1: UARTRXINTR (receive FIFO interrupt line) + * + sysbus IRQ 2: UARTTXINTR (transmit FIFO interrupt line) + * + sysbus IRQ 3: UARTRTINTR (receive timeout interrupt line) + * + sysbus IRQ 4: UARTMSINTR (momem status interrupt line) + * + sysbus IRQ 5: UARTEINTR (error interrupt line) + */ + #include "qemu/osdep.h" #include "hw/char/pl011.h" #include "hw/sysbus.h" @@ -XXX,XX +XXX,XX @@ #define PL011_FLAG_TXFF 0x20 #define PL011_FLAG_RXFE 0x10 +/* Interrupt status bits in UARTRIS, UARTMIS, UARTIMSC */ +#define INT_OE (1 << 10) +#define INT_BE (1 << 9) +#define INT_PE (1 << 8) +#define INT_FE (1 << 7) +#define INT_RT (1 << 6) +#define INT_TX (1 << 5) +#define INT_RX (1 << 4) +#define INT_DSR (1 << 3) +#define INT_DCD (1 << 2) +#define INT_CTS (1 << 1) +#define INT_RI (1 << 0) +#define INT_E (INT_OE | INT_BE | INT_PE | INT_FE) +#define INT_MS (INT_RI | INT_DSR | INT_DCD | INT_CTS) + static const unsigned char pl011_id_arm[8] = { 0x11, 0x10, 0x14, 0x00, 0x0d, 0xf0, 0x05, 0xb1 }; static const unsigned char pl011_id_luminary[8] = { 0x11, 0x00, 0x18, 0x01, 0x0d, 0xf0, 0x05, 0xb1 }; +/* Which bits in the interrupt status matter for each outbound IRQ line ? */ +static const uint32_t irqmask[] = { + INT_E | INT_MS | INT_RT | INT_TX | INT_RX, /* combined IRQ */ + INT_RX, + INT_TX, + INT_RT, + INT_MS, + INT_E, +}; + static void pl011_update(PL011State *s) { uint32_t flags; + int i; flags = s->int_level & s->int_enabled; trace_pl011_irq_state(flags != 0); - qemu_set_irq(s->irq, flags != 0); + for (i = 0; i < ARRAY_SIZE(s->irq); i++) { + qemu_set_irq(s->irq[i], (flags & irqmask[i]) != 0); + } } static uint64_t pl011_read(void *opaque, hwaddr offset, @@ -XXX,XX +XXX,XX @@ static void pl011_init(Object *obj) { SysBusDevice *sbd = SYS_BUS_DEVICE(obj); PL011State *s = PL011(obj); + int i; memory_region_init_io(&s->iomem, OBJECT(s), &pl011_ops, s, "pl011", 0x1000); sysbus_init_mmio(sbd, &s->iomem); - sysbus_init_irq(sbd, &s->irq); + for (i = 0; i < ARRAY_SIZE(s->irq); i++) { + sysbus_init_irq(sbd, &s->irq[i]); + } s->read_trigger = 1; s->ifl = 0x12; -- 2.20.1 The pl011 logs when the guest makes a bad access. It prints the address offset in hex but confusingly omits the '0x' prefix; add it. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/char/pl011.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/hw/char/pl011.c b/hw/char/pl011.c index XXXXXXX..XXXXXXX 100644 --- a/hw/char/pl011.c +++ b/hw/char/pl011.c @@ -XXX,XX +XXX,XX @@ static uint64_t pl011_read(void *opaque, hwaddr offset, break; default: qemu_log_mask(LOG_GUEST_ERROR, - "pl011_read: Bad offset %x\n", (int)offset); + "pl011_read: Bad offset 0x%x\n", (int)offset); r = 0; break; } @@ -XXX,XX +XXX,XX @@ static void pl011_write(void *opaque, hwaddr offset, break; default: qemu_log_mask(LOG_GUEST_ERROR, - "pl011_write: Bad offset %x\n", (int)offset); + "pl011_write: Bad offset 0x%x\n", (int)offset); } } -- 2.20.1 In commit 4b635cf7a95e501211 we added a QOM property to the ARMSSE object, but forgot to add it to the documentation comment in the header. Correct the omission. Fixes: 4b635cf7a95e501211 ("hw/arm/armsse: Make SRAM bank size configurable") Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/arm/armsse.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/include/hw/arm/armsse.h b/include/hw/arm/armsse.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/armsse.h +++ b/include/hw/arm/armsse.h @@ -XXX,XX +XXX,XX @@ * being the same for both, to avoid having to have separate Property * lists for different variants. This restriction can be relaxed later * if necessary.) + * + QOM property "SRAM_ADDR_WIDTH" sets the number of bits used for the + * address of each SRAM bank (and thus the total amount of internal SRAM) * + Named GPIO inputs "EXP_IRQ" 0..n are the expansion interrupts for CPU 0, * which are wired to its NVIC lines 32 .. n+32 * + Named GPIO inputs "EXP_CPU1_IRQ" 0..n are the expansion interrupts for -- 2.20.1 The Musca boards have DAPLink firmware that sets the initial secure VTOR value (the location of the vector table) differently depending on the boot mode (from flash, from RAM, etc). Export the init-svtor as a QOM property of the ARMSSE object so that the board can change it. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- include/hw/arm/armsse.h | 3 +++ hw/arm/armsse.c | 8 ++++---- 2 files changed, 7 insertions(+), 4 deletions(-) diff --git a/include/hw/arm/armsse.h b/include/hw/arm/armsse.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/armsse.h +++ b/include/hw/arm/armsse.h @@ -XXX,XX +XXX,XX @@ * if necessary.) * + QOM property "SRAM_ADDR_WIDTH" sets the number of bits used for the * address of each SRAM bank (and thus the total amount of internal SRAM) + * + QOM property "init-svtor" sets the initial value of the CPU SVTOR register + * (where it expects to load the PC and SP from the vector table on reset) * + Named GPIO inputs "EXP_IRQ" 0..n are the expansion interrupts for CPU 0, * which are wired to its NVIC lines 32 .. n+32 * + Named GPIO inputs "EXP_CPU1_IRQ" 0..n are the expansion interrupts for @@ -XXX,XX +XXX,XX @@ typedef struct ARMSSE { uint32_t exp_numirq; uint32_t mainclk_frq; uint32_t sram_addr_width; + uint32_t init_svtor; } ARMSSE; typedef struct ARMSSEInfo ARMSSEInfo; diff --git a/hw/arm/armsse.c b/hw/arm/armsse.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/armsse.c +++ b/hw/arm/armsse.c @@ -XXX,XX +XXX,XX @@ static void armsse_realize(DeviceState *dev, Error **errp) * the INITSVTOR* registers before powering up the CPUs in any case, * so the hardware's default value doesn't matter. QEMU doesn't emulate * the control processor, so instead we behave in the way that the - * firmware does. All boards currently known about have firmware that - * sets the INITSVTOR0 and INITSVTOR1 registers to 0x10000000, like the - * IoTKit default. We can make this more configurable if necessary. + * firmware does. The initial value is configurable by the board code + * to match whatever its firmware does. */ - qdev_prop_set_uint32(cpudev, "init-svtor", 0x10000000); + qdev_prop_set_uint32(cpudev, "init-svtor", s->init_svtor); /* * Start all CPUs except CPU0 powered down. In real hardware it is * a configurable property of the SSE-200 which CPUs start powered up @@ -XXX,XX +XXX,XX @@ static Property armsse_properties[] = { DEFINE_PROP_UINT32("EXP_NUMIRQ", ARMSSE, exp_numirq, 64), DEFINE_PROP_UINT32("MAINCLK", ARMSSE, mainclk_frq, 0), DEFINE_PROP_UINT32("SRAM_ADDR_WIDTH", ARMSSE, sram_addr_width, 15), + DEFINE_PROP_UINT32("init-svtor", ARMSSE, init_svtor, 0x10000000), DEFINE_PROP_END_OF_LIST() }; -- 2.20.1 The Musca-A and Musca-B1 development boards are based on the SSE-200 subsystem for embedded. Implement an initial skeleton model of these boards, which are similar but not identical. This commit creates the board model with the SSE and the IRQ splitters to wire IRQs up to its two CPUs. As yet there are no devices and no memory: these will be added later. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/arm/Makefile.objs | 1 + hw/arm/musca.c | 197 ++++++++++++++++++++++++++++++++ MAINTAINERS | 6 + default-configs/arm-softmmu.mak | 1 + 4 files changed, 205 insertions(+) create mode 100644 hw/arm/musca.c diff --git a/hw/arm/Makefile.objs b/hw/arm/Makefile.objs index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/Makefile.objs +++ b/hw/arm/Makefile.objs @@ -XXX,XX +XXX,XX @@ obj-$(CONFIG_ASPEED_SOC) += aspeed_soc.o aspeed.o obj-$(CONFIG_MPS2) += mps2.o obj-$(CONFIG_MPS2) += mps2-tz.o obj-$(CONFIG_MSF2) += msf2-soc.o msf2-som.o +obj-$(CONFIG_MUSCA) += musca.o obj-$(CONFIG_ARMSSE) += armsse.o obj-$(CONFIG_FSL_IMX7) += fsl-imx7.o mcimx7d-sabre.o obj-$(CONFIG_ARM_SMMUV3) += smmu-common.o smmuv3.o diff --git a/hw/arm/musca.c b/hw/arm/musca.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/hw/arm/musca.c @@ -XXX,XX +XXX,XX @@ +/* + * Arm Musca-B1 test chip board emulation + * + * Copyright (c) 2019 Linaro Limited + * Written by Peter Maydell + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 or + * (at your option) any later version. + */ + +/* + * The Musca boards are a reference implementation of a system using + * the SSE-200 subsystem for embedded: + * https://developer.arm.com/products/system-design/development-boards/iot-test-chips-and-boards/musca-a-test-chip-board + * https://developer.arm.com/products/system-design/development-boards/iot-test-chips-and-boards/musca-b-test-chip-board + * We model the A and B1 variants of this board, as described in the TRMs: + * http://infocenter.arm.com/help/topic/com.arm.doc.101107_0000_00_en/index.html + * http://infocenter.arm.com/help/topic/com.arm.doc.101312_0000_00_en/index.html + */ + +#include "qemu/osdep.h" +#include "qemu/error-report.h" +#include "qapi/error.h" +#include "exec/address-spaces.h" +#include "hw/arm/arm.h" +#include "hw/arm/armsse.h" +#include "hw/boards.h" +#include "hw/core/split-irq.h" + +#define MUSCA_NUMIRQ_MAX 96 + +typedef enum MuscaType { + MUSCA_A, + MUSCA_B1, +} MuscaType; + +typedef struct { + MachineClass parent; + MuscaType type; + uint32_t init_svtor; + int sram_addr_width; + int num_irqs; +} MuscaMachineClass; + +typedef struct { + MachineState parent; + + ARMSSE sse; + SplitIRQ cpu_irq_splitter[MUSCA_NUMIRQ_MAX]; +} MuscaMachineState; + +#define TYPE_MUSCA_MACHINE "musca" +#define TYPE_MUSCA_A_MACHINE MACHINE_TYPE_NAME("musca-a") +#define TYPE_MUSCA_B1_MACHINE MACHINE_TYPE_NAME("musca-b1") + +#define MUSCA_MACHINE(obj) \ + OBJECT_CHECK(MuscaMachineState, obj, TYPE_MUSCA_MACHINE) +#define MUSCA_MACHINE_GET_CLASS(obj) \ + OBJECT_GET_CLASS(MuscaMachineClass, obj, TYPE_MUSCA_MACHINE) +#define MUSCA_MACHINE_CLASS(klass) \ + OBJECT_CLASS_CHECK(MuscaMachineClass, klass, TYPE_MUSCA_MACHINE) + +/* + * Main SYSCLK frequency in Hz + * TODO this should really be different for the two cores, but we + * don't model that in our SSE-200 model yet. + */ +#define SYSCLK_FRQ 40000000 + +static void musca_init(MachineState *machine) +{ + MuscaMachineState *mms = MUSCA_MACHINE(machine); + MuscaMachineClass *mmc = MUSCA_MACHINE_GET_CLASS(mms); + MachineClass *mc = MACHINE_GET_CLASS(machine); + MemoryRegion *system_memory = get_system_memory(); + DeviceState *ssedev; + int i; + + assert(mmc->num_irqs <= MUSCA_NUMIRQ_MAX); + + if (strcmp(machine->cpu_type, mc->default_cpu_type) != 0) { + error_report("This board can only be used with CPU %s", + mc->default_cpu_type); + exit(1); + } + + sysbus_init_child_obj(OBJECT(machine), "sse-200", &mms->sse, + sizeof(mms->sse), TYPE_SSE200); + ssedev = DEVICE(&mms->sse); + object_property_set_link(OBJECT(&mms->sse), OBJECT(system_memory), + "memory", &error_fatal); + qdev_prop_set_uint32(ssedev, "EXP_NUMIRQ", mmc->num_irqs); + qdev_prop_set_uint32(ssedev, "init-svtor", mmc->init_svtor); + qdev_prop_set_uint32(ssedev, "SRAM_ADDR_WIDTH", mmc->sram_addr_width); + qdev_prop_set_uint32(ssedev, "MAINCLK", SYSCLK_FRQ); + object_property_set_bool(OBJECT(&mms->sse), true, "realized", + &error_fatal); + + /* + * We need to create splitters to feed the IRQ inputs + * for each CPU in the SSE-200 from each device in the board. + */ + for (i = 0; i < mmc->num_irqs; i++) { + char *name = g_strdup_printf("musca-irq-splitter%d", i); + SplitIRQ *splitter = &mms->cpu_irq_splitter[i]; + + object_initialize_child(OBJECT(machine), name, + splitter, sizeof(*splitter), + TYPE_SPLIT_IRQ, &error_fatal, NULL); + g_free(name); + + object_property_set_int(OBJECT(splitter), 2, "num-lines", + &error_fatal); + object_property_set_bool(OBJECT(splitter), true, "realized", + &error_fatal); + qdev_connect_gpio_out(DEVICE(splitter), 0, + qdev_get_gpio_in_named(ssedev, "EXP_IRQ", i)); + qdev_connect_gpio_out(DEVICE(splitter), 1, + qdev_get_gpio_in_named(ssedev, + "EXP_CPU1_IRQ", i)); + } + + armv7m_load_kernel(ARM_CPU(first_cpu), machine->kernel_filename, 0x2000000); +} + +static void musca_class_init(ObjectClass *oc, void *data) +{ + MachineClass *mc = MACHINE_CLASS(oc); + + mc->default_cpus = 2; + mc->min_cpus = mc->default_cpus; + mc->max_cpus = mc->default_cpus; + mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-m33"); + mc->init = musca_init; +} + +static void musca_a_class_init(ObjectClass *oc, void *data) +{ + MachineClass *mc = MACHINE_CLASS(oc); + MuscaMachineClass *mmc = MUSCA_MACHINE_CLASS(oc); + + mc->desc = "ARM Musca-A board (dual Cortex-M33)"; + mmc->type = MUSCA_A; + mmc->init_svtor = 0x10200000; + mmc->sram_addr_width = 15; + mmc->num_irqs = 64; +} + +static void musca_b1_class_init(ObjectClass *oc, void *data) +{ + MachineClass *mc = MACHINE_CLASS(oc); + MuscaMachineClass *mmc = MUSCA_MACHINE_CLASS(oc); + + mc->desc = "ARM Musca-B1 board (dual Cortex-M33)"; + mmc->type = MUSCA_B1; + /* + * This matches the DAPlink firmware which boots from QSPI. There + * is also a firmware blob which boots from the eFlash, which + * uses init_svtor = 0x1A000000. QEMU doesn't currently support that, + * though we could in theory expose a machine property on the command + * line to allow the user to request eFlash boot. + */ + mmc->init_svtor = 0x10000000; + mmc->sram_addr_width = 17; + mmc->num_irqs = 96; +} + +static const TypeInfo musca_info = { + .name = TYPE_MUSCA_MACHINE, + .parent = TYPE_MACHINE, + .abstract = true, + .instance_size = sizeof(MuscaMachineState), + .class_size = sizeof(MuscaMachineClass), + .class_init = musca_class_init, +}; + +static const TypeInfo musca_a_info = { + .name = TYPE_MUSCA_A_MACHINE, + .parent = TYPE_MUSCA_MACHINE, + .class_init = musca_a_class_init, +}; + +static const TypeInfo musca_b1_info = { + .name = TYPE_MUSCA_B1_MACHINE, + .parent = TYPE_MUSCA_MACHINE, + .class_init = musca_b1_class_init, +}; + +static void musca_machine_init(void) +{ + type_register_static(&musca_info); + type_register_static(&musca_a_info); + type_register_static(&musca_b1_info); +} + +type_init(musca_machine_init); diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: include/hw/misc/iotkit-sysinfo.h F: hw/misc/armsse-cpuid.c F: include/hw/misc/armsse-cpuid.h +Musca +M: Peter Maydell <peter.maydell@linaro.org> +L: qemu-arm@nongnu.org +S: Maintained +F: hw/arm/musca.c + Musicpal M: Jan Kiszka <jan.kiszka@web.de> M: Peter Maydell <peter.maydell@linaro.org> diff --git a/default-configs/arm-softmmu.mak b/default-configs/arm-softmmu.mak index XXXXXXX..XXXXXXX 100644 --- a/default-configs/arm-softmmu.mak +++ b/default-configs/arm-softmmu.mak @@ -XXX,XX +XXX,XX @@ CONFIG_TUSB6010=y CONFIG_IMX=y CONFIG_MAINSTONE=y CONFIG_MPS2=y +CONFIG_MUSCA=y CONFIG_NSERIES=y CONFIG_RASPI=y CONFIG_REALVIEW=y -- 2.20.1 Many of the devices on the Musca board live behind TrustZone Peripheral Protection Controllers (PPCs); add models of the PPCs, using a similar scheme to the MPS2 board models. This commit wires up the PPCs with "unimplemented device" stubs behind them in the correct places in the address map. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/arm/musca.c | 289 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 289 insertions(+) diff --git a/hw/arm/musca.c b/hw/arm/musca.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/musca.c +++ b/hw/arm/musca.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/armsse.h" #include "hw/boards.h" #include "hw/core/split-irq.h" +#include "hw/misc/tz-ppc.h" +#include "hw/misc/unimp.h" #define MUSCA_NUMIRQ_MAX 96 +#define MUSCA_PPC_MAX 3 typedef enum MuscaType { MUSCA_A, @@ -XXX,XX +XXX,XX @@ typedef struct { ARMSSE sse; SplitIRQ cpu_irq_splitter[MUSCA_NUMIRQ_MAX]; + SplitIRQ sec_resp_splitter; + TZPPC ppc[MUSCA_PPC_MAX]; + MemoryRegion container; + UnimplementedDeviceState eflash[2]; + UnimplementedDeviceState qspi; + UnimplementedDeviceState mpc[5]; + UnimplementedDeviceState mhu[2]; + UnimplementedDeviceState pwm[3]; + UnimplementedDeviceState i2s; + UnimplementedDeviceState uart[2]; + UnimplementedDeviceState i2c[2]; + UnimplementedDeviceState spi; + UnimplementedDeviceState scc; + UnimplementedDeviceState timer; + UnimplementedDeviceState rtc; + UnimplementedDeviceState pvt; + UnimplementedDeviceState sdio; + UnimplementedDeviceState gpio; } MuscaMachineState; #define TYPE_MUSCA_MACHINE "musca" @@ -XXX,XX +XXX,XX @@ typedef struct { */ #define SYSCLK_FRQ 40000000 +/* + * Most of the devices in the Musca board sit behind Peripheral Protection + * Controllers. These data structures define the layout of which devices + * sit behind which PPCs. + * The devfn for each port is a function which creates, configures + * and initializes the device, returning the MemoryRegion which + * needs to be plugged into the downstream end of the PPC port. + */ +typedef MemoryRegion *MakeDevFn(MuscaMachineState *mms, void *opaque, + const char *name, hwaddr size); + +typedef struct PPCPortInfo { + const char *name; + MakeDevFn *devfn; + void *opaque; + hwaddr addr; + hwaddr size; +} PPCPortInfo; + +typedef struct PPCInfo { + const char *name; + PPCPortInfo ports[TZ_NUM_PORTS]; +} PPCInfo; + +static MemoryRegion *make_unimp_dev(MuscaMachineState *mms, + void *opaque, const char *name, hwaddr size) +{ + /* + * Initialize, configure and realize a TYPE_UNIMPLEMENTED_DEVICE, + * and return a pointer to its MemoryRegion. + */ + UnimplementedDeviceState *uds = opaque; + + sysbus_init_child_obj(OBJECT(mms), name, uds, + sizeof(UnimplementedDeviceState), + TYPE_UNIMPLEMENTED_DEVICE); + qdev_prop_set_string(DEVICE(uds), "name", name); + qdev_prop_set_uint64(DEVICE(uds), "size", size); + object_property_set_bool(OBJECT(uds), true, "realized", &error_fatal); + return sysbus_mmio_get_region(SYS_BUS_DEVICE(uds), 0); +} + +static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, + const char *name, hwaddr size) +{ + /* + * Create the container MemoryRegion for all the devices that live + * behind the Musca-A PPC's single port. These devices don't have a PPC + * port each, but we use the PPCPortInfo struct as a convenient way + * to describe them. Note that addresses here are relative to the base + * address of the PPC port region: 0x40100000, and devices appear both + * at the 0x4... NS region and the 0x5... S region. + */ + int i; + MemoryRegion *container = &mms->container; + + const PPCPortInfo devices[] = { + { "uart0", make_unimp_dev, &mms->uart[0], 0x1000, 0x1000 }, + { "uart1", make_unimp_dev, &mms->uart[1], 0x2000, 0x1000 }, + { "spi", make_unimp_dev, &mms->spi, 0x3000, 0x1000 }, + { "i2c0", make_unimp_dev, &mms->i2c[0], 0x4000, 0x1000 }, + { "i2c1", make_unimp_dev, &mms->i2c[1], 0x5000, 0x1000 }, + { "i2s", make_unimp_dev, &mms->i2s, 0x6000, 0x1000 }, + { "pwm0", make_unimp_dev, &mms->pwm[0], 0x7000, 0x1000 }, + { "rtc", make_unimp_dev, &mms->rtc, 0x8000, 0x1000 }, + { "qspi", make_unimp_dev, &mms->qspi, 0xa000, 0x1000 }, + { "timer", make_unimp_dev, &mms->timer, 0xb000, 0x1000 }, + { "scc", make_unimp_dev, &mms->scc, 0xc000, 0x1000 }, + { "pwm1", make_unimp_dev, &mms->pwm[1], 0xe000, 0x1000 }, + { "pwm2", make_unimp_dev, &mms->pwm[2], 0xf000, 0x1000 }, + { "gpio", make_unimp_dev, &mms->gpio, 0x10000, 0x1000 }, + { "mpc0", make_unimp_dev, &mms->mpc[0], 0x12000, 0x1000 }, + { "mpc1", make_unimp_dev, &mms->mpc[1], 0x13000, 0x1000 }, + }; + + memory_region_init(container, OBJECT(mms), "musca-device-container", size); + + for (i = 0; i < ARRAY_SIZE(devices); i++) { + const PPCPortInfo *pinfo = &devices[i]; + MemoryRegion *mr; + + mr = pinfo->devfn(mms, pinfo->opaque, pinfo->name, pinfo->size); + memory_region_add_subregion(container, pinfo->addr, mr); + } + + return &mms->container; +} + static void musca_init(MachineState *machine) { MuscaMachineState *mms = MUSCA_MACHINE(machine); @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) MachineClass *mc = MACHINE_GET_CLASS(machine); MemoryRegion *system_memory = get_system_memory(); DeviceState *ssedev; + DeviceState *dev_splitter; + const PPCInfo *ppcs; + int num_ppcs; int i; assert(mmc->num_irqs <= MUSCA_NUMIRQ_MAX); @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) "EXP_CPU1_IRQ", i)); } + /* + * The sec_resp_cfg output from the SSE-200 must be split into multiple + * lines, one for each of the PPCs we create here. + */ + object_initialize(&mms->sec_resp_splitter, sizeof(mms->sec_resp_splitter), + TYPE_SPLIT_IRQ); + object_property_add_child(OBJECT(machine), "sec-resp-splitter", + OBJECT(&mms->sec_resp_splitter), &error_fatal); + object_property_set_int(OBJECT(&mms->sec_resp_splitter), + ARRAY_SIZE(mms->ppc), "num-lines", &error_fatal); + object_property_set_bool(OBJECT(&mms->sec_resp_splitter), true, + "realized", &error_fatal); + dev_splitter = DEVICE(&mms->sec_resp_splitter); + qdev_connect_gpio_out_named(ssedev, "sec_resp_cfg", 0, + qdev_get_gpio_in(dev_splitter, 0)); + + /* + * Most of the devices in the board are behind Peripheral Protection + * Controllers. The required order for initializing things is: + * + initialize the PPC + * + initialize, configure and realize downstream devices + * + connect downstream device MemoryRegions to the PPC + * + realize the PPC + * + map the PPC's MemoryRegions to the places in the address map + * where the downstream devices should appear + * + wire up the PPC's control lines to the SSE object + * + * The PPC mapping differs for the -A and -B1 variants; the -A version + * is much simpler, using only a single port of a single PPC and putting + * all the devices behind that. + */ + const PPCInfo a_ppcs[] = { { + .name = "ahb_ppcexp0", + .ports = { + { "musca-devices", make_musca_a_devs, 0, 0x40100000, 0x100000 }, + }, + }, + }; + + /* + * Devices listed with an 0x4.. address appear in both the NS 0x4.. region + * and the 0x5.. S region. Devices listed with an 0x5.. address appear + * only in the S region. + */ + const PPCInfo b1_ppcs[] = { { + .name = "apb_ppcexp0", + .ports = { + { "eflash0", make_unimp_dev, &mms->eflash[0], + 0x52400000, 0x1000 }, + { "eflash1", make_unimp_dev, &mms->eflash[1], + 0x52500000, 0x1000 }, + { "qspi", make_unimp_dev, &mms->qspi, 0x42800000, 0x100000 }, + { "mpc0", make_unimp_dev, &mms->mpc[0], 0x52000000, 0x1000 }, + { "mpc1", make_unimp_dev, &mms->mpc[1], 0x52100000, 0x1000 }, + { "mpc2", make_unimp_dev, &mms->mpc[2], 0x52200000, 0x1000 }, + { "mpc3", make_unimp_dev, &mms->mpc[3], 0x52300000, 0x1000 }, + { "mhu0", make_unimp_dev, &mms->mhu[0], 0x42600000, 0x100000 }, + { "mhu1", make_unimp_dev, &mms->mhu[1], 0x42700000, 0x100000 }, + { }, /* port 9: unused */ + { }, /* port 10: unused */ + { }, /* port 11: unused */ + { }, /* port 12: unused */ + { }, /* port 13: unused */ + { "mpc4", make_unimp_dev, &mms->mpc[4], 0x52e00000, 0x1000 }, + }, + }, { + .name = "apb_ppcexp1", + .ports = { + { "pwm0", make_unimp_dev, &mms->pwm[0], 0x40101000, 0x1000 }, + { "pwm1", make_unimp_dev, &mms->pwm[1], 0x40102000, 0x1000 }, + { "pwm2", make_unimp_dev, &mms->pwm[2], 0x40103000, 0x1000 }, + { "i2s", make_unimp_dev, &mms->i2s, 0x40104000, 0x1000 }, + { "uart0", make_unimp_dev, &mms->uart[0], 0x40105000, 0x1000 }, + { "uart1", make_unimp_dev, &mms->uart[1], 0x40106000, 0x1000 }, + { "i2c0", make_unimp_dev, &mms->i2c[0], 0x40108000, 0x1000 }, + { "i2c1", make_unimp_dev, &mms->i2c[1], 0x40109000, 0x1000 }, + { "spi", make_unimp_dev, &mms->spi, 0x4010a000, 0x1000 }, + { "scc", make_unimp_dev, &mms->scc, 0x5010b000, 0x1000 }, + { "timer", make_unimp_dev, &mms->timer, 0x4010c000, 0x1000 }, + { "rtc", make_unimp_dev, &mms->rtc, 0x4010d000, 0x1000 }, + { "pvt", make_unimp_dev, &mms->pvt, 0x4010e000, 0x1000 }, + { "sdio", make_unimp_dev, &mms->sdio, 0x4010f000, 0x1000 }, + }, + }, { + .name = "ahb_ppcexp0", + .ports = { + { }, /* port 0: unused */ + { "gpio", make_unimp_dev, &mms->gpio, 0x41000000, 0x1000 }, + }, + }, + }; + + switch (mmc->type) { + case MUSCA_A: + ppcs = a_ppcs; + num_ppcs = ARRAY_SIZE(a_ppcs); + break; + case MUSCA_B1: + ppcs = b1_ppcs; + num_ppcs = ARRAY_SIZE(b1_ppcs); + break; + default: + g_assert_not_reached(); + } + assert(num_ppcs <= MUSCA_PPC_MAX); + + for (i = 0; i < num_ppcs; i++) { + const PPCInfo *ppcinfo = &ppcs[i]; + TZPPC *ppc = &mms->ppc[i]; + DeviceState *ppcdev; + int port; + char *gpioname; + + sysbus_init_child_obj(OBJECT(machine), ppcinfo->name, ppc, + sizeof(TZPPC), TYPE_TZ_PPC); + ppcdev = DEVICE(ppc); + + for (port = 0; port < TZ_NUM_PORTS; port++) { + const PPCPortInfo *pinfo = &ppcinfo->ports[port]; + MemoryRegion *mr; + char *portname; + + if (!pinfo->devfn) { + continue; + } + + mr = pinfo->devfn(mms, pinfo->opaque, pinfo->name, pinfo->size); + portname = g_strdup_printf("port[%d]", port); + object_property_set_link(OBJECT(ppc), OBJECT(mr), + portname, &error_fatal); + g_free(portname); + } + + object_property_set_bool(OBJECT(ppc), true, "realized", &error_fatal); + + for (port = 0; port < TZ_NUM_PORTS; port++) { + const PPCPortInfo *pinfo = &ppcinfo->ports[port]; + + if (!pinfo->devfn) { + continue; + } + sysbus_mmio_map(SYS_BUS_DEVICE(ppc), port, pinfo->addr); + + gpioname = g_strdup_printf("%s_nonsec", ppcinfo->name); + qdev_connect_gpio_out_named(ssedev, gpioname, port, + qdev_get_gpio_in_named(ppcdev, + "cfg_nonsec", + port)); + g_free(gpioname); + gpioname = g_strdup_printf("%s_ap", ppcinfo->name); + qdev_connect_gpio_out_named(ssedev, gpioname, port, + qdev_get_gpio_in_named(ppcdev, + "cfg_ap", port)); + g_free(gpioname); + } + + gpioname = g_strdup_printf("%s_irq_enable", ppcinfo->name); + qdev_connect_gpio_out_named(ssedev, gpioname, 0, + qdev_get_gpio_in_named(ppcdev, + "irq_enable", 0)); + g_free(gpioname); + gpioname = g_strdup_printf("%s_irq_clear", ppcinfo->name); + qdev_connect_gpio_out_named(ssedev, gpioname, 0, + qdev_get_gpio_in_named(ppcdev, + "irq_clear", 0)); + g_free(gpioname); + gpioname = g_strdup_printf("%s_irq_status", ppcinfo->name); + qdev_connect_gpio_out_named(ppcdev, "irq", 0, + qdev_get_gpio_in_named(ssedev, + gpioname, 0)); + g_free(gpioname); + + qdev_connect_gpio_out(dev_splitter, i, + qdev_get_gpio_in_named(ppcdev, + "cfg_sec_resp", 0)); + } + armv7m_load_kernel(ARM_CPU(first_cpu), machine->kernel_filename, 0x2000000); } -- 2.20.1 The Musca board puts its SRAM and flash behind TrustZone Memory Protection Controllers (MPCs). Each MPC sits between the CPU and the RAM/flash, and also has a set of memory mapped control registers. Wire up the MPCs, and the memory behind them. For the moment we implement the flash as simple ROM, which cannot be reprogrammed by the guest. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/arm/musca.c | 155 ++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 147 insertions(+), 8 deletions(-) diff --git a/hw/arm/musca.c b/hw/arm/musca.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/musca.c +++ b/hw/arm/musca.c @@ -XXX,XX +XXX,XX @@ #include "hw/arm/armsse.h" #include "hw/boards.h" #include "hw/core/split-irq.h" +#include "hw/misc/tz-mpc.h" #include "hw/misc/tz-ppc.h" #include "hw/misc/unimp.h" #define MUSCA_NUMIRQ_MAX 96 #define MUSCA_PPC_MAX 3 +#define MUSCA_MPC_MAX 5 + +typedef struct MPCInfo MPCInfo; typedef enum MuscaType { MUSCA_A, @@ -XXX,XX +XXX,XX @@ typedef struct { uint32_t init_svtor; int sram_addr_width; int num_irqs; + const MPCInfo *mpc_info; + int num_mpcs; } MuscaMachineClass; typedef struct { MachineState parent; ARMSSE sse; + /* RAM and flash */ + MemoryRegion ram[MUSCA_MPC_MAX]; SplitIRQ cpu_irq_splitter[MUSCA_NUMIRQ_MAX]; SplitIRQ sec_resp_splitter; TZPPC ppc[MUSCA_PPC_MAX]; MemoryRegion container; UnimplementedDeviceState eflash[2]; UnimplementedDeviceState qspi; - UnimplementedDeviceState mpc[5]; + TZMPC mpc[MUSCA_MPC_MAX]; UnimplementedDeviceState mhu[2]; UnimplementedDeviceState pwm[3]; UnimplementedDeviceState i2s; @@ -XXX,XX +XXX,XX @@ typedef struct { UnimplementedDeviceState pvt; UnimplementedDeviceState sdio; UnimplementedDeviceState gpio; + UnimplementedDeviceState cryptoisland; } MuscaMachineState; #define TYPE_MUSCA_MACHINE "musca" @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_unimp_dev(MuscaMachineState *mms, return sysbus_mmio_get_region(SYS_BUS_DEVICE(uds), 0); } +typedef enum MPCInfoType { + MPC_RAM, + MPC_ROM, + MPC_CRYPTOISLAND, +} MPCInfoType; + +struct MPCInfo { + const char *name; + hwaddr addr; + hwaddr size; + MPCInfoType type; +}; + +/* Order of the MPCs here must match the order of the bits in SECMPCINTSTATUS */ +static const MPCInfo a_mpc_info[] = { { + .name = "qspi", + .type = MPC_ROM, + .addr = 0x00200000, + .size = 0x00800000, + }, { + .name = "sram", + .type = MPC_RAM, + .addr = 0x00000000, + .size = 0x00200000, + } +}; + +static const MPCInfo b1_mpc_info[] = { { + .name = "qspi", + .type = MPC_ROM, + .addr = 0x00000000, + .size = 0x02000000, + }, { + .name = "sram", + .type = MPC_RAM, + .addr = 0x0a400000, + .size = 0x00080000, + }, { + .name = "eflash0", + .type = MPC_ROM, + .addr = 0x0a000000, + .size = 0x00200000, + }, { + .name = "eflash1", + .type = MPC_ROM, + .addr = 0x0a200000, + .size = 0x00200000, + }, { + .name = "cryptoisland", + .type = MPC_CRYPTOISLAND, + .addr = 0x0a000000, + .size = 0x00200000, + } +}; + +static MemoryRegion *make_mpc(MuscaMachineState *mms, void *opaque, + const char *name, hwaddr size) +{ + /* + * Create an MPC and the RAM or flash behind it. + * MPC 0: eFlash 0 + * MPC 1: eFlash 1 + * MPC 2: SRAM + * MPC 3: QSPI flash + * MPC 4: CryptoIsland + * For now we implement the flash regions as ROM (ie not programmable) + * (with their control interface memory regions being unimplemented + * stubs behind the PPCs). + * The whole CryptoIsland region behind its MPC is an unimplemented stub. + */ + MuscaMachineClass *mmc = MUSCA_MACHINE_GET_CLASS(mms); + TZMPC *mpc = opaque; + int i = mpc - &mms->mpc[0]; + MemoryRegion *downstream; + MemoryRegion *upstream; + UnimplementedDeviceState *uds; + char *mpcname; + const MPCInfo *mpcinfo = mmc->mpc_info; + + mpcname = g_strdup_printf("%s-mpc", mpcinfo[i].name); + + switch (mpcinfo[i].type) { + case MPC_ROM: + downstream = &mms->ram[i]; + memory_region_init_rom(downstream, NULL, mpcinfo[i].name, + mpcinfo[i].size, &error_fatal); + break; + case MPC_RAM: + downstream = &mms->ram[i]; + memory_region_init_ram(downstream, NULL, mpcinfo[i].name, + mpcinfo[i].size, &error_fatal); + break; + case MPC_CRYPTOISLAND: + /* We don't implement the CryptoIsland yet */ + uds = &mms->cryptoisland; + sysbus_init_child_obj(OBJECT(mms), name, uds, + sizeof(UnimplementedDeviceState), + TYPE_UNIMPLEMENTED_DEVICE); + qdev_prop_set_string(DEVICE(uds), "name", mpcinfo[i].name); + qdev_prop_set_uint64(DEVICE(uds), "size", mpcinfo[i].size); + object_property_set_bool(OBJECT(uds), true, "realized", &error_fatal); + downstream = sysbus_mmio_get_region(SYS_BUS_DEVICE(uds), 0); + break; + default: + g_assert_not_reached(); + } + + sysbus_init_child_obj(OBJECT(mms), mpcname, mpc, sizeof(mms->mpc[0]), + TYPE_TZ_MPC); + object_property_set_link(OBJECT(mpc), OBJECT(downstream), + "downstream", &error_fatal); + object_property_set_bool(OBJECT(mpc), true, "realized", &error_fatal); + /* Map the upstream end of the MPC into system memory */ + upstream = sysbus_mmio_get_region(SYS_BUS_DEVICE(mpc), 1); + memory_region_add_subregion(get_system_memory(), mpcinfo[i].addr, upstream); + /* and connect its interrupt to the SSE-200 */ + qdev_connect_gpio_out_named(DEVICE(mpc), "irq", 0, + qdev_get_gpio_in_named(DEVICE(&mms->sse), + "mpcexp_status", i)); + + g_free(mpcname); + /* Return the register interface MR for our caller to map behind the PPC */ + return sysbus_mmio_get_region(SYS_BUS_DEVICE(mpc), 0); +} + static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, const char *name, hwaddr size) { @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, { "pwm1", make_unimp_dev, &mms->pwm[1], 0xe000, 0x1000 }, { "pwm2", make_unimp_dev, &mms->pwm[2], 0xf000, 0x1000 }, { "gpio", make_unimp_dev, &mms->gpio, 0x10000, 0x1000 }, - { "mpc0", make_unimp_dev, &mms->mpc[0], 0x12000, 0x1000 }, - { "mpc1", make_unimp_dev, &mms->mpc[1], 0x13000, 0x1000 }, + { "mpc0", make_mpc, &mms->mpc[0], 0x12000, 0x1000 }, + { "mpc1", make_mpc, &mms->mpc[1], 0x13000, 0x1000 }, }; memory_region_init(container, OBJECT(mms), "musca-device-container", size); @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) int i; assert(mmc->num_irqs <= MUSCA_NUMIRQ_MAX); + assert(mmc->num_mpcs <= MUSCA_MPC_MAX); if (strcmp(machine->cpu_type, mc->default_cpu_type) != 0) { error_report("This board can only be used with CPU %s", @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) { "eflash1", make_unimp_dev, &mms->eflash[1], 0x52500000, 0x1000 }, { "qspi", make_unimp_dev, &mms->qspi, 0x42800000, 0x100000 }, - { "mpc0", make_unimp_dev, &mms->mpc[0], 0x52000000, 0x1000 }, - { "mpc1", make_unimp_dev, &mms->mpc[1], 0x52100000, 0x1000 }, - { "mpc2", make_unimp_dev, &mms->mpc[2], 0x52200000, 0x1000 }, - { "mpc3", make_unimp_dev, &mms->mpc[3], 0x52300000, 0x1000 }, + { "mpc0", make_mpc, &mms->mpc[0], 0x52000000, 0x1000 }, + { "mpc1", make_mpc, &mms->mpc[1], 0x52100000, 0x1000 }, + { "mpc2", make_mpc, &mms->mpc[2], 0x52200000, 0x1000 }, + { "mpc3", make_mpc, &mms->mpc[3], 0x52300000, 0x1000 }, { "mhu0", make_unimp_dev, &mms->mhu[0], 0x42600000, 0x100000 }, { "mhu1", make_unimp_dev, &mms->mhu[1], 0x42700000, 0x100000 }, { }, /* port 9: unused */ @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) { }, /* port 11: unused */ { }, /* port 12: unused */ { }, /* port 13: unused */ - { "mpc4", make_unimp_dev, &mms->mpc[4], 0x52e00000, 0x1000 }, + { "mpc4", make_mpc, &mms->mpc[4], 0x52e00000, 0x1000 }, }, }, { .name = "apb_ppcexp1", @@ -XXX,XX +XXX,XX @@ static void musca_a_class_init(ObjectClass *oc, void *data) mmc->init_svtor = 0x10200000; mmc->sram_addr_width = 15; mmc->num_irqs = 64; + mmc->mpc_info = a_mpc_info; + mmc->num_mpcs = ARRAY_SIZE(a_mpc_info); } static void musca_b1_class_init(ObjectClass *oc, void *data) @@ -XXX,XX +XXX,XX @@ static void musca_b1_class_init(ObjectClass *oc, void *data) mmc->init_svtor = 0x10000000; mmc->sram_addr_width = 17; mmc->num_irqs = 96; + mmc->mpc_info = b1_mpc_info; + mmc->num_mpcs = ARRAY_SIZE(b1_mpc_info); } static const TypeInfo musca_info = { -- 2.20.1 Wire up the PL031 RTC for the Musca board. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/arm/musca.c | 26 +++++++++++++++++++++++--- 1 file changed, 23 insertions(+), 3 deletions(-) diff --git a/hw/arm/musca.c b/hw/arm/musca.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/musca.c +++ b/hw/arm/musca.c @@ -XXX,XX +XXX,XX @@ #include "hw/misc/tz-mpc.h" #include "hw/misc/tz-ppc.h" #include "hw/misc/unimp.h" +#include "hw/timer/pl031.h" #define MUSCA_NUMIRQ_MAX 96 #define MUSCA_PPC_MAX 3 @@ -XXX,XX +XXX,XX @@ typedef struct { UnimplementedDeviceState spi; UnimplementedDeviceState scc; UnimplementedDeviceState timer; - UnimplementedDeviceState rtc; + PL031State rtc; UnimplementedDeviceState pvt; UnimplementedDeviceState sdio; UnimplementedDeviceState gpio; @@ -XXX,XX +XXX,XX @@ typedef struct { */ #define SYSCLK_FRQ 40000000 +static qemu_irq get_sse_irq_in(MuscaMachineState *mms, int irqno) +{ + /* Return a qemu_irq which will signal IRQ n to all CPUs in the SSE. */ + assert(irqno < MUSCA_NUMIRQ_MAX); + + return qdev_get_gpio_in(DEVICE(&mms->cpu_irq_splitter[irqno]), 0); +} + /* * Most of the devices in the Musca board sit behind Peripheral Protection * Controllers. These data structures define the layout of which devices @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_mpc(MuscaMachineState *mms, void *opaque, return sysbus_mmio_get_region(SYS_BUS_DEVICE(mpc), 0); } +static MemoryRegion *make_rtc(MuscaMachineState *mms, void *opaque, + const char *name, hwaddr size) +{ + PL031State *rtc = opaque; + + sysbus_init_child_obj(OBJECT(mms), name, rtc, sizeof(mms->rtc), TYPE_PL031); + object_property_set_bool(OBJECT(rtc), true, "realized", &error_fatal); + sysbus_connect_irq(SYS_BUS_DEVICE(rtc), 0, get_sse_irq_in(mms, 39)); + return sysbus_mmio_get_region(SYS_BUS_DEVICE(rtc), 0); +} + static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, const char *name, hwaddr size) { @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, { "i2c1", make_unimp_dev, &mms->i2c[1], 0x5000, 0x1000 }, { "i2s", make_unimp_dev, &mms->i2s, 0x6000, 0x1000 }, { "pwm0", make_unimp_dev, &mms->pwm[0], 0x7000, 0x1000 }, - { "rtc", make_unimp_dev, &mms->rtc, 0x8000, 0x1000 }, + { "rtc", make_rtc, &mms->rtc, 0x8000, 0x1000 }, { "qspi", make_unimp_dev, &mms->qspi, 0xa000, 0x1000 }, { "timer", make_unimp_dev, &mms->timer, 0xb000, 0x1000 }, { "scc", make_unimp_dev, &mms->scc, 0xc000, 0x1000 }, @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) { "spi", make_unimp_dev, &mms->spi, 0x4010a000, 0x1000 }, { "scc", make_unimp_dev, &mms->scc, 0x5010b000, 0x1000 }, { "timer", make_unimp_dev, &mms->timer, 0x4010c000, 0x1000 }, - { "rtc", make_unimp_dev, &mms->rtc, 0x4010d000, 0x1000 }, + { "rtc", make_rtc, &mms->rtc, 0x4010d000, 0x1000 }, { "pvt", make_unimp_dev, &mms->pvt, 0x4010e000, 0x1000 }, { "sdio", make_unimp_dev, &mms->sdio, 0x4010f000, 0x1000 }, }, -- 2.20.1 Wire up the two PL011 UARTs in the Musca board. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- hw/arm/musca.c | 34 +++++++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 5 deletions(-) diff --git a/hw/arm/musca.c b/hw/arm/musca.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/musca.c +++ b/hw/arm/musca.c @@ -XXX,XX +XXX,XX @@ #include "qemu/error-report.h" #include "qapi/error.h" #include "exec/address-spaces.h" +#include "sysemu/sysemu.h" #include "hw/arm/arm.h" #include "hw/arm/armsse.h" #include "hw/boards.h" +#include "hw/char/pl011.h" #include "hw/core/split-irq.h" #include "hw/misc/tz-mpc.h" #include "hw/misc/tz-ppc.h" @@ -XXX,XX +XXX,XX @@ typedef struct { UnimplementedDeviceState mhu[2]; UnimplementedDeviceState pwm[3]; UnimplementedDeviceState i2s; - UnimplementedDeviceState uart[2]; + PL011State uart[2]; UnimplementedDeviceState i2c[2]; UnimplementedDeviceState spi; UnimplementedDeviceState scc; @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_rtc(MuscaMachineState *mms, void *opaque, return sysbus_mmio_get_region(SYS_BUS_DEVICE(rtc), 0); } +static MemoryRegion *make_uart(MuscaMachineState *mms, void *opaque, + const char *name, hwaddr size) +{ + PL011State *uart = opaque; + int i = uart - &mms->uart[0]; + int irqbase = 7 + i * 6; + SysBusDevice *s; + + sysbus_init_child_obj(OBJECT(mms), name, uart, sizeof(mms->uart[0]), + TYPE_PL011); + qdev_prop_set_chr(DEVICE(uart), "chardev", serial_hd(i)); + object_property_set_bool(OBJECT(uart), true, "realized", &error_fatal); + s = SYS_BUS_DEVICE(uart); + sysbus_connect_irq(s, 0, get_sse_irq_in(mms, irqbase + 5)); /* combined */ + sysbus_connect_irq(s, 1, get_sse_irq_in(mms, irqbase + 0)); /* RX */ + sysbus_connect_irq(s, 2, get_sse_irq_in(mms, irqbase + 1)); /* TX */ + sysbus_connect_irq(s, 3, get_sse_irq_in(mms, irqbase + 2)); /* RT */ + sysbus_connect_irq(s, 4, get_sse_irq_in(mms, irqbase + 3)); /* MS */ + sysbus_connect_irq(s, 5, get_sse_irq_in(mms, irqbase + 4)); /* E */ + return sysbus_mmio_get_region(SYS_BUS_DEVICE(uart), 0); +} + static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, const char *name, hwaddr size) { @@ -XXX,XX +XXX,XX @@ static MemoryRegion *make_musca_a_devs(MuscaMachineState *mms, void *opaque, MemoryRegion *container = &mms->container; const PPCPortInfo devices[] = { - { "uart0", make_unimp_dev, &mms->uart[0], 0x1000, 0x1000 }, - { "uart1", make_unimp_dev, &mms->uart[1], 0x2000, 0x1000 }, + { "uart0", make_uart, &mms->uart[0], 0x1000, 0x1000 }, + { "uart1", make_uart, &mms->uart[1], 0x2000, 0x1000 }, { "spi", make_unimp_dev, &mms->spi, 0x3000, 0x1000 }, { "i2c0", make_unimp_dev, &mms->i2c[0], 0x4000, 0x1000 }, { "i2c1", make_unimp_dev, &mms->i2c[1], 0x5000, 0x1000 }, @@ -XXX,XX +XXX,XX @@ static void musca_init(MachineState *machine) { "pwm1", make_unimp_dev, &mms->pwm[1], 0x40102000, 0x1000 }, { "pwm2", make_unimp_dev, &mms->pwm[2], 0x40103000, 0x1000 }, { "i2s", make_unimp_dev, &mms->i2s, 0x40104000, 0x1000 }, - { "uart0", make_unimp_dev, &mms->uart[0], 0x40105000, 0x1000 }, - { "uart1", make_unimp_dev, &mms->uart[1], 0x40106000, 0x1000 }, + { "uart0", make_uart, &mms->uart[0], 0x40105000, 0x1000 }, + { "uart1", make_uart, &mms->uart[1], 0x40106000, 0x1000 }, { "i2c0", make_unimp_dev, &mms->i2c[0], 0x40108000, 0x1000 }, { "i2c1", make_unimp_dev, &mms->i2c[1], 0x40109000, 0x1000 }, { "spi", make_unimp_dev, &mms->spi, 0x4010a000, 0x1000 }, -- 2.20.1 The region 0x40010000 .. 0x4001ffff and its secure-only alias at 0x50010000... are for per-CPU devices. We implement this by giving each CPU its own container memory region, where the per-CPU devices live. Unfortunately, the alias region which makes devices mapped at 0x4... addresses also appear at 0x5... is only implemented in the overall "all CPUs" container. The effect of this bug is that the CPU_IDENTITY register block appears only at 0x4001f000, but not at the 0x5001f000 alias where it should also appear. Guests (like very recent Arm Trusted Firmware-M) which try to access it at 0x5001f000 will crash. Fix this by moving the handling for this alias from the "all CPUs" container to the per-CPU container. (We leave the aliases for 0x1... and 0x3... in the overall container, because there are no per-CPU devices there.) Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Message-id: 20190215180500.6906-1-peter.maydell@linaro.org Reviewed-by: Alex Bennée <alex.bennee@linaro.org> --- include/hw/arm/armsse.h | 2 +- hw/arm/armsse.c | 26 ++++++++++++++++---------- 2 files changed, 17 insertions(+), 11 deletions(-) diff --git a/include/hw/arm/armsse.h b/include/hw/arm/armsse.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/arm/armsse.h +++ b/include/hw/arm/armsse.h @@ -XXX,XX +XXX,XX @@ typedef struct ARMSSE { MemoryRegion cpu_container[SSE_MAX_CPUS]; MemoryRegion alias1; MemoryRegion alias2; - MemoryRegion alias3; + MemoryRegion alias3[SSE_MAX_CPUS]; MemoryRegion sram[MAX_SRAM_BANKS]; qemu_irq *exp_irqs[SSE_MAX_CPUS]; diff --git a/hw/arm/armsse.c b/hw/arm/armsse.c index XXXXXXX..XXXXXXX 100644 --- a/hw/arm/armsse.c +++ b/hw/arm/armsse.c @@ -XXX,XX +XXX,XX @@ static bool irq_is_common[32] = { /* 30, 31: reserved */ }; -/* Create an alias region of @size bytes starting at @base +/* + * Create an alias region in @container of @size bytes starting at @base * which mirrors the memory starting at @orig. */ -static void make_alias(ARMSSE *s, MemoryRegion *mr, const char *name, - hwaddr base, hwaddr size, hwaddr orig) +static void make_alias(ARMSSE *s, MemoryRegion *mr, MemoryRegion *container, + const char *name, hwaddr base, hwaddr size, hwaddr orig) { - memory_region_init_alias(mr, NULL, name, &s->container, orig, size); + memory_region_init_alias(mr, NULL, name, container, orig, size); /* The alias is even lower priority than unimplemented_device regions */ - memory_region_add_subregion_overlap(&s->container, base, mr, -1500); + memory_region_add_subregion_overlap(container, base, mr, -1500); } static void irq_status_forwarder(void *opaque, int n, int level) @@ -XXX,XX +XXX,XX @@ static void armsse_realize(DeviceState *dev, Error **errp) } /* Set up the big aliases first */ - make_alias(s, &s->alias1, "alias 1", 0x10000000, 0x10000000, 0x00000000); - make_alias(s, &s->alias2, "alias 2", 0x30000000, 0x10000000, 0x20000000); + make_alias(s, &s->alias1, &s->container, "alias 1", + 0x10000000, 0x10000000, 0x00000000); + make_alias(s, &s->alias2, &s->container, + "alias 2", 0x30000000, 0x10000000, 0x20000000); /* The 0x50000000..0x5fffffff region is not a pure alias: it has * a few extra devices that only appear there (generally the * control interfaces for the protection controllers). * We implement this by mapping those devices over the top of this - * alias MR at a higher priority. + * alias MR at a higher priority. Some of the devices in this range + * are per-CPU, so we must put this alias in the per-cpu containers. */ - make_alias(s, &s->alias3, "alias 3", 0x50000000, 0x10000000, 0x40000000); - + for (i = 0; i < info->num_cpus; i++) { + make_alias(s, &s->alias3[i], &s->cpu_container[i], + "alias 3", 0x50000000, 0x10000000, 0x40000000); + } /* Security controller */ object_property_set_bool(OBJECT(&s->secctl), true, "realized", &err); -- 2.20.1 I don't have anything else queued up at the moment, so this is just Richard's SME patches. -- PMM The following changes since commit 63b38f6c85acd312c2cab68554abf33adf4ee2b3: Merge tag 'pull-target-arm-20220707' of https://git.linaro.org/people/pmaydell/qemu-arm into staging (2022-07-08 06:17:11 +0530) are available in the Git repository at: https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20220711 for you to fetch changes up to f9982ceaf26df27d15547a3a7990a95019e9e3a8: linux-user/aarch64: Add SME related hwcap entries (2022-07-11 13:43:52 +0100) ---------------------------------------------------------------- target-arm: * Implement SME emulation, for both system and linux-user ---------------------------------------------------------------- Richard Henderson (45): target/arm: Handle SME in aarch64_cpu_dump_state target/arm: Add infrastructure for disas_sme target/arm: Trap non-streaming usage when Streaming SVE is active target/arm: Mark ADR as non-streaming target/arm: Mark RDFFR, WRFFR, SETFFR as non-streaming target/arm: Mark BDEP, BEXT, BGRP, COMPACT, FEXPA, FTSSEL as non-streaming target/arm: Mark PMULL, FMMLA as non-streaming target/arm: Mark FTSMUL, FTMAD, FADDA as non-streaming target/arm: Mark SMMLA, UMMLA, USMMLA as non-streaming target/arm: Mark string/histo/crypto as non-streaming target/arm: Mark gather/scatter load/store as non-streaming target/arm: Mark gather prefetch as non-streaming target/arm: Mark LDFF1 and LDNF1 as non-streaming target/arm: Mark LD1RO as non-streaming target/arm: Add SME enablement checks target/arm: Handle SME in sve_access_check target/arm: Implement SME RDSVL, ADDSVL, ADDSPL target/arm: Implement SME ZERO target/arm: Implement SME MOVA target/arm: Implement SME LD1, ST1 target/arm: Export unpredicated ld/st from translate-sve.c target/arm: Implement SME LDR, STR target/arm: Implement SME ADDHA, ADDVA target/arm: Implement FMOPA, FMOPS (non-widening) target/arm: Implement BFMOPA, BFMOPS target/arm: Implement FMOPA, FMOPS (widening) target/arm: Implement SME integer outer product target/arm: Implement PSEL target/arm: Implement REVD target/arm: Implement SCLAMP, UCLAMP target/arm: Reset streaming sve state on exception boundaries target/arm: Enable SME for -cpu max linux-user/aarch64: Clear tpidr2_el0 if CLONE_SETTLS linux-user/aarch64: Reset PSTATE.SM on syscalls linux-user/aarch64: Add SM bit to SVE signal context linux-user/aarch64: Tidy target_restore_sigframe error return linux-user/aarch64: Do not allow duplicate or short sve records linux-user/aarch64: Verify extra record lock succeeded linux-user/aarch64: Move sve record checks into restore linux-user/aarch64: Implement SME signal handling linux-user: Rename sve prctls linux-user/aarch64: Implement PR_SME_GET_VL, PR_SME_SET_VL target/arm: Only set ZEN in reset if SVE present target/arm: Enable SME for user-only linux-user/aarch64: Add SME related hwcap entries docs/system/arm/emulation.rst | 4 + linux-user/aarch64/target_cpu.h | 5 +- linux-user/aarch64/target_prctl.h | 62 +- target/arm/cpu.h | 7 + target/arm/helper-sme.h | 126 ++++ target/arm/helper-sve.h | 4 + target/arm/helper.h | 18 + target/arm/translate-a64.h | 45 ++ target/arm/translate.h | 16 + target/arm/sme-fa64.decode | 60 ++ target/arm/sme.decode | 88 +++ target/arm/sve.decode | 41 +- linux-user/aarch64/cpu_loop.c | 9 + linux-user/aarch64/signal.c | 243 ++++++-- linux-user/elfload.c | 20 + linux-user/syscall.c | 28 +- target/arm/cpu.c | 35 +- target/arm/cpu64.c | 11 + target/arm/helper.c | 56 +- target/arm/sme_helper.c | 1140 +++++++++++++++++++++++++++++++++++++ target/arm/sve_helper.c | 28 + target/arm/translate-a64.c | 103 +++- target/arm/translate-sme.c | 373 ++++++++++++ target/arm/translate-sve.c | 393 ++++++++++--- target/arm/translate-vfp.c | 12 + target/arm/translate.c | 2 + target/arm/vec_helper.c | 24 + target/arm/meson.build | 3 + 28 files changed, 2821 insertions(+), 135 deletions(-) create mode 100644 target/arm/sme-fa64.decode create mode 100644 target/arm/sme.decode create mode 100644 target/arm/translate-sme.c From: Richard Henderson <richard.henderson@linaro.org> Dump SVCR, plus use the correct access check for Streaming Mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-2-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) int i; int el = arm_current_el(env); const char *ns_status; + bool sve; qemu_fprintf(f, " PC=%016" PRIx64 " ", env->pc); for (i = 0; i < 32; i++) { @@ -XXX,XX +XXX,XX @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) el, psr & PSTATE_SP ? 'h' : 't'); + if (cpu_isar_feature(aa64_sme, cpu)) { + qemu_fprintf(f, " SVCR=%08" PRIx64 " %c%c", + env->svcr, + (FIELD_EX64(env->svcr, SVCR, ZA) ? 'Z' : '-'), + (FIELD_EX64(env->svcr, SVCR, SM) ? 'S' : '-')); + } if (cpu_isar_feature(aa64_bti, cpu)) { qemu_fprintf(f, " BTYPE=%d", (psr & PSTATE_BTYPE) >> 10); } @@ -XXX,XX +XXX,XX @@ static void aarch64_cpu_dump_state(CPUState *cs, FILE *f, int flags) qemu_fprintf(f, " FPCR=%08x FPSR=%08x\n", vfp_get_fpcr(env), vfp_get_fpsr(env)); - if (cpu_isar_feature(aa64_sve, cpu) && sve_exception_el(env, el) == 0) { + if (cpu_isar_feature(aa64_sme, cpu) && FIELD_EX64(env->svcr, SVCR, SM)) { + sve = sme_exception_el(env, el) == 0; + } else if (cpu_isar_feature(aa64_sve, cpu)) { + sve = sve_exception_el(env, el) == 0; + } else { + sve = false; + } + + if (sve) { int j, zcr_len = sve_vqm1_for_el(env, el); for (i = 0; i <= FFR_PRED_NUM; i++) { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This includes the build rules for the decoder, and the new file for translation, but excludes any instructions. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-3-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 1 + target/arm/sme.decode | 20 ++++++++++++++++++++ target/arm/translate-a64.c | 7 ++++++- target/arm/translate-sme.c | 35 +++++++++++++++++++++++++++++++++++ target/arm/meson.build | 2 ++ 5 files changed, 64 insertions(+), 1 deletion(-) create mode 100644 target/arm/sme.decode create mode 100644 target/arm/translate-sme.c diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline int pred_gvec_reg_size(DisasContext *s) } bool disas_sve(DisasContext *, uint32_t); +bool disas_sme(DisasContext *, uint32_t); void gen_gvec_rax1(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, uint32_t rm_ofs, uint32_t opr_sz, uint32_t max_sz); diff --git a/target/arm/sme.decode b/target/arm/sme.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ +# AArch64 SME instruction descriptions +# +# Copyright (c) 2022 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/>. + +# +# This file is processed by scripts/decodetree.py +# diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) } switch (extract32(insn, 25, 4)) { - case 0x0: case 0x1: case 0x3: /* UNALLOCATED */ + case 0x0: + if (!extract32(insn, 31, 1) || !disas_sme(s, insn)) { + unallocated_encoding(s); + } + break; + case 0x1: case 0x3: /* UNALLOCATED */ unallocated_encoding(s); break; case 0x2: diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ +/* + * AArch64 SME translation + * + * Copyright (c) 2022 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 "tcg/tcg-op.h" +#include "tcg/tcg-op-gvec.h" +#include "tcg/tcg-gvec-desc.h" +#include "translate.h" +#include "exec/helper-gen.h" +#include "translate-a64.h" +#include "fpu/softfloat.h" + + +/* + * Include the generated decoder. + */ + +#include "decode-sme.c.inc" 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 @@ gen = [ decodetree.process('sve.decode', extra_args: '--decode=disas_sve'), + decodetree.process('sme.decode', extra_args: '--decode=disas_sme'), decodetree.process('neon-shared.decode', extra_args: '--decode=disas_neon_shared'), decodetree.process('neon-dp.decode', extra_args: '--decode=disas_neon_dp'), decodetree.process('neon-ls.decode', extra_args: '--decode=disas_neon_ls'), @@ -XXX,XX +XXX,XX @@ arm_ss.add(when: 'TARGET_AARCH64', if_true: files( 'sme_helper.c', 'translate-a64.c', 'translate-sve.c', + 'translate-sme.c', )) arm_softmmu_ss = ss.source_set() -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This new behaviour is in the ARM pseudocode function AArch64.CheckFPAdvSIMDEnabled, which applies to AArch32 via AArch32.CheckAdvSIMDOrFPEnabled when the EL to which the trap would be delivered is in AArch64 mode. Given that ARMv9 drops support for AArch32 outside EL0, the trap EL detection ought to be trivially true, but the pseudocode still contains a number of conditions, and QEMU has not yet committed to dropping A32 support for EL[12] when v9 features are present. Since the computation of SME_TRAP_NONSTREAMING is necessarily different for the two modes, we might as well preserve bits within TBFLAG_ANY and allocate separate bits within TBFLAG_A32 and TBFLAG_A64 instead. Note that DDI0616A.a has typos for bits [22:21] of LD1RO in the table of instructions illegal in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-4-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.h | 7 +++ target/arm/translate.h | 4 ++ target/arm/sme-fa64.decode | 90 ++++++++++++++++++++++++++++++++++++++ target/arm/helper.c | 41 +++++++++++++++++ target/arm/translate-a64.c | 40 ++++++++++++++++- target/arm/translate-vfp.c | 12 +++++ target/arm/translate.c | 2 + target/arm/meson.build | 1 + 8 files changed, 195 insertions(+), 2 deletions(-) create mode 100644 target/arm/sme-fa64.decode diff --git a/target/arm/cpu.h b/target/arm/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.h +++ b/target/arm/cpu.h @@ -XXX,XX +XXX,XX @@ FIELD(TBFLAG_A32, HSTR_ACTIVE, 9, 1) * the same thing as the current security state of the processor! */ FIELD(TBFLAG_A32, NS, 10, 1) +/* + * Indicates that SME Streaming mode is active, and SMCR_ELx.FA64 is not. + * This requires an SME trap from AArch32 mode when using NEON. + */ +FIELD(TBFLAG_A32, SME_TRAP_NONSTREAMING, 11, 1) /* * Bit usage when in AArch32 state, for M-profile only. @@ -XXX,XX +XXX,XX @@ FIELD(TBFLAG_A64, SMEEXC_EL, 20, 2) FIELD(TBFLAG_A64, PSTATE_SM, 22, 1) FIELD(TBFLAG_A64, PSTATE_ZA, 23, 1) FIELD(TBFLAG_A64, SVL, 24, 4) +/* Indicates that SME Streaming mode is active, and SMCR_ELx.FA64 is not. */ +FIELD(TBFLAG_A64, SME_TRAP_NONSTREAMING, 28, 1) /* * Helpers for using the above. diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ typedef struct DisasContext { bool pstate_sm; /* True if PSTATE.ZA is set. */ bool pstate_za; + /* True if non-streaming insns should raise an SME Streaming exception. */ + bool sme_trap_nonstreaming; + /* True if the current instruction is non-streaming. */ + bool is_nonstreaming; /* True if MVE insns are definitely not predicated by VPR or LTPSIZE */ bool mve_no_pred; /* diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ +# AArch64 SME allowed instruction decoding +# +# Copyright (c) 2022 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/>. + +# +# This file is processed by scripts/decodetree.py +# + +# These patterns are taken from Appendix E1.1 of DDI0616 A.a, +# Arm Architecture Reference Manual Supplement, +# The Scalable Matrix Extension (SME), for Armv9-A + +{ + [ + OK 0-00 1110 0000 0001 0010 11-- ---- ---- # SMOV W|Xd,Vn.B[0] + OK 0-00 1110 0000 0010 0010 11-- ---- ---- # SMOV W|Xd,Vn.H[0] + OK 0100 1110 0000 0100 0010 11-- ---- ---- # SMOV Xd,Vn.S[0] + OK 0000 1110 0000 0001 0011 11-- ---- ---- # UMOV Wd,Vn.B[0] + OK 0000 1110 0000 0010 0011 11-- ---- ---- # UMOV Wd,Vn.H[0] + OK 0000 1110 0000 0100 0011 11-- ---- ---- # UMOV Wd,Vn.S[0] + OK 0100 1110 0000 1000 0011 11-- ---- ---- # UMOV Xd,Vn.D[0] + ] + FAIL 0--0 111- ---- ---- ---- ---- ---- ---- # Advanced SIMD vector operations +} + +{ + [ + OK 0101 1110 --1- ---- 11-1 11-- ---- ---- # FMULX/FRECPS/FRSQRTS (scalar) + OK 0101 1110 -10- ---- 00-1 11-- ---- ---- # FMULX/FRECPS/FRSQRTS (scalar, FP16) + OK 01-1 1110 1-10 0001 11-1 10-- ---- ---- # FRECPE/FRSQRTE/FRECPX (scalar) + OK 01-1 1110 1111 1001 11-1 10-- ---- ---- # FRECPE/FRSQRTE/FRECPX (scalar, FP16) + ] + FAIL 01-1 111- ---- ---- ---- ---- ---- ---- # Advanced SIMD single-element operations +} + +FAIL 0-00 110- ---- ---- ---- ---- ---- ---- # Advanced SIMD structure load/store +FAIL 1100 1110 ---- ---- ---- ---- ---- ---- # Advanced SIMD cryptography extensions +FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS + +# These are the "avoidance of doubt" final table of Illegal Advanced SIMD instructions +# We don't actually need to include these, as the default is OK. +# -001 111- ---- ---- ---- ---- ---- ---- # Scalar floating-point operations +# --10 110- ---- ---- ---- ---- ---- ---- # Load/store pair of FP registers +# --01 1100 ---- ---- ---- ---- ---- ---- # Load FP register (PC-relative literal) +# --11 1100 --0- ---- ---- ---- ---- ---- # Load/store FP register (unscaled imm) +# --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) +# --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) + +FAIL 0000 0100 --1- ---- 1010 ---- ---- ---- # ADR +FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA +FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT +FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS +FAIL 0010 0101 --10 1--- 1001 ---- ---- ---- # WRFFR, SETFFR +FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP +FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) +FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA +FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL +FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD +FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA +FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA +FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions +FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) +FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) +FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) +FAIL 1000 010- -01- ---- 1--- ---- ---- ---- # SVE 32-bit gather load (vector+imm) +FAIL 1000 0100 0-0- ---- 0--- ---- ---- ---- # SVE 32-bit gather load byte (scalar+vector) +FAIL 1000 0100 1--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load half (scalar+vector) +FAIL 1000 0101 0--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load word (scalar+vector) +FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) +FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) +FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) +FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) +FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch +FAIL 1110 010- -00- ---- 001- ---- ---- ---- # SVE2 64-bit scatter NT store (vector+scalar) +FAIL 1110 010- -10- ---- 001- ---- ---- ---- # SVE2 32-bit scatter NT store (vector+scalar) +FAIL 1110 010- ---- ---- 1-0- ---- ---- ---- # SVE scatter store (scalar+32-bit vector) +FAIL 1110 010- ---- ---- 101- ---- ---- ---- # SVE scatter store (misc) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ int sme_exception_el(CPUARMState *env, int el) return 0; } +/* This corresponds to the ARM pseudocode function IsFullA64Enabled(). */ +static bool sme_fa64(CPUARMState *env, int el) +{ + if (!cpu_isar_feature(aa64_sme_fa64, env_archcpu(env))) { + return false; + } + + if (el <= 1 && !el_is_in_host(env, el)) { + if (!FIELD_EX64(env->vfp.smcr_el[1], SMCR, FA64)) { + return false; + } + } + if (el <= 2 && arm_is_el2_enabled(env)) { + if (!FIELD_EX64(env->vfp.smcr_el[2], SMCR, FA64)) { + return false; + } + } + if (arm_feature(env, ARM_FEATURE_EL3)) { + if (!FIELD_EX64(env->vfp.smcr_el[3], SMCR, FA64)) { + return false; + } + } + + return true; +} + /* * Given that SVE is enabled, return the vector length for EL. */ @@ -XXX,XX +XXX,XX @@ static CPUARMTBFlags rebuild_hflags_a32(CPUARMState *env, int fp_el, DP_TBFLAG_ANY(flags, PSTATE__IL, 1); } + /* + * The SME exception we are testing for is raised via + * AArch64.CheckFPAdvSIMDEnabled(), as called from + * AArch32.CheckAdvSIMDOrFPEnabled(). + */ + if (el == 0 + && FIELD_EX64(env->svcr, SVCR, SM) + && (!arm_is_el2_enabled(env) + || (arm_el_is_aa64(env, 2) && !(env->cp15.hcr_el2 & HCR_TGE))) + && arm_el_is_aa64(env, 1) + && !sme_fa64(env, el)) { + DP_TBFLAG_A32(flags, SME_TRAP_NONSTREAMING, 1); + } + return rebuild_hflags_common_32(env, fp_el, mmu_idx, flags); } @@ -XXX,XX +XXX,XX @@ static CPUARMTBFlags rebuild_hflags_a64(CPUARMState *env, int el, int fp_el, } if (FIELD_EX64(env->svcr, SVCR, SM)) { DP_TBFLAG_A64(flags, PSTATE_SM, 1); + DP_TBFLAG_A64(flags, SME_TRAP_NONSTREAMING, !sme_fa64(env, el)); } DP_TBFLAG_A64(flags, PSTATE_ZA, FIELD_EX64(env->svcr, SVCR, ZA)); } diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static void do_vec_ld(DisasContext *s, int destidx, int element, * unallocated-encoding checks (otherwise the syndrome information * for the resulting exception will be incorrect). */ -static bool fp_access_check(DisasContext *s) +static bool fp_access_check_only(DisasContext *s) { if (s->fp_excp_el) { assert(!s->fp_access_checked); @@ -XXX,XX +XXX,XX @@ static bool fp_access_check(DisasContext *s) return true; } +static bool fp_access_check(DisasContext *s) +{ + if (!fp_access_check_only(s)) { + return false; + } + if (s->sme_trap_nonstreaming && s->is_nonstreaming) { + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_smetrap(SME_ET_Streaming, false)); + return false; + } + return true; +} + /* Check that SVE access is enabled. If it is, return true. * If not, emit code to generate an appropriate exception and return false. */ @@ -XXX,XX +XXX,XX @@ static void handle_sys(DisasContext *s, uint32_t insn, bool isread, default: g_assert_not_reached(); } - if ((ri->type & ARM_CP_FPU) && !fp_access_check(s)) { + if ((ri->type & ARM_CP_FPU) && !fp_access_check_only(s)) { return; } else if ((ri->type & ARM_CP_SVE) && !sve_access_check(s)) { return; @@ -XXX,XX +XXX,XX @@ static void disas_data_proc_simd_fp(DisasContext *s, uint32_t insn) } } +/* + * Include the generated SME FA64 decoder. + */ + +#include "decode-sme-fa64.c.inc" + +static bool trans_OK(DisasContext *s, arg_OK *a) +{ + return true; +} + +static bool trans_FAIL(DisasContext *s, arg_OK *a) +{ + s->is_nonstreaming = true; + return true; +} + /** * is_guarded_page: * @env: The cpu environment @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_init_disas_context(DisasContextBase *dcbase, dc->mte_active[1] = EX_TBFLAG_A64(tb_flags, MTE0_ACTIVE); dc->pstate_sm = EX_TBFLAG_A64(tb_flags, PSTATE_SM); dc->pstate_za = EX_TBFLAG_A64(tb_flags, PSTATE_ZA); + dc->sme_trap_nonstreaming = EX_TBFLAG_A64(tb_flags, SME_TRAP_NONSTREAMING); dc->vec_len = 0; dc->vec_stride = 0; dc->cp_regs = arm_cpu->cp_regs; @@ -XXX,XX +XXX,XX @@ static void aarch64_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu) } } + s->is_nonstreaming = false; + if (s->sme_trap_nonstreaming) { + disas_sme_fa64(s, insn); + } + switch (extract32(insn, 25, 4)) { case 0x0: if (!extract32(insn, 31, 1) || !disas_sme(s, insn)) { 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 vfp_access_check_a(DisasContext *s, bool ignore_vfp_enabled) return false; } + /* + * Note that rebuild_hflags_a32 has already accounted for being in EL0 + * and the higher EL in A64 mode, etc. Unlike A64 mode, there do not + * appear to be any insns which touch VFP which are allowed. + */ + if (s->sme_trap_nonstreaming) { + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_smetrap(SME_ET_Streaming, + s->base.pc_next - s->pc_curr == 2)); + return false; + } + if (!s->vfp_enabled && !ignore_vfp_enabled) { assert(!arm_dc_feature(s, ARM_FEATURE_M)); unallocated_encoding(s); diff --git a/target/arm/translate.c b/target/arm/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -XXX,XX +XXX,XX @@ static void arm_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs) dc->vec_len = EX_TBFLAG_A32(tb_flags, VECLEN); dc->vec_stride = EX_TBFLAG_A32(tb_flags, VECSTRIDE); } + dc->sme_trap_nonstreaming = + EX_TBFLAG_A32(tb_flags, SME_TRAP_NONSTREAMING); } dc->cp_regs = cpu->cp_regs; dc->features = env->features; 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 @@ gen = [ decodetree.process('sve.decode', extra_args: '--decode=disas_sve'), decodetree.process('sme.decode', extra_args: '--decode=disas_sme'), + decodetree.process('sme-fa64.decode', extra_args: '--static-decode=disas_sme_fa64'), decodetree.process('neon-shared.decode', extra_args: '--decode=disas_neon_shared'), decodetree.process('neon-dp.decode', extra_args: '--decode=disas_neon_dp'), decodetree.process('neon-ls.decode', extra_args: '--decode=disas_neon_ls'), -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark ADR as a non-streaming instruction, which should trap if full a64 support is not enabled in streaming mode. Removing entries from sme-fa64.decode is an easy way to see what remains to be done. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-5-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate.h | 7 +++++++ target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 8 ++++---- 3 files changed, 11 insertions(+), 5 deletions(-) diff --git a/target/arm/translate.h b/target/arm/translate.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate.h +++ b/target/arm/translate.h @@ -XXX,XX +XXX,XX @@ uint64_t asimd_imm_const(uint32_t imm, int cmode, int op); static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \ { return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__); } +#define TRANS_FEAT_NONSTREAMING(NAME, FEAT, FUNC, ...) \ + static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \ + { \ + s->is_nonstreaming = true; \ + return dc_isar_feature(FEAT, s) && FUNC(s, __VA_ARGS__); \ + } + #endif /* TARGET_ARM_TRANSLATE_H */ diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0000 0100 --1- ---- 1010 ---- ---- ---- # ADR FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_adr(DisasContext *s, arg_rrri *a, gen_helper_gvec_3 *fn) return gen_gvec_ool_zzz(s, fn, a->rd, a->rn, a->rm, a->imm); } -TRANS_FEAT(ADR_p32, aa64_sve, do_adr, a, gen_helper_sve_adr_p32) -TRANS_FEAT(ADR_p64, aa64_sve, do_adr, a, gen_helper_sve_adr_p64) -TRANS_FEAT(ADR_s32, aa64_sve, do_adr, a, gen_helper_sve_adr_s32) -TRANS_FEAT(ADR_u32, aa64_sve, do_adr, a, gen_helper_sve_adr_u32) +TRANS_FEAT_NONSTREAMING(ADR_p32, aa64_sve, do_adr, a, gen_helper_sve_adr_p32) +TRANS_FEAT_NONSTREAMING(ADR_p64, aa64_sve, do_adr, a, gen_helper_sve_adr_p64) +TRANS_FEAT_NONSTREAMING(ADR_s32, aa64_sve, do_adr, a, gen_helper_sve_adr_s32) +TRANS_FEAT_NONSTREAMING(ADR_u32, aa64_sve, do_adr, a, gen_helper_sve_adr_u32) /* *** SVE Integer Misc - Unpredicated Group -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-6-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 9 ++++++--- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT -FAIL 0010 0101 --01 100- 1111 000- ---0 ---- # RDFFR, RDFFRS -FAIL 0010 0101 --10 1--- 1001 ---- ---- ---- # WRFFR, SETFFR FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_predset(DisasContext *s, int esz, int rd, int pat, bool setflag) TRANS_FEAT(PTRUE, aa64_sve, do_predset, a->esz, a->rd, a->pat, a->s) /* Note pat == 31 is #all, to set all elements. */ -TRANS_FEAT(SETFFR, aa64_sve, do_predset, 0, FFR_PRED_NUM, 31, false) +TRANS_FEAT_NONSTREAMING(SETFFR, aa64_sve, + do_predset, 0, FFR_PRED_NUM, 31, false) /* Note pat == 32 is #unimp, to set no elements. */ TRANS_FEAT(PFALSE, aa64_sve, do_predset, 0, a->rd, 32, false) @@ -XXX,XX +XXX,XX @@ static bool trans_RDFFR_p(DisasContext *s, arg_RDFFR_p *a) .rd = a->rd, .pg = a->pg, .s = a->s, .rn = FFR_PRED_NUM, .rm = FFR_PRED_NUM, }; + + s->is_nonstreaming = true; return trans_AND_pppp(s, &alt_a); } -TRANS_FEAT(RDFFR, aa64_sve, do_mov_p, a->rd, FFR_PRED_NUM) -TRANS_FEAT(WRFFR, aa64_sve, do_mov_p, FFR_PRED_NUM, a->rn) +TRANS_FEAT_NONSTREAMING(RDFFR, aa64_sve, do_mov_p, a->rd, FFR_PRED_NUM) +TRANS_FEAT_NONSTREAMING(WRFFR, aa64_sve, do_mov_p, FFR_PRED_NUM, a->rn) static bool do_pfirst_pnext(DisasContext *s, arg_rr_esz *a, void (*gen_fn)(TCGv_i32, TCGv_ptr, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-7-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 22 ++++++++++++---------- 2 files changed, 12 insertions(+), 13 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0000 0100 --1- ---- 1011 -0-- ---- ---- # FTSSEL, FEXPA -FAIL 0000 0101 --10 0001 100- ---- ---- ---- # COMPACT -FAIL 0100 0101 --0- ---- 1011 ---- ---- ---- # BDEP, BEXT, BGRP FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_2 * const fexpa_fns[4] = { NULL, gen_helper_sve_fexpa_h, gen_helper_sve_fexpa_s, gen_helper_sve_fexpa_d, }; -TRANS_FEAT(FEXPA, aa64_sve, gen_gvec_ool_zz, - fexpa_fns[a->esz], a->rd, a->rn, 0) +TRANS_FEAT_NONSTREAMING(FEXPA, aa64_sve, gen_gvec_ool_zz, + fexpa_fns[a->esz], a->rd, a->rn, 0) static gen_helper_gvec_3 * const ftssel_fns[4] = { NULL, gen_helper_sve_ftssel_h, gen_helper_sve_ftssel_s, gen_helper_sve_ftssel_d, }; -TRANS_FEAT(FTSSEL, aa64_sve, gen_gvec_ool_arg_zzz, ftssel_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(FTSSEL, aa64_sve, gen_gvec_ool_arg_zzz, + ftssel_fns[a->esz], a, 0) /* *** SVE Predicate Logical Operations Group @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(TRN2_q, aa64_sve_f64mm, gen_gvec_ool_arg_zzz, static gen_helper_gvec_3 * const compact_fns[4] = { NULL, NULL, gen_helper_sve_compact_s, gen_helper_sve_compact_d }; -TRANS_FEAT(COMPACT, aa64_sve, gen_gvec_ool_arg_zpz, compact_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(COMPACT, aa64_sve, gen_gvec_ool_arg_zpz, + compact_fns[a->esz], a, 0) /* Call the helper that computes the ARM LastActiveElement pseudocode * function, scaled by the element size. This includes the not found @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_3 * const bext_fns[4] = { gen_helper_sve2_bext_b, gen_helper_sve2_bext_h, gen_helper_sve2_bext_s, gen_helper_sve2_bext_d, }; -TRANS_FEAT(BEXT, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bext_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BEXT, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bext_fns[a->esz], a, 0) static gen_helper_gvec_3 * const bdep_fns[4] = { gen_helper_sve2_bdep_b, gen_helper_sve2_bdep_h, gen_helper_sve2_bdep_s, gen_helper_sve2_bdep_d, }; -TRANS_FEAT(BDEP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bdep_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BDEP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bdep_fns[a->esz], a, 0) static gen_helper_gvec_3 * const bgrp_fns[4] = { gen_helper_sve2_bgrp_b, gen_helper_sve2_bgrp_h, gen_helper_sve2_bgrp_s, gen_helper_sve2_bgrp_d, }; -TRANS_FEAT(BGRP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, - bgrp_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(BGRP, aa64_sve2_bitperm, gen_gvec_ool_arg_zzz, + bgrp_fns[a->esz], a, 0) static gen_helper_gvec_3 * const cadd_fns[4] = { gen_helper_sve2_cadd_b, gen_helper_sve2_cadd_h, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-8-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 24 +++++++++++++++--------- 2 files changed, 15 insertions(+), 11 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 000- ---- 0110 1--- ---- ---- # PMULLB, PMULLT (128b result) -FAIL 0110 0100 --1- ---- 1110 01-- ---- ---- # FMMLA, BFMMLA FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_trans_pmull(DisasContext *s, arg_rrr_esz *a, bool sel) gen_helper_gvec_pmull_q, gen_helper_sve2_pmull_h, NULL, gen_helper_sve2_pmull_d, }; - if (a->esz == 0 - ? !dc_isar_feature(aa64_sve2_pmull128, s) - : !dc_isar_feature(aa64_sve, s)) { + + if (a->esz == 0) { + if (!dc_isar_feature(aa64_sve2_pmull128, s)) { + return false; + } + s->is_nonstreaming = true; + } else if (!dc_isar_feature(aa64_sve, s)) { return false; } return gen_gvec_ool_arg_zzz(s, fns[a->esz], a, sel); @@ -XXX,XX +XXX,XX @@ DO_ZPZZ_FP(FMINP, aa64_sve2, sve2_fminp_zpzz) * SVE Integer Multiply-Add (unpredicated) */ -TRANS_FEAT(FMMLA_s, aa64_sve_f32mm, gen_gvec_fpst_zzzz, gen_helper_fmmla_s, - a->rd, a->rn, a->rm, a->ra, 0, FPST_FPCR) -TRANS_FEAT(FMMLA_d, aa64_sve_f64mm, gen_gvec_fpst_zzzz, gen_helper_fmmla_d, - a->rd, a->rn, a->rm, a->ra, 0, FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FMMLA_s, aa64_sve_f32mm, gen_gvec_fpst_zzzz, + gen_helper_fmmla_s, a->rd, a->rn, a->rm, a->ra, + 0, FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FMMLA_d, aa64_sve_f64mm, gen_gvec_fpst_zzzz, + gen_helper_fmmla_d, a->rd, a->rn, a->rm, a->ra, + 0, FPST_FPCR) static gen_helper_gvec_4 * const sqdmlal_zzzw_fns[] = { NULL, gen_helper_sve2_sqdmlal_zzzw_h, @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(BFDOT_zzzz, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, TRANS_FEAT(BFDOT_zzxz, aa64_sve_bf16, gen_gvec_ool_arg_zzxz, gen_helper_gvec_bfdot_idx, a) -TRANS_FEAT(BFMMLA, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_bfmmla, a, 0) +TRANS_FEAT_NONSTREAMING(BFMMLA, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_bfmmla, a, 0) static bool do_BFMLAL_zzzw(DisasContext *s, arg_rrrr_esz *a, bool sel) { -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-9-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 15 +++++++++++---- 2 files changed, 11 insertions(+), 7 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0110 0101 --0- ---- 0000 11-- ---- ---- # FTSMUL -FAIL 0110 0101 --01 0--- 100- ---- ---- ---- # FTMAD -FAIL 0110 0101 --01 1--- 001- ---- ---- ---- # FADDA FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static gen_helper_gvec_3_ptr * const ftmad_fns[4] = { NULL, gen_helper_sve_ftmad_h, gen_helper_sve_ftmad_s, gen_helper_sve_ftmad_d, }; -TRANS_FEAT(FTMAD, aa64_sve, gen_gvec_fpst_zzz, - ftmad_fns[a->esz], a->rd, a->rn, a->rm, a->imm, - a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR) +TRANS_FEAT_NONSTREAMING(FTMAD, aa64_sve, gen_gvec_fpst_zzz, + ftmad_fns[a->esz], a->rd, a->rn, a->rm, a->imm, + a->esz == MO_16 ? FPST_FPCR_F16 : FPST_FPCR) /* *** SVE Floating Point Accumulating Reduction Group @@ -XXX,XX +XXX,XX @@ static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a) if (a->esz == 0 || !dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_FADDA(DisasContext *s, arg_rprr_esz *a) DO_FP3(FADD_zzz, fadd) DO_FP3(FSUB_zzz, fsub) DO_FP3(FMUL_zzz, fmul) -DO_FP3(FTSMUL, ftsmul) DO_FP3(FRECPS, recps) DO_FP3(FRSQRTS, rsqrts) #undef DO_FP3 +static gen_helper_gvec_3_ptr * const ftsmul_fns[4] = { + NULL, gen_helper_gvec_ftsmul_h, + gen_helper_gvec_ftsmul_s, gen_helper_gvec_ftsmul_d +}; +TRANS_FEAT_NONSTREAMING(FTSMUL, aa64_sve, gen_gvec_fpst_arg_zzz, + ftsmul_fns[a->esz], a, 0) + /* *** SVE Floating Point Arithmetic - Predicated Group */ -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-10-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 12 ++++++------ 2 files changed, 6 insertions(+), 7 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 --0- ---- 1001 10-- ---- ---- # SMMLA, UMMLA, USMMLA FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(FMLALT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, false, true) TRANS_FEAT(FMLSLB_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, false) TRANS_FEAT(FMLSLT_zzxw, aa64_sve2, do_FMLAL_zzxw, a, true, true) -TRANS_FEAT(SMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_smmla_b, a, 0) -TRANS_FEAT(USMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_usmmla_b, a, 0) -TRANS_FEAT(UMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, - gen_helper_gvec_ummla_b, a, 0) +TRANS_FEAT_NONSTREAMING(SMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_smmla_b, a, 0) +TRANS_FEAT_NONSTREAMING(USMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_usmmla_b, a, 0) +TRANS_FEAT_NONSTREAMING(UMMLA, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, + gen_helper_gvec_ummla_b, a, 0) TRANS_FEAT(BFDOT_zzzz, aa64_sve_bf16, gen_gvec_ool_arg_zzzz, gen_helper_gvec_bfdot, a, 0) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-11-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 1 - target/arm/translate-sve.c | 35 ++++++++++++++++++----------------- 2 files changed, 18 insertions(+), 18 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 0100 0101 --1- ---- 1--- ---- ---- ---- # SVE2 string/histo/crypto instructions FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ DO_SVE2_ZZZ_NARROW(RSUBHNT, rsubhnt) static gen_helper_gvec_flags_4 * const match_fns[4] = { gen_helper_sve2_match_ppzz_b, gen_helper_sve2_match_ppzz_h, NULL, NULL }; -TRANS_FEAT(MATCH, aa64_sve2, do_ppzz_flags, a, match_fns[a->esz]) +TRANS_FEAT_NONSTREAMING(MATCH, aa64_sve2, do_ppzz_flags, a, match_fns[a->esz]) static gen_helper_gvec_flags_4 * const nmatch_fns[4] = { gen_helper_sve2_nmatch_ppzz_b, gen_helper_sve2_nmatch_ppzz_h, NULL, NULL }; -TRANS_FEAT(NMATCH, aa64_sve2, do_ppzz_flags, a, nmatch_fns[a->esz]) +TRANS_FEAT_NONSTREAMING(NMATCH, aa64_sve2, do_ppzz_flags, a, nmatch_fns[a->esz]) static gen_helper_gvec_4 * const histcnt_fns[4] = { NULL, NULL, gen_helper_sve2_histcnt_s, gen_helper_sve2_histcnt_d }; -TRANS_FEAT(HISTCNT, aa64_sve2, gen_gvec_ool_arg_zpzz, - histcnt_fns[a->esz], a, 0) +TRANS_FEAT_NONSTREAMING(HISTCNT, aa64_sve2, gen_gvec_ool_arg_zpzz, + histcnt_fns[a->esz], a, 0) -TRANS_FEAT(HISTSEG, aa64_sve2, gen_gvec_ool_arg_zzz, - a->esz == 0 ? gen_helper_sve2_histseg : NULL, a, 0) +TRANS_FEAT_NONSTREAMING(HISTSEG, aa64_sve2, gen_gvec_ool_arg_zzz, + a->esz == 0 ? gen_helper_sve2_histseg : NULL, a, 0) DO_ZPZZ_FP(FADDP, aa64_sve2, sve2_faddp_zpzz) DO_ZPZZ_FP(FMAXNMP, aa64_sve2, sve2_fmaxnmp_zpzz) @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(SQRDCMLAH_zzzz, aa64_sve2, gen_gvec_ool_zzzz, TRANS_FEAT(USDOT_zzzz, aa64_sve_i8mm, gen_gvec_ool_arg_zzzz, a->esz == 2 ? gen_helper_gvec_usdot_b : NULL, a, 0) -TRANS_FEAT(AESMC, aa64_sve2_aes, gen_gvec_ool_zz, - gen_helper_crypto_aesmc, a->rd, a->rd, a->decrypt) +TRANS_FEAT_NONSTREAMING(AESMC, aa64_sve2_aes, gen_gvec_ool_zz, + gen_helper_crypto_aesmc, a->rd, a->rd, a->decrypt) -TRANS_FEAT(AESE, aa64_sve2_aes, gen_gvec_ool_arg_zzz, - gen_helper_crypto_aese, a, false) -TRANS_FEAT(AESD, aa64_sve2_aes, gen_gvec_ool_arg_zzz, - gen_helper_crypto_aese, a, true) +TRANS_FEAT_NONSTREAMING(AESE, aa64_sve2_aes, gen_gvec_ool_arg_zzz, + gen_helper_crypto_aese, a, false) +TRANS_FEAT_NONSTREAMING(AESD, aa64_sve2_aes, gen_gvec_ool_arg_zzz, + gen_helper_crypto_aese, a, true) -TRANS_FEAT(SM4E, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, - gen_helper_crypto_sm4e, a, 0) -TRANS_FEAT(SM4EKEY, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, - gen_helper_crypto_sm4ekey, a, 0) +TRANS_FEAT_NONSTREAMING(SM4E, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, + gen_helper_crypto_sm4e, a, 0) +TRANS_FEAT_NONSTREAMING(SM4EKEY, aa64_sve2_sm4, gen_gvec_ool_arg_zzz, + gen_helper_crypto_sm4ekey, a, 0) -TRANS_FEAT(RAX1, aa64_sve2_sha3, gen_gvec_fn_arg_zzz, gen_gvec_rax1, a) +TRANS_FEAT_NONSTREAMING(RAX1, aa64_sve2_sha3, gen_gvec_fn_arg_zzz, + gen_gvec_rax1, a) TRANS_FEAT(FCVTNT_sh, aa64_sve2, gen_gvec_fpst_arg_zpz, gen_helper_sve2_fcvtnt_sh, a, 0, FPST_FPCR) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-12-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 9 --------- target/arm/translate-sve.c | 6 ++++++ 2 files changed, 6 insertions(+), 9 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1000 010- -00- ---- 10-- ---- ---- ---- # SVE2 32-bit gather NT load (vector+scalar) FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) -FAIL 1000 010- -01- ---- 1--- ---- ---- ---- # SVE 32-bit gather load (vector+imm) -FAIL 1000 0100 0-0- ---- 0--- ---- ---- ---- # SVE 32-bit gather load byte (scalar+vector) -FAIL 1000 0100 1--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load half (scalar+vector) -FAIL 1000 0101 0--- ---- 0--- ---- ---- ---- # SVE 32-bit gather load word (scalar+vector) FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch -FAIL 1110 010- -00- ---- 001- ---- ---- ---- # SVE2 64-bit scatter NT store (vector+scalar) -FAIL 1110 010- -10- ---- 001- ---- ---- ---- # SVE2 32-bit scatter NT store (vector+scalar) -FAIL 1110 010- ---- ---- 1-0- ---- ---- ---- # SVE scatter store (scalar+32-bit vector) -FAIL 1110 010- ---- ---- 101- ---- ---- ---- # SVE scatter store (misc) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD1_zprz(DisasContext *s, arg_LD1_zprz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LD1_zpiz(DisasContext *s, arg_LD1_zpiz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LDNT1_zprz(DisasContext *s, arg_LD1_zprz *a) if (!dc_isar_feature(aa64_sve2, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zprz(DisasContext *s, arg_ST1_zprz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_ST1_zpiz(DisasContext *s, arg_ST1_zpiz *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STNT1_zprz(DisasContext *s, arg_ST1_zprz *a) if (!dc_isar_feature(aa64_sve2, s)) { return false; } + s->is_nonstreaming = true; if (!sve_access_check(s)) { return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. In this case, introduce PRF_ns (prefetch non-streaming) to handle the checks. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-13-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/sve.decode | 10 +++++----- target/arm/translate-sve.c | 11 +++++++++++ 3 files changed, 16 insertions(+), 8 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1000 010- -00- ---- 111- ---- ---- ---- # SVE 32-bit gather prefetch (vector+imm) -FAIL 1000 0100 0-1- ---- 0--- ---- ---- ---- # SVE 32-bit gather prefetch (scalar+vector) FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) -FAIL 1100 010- ---- ---- ---- ---- ---- ---- # SVE 64-bit gather load/prefetch diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ LD1RO_zpri 1010010 .. 01 0.... 001 ... ..... ..... \ @rpri_load_msz nreg=0 # SVE 32-bit gather prefetch (scalar plus 32-bit scaled offsets) -PRF 1000010 00 -1 ----- 0-- --- ----- 0 ---- +PRF_ns 1000010 00 -1 ----- 0-- --- ----- 0 ---- # SVE 32-bit gather prefetch (vector plus immediate) -PRF 1000010 -- 00 ----- 111 --- ----- 0 ---- +PRF_ns 1000010 -- 00 ----- 111 --- ----- 0 ---- # SVE contiguous prefetch (scalar plus immediate) PRF 1000010 11 1- ----- 0-- --- ----- 0 ---- @@ -XXX,XX +XXX,XX @@ LD1_zpiz 1100010 .. 01 ..... 1.. ... ..... ..... \ @rpri_g_load esz=3 # SVE 64-bit gather prefetch (scalar plus 64-bit scaled offsets) -PRF 1100010 00 11 ----- 1-- --- ----- 0 ---- +PRF_ns 1100010 00 11 ----- 1-- --- ----- 0 ---- # SVE 64-bit gather prefetch (scalar plus unpacked 32-bit scaled offsets) -PRF 1100010 00 -1 ----- 0-- --- ----- 0 ---- +PRF_ns 1100010 00 -1 ----- 0-- --- ----- 0 ---- # SVE 64-bit gather prefetch (vector plus immediate) -PRF 1100010 -- 00 ----- 111 --- ----- 0 ---- +PRF_ns 1100010 -- 00 ----- 111 --- ----- 0 ---- ### SVE Memory Store Group diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_PRF_rr(DisasContext *s, arg_PRF_rr *a) return true; } +static bool trans_PRF_ns(DisasContext *s, arg_PRF_ns *a) +{ + if (!dc_isar_feature(aa64_sve, s)) { + return false; + } + /* Prefetch is a nop within QEMU. */ + s->is_nonstreaming = true; + (void)sve_access_check(s); + return true; +} + /* * Move Prefix * -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-14-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 2 -- target/arm/translate-sve.c | 2 ++ 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) -FAIL 1010 010- ---- ---- 011- ---- ---- ---- # SVE contiguous FF load (scalar+scalar) -FAIL 1010 010- ---1 ---- 101- ---- ---- ---- # SVE contiguous NF load (scalar+imm) FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDFF1_zprr(DisasContext *s, arg_rprr_load *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype)); @@ -XXX,XX +XXX,XX @@ static bool trans_LDNF1_zpri(DisasContext *s, arg_rpri_load *a) if (!dc_isar_feature(aa64_sve, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { int vsz = vec_full_reg_size(s); int elements = vsz >> dtype_esz[a->dtype]; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Mark these as a non-streaming instructions, which should trap if full a64 support is not enabled in streaming mode. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-15-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme-fa64.decode | 3 --- target/arm/translate-sve.c | 2 ++ 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/target/arm/sme-fa64.decode b/target/arm/sme-fa64.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme-fa64.decode +++ b/target/arm/sme-fa64.decode @@ -XXX,XX +XXX,XX @@ FAIL 0001 1110 0111 1110 0000 00-- ---- ---- # FJCVTZS # --11 1100 --0- ---- ---- ---- ---- ---- # Load/store FP register (unscaled imm) # --11 1100 --1- ---- ---- ---- ---- --10 # Load/store FP register (register offset) # --11 1101 ---- ---- ---- ---- ---- ---- # Load/store FP register (scaled imm) - -FAIL 1010 010- -01- ---- 000- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+scalar) -FAIL 1010 010- -010 ---- 001- ---- ---- ---- # SVE load & replicate 32 bytes (scalar+imm) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_LD1RO_zprr(DisasContext *s, arg_rprr_load *a) if (a->rm == 31) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), dtype_msz(a->dtype)); @@ -XXX,XX +XXX,XX @@ static bool trans_LD1RO_zpri(DisasContext *s, arg_rpri_load *a) if (!dc_isar_feature(aa64_sve_f64mm, s)) { return false; } + s->is_nonstreaming = true; if (sve_access_check(s)) { TCGv_i64 addr = new_tmp_a64(s); tcg_gen_addi_i64(addr, cpu_reg_sp(s, a->rn), a->imm * 32); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These functions will be used to verify that the cpu is in the correct state for a given instruction. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-16-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 21 +++++++++++++++++++++ target/arm/translate-a64.c | 34 ++++++++++++++++++++++++++++++++++ 2 files changed, 55 insertions(+) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ void write_fp_dreg(DisasContext *s, int reg, TCGv_i64 v); bool logic_imm_decode_wmask(uint64_t *result, unsigned int immn, unsigned int imms, unsigned int immr); bool sve_access_check(DisasContext *s); +bool sme_enabled_check(DisasContext *s); +bool sme_enabled_check_with_svcr(DisasContext *s, unsigned); + +/* This function corresponds to CheckStreamingSVEEnabled. */ +static inline bool sme_sm_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_SM_MASK); +} + +/* This function corresponds to CheckSMEAndZAEnabled. */ +static inline bool sme_za_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_ZA_MASK); +} + +/* Note that this function corresponds to CheckStreamingSVEAndZAEnabled. */ +static inline bool sme_smza_enabled_check(DisasContext *s) +{ + return sme_enabled_check_with_svcr(s, R_SVCR_SM_MASK | R_SVCR_ZA_MASK); +} + TCGv_i64 clean_data_tbi(DisasContext *s, TCGv_i64 addr); TCGv_i64 gen_mte_check1(DisasContext *s, TCGv_i64 addr, bool is_write, bool tag_checked, int log2_size); diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static bool sme_access_check(DisasContext *s) return true; } +/* This function corresponds to CheckSMEEnabled. */ +bool sme_enabled_check(DisasContext *s) +{ + /* + * Note that unlike sve_excp_el, we have not constrained sme_excp_el + * to be zero when fp_excp_el has priority. This is because we need + * sme_excp_el by itself for cpregs access checks. + */ + if (!s->fp_excp_el || s->sme_excp_el < s->fp_excp_el) { + s->fp_access_checked = true; + return sme_access_check(s); + } + return fp_access_check_only(s); +} + +/* Common subroutine for CheckSMEAnd*Enabled. */ +bool sme_enabled_check_with_svcr(DisasContext *s, unsigned req) +{ + if (!sme_enabled_check(s)) { + return false; + } + if (FIELD_EX64(req, SVCR, SM) && !s->pstate_sm) { + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_smetrap(SME_ET_NotStreaming, false)); + return false; + } + if (FIELD_EX64(req, SVCR, ZA) && !s->pstate_za) { + gen_exception_insn(s, s->pc_curr, EXCP_UDEF, + syn_smetrap(SME_ET_InactiveZA, false)); + return false; + } + return true; +} + /* * This utility function is for doing register extension with an * optional shift. You will likely want to pass a temporary for the -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> The pseudocode for CheckSVEEnabled gains a check for Streaming SVE mode, and for SME present but SVE absent. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-17-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -XXX,XX +XXX,XX @@ static bool fp_access_check(DisasContext *s) return true; } -/* Check that SVE access is enabled. If it is, return true. +/* + * Check that SVE access is enabled. If it is, return true. * If not, emit code to generate an appropriate exception and return false. + * This function corresponds to CheckSVEEnabled(). */ bool sve_access_check(DisasContext *s) { - if (s->sve_excp_el) { - assert(!s->sve_access_checked); - s->sve_access_checked = true; - + if (s->pstate_sm || !dc_isar_feature(aa64_sve, s)) { + assert(dc_isar_feature(aa64_sme, s)); + if (!sme_sm_enabled_check(s)) { + goto fail_exit; + } + } else if (s->sve_excp_el) { gen_exception_insn_el(s, s->pc_curr, EXCP_UDEF, syn_sve_access_trap(), s->sve_excp_el); - return false; + goto fail_exit; } s->sve_access_checked = true; return fp_access_check(s); + + fail_exit: + /* Assert that we only raise one exception per instruction. */ + assert(!s->sve_access_checked); + s->sve_access_checked = true; + return false; } /* -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These SME instructions are nominally within the SVE decode space, so we add them to sve.decode and translate-sve.c. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-18-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 12 ++++++++++++ target/arm/sve.decode | 5 ++++- target/arm/translate-sve.c | 38 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 54 insertions(+), 1 deletion(-) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline int vec_full_reg_size(DisasContext *s) return s->vl; } +/* Return the byte size of the vector register, SVL / 8. */ +static inline int streaming_vec_reg_size(DisasContext *s) +{ + return s->svl; +} + /* * Return the offset info CPUARMState of the predicate vector register Pn. * Note for this purpose, FFR is P16. @@ -XXX,XX +XXX,XX @@ static inline int pred_full_reg_size(DisasContext *s) return s->vl >> 3; } +/* Return the byte size of the predicate register, SVL / 64. */ +static inline int streaming_pred_reg_size(DisasContext *s) +{ + return s->svl >> 3; +} + /* * Round up the size of a register to a size allowed by * the tcg vector infrastructure. Any operation which uses this diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ INDEX_ri 00000100 esz:2 1 imm:s5 010001 rn:5 rd:5 # SVE index generation (register start, register increment) INDEX_rr 00000100 .. 1 ..... 010011 ..... ..... @rd_rn_rm -### SVE Stack Allocation Group +### SVE / Streaming SVE Stack Allocation Group # SVE stack frame adjustment ADDVL 00000100 001 ..... 01010 ...... ..... @rd_rn_i6 +ADDSVL 00000100 001 ..... 01011 ...... ..... @rd_rn_i6 ADDPL 00000100 011 ..... 01010 ...... ..... @rd_rn_i6 +ADDSPL 00000100 011 ..... 01011 ...... ..... @rd_rn_i6 # SVE stack frame size RDVL 00000100 101 11111 01010 imm:s6 rd:5 +RDSVL 00000100 101 11111 01011 imm:s6 rd:5 ### SVE Bitwise Shift - Unpredicated Group diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_ADDVL(DisasContext *s, arg_ADDVL *a) return true; } +static bool trans_ADDSVL(DisasContext *s, arg_ADDSVL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 rd = cpu_reg_sp(s, a->rd); + TCGv_i64 rn = cpu_reg_sp(s, a->rn); + tcg_gen_addi_i64(rd, rn, a->imm * streaming_vec_reg_size(s)); + } + return true; +} + static bool trans_ADDPL(DisasContext *s, arg_ADDPL *a) { if (!dc_isar_feature(aa64_sve, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_ADDPL(DisasContext *s, arg_ADDPL *a) return true; } +static bool trans_ADDSPL(DisasContext *s, arg_ADDSPL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 rd = cpu_reg_sp(s, a->rd); + TCGv_i64 rn = cpu_reg_sp(s, a->rn); + tcg_gen_addi_i64(rd, rn, a->imm * streaming_pred_reg_size(s)); + } + return true; +} + static bool trans_RDVL(DisasContext *s, arg_RDVL *a) { if (!dc_isar_feature(aa64_sve, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_RDVL(DisasContext *s, arg_RDVL *a) return true; } +static bool trans_RDSVL(DisasContext *s, arg_RDSVL *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_enabled_check(s)) { + TCGv_i64 reg = cpu_reg(s, a->rd); + tcg_gen_movi_i64(reg, a->imm * streaming_vec_reg_size(s)); + } + return true; +} + /* *** SVE Compute Vector Address Group */ -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-19-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 4 ++++ target/arm/sme_helper.c | 25 +++++++++++++++++++++++++ target/arm/translate-sme.c | 13 +++++++++++++ 4 files changed, 44 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_2(set_pstate_sm, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_2(set_pstate_za, TCG_CALL_NO_RWG, void, env, i32) + +DEF_HELPER_FLAGS_3(sme_zero, TCG_CALL_NO_RWG, void, env, i32, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ # # This file is processed by scripts/decodetree.py # + +### SME Misc + +ZERO 11000000 00 001 00000000000 imm:8 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void helper_set_pstate_za(CPUARMState *env, uint32_t i) memset(env->zarray, 0, sizeof(env->zarray)); } } + +void helper_sme_zero(CPUARMState *env, uint32_t imm, uint32_t svl) +{ + uint32_t i; + + /* + * Special case clearing the entire ZA space. + * This falls into the CONSTRAINED UNPREDICTABLE zeroing of any + * parts of the ZA storage outside of SVL. + */ + if (imm == 0xff) { + memset(env->zarray, 0, sizeof(env->zarray)); + return; + } + + /* + * Recall that ZAnH.D[m] is spread across ZA[n+8*m], + * so each row is discontiguous within ZA[]. + */ + for (i = 0; i < svl; i++) { + if (imm & (1 << (i % 8))) { + memset(&env->zarray[i], 0, svl); + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ */ #include "decode-sme.c.inc" + + +static bool trans_ZERO(DisasContext *s, arg_ZERO *a) +{ + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (sme_za_enabled_check(s)) { + gen_helper_sme_zero(cpu_env, tcg_constant_i32(a->imm), + tcg_constant_i32(streaming_vec_reg_size(s))); + } + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can reuse the SVE functions for implementing moves to/from horizontal tile slices, but we need new ones for moves to/from vertical tile slices. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-20-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 12 +++ target/arm/helper-sve.h | 2 + target/arm/translate-a64.h | 8 ++ target/arm/translate.h | 5 ++ target/arm/sme.decode | 15 ++++ target/arm/sme_helper.c | 151 ++++++++++++++++++++++++++++++++++++- target/arm/sve_helper.c | 12 +++ target/arm/translate-sme.c | 127 +++++++++++++++++++++++++++++++ 8 files changed, 331 insertions(+), 1 deletion(-) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_2(set_pstate_sm, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_2(set_pstate_za, TCG_CALL_NO_RWG, void, env, i32) DEF_HELPER_FLAGS_3(sme_zero, TCG_CALL_NO_RWG, void, env, i32, i32) + +/* Move to/from vertical array slices, i.e. columns, so 'c'. */ +DEF_HELPER_FLAGS_4(sme_mova_cz_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_cz_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_mova_zc_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sve_sel_zpzz_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve_sel_zpzz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve_sel_zpzz_q, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sve2_addp_zpzz_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ static inline int pred_gvec_reg_size(DisasContext *s) return size_for_gvec(pred_full_reg_size(s)); } +/* Return a newly allocated pointer to the predicate register. */ +static inline TCGv_ptr pred_full_reg_ptr(DisasContext *s, int regno) +{ + TCGv_ptr ret = tcg_temp_new_ptr(); + tcg_gen_addi_ptr(ret, cpu_env, pred_full_reg_offset(s, regno)); + return ret; +} + bool disas_sve(DisasContext *, uint32_t); bool disas_sme(DisasContext *, uint32_t); 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 plus_2(DisasContext *s, int x) return x + 2; } +static inline int plus_12(DisasContext *s, int x) +{ + return x + 12; +} + static inline int times_2(DisasContext *s, int x) { return x * 2; diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ### SME Misc ZERO 11000000 00 001 00000000000 imm:8 + +### SME Move into/from Array + +%mova_rs 13:2 !function=plus_12 +&mova esz rs pg zr za_imm v:bool to_vec:bool + +MOVA 11000000 esz:2 00000 0 v:1 .. pg:3 zr:5 0 za_imm:4 \ + &mova to_vec=0 rs=%mova_rs +MOVA 11000000 11 00000 1 v:1 .. pg:3 zr:5 0 za_imm:4 \ + &mova to_vec=0 rs=%mova_rs esz=4 + +MOVA 11000000 esz:2 00001 0 v:1 .. pg:3 0 za_imm:4 zr:5 \ + &mova to_vec=1 rs=%mova_rs +MOVA 11000000 11 00001 1 v:1 .. pg:3 0 za_imm:4 zr:5 \ + &mova to_vec=1 rs=%mova_rs esz=4 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "cpu.h" -#include "internals.h" +#include "tcg/tcg-gvec-desc.h" #include "exec/helper-proto.h" +#include "qemu/int128.h" +#include "vec_internal.h" /* ResetSVEState */ void arm_reset_sve_state(CPUARMState *env) @@ -XXX,XX +XXX,XX @@ void helper_sme_zero(CPUARMState *env, uint32_t imm, uint32_t svl) } } } + + +/* + * When considering the ZA storage as an array of elements of + * type T, the index within that array of the Nth element of + * a vertical slice of a tile can be calculated like this, + * regardless of the size of type T. This is because the tiles + * are interleaved, so if type T is size N bytes then row 1 of + * the tile is N rows away from row 0. The division by N to + * convert a byte offset into an array index and the multiplication + * by N to convert from vslice-index-within-the-tile to + * the index within the ZA storage cancel out. + */ +#define tile_vslice_index(i) ((i) * sizeof(ARMVectorReg)) + +/* + * When doing byte arithmetic on the ZA storage, the element + * byteoff bytes away in a tile vertical slice is always this + * many bytes away in the ZA storage, regardless of the + * size of the tile element, assuming that byteoff is a multiple + * of the element size. Again this is because of the interleaving + * of the tiles. For instance if we have 1 byte per element then + * each row of the ZA storage has one byte of the vslice data, + * and (counting from 0) byte 8 goes in row 8 of the storage + * at offset (8 * row-size-in-bytes). + * If we have 8 bytes per element then each row of the ZA storage + * has 8 bytes of the data, but there are 8 interleaved tiles and + * so byte 8 of the data goes into row 1 of the tile, + * which is again row 8 of the storage, so the offset is still + * (8 * row-size-in-bytes). Similarly for other element sizes. + */ +#define tile_vslice_offset(byteoff) ((byteoff) * sizeof(ARMVectorReg)) + + +/* + * Move Zreg vector to ZArray column. + */ +#define DO_MOVA_C(NAME, TYPE, H) \ +void HELPER(NAME)(void *za, void *vn, void *vg, uint32_t desc) \ +{ \ + int i, oprsz = simd_oprsz(desc); \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + *(TYPE *)(za + tile_vslice_offset(i)) = *(TYPE *)(vn + H(i)); \ + } \ + i += sizeof(TYPE); \ + pg >>= sizeof(TYPE); \ + } while (i & 15); \ + } \ +} + +DO_MOVA_C(sme_mova_cz_b, uint8_t, H1) +DO_MOVA_C(sme_mova_cz_h, uint16_t, H1_2) +DO_MOVA_C(sme_mova_cz_s, uint32_t, H1_4) + +void HELPER(sme_mova_cz_d)(void *za, void *vn, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 8; + uint8_t *pg = vg; + uint64_t *n = vn; + uint64_t *a = za; + + for (i = 0; i < oprsz; i++) { + if (pg[H1(i)] & 1) { + a[tile_vslice_index(i)] = n[i]; + } + } +} + +void HELPER(sme_mova_cz_q)(void *za, void *vn, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 16; + uint16_t *pg = vg; + Int128 *n = vn; + Int128 *a = za; + + /* + * Int128 is used here simply to copy 16 bytes, and to simplify + * the address arithmetic. + */ + for (i = 0; i < oprsz; i++) { + if (pg[H2(i)] & 1) { + a[tile_vslice_index(i)] = n[i]; + } + } +} + +#undef DO_MOVA_C + +/* + * Move ZArray column to Zreg vector. + */ +#define DO_MOVA_Z(NAME, TYPE, H) \ +void HELPER(NAME)(void *vd, void *za, void *vg, uint32_t desc) \ +{ \ + int i, oprsz = simd_oprsz(desc); \ + for (i = 0; i < oprsz; ) { \ + uint16_t pg = *(uint16_t *)(vg + H1_2(i >> 3)); \ + do { \ + if (pg & 1) { \ + *(TYPE *)(vd + H(i)) = *(TYPE *)(za + tile_vslice_offset(i)); \ + } \ + i += sizeof(TYPE); \ + pg >>= sizeof(TYPE); \ + } while (i & 15); \ + } \ +} + +DO_MOVA_Z(sme_mova_zc_b, uint8_t, H1) +DO_MOVA_Z(sme_mova_zc_h, uint16_t, H1_2) +DO_MOVA_Z(sme_mova_zc_s, uint32_t, H1_4) + +void HELPER(sme_mova_zc_d)(void *vd, void *za, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 8; + uint8_t *pg = vg; + uint64_t *d = vd; + uint64_t *a = za; + + for (i = 0; i < oprsz; i++) { + if (pg[H1(i)] & 1) { + d[i] = a[tile_vslice_index(i)]; + } + } +} + +void HELPER(sme_mova_zc_q)(void *vd, void *za, void *vg, uint32_t desc) +{ + int i, oprsz = simd_oprsz(desc) / 16; + uint16_t *pg = vg; + Int128 *d = vd; + Int128 *a = za; + + /* + * Int128 is used here simply to copy 16 bytes, and to simplify + * the address arithmetic. + */ + for (i = 0; i < oprsz; i++, za += sizeof(ARMVectorReg)) { + if (pg[H2(i)] & 1) { + d[i] = a[tile_vslice_index(i)]; + } + } +} + +#undef DO_MOVA_Z diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sve_sel_zpzz_d)(void *vd, void *vn, void *vm, } } +void HELPER(sve_sel_zpzz_q)(void *vd, void *vn, void *vm, + void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 16; + Int128 *d = vd, *n = vn, *m = vm; + uint16_t *pg = vg; + + for (i = 0; i < opr_sz; i += 1) { + d[i] = (pg[H2(i)] & 1 ? n : m)[i]; + } +} + /* Two operand comparison controlled by a predicate. * ??? It is very tempting to want to be able to expand this inline * with x86 instructions, e.g. diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ #include "decode-sme.c.inc" +/* + * Resolve tile.size[index] to a host pointer, where tile and index + * are always decoded together, dependent on the element size. + */ +static TCGv_ptr get_tile_rowcol(DisasContext *s, int esz, int rs, + int tile_index, bool vertical) +{ + int tile = tile_index >> (4 - esz); + int index = esz == MO_128 ? 0 : extract32(tile_index, 0, 4 - esz); + int pos, len, offset; + TCGv_i32 tmp; + TCGv_ptr addr; + + /* Compute the final index, which is Rs+imm. */ + tmp = tcg_temp_new_i32(); + tcg_gen_trunc_tl_i32(tmp, cpu_reg(s, rs)); + tcg_gen_addi_i32(tmp, tmp, index); + + /* Prepare a power-of-two modulo via extraction of @len bits. */ + len = ctz32(streaming_vec_reg_size(s)) - esz; + + if (vertical) { + /* + * Compute the byte offset of the index within the tile: + * (index % (svl / size)) * size + * = (index % (svl >> esz)) << esz + * Perform the power-of-two modulo via extraction of the low @len bits. + * Perform the multiply by shifting left by @pos bits. + * Perform these operations simultaneously via deposit into zero. + */ + pos = esz; + tcg_gen_deposit_z_i32(tmp, tmp, pos, len); + + /* + * For big-endian, adjust the indexed column byte offset within + * the uint64_t host words that make up env->zarray[]. + */ + if (HOST_BIG_ENDIAN && esz < MO_64) { + tcg_gen_xori_i32(tmp, tmp, 8 - (1 << esz)); + } + } else { + /* + * Compute the byte offset of the index within the tile: + * (index % (svl / size)) * (size * sizeof(row)) + * = (index % (svl >> esz)) << (esz + log2(sizeof(row))) + */ + pos = esz + ctz32(sizeof(ARMVectorReg)); + tcg_gen_deposit_z_i32(tmp, tmp, pos, len); + + /* Row slices are always aligned and need no endian adjustment. */ + } + + /* The tile byte offset within env->zarray is the row. */ + offset = tile * sizeof(ARMVectorReg); + + /* Include the byte offset of zarray to make this relative to env. */ + offset += offsetof(CPUARMState, zarray); + tcg_gen_addi_i32(tmp, tmp, offset); + + /* Add the byte offset to env to produce the final pointer. */ + addr = tcg_temp_new_ptr(); + tcg_gen_ext_i32_ptr(addr, tmp); + tcg_temp_free_i32(tmp); + tcg_gen_add_ptr(addr, addr, cpu_env); + + return addr; +} + static bool trans_ZERO(DisasContext *s, arg_ZERO *a) { if (!dc_isar_feature(aa64_sme, s)) { @@ -XXX,XX +XXX,XX @@ static bool trans_ZERO(DisasContext *s, arg_ZERO *a) } return true; } + +static bool trans_MOVA(DisasContext *s, arg_MOVA *a) +{ + static gen_helper_gvec_4 * const h_fns[5] = { + gen_helper_sve_sel_zpzz_b, gen_helper_sve_sel_zpzz_h, + gen_helper_sve_sel_zpzz_s, gen_helper_sve_sel_zpzz_d, + gen_helper_sve_sel_zpzz_q + }; + static gen_helper_gvec_3 * const cz_fns[5] = { + gen_helper_sme_mova_cz_b, gen_helper_sme_mova_cz_h, + gen_helper_sme_mova_cz_s, gen_helper_sme_mova_cz_d, + gen_helper_sme_mova_cz_q, + }; + static gen_helper_gvec_3 * const zc_fns[5] = { + gen_helper_sme_mova_zc_b, gen_helper_sme_mova_zc_h, + gen_helper_sme_mova_zc_s, gen_helper_sme_mova_zc_d, + gen_helper_sme_mova_zc_q, + }; + + TCGv_ptr t_za, t_zr, t_pg; + TCGv_i32 t_desc; + int svl; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sme_smza_enabled_check(s)) { + return true; + } + + t_za = get_tile_rowcol(s, a->esz, a->rs, a->za_imm, a->v); + t_zr = vec_full_reg_ptr(s, a->zr); + t_pg = pred_full_reg_ptr(s, a->pg); + + svl = streaming_vec_reg_size(s); + t_desc = tcg_constant_i32(simd_desc(svl, svl, 0)); + + if (a->v) { + /* Vertical slice -- use sme mova helpers. */ + if (a->to_vec) { + zc_fns[a->esz](t_zr, t_za, t_pg, t_desc); + } else { + cz_fns[a->esz](t_za, t_zr, t_pg, t_desc); + } + } else { + /* Horizontal slice -- reuse sve sel helpers. */ + if (a->to_vec) { + h_fns[a->esz](t_zr, t_za, t_zr, t_pg, t_desc); + } else { + h_fns[a->esz](t_za, t_zr, t_za, t_pg, t_desc); + } + } + + tcg_temp_free_ptr(t_za); + tcg_temp_free_ptr(t_zr); + tcg_temp_free_ptr(t_pg); + + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We cannot reuse the SVE functions for LD[1-4] and ST[1-4], because those functions accept only a Zreg register number. For SME, we want to pass a pointer into ZA storage. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-21-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 82 +++++ target/arm/sme.decode | 9 + target/arm/sme_helper.c | 595 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 70 +++++ 4 files changed, 756 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sme_mova_cz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_zc_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_cz_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sme_mova_zc_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(sme_ld1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_ld1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_ld1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_st1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) +DEF_HELPER_FLAGS_5(sme_st1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ MOVA 11000000 esz:2 00001 0 v:1 .. pg:3 0 za_imm:4 zr:5 \ &mova to_vec=1 rs=%mova_rs MOVA 11000000 11 00001 1 v:1 .. pg:3 0 za_imm:4 zr:5 \ &mova to_vec=1 rs=%mova_rs esz=4 + +### SME Memory + +&ldst esz rs pg rn rm za_imm v:bool st:bool + +LDST1 1110000 0 esz:2 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ + &ldst rs=%mova_rs +LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ + &ldst esz=4 rs=%mova_rs diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "cpu.h" +#include "internals.h" #include "tcg/tcg-gvec-desc.h" #include "exec/helper-proto.h" +#include "exec/cpu_ldst.h" +#include "exec/exec-all.h" #include "qemu/int128.h" #include "vec_internal.h" +#include "sve_ldst_internal.h" /* ResetSVEState */ void arm_reset_sve_state(CPUARMState *env) @@ -XXX,XX +XXX,XX @@ void HELPER(sme_mova_zc_q)(void *vd, void *za, void *vg, uint32_t desc) } #undef DO_MOVA_Z + +/* + * Clear elements in a tile slice comprising len bytes. + */ + +typedef void ClearFn(void *ptr, size_t off, size_t len); + +static void clear_horizontal(void *ptr, size_t off, size_t len) +{ + memset(ptr + off, 0, len); +} + +static void clear_vertical_b(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; ++i) { + *(uint8_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_h(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 2) { + *(uint16_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_s(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 4) { + *(uint32_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_d(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 8) { + *(uint64_t *)(vptr + tile_vslice_offset(i + off)) = 0; + } +} + +static void clear_vertical_q(void *vptr, size_t off, size_t len) +{ + for (size_t i = 0; i < len; i += 16) { + memset(vptr + tile_vslice_offset(i + off), 0, 16); + } +} + +/* + * Copy elements from an array into a tile slice comprising len bytes. + */ + +typedef void CopyFn(void *dst, const void *src, size_t len); + +static void copy_horizontal(void *dst, const void *src, size_t len) +{ + memcpy(dst, src, len); +} + +static void copy_vertical_b(void *vdst, const void *vsrc, size_t len) +{ + const uint8_t *src = vsrc; + uint8_t *dst = vdst; + size_t i; + + for (i = 0; i < len; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_h(void *vdst, const void *vsrc, size_t len) +{ + const uint16_t *src = vsrc; + uint16_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 2; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_s(void *vdst, const void *vsrc, size_t len) +{ + const uint32_t *src = vsrc; + uint32_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 4; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_d(void *vdst, const void *vsrc, size_t len) +{ + const uint64_t *src = vsrc; + uint64_t *dst = vdst; + size_t i; + + for (i = 0; i < len / 8; ++i) { + dst[tile_vslice_index(i)] = src[i]; + } +} + +static void copy_vertical_q(void *vdst, const void *vsrc, size_t len) +{ + for (size_t i = 0; i < len; i += 16) { + memcpy(vdst + tile_vslice_offset(i), vsrc + i, 16); + } +} + +/* + * Host and TLB primitives for vertical tile slice addressing. + */ + +#define DO_LD(NAME, TYPE, HOST, TLB) \ +static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + TYPE val = HOST(host); \ + *(TYPE *)(za + tile_vslice_offset(off)) = val; \ +} \ +static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \ + intptr_t off, target_ulong addr, uintptr_t ra) \ +{ \ + TYPE val = TLB(env, useronly_clean_ptr(addr), ra); \ + *(TYPE *)(za + tile_vslice_offset(off)) = val; \ +} + +#define DO_ST(NAME, TYPE, HOST, TLB) \ +static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \ + HOST(host, val); \ +} \ +static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \ + intptr_t off, target_ulong addr, uintptr_t ra) \ +{ \ + TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \ + TLB(env, useronly_clean_ptr(addr), val, ra); \ +} + +/* + * The ARMVectorReg elements are stored in host-endian 64-bit units. + * For 128-bit quantities, the sequence defined by the Elem[] pseudocode + * corresponds to storing the two 64-bit pieces in little-endian order. + */ +#define DO_LDQ(HNAME, VNAME, BE, HOST, TLB) \ +static inline void HNAME##_host(void *za, intptr_t off, void *host) \ +{ \ + uint64_t val0 = HOST(host), val1 = HOST(host + 8); \ + uint64_t *ptr = za + off; \ + ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \ +} \ +static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + HNAME##_host(za, tile_vslice_offset(off), host); \ +} \ +static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + uint64_t val0 = TLB(env, useronly_clean_ptr(addr), ra); \ + uint64_t val1 = TLB(env, useronly_clean_ptr(addr + 8), ra); \ + uint64_t *ptr = za + off; \ + ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \ +} \ +static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \ +} + +#define DO_STQ(HNAME, VNAME, BE, HOST, TLB) \ +static inline void HNAME##_host(void *za, intptr_t off, void *host) \ +{ \ + uint64_t *ptr = za + off; \ + HOST(host, ptr[BE]); \ + HOST(host + 1, ptr[!BE]); \ +} \ +static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \ +{ \ + HNAME##_host(za, tile_vslice_offset(off), host); \ +} \ +static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + uint64_t *ptr = za + off; \ + TLB(env, useronly_clean_ptr(addr), ptr[BE], ra); \ + TLB(env, useronly_clean_ptr(addr + 8), ptr[!BE], ra); \ +} \ +static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \ + target_ulong addr, uintptr_t ra) \ +{ \ + HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \ +} + +DO_LD(ld1b, uint8_t, ldub_p, cpu_ldub_data_ra) +DO_LD(ld1h_be, uint16_t, lduw_be_p, cpu_lduw_be_data_ra) +DO_LD(ld1h_le, uint16_t, lduw_le_p, cpu_lduw_le_data_ra) +DO_LD(ld1s_be, uint32_t, ldl_be_p, cpu_ldl_be_data_ra) +DO_LD(ld1s_le, uint32_t, ldl_le_p, cpu_ldl_le_data_ra) +DO_LD(ld1d_be, uint64_t, ldq_be_p, cpu_ldq_be_data_ra) +DO_LD(ld1d_le, uint64_t, ldq_le_p, cpu_ldq_le_data_ra) + +DO_LDQ(sve_ld1qq_be, sme_ld1q_be, 1, ldq_be_p, cpu_ldq_be_data_ra) +DO_LDQ(sve_ld1qq_le, sme_ld1q_le, 0, ldq_le_p, cpu_ldq_le_data_ra) + +DO_ST(st1b, uint8_t, stb_p, cpu_stb_data_ra) +DO_ST(st1h_be, uint16_t, stw_be_p, cpu_stw_be_data_ra) +DO_ST(st1h_le, uint16_t, stw_le_p, cpu_stw_le_data_ra) +DO_ST(st1s_be, uint32_t, stl_be_p, cpu_stl_be_data_ra) +DO_ST(st1s_le, uint32_t, stl_le_p, cpu_stl_le_data_ra) +DO_ST(st1d_be, uint64_t, stq_be_p, cpu_stq_be_data_ra) +DO_ST(st1d_le, uint64_t, stq_le_p, cpu_stq_le_data_ra) + +DO_STQ(sve_st1qq_be, sme_st1q_be, 1, stq_be_p, cpu_stq_be_data_ra) +DO_STQ(sve_st1qq_le, sme_st1q_le, 0, stq_le_p, cpu_stq_le_data_ra) + +#undef DO_LD +#undef DO_ST +#undef DO_LDQ +#undef DO_STQ + +/* + * Common helper for all contiguous predicated loads. + */ + +static inline QEMU_ALWAYS_INLINE +void sme_ld1(CPUARMState *env, void *za, uint64_t *vg, + const target_ulong addr, uint32_t desc, const uintptr_t ra, + const int esz, uint32_t mtedesc, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn, + ClearFn *clr_fn, + CopyFn *cpy_fn) +{ + const intptr_t reg_max = simd_oprsz(desc); + const intptr_t esize = 1 << esz; + intptr_t reg_off, reg_last; + SVEContLdSt info; + void *host; + int flags; + + /* Find the active elements. */ + if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) { + /* The entire predicate was false; no load occurs. */ + clr_fn(za, 0, reg_max); + return; + } + + /* Probe the page(s). Exit with exception for any invalid page. */ + sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_LOAD, ra); + + /* Handle watchpoints for all active elements. */ + sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize, + BP_MEM_READ, ra); + + /* + * Handle mte checks for all active elements. + * Since TBI must be set for MTE, !mtedesc => !mte_active. + */ + if (mtedesc) { + sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize, + mtedesc, ra); + } + + flags = info.page[0].flags | info.page[1].flags; + if (unlikely(flags != 0)) { +#ifdef CONFIG_USER_ONLY + g_assert_not_reached(); +#else + /* + * At least one page includes MMIO. + * Any bus operation can fail with cpu_transaction_failed, + * which for ARM will raise SyncExternal. Perform the load + * into scratch memory to preserve register state until the end. + */ + ARMVectorReg scratch = { }; + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[1]; + if (reg_last < 0) { + reg_last = info.reg_off_split; + if (reg_last < 0) { + reg_last = info.reg_off_last[0]; + } + } + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + tlb_fn(env, &scratch, reg_off, addr + reg_off, ra); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + + cpy_fn(za, &scratch, reg_max); + return; +#endif + } + + /* The entire operation is in RAM, on valid pages. */ + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[0]; + host = info.page[0].host; + + if (!vertical) { + memset(za, 0, reg_max); + } else if (reg_off) { + clr_fn(za, 0, reg_off); + } + + while (reg_off <= reg_last) { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } else if (vertical) { + clr_fn(za, reg_off, esize); + } + reg_off += esize; + } while (reg_off <= reg_last && (reg_off & 63)); + } + + /* + * Use the slow path to manage the cross-page misalignment. + * But we know this is RAM and cannot trap. + */ + reg_off = info.reg_off_split; + if (unlikely(reg_off >= 0)) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + + reg_off = info.reg_off_first[1]; + if (unlikely(reg_off >= 0)) { + reg_last = info.reg_off_last[1]; + host = info.page[1].host; + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } else if (vertical) { + clr_fn(za, reg_off, esize); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + } +} + +static inline QEMU_ALWAYS_INLINE +void sme_ld1_mte(CPUARMState *env, void *za, uint64_t *vg, + target_ulong addr, uint32_t desc, uintptr_t ra, + const int esz, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn, + ClearFn *clr_fn, + CopyFn *cpy_fn) +{ + uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + int bit55 = extract64(addr, 55, 1); + + /* Remove mtedesc from the normal sve descriptor. */ + desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + + /* Perform gross MTE suppression early. */ + if (!tbi_check(desc, bit55) || + tcma_check(desc, bit55, allocation_tag_from_addr(addr))) { + mtedesc = 0; + } + + sme_ld1(env, za, vg, addr, desc, ra, esz, mtedesc, vertical, + host_fn, tlb_fn, clr_fn, cpy_fn); +} + +#define DO_LD(L, END, ESZ) \ +void HELPER(sme_ld1##L##END##_h)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \ + sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \ + clear_horizontal, copy_horizontal); \ +} \ +void HELPER(sme_ld1##L##END##_v)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \ + sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \ + clear_vertical_##L, copy_vertical_##L); \ +} \ +void HELPER(sme_ld1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \ + sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \ + clear_horizontal, copy_horizontal); \ +} \ +void HELPER(sme_ld1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \ + sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \ + clear_vertical_##L, copy_vertical_##L); \ +} + +DO_LD(b, , MO_8) +DO_LD(h, _be, MO_16) +DO_LD(h, _le, MO_16) +DO_LD(s, _be, MO_32) +DO_LD(s, _le, MO_32) +DO_LD(d, _be, MO_64) +DO_LD(d, _le, MO_64) +DO_LD(q, _be, MO_128) +DO_LD(q, _le, MO_128) + +#undef DO_LD + +/* + * Common helper for all contiguous predicated stores. + */ + +static inline QEMU_ALWAYS_INLINE +void sme_st1(CPUARMState *env, void *za, uint64_t *vg, + const target_ulong addr, uint32_t desc, const uintptr_t ra, + const int esz, uint32_t mtedesc, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn) +{ + const intptr_t reg_max = simd_oprsz(desc); + const intptr_t esize = 1 << esz; + intptr_t reg_off, reg_last; + SVEContLdSt info; + void *host; + int flags; + + /* Find the active elements. */ + if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) { + /* The entire predicate was false; no store occurs. */ + return; + } + + /* Probe the page(s). Exit with exception for any invalid page. */ + sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_STORE, ra); + + /* Handle watchpoints for all active elements. */ + sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize, + BP_MEM_WRITE, ra); + + /* + * Handle mte checks for all active elements. + * Since TBI must be set for MTE, !mtedesc => !mte_active. + */ + if (mtedesc) { + sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize, + mtedesc, ra); + } + + flags = info.page[0].flags | info.page[1].flags; + if (unlikely(flags != 0)) { +#ifdef CONFIG_USER_ONLY + g_assert_not_reached(); +#else + /* + * At least one page includes MMIO. + * Any bus operation can fail with cpu_transaction_failed, + * which for ARM will raise SyncExternal. We cannot avoid + * this fault and will leave with the store incomplete. + */ + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[1]; + if (reg_last < 0) { + reg_last = info.reg_off_split; + if (reg_last < 0) { + reg_last = info.reg_off_last[0]; + } + } + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + reg_off += esize; + } while (reg_off & 63); + } while (reg_off <= reg_last); + return; +#endif + } + + reg_off = info.reg_off_first[0]; + reg_last = info.reg_off_last[0]; + host = info.page[0].host; + + while (reg_off <= reg_last) { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } + reg_off += 1 << esz; + } while (reg_off <= reg_last && (reg_off & 63)); + } + + /* + * Use the slow path to manage the cross-page misalignment. + * But we know this is RAM and cannot trap. + */ + reg_off = info.reg_off_split; + if (unlikely(reg_off >= 0)) { + tlb_fn(env, za, reg_off, addr + reg_off, ra); + } + + reg_off = info.reg_off_first[1]; + if (unlikely(reg_off >= 0)) { + reg_last = info.reg_off_last[1]; + host = info.page[1].host; + + do { + uint64_t pg = vg[reg_off >> 6]; + do { + if ((pg >> (reg_off & 63)) & 1) { + host_fn(za, reg_off, host + reg_off); + } + reg_off += 1 << esz; + } while (reg_off & 63); + } while (reg_off <= reg_last); + } +} + +static inline QEMU_ALWAYS_INLINE +void sme_st1_mte(CPUARMState *env, void *za, uint64_t *vg, target_ulong addr, + uint32_t desc, uintptr_t ra, int esz, bool vertical, + sve_ldst1_host_fn *host_fn, + sve_ldst1_tlb_fn *tlb_fn) +{ + uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + int bit55 = extract64(addr, 55, 1); + + /* Remove mtedesc from the normal sve descriptor. */ + desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT); + + /* Perform gross MTE suppression early. */ + if (!tbi_check(desc, bit55) || + tcma_check(desc, bit55, allocation_tag_from_addr(addr))) { + mtedesc = 0; + } + + sme_st1(env, za, vg, addr, desc, ra, esz, mtedesc, + vertical, host_fn, tlb_fn); +} + +#define DO_ST(L, END, ESZ) \ +void HELPER(sme_st1##L##END##_h)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \ + sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \ +} \ +void HELPER(sme_st1##L##END##_v)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \ + sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \ +} \ +void HELPER(sme_st1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \ + sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \ +} \ +void HELPER(sme_st1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \ + target_ulong addr, uint32_t desc) \ +{ \ + sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \ + sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \ +} + +DO_ST(b, , MO_8) +DO_ST(h, _be, MO_16) +DO_ST(h, _le, MO_16) +DO_ST(s, _be, MO_32) +DO_ST(s, _le, MO_32) +DO_ST(d, _be, MO_64) +DO_ST(d, _le, MO_64) +DO_ST(q, _be, MO_128) +DO_ST(q, _le, MO_128) + +#undef DO_ST diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool trans_MOVA(DisasContext *s, arg_MOVA *a) return true; } + +static bool trans_LDST1(DisasContext *s, arg_LDST1 *a) +{ + typedef void GenLdSt1(TCGv_env, TCGv_ptr, TCGv_ptr, TCGv, TCGv_i32); + + /* + * Indexed by [esz][be][v][mte][st], which is (except for load/store) + * also the order in which the elements appear in the function names, + * and so how we must concatenate the pieces. + */ + +#define FN_LS(F) { gen_helper_sme_ld1##F, gen_helper_sme_st1##F } +#define FN_MTE(F) { FN_LS(F), FN_LS(F##_mte) } +#define FN_HV(F) { FN_MTE(F##_h), FN_MTE(F##_v) } +#define FN_END(L, B) { FN_HV(L), FN_HV(B) } + + static GenLdSt1 * const fns[5][2][2][2][2] = { + FN_END(b, b), + FN_END(h_le, h_be), + FN_END(s_le, s_be), + FN_END(d_le, d_be), + FN_END(q_le, q_be), + }; + +#undef FN_LS +#undef FN_MTE +#undef FN_HV +#undef FN_END + + TCGv_ptr t_za, t_pg; + TCGv_i64 addr; + int svl, desc = 0; + bool be = s->be_data == MO_BE; + bool mte = s->mte_active[0]; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sme_smza_enabled_check(s)) { + return true; + } + + t_za = get_tile_rowcol(s, a->esz, a->rs, a->za_imm, a->v); + t_pg = pred_full_reg_ptr(s, a->pg); + addr = tcg_temp_new_i64(); + + tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), a->esz); + tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn)); + + if (mte) { + desc = FIELD_DP32(desc, MTEDESC, MIDX, get_mem_index(s)); + desc = FIELD_DP32(desc, MTEDESC, TBI, s->tbid); + desc = FIELD_DP32(desc, MTEDESC, TCMA, s->tcma); + desc = FIELD_DP32(desc, MTEDESC, WRITE, a->st); + desc = FIELD_DP32(desc, MTEDESC, SIZEM1, (1 << a->esz) - 1); + desc <<= SVE_MTEDESC_SHIFT; + } else { + addr = clean_data_tbi(s, addr); + } + svl = streaming_vec_reg_size(s); + desc = simd_desc(svl, svl, desc); + + fns[a->esz][be][a->v][mte][a->st](cpu_env, t_za, t_pg, addr, + tcg_constant_i32(desc)); + + tcg_temp_free_ptr(t_za); + tcg_temp_free_ptr(t_pg); + tcg_temp_free_i64(addr); + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Add a TCGv_ptr base argument, which will be cpu_env for SVE. We will reuse this for SME save and restore array insns. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-22-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/translate-a64.h | 3 +++ target/arm/translate-sve.c | 48 ++++++++++++++++++++++++++++---------- 2 files changed, 39 insertions(+), 12 deletions(-) diff --git a/target/arm/translate-a64.h b/target/arm/translate-a64.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-a64.h +++ b/target/arm/translate-a64.h @@ -XXX,XX +XXX,XX @@ void gen_gvec_xar(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs, uint32_t rm_ofs, int64_t shift, uint32_t opr_sz, uint32_t max_sz); +void gen_sve_ldr(DisasContext *s, TCGv_ptr, int vofs, int len, int rn, int imm); +void gen_sve_str(DisasContext *s, TCGv_ptr, int vofs, int len, int rn, int imm); + #endif /* TARGET_ARM_TRANSLATE_A64_H */ diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(UCVTF_dd, aa64_sve, gen_gvec_fpst_arg_zpz, * The load should begin at the address Rn + IMM. */ -static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) +void gen_sve_ldr(DisasContext *s, TCGv_ptr base, int vofs, + int len, int rn, int imm) { int len_align = QEMU_ALIGN_DOWN(len, 8); int len_remain = len % 8; @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) t0 = tcg_temp_new_i64(); for (i = 0; i < len_align; i += 8) { tcg_gen_qemu_ld_i64(t0, clean_addr, midx, MO_LEUQ); - tcg_gen_st_i64(t0, cpu_env, vofs + i); + tcg_gen_st_i64(t0, base, vofs + i); tcg_gen_addi_i64(clean_addr, clean_addr, 8); } tcg_temp_free_i64(t0); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) clean_addr = new_tmp_a64_local(s); tcg_gen_mov_i64(clean_addr, t0); + if (base != cpu_env) { + TCGv_ptr b = tcg_temp_local_new_ptr(); + tcg_gen_mov_ptr(b, base); + base = b; + } + gen_set_label(loop); t0 = tcg_temp_new_i64(); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_addi_i64(clean_addr, clean_addr, 8); tp = tcg_temp_new_ptr(); - tcg_gen_add_ptr(tp, cpu_env, i); + tcg_gen_add_ptr(tp, base, i); tcg_gen_addi_ptr(i, i, 8); tcg_gen_st_i64(t0, tp, vofs); tcg_temp_free_ptr(tp); @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop); tcg_temp_free_ptr(i); + + if (base != cpu_env) { + tcg_temp_free_ptr(base); + assert(len_remain == 0); + } } /* @@ -XXX,XX +XXX,XX @@ static void do_ldr(DisasContext *s, uint32_t vofs, int len, int rn, int imm) default: g_assert_not_reached(); } - tcg_gen_st_i64(t0, cpu_env, vofs + len_align); + tcg_gen_st_i64(t0, base, vofs + len_align); tcg_temp_free_i64(t0); } } /* Similarly for stores. */ -static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) +void gen_sve_str(DisasContext *s, TCGv_ptr base, int vofs, + int len, int rn, int imm) { int len_align = QEMU_ALIGN_DOWN(len, 8); int len_remain = len % 8; @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) t0 = tcg_temp_new_i64(); for (i = 0; i < len_align; i += 8) { - tcg_gen_ld_i64(t0, cpu_env, vofs + i); + tcg_gen_ld_i64(t0, base, vofs + i); tcg_gen_qemu_st_i64(t0, clean_addr, midx, MO_LEUQ); tcg_gen_addi_i64(clean_addr, clean_addr, 8); } @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) clean_addr = new_tmp_a64_local(s); tcg_gen_mov_i64(clean_addr, t0); + if (base != cpu_env) { + TCGv_ptr b = tcg_temp_local_new_ptr(); + tcg_gen_mov_ptr(b, base); + base = b; + } + gen_set_label(loop); t0 = tcg_temp_new_i64(); tp = tcg_temp_new_ptr(); - tcg_gen_add_ptr(tp, cpu_env, i); + tcg_gen_add_ptr(tp, base, i); tcg_gen_ld_i64(t0, tp, vofs); tcg_gen_addi_ptr(i, i, 8); tcg_temp_free_ptr(tp); @@ -XXX,XX +XXX,XX @@ static void do_str(DisasContext *s, uint32_t vofs, int len, int rn, int imm) tcg_gen_brcondi_ptr(TCG_COND_LTU, i, len_align, loop); tcg_temp_free_ptr(i); + + if (base != cpu_env) { + tcg_temp_free_ptr(base); + assert(len_remain == 0); + } } /* Predicate register stores can be any multiple of 2. */ if (len_remain) { t0 = tcg_temp_new_i64(); - tcg_gen_ld_i64(t0, cpu_env, vofs + len_align); + tcg_gen_ld_i64(t0, base, vofs + len_align); switch (len_remain) { case 2: @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_zri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = vec_full_reg_size(s); int off = vec_full_reg_offset(s, a->rd); - do_ldr(s, off, size, a->rn, a->imm * size); + gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_LDR_pri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = pred_full_reg_size(s); int off = pred_full_reg_offset(s, a->rd); - do_ldr(s, off, size, a->rn, a->imm * size); + gen_sve_ldr(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STR_zri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = vec_full_reg_size(s); int off = vec_full_reg_offset(s, a->rd); - do_str(s, off, size, a->rn, a->imm * size); + gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } @@ -XXX,XX +XXX,XX @@ static bool trans_STR_pri(DisasContext *s, arg_rri *a) if (sve_access_check(s)) { int size = pred_full_reg_size(s); int off = pred_full_reg_offset(s, a->rd); - do_str(s, off, size, a->rn, a->imm * size); + gen_sve_str(s, cpu_env, off, size, a->rn, a->imm * size); } return true; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can reuse the SVE functions for LDR and STR, passing in the base of the ZA vector and a zero offset. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-23-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sme.decode | 7 +++++++ target/arm/translate-sme.c | 24 ++++++++++++++++++++++++ 2 files changed, 31 insertions(+) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ LDST1 1110000 0 esz:2 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ &ldst rs=%mova_rs LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ &ldst esz=4 rs=%mova_rs + +&ldstr rv rn imm +@ldstr ....... ... . ...... .. ... rn:5 . imm:4 \ + &ldstr rv=%mova_rs + +LDR 1110000 100 0 000000 .. 000 ..... 0 .... @ldstr +STR 1110000 100 1 000000 .. 000 ..... 0 .... @ldstr diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool trans_LDST1(DisasContext *s, arg_LDST1 *a) tcg_temp_free_i64(addr); return true; } + +typedef void GenLdStR(DisasContext *, TCGv_ptr, int, int, int, int); + +static bool do_ldst_r(DisasContext *s, arg_ldstr *a, GenLdStR *fn) +{ + int svl = streaming_vec_reg_size(s); + int imm = a->imm; + TCGv_ptr base; + + if (!sme_za_enabled_check(s)) { + return true; + } + + /* ZA[n] equates to ZA0H.B[n]. */ + base = get_tile_rowcol(s, MO_8, a->rv, imm, false); + + fn(s, base, 0, svl, a->rn, imm * svl); + + tcg_temp_free_ptr(base); + return true; +} + +TRANS_FEAT(LDR, aa64_sme, do_ldst_r, a, gen_sve_ldr) +TRANS_FEAT(STR, aa64_sme, do_ldst_r, a, gen_sve_str) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-24-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 5 +++ target/arm/sme.decode | 11 +++++ target/arm/sme_helper.c | 90 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 31 +++++++++++++ 4 files changed, 137 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_st1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i DEF_HELPER_FLAGS_5(sme_st1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_5(sme_st1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) DEF_HELPER_FLAGS_5(sme_st1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32) + +DEF_HELPER_FLAGS_5(sme_addha_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \ LDR 1110000 100 0 000000 .. 000 ..... 0 .... @ldstr STR 1110000 100 1 000000 .. 000 ..... 0 .... @ldstr + +### SME Add Vector to Array + +&adda zad zn pm pn +@adda_32 ........ .. ..... . pm:3 pn:3 zn:5 ... zad:2 &adda +@adda_64 ........ .. ..... . pm:3 pn:3 zn:5 .. zad:3 &adda + +ADDHA_s 11000000 10 01000 0 ... ... ..... 000 .. @adda_32 +ADDVA_s 11000000 10 01000 1 ... ... ..... 000 .. @adda_32 +ADDHA_d 11000000 11 01000 0 ... ... ..... 00 ... @adda_64 +ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ DO_ST(q, _be, MO_128) DO_ST(q, _le, MO_128) #undef DO_ST + +void HELPER(sme_addha_s)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 4; + uint64_t *pn = vpn, *pm = vpm; + uint32_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ) { + uint64_t pa = pn[row >> 4]; + do { + if (pa & 1) { + for (col = 0; col < oprsz; ) { + uint64_t pb = pm[col >> 4]; + do { + if (pb & 1) { + zda[tile_vslice_index(row) + H4(col)] += zn[H4(col)]; + } + pb >>= 4; + } while (++col & 15); + } + } + pa >>= 4; + } while (++row & 15); + } +} + +void HELPER(sme_addha_d)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint8_t *pn = vpn, *pm = vpm; + uint64_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + zda[tile_vslice_index(row) + col] += zn[col]; + } + } + } + } +} + +void HELPER(sme_addva_s)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 4; + uint64_t *pn = vpn, *pm = vpm; + uint32_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ) { + uint64_t pa = pn[row >> 4]; + do { + if (pa & 1) { + uint32_t zn_row = zn[H4(row)]; + for (col = 0; col < oprsz; ) { + uint64_t pb = pm[col >> 4]; + do { + if (pb & 1) { + zda[tile_vslice_index(row) + H4(col)] += zn_row; + } + pb >>= 4; + } while (++col & 15); + } + } + pa >>= 4; + } while (++row & 15); + } +} + +void HELPER(sme_addva_d)(void *vzda, void *vzn, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint8_t *pn = vpn, *pm = vpm; + uint64_t *zda = vzda, *zn = vzn; + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + uint64_t zn_row = zn[row]; + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + zda[tile_vslice_index(row) + col] += zn_row; + } + } + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool do_ldst_r(DisasContext *s, arg_ldstr *a, GenLdStR *fn) TRANS_FEAT(LDR, aa64_sme, do_ldst_r, a, gen_sve_ldr) TRANS_FEAT(STR, aa64_sme, do_ldst_r, a, gen_sve_str) + +static bool do_adda(DisasContext *s, arg_adda *a, MemOp esz, + gen_helper_gvec_4 *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, 0); + TCGv_ptr za, zn, pn, pm; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + + fn(za, zn, pn, pm, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + return true; +} + +TRANS_FEAT(ADDHA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addha_s) +TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) +TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) +TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-25-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 5 +++ target/arm/sme.decode | 9 +++++ target/arm/sme_helper.c | 69 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 32 ++++++++++++++++++ 4 files changed, 115 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_addha_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ADDHA_s 11000000 10 01000 0 ... ... ..... 000 .. @adda_32 ADDVA_s 11000000 10 01000 1 ... ... ..... 000 .. @adda_32 ADDHA_d 11000000 11 01000 0 ... ... ..... 00 ... @adda_64 ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 + +### SME Outer Product + +&op zad zn zm pm pn sub:bool +@op_32 ........ ... zm:5 pm:3 pn:3 zn:5 sub:1 .. zad:2 &op +@op_64 ........ ... zm:5 pm:3 pn:3 zn:5 sub:1 . zad:3 &op + +FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 +FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ #include "exec/cpu_ldst.h" #include "exec/exec-all.h" #include "qemu/int128.h" +#include "fpu/softfloat.h" #include "vec_internal.h" #include "sve_ldst_internal.h" @@ -XXX,XX +XXX,XX @@ void HELPER(sme_addva_d)(void *vzda, void *vzn, void *vpn, } } } + +void HELPER(sme_fmopa_s)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) << 31; + uint16_t *pn = vpn, *pm = vpm; + float_status fpst; + + /* + * Make a copy of float_status because this operation does not + * update the cumulative fp exception status. It also produces + * default nans. + */ + fpst = *(float_status *)vst; + set_default_nan_mode(true, &fpst); + + for (row = 0; row < oprsz; ) { + uint16_t pa = pn[H2(row >> 4)]; + do { + if (pa & 1) { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)) ^ neg; + + for (col = 0; col < oprsz; ) { + uint16_t pb = pm[H2(col >> 4)]; + do { + if (pb & 1) { + uint32_t *a = vza_row + H1_4(col); + uint32_t *m = vzm + H1_4(col); + *a = float32_muladd(n, *m, *a, 0, vst); + } + col += 4; + pb >>= 4; + } while (col & 15); + } + } + row += 4; + pa >>= 4; + } while (row & 15); + } +} + +void HELPER(sme_fmopa_d)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + uint64_t neg = (uint64_t)simd_data(desc) << 63; + uint64_t *za = vza, *zn = vzn, *zm = vzm; + uint8_t *pn = vpn, *pm = vpm; + float_status fpst = *(float_status *)vst; + + set_default_nan_mode(true, &fpst); + + for (row = 0; row < oprsz; ++row) { + if (pn[H1(row)] & 1) { + uint64_t *za_row = &za[tile_vslice_index(row)]; + uint64_t n = zn[row] ^ neg; + + for (col = 0; col < oprsz; ++col) { + if (pm[H1(col)] & 1) { + uint64_t *a = &za_row[col]; + *a = float64_muladd(n, zm[col], *a, 0, &fpst); + } + } + } + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(ADDHA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addha_s) TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) + +static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, + gen_helper_gvec_5_ptr *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, a->sub); + TCGv_ptr za, zn, zm, pn, pm, fpst; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + zm = vec_full_reg_ptr(s, a->zm); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + fpst = fpstatus_ptr(FPST_FPCR); + + fn(za, zn, zm, pn, pm, fpst, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + tcg_temp_free_ptr(fpst); + return true; +} + +TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) +TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-26-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 2 ++ target/arm/sme_helper.c | 56 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 30 ++++++++++++++++++++ 4 files changed, 90 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_bfmopa, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ ADDVA_d 11000000 11 01000 1 ... ... ..... 00 ... @adda_64 FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 + +BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sme_fmopa_d)(void *vza, void *vzn, void *vzm, void *vpn, } } } + +/* + * Alter PAIR as needed for controlling predicates being false, + * and for NEG on an enabled row element. + */ +static inline uint32_t f16mop_adj_pair(uint32_t pair, uint32_t pg, uint32_t neg) +{ + /* + * The pseudocode uses a conditional negate after the conditional zero. + * It is simpler here to unconditionally negate before conditional zero. + */ + pair ^= neg; + if (!(pg & 1)) { + pair &= 0xffff0000u; + } + if (!(pg & 4)) { + pair &= 0x0000ffffu; + } + return pair; +} + +void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) * 0x80008000u; + uint16_t *pn = vpn, *pm = vpm; + + for (row = 0; row < oprsz; ) { + uint16_t prow = pn[H2(row >> 4)]; + do { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)); + + n = f16mop_adj_pair(n, prow, neg); + + for (col = 0; col < oprsz; ) { + uint16_t pcol = pm[H2(col >> 4)]; + do { + if (prow & pcol & 0b0101) { + uint32_t *a = vza_row + H1_4(col); + uint32_t m = *(uint32_t *)(vzm + H1_4(col)); + + m = f16mop_adj_pair(m, pcol, 0); + *a = bfdotadd(*a, n, m); + + col += 4; + pcol >>= 4; + } + } while (col & 15); + } + row += 4; + prow >>= 4; + } while (row & 15); + } +} diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(ADDVA_s, aa64_sme, do_adda, a, MO_32, gen_helper_sme_addva_s) TRANS_FEAT(ADDHA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addha_d) TRANS_FEAT(ADDVA_d, aa64_sme_i16i64, do_adda, a, MO_64, gen_helper_sme_addva_d) +static bool do_outprod(DisasContext *s, arg_op *a, MemOp esz, + gen_helper_gvec_5 *fn) +{ + int svl = streaming_vec_reg_size(s); + uint32_t desc = simd_desc(svl, svl, a->sub); + TCGv_ptr za, zn, zm, pn, pm; + + if (!sme_smza_enabled_check(s)) { + return true; + } + + /* Sum XZR+zad to find ZAd. */ + za = get_tile_rowcol(s, esz, 31, a->zad, false); + zn = vec_full_reg_ptr(s, a->zn); + zm = vec_full_reg_ptr(s, a->zm); + pn = pred_full_reg_ptr(s, a->pn); + pm = pred_full_reg_ptr(s, a->pm); + + fn(za, zn, zm, pn, pm, tcg_constant_i32(desc)); + + tcg_temp_free_ptr(za); + tcg_temp_free_ptr(zn); + tcg_temp_free_ptr(pn); + tcg_temp_free_ptr(pm); + return true; +} + static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, gen_helper_gvec_5_ptr *fn) { @@ -XXX,XX +XXX,XX @@ static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) + +/* TODO: FEAT_EBF16 */ +TRANS_FEAT(BFMOPA, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_bfmopa) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-27-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 2 ++ target/arm/sme.decode | 1 + target/arm/sme_helper.c | 74 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 1 + 4 files changed, 78 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_5(sme_addva_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addha_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(sme_addva_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_7(sme_fmopa_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ FMOPA_s 10000000 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 +FMOPA_h 10000001 101 ..... ... ... ..... . 00 .. @op_32 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ static inline uint32_t f16mop_adj_pair(uint32_t pair, uint32_t pg, uint32_t neg) return pair; } +static float32 f16_dotadd(float32 sum, uint32_t e1, uint32_t e2, + float_status *s_std, float_status *s_odd) +{ + float64 e1r = float16_to_float64(e1 & 0xffff, true, s_std); + float64 e1c = float16_to_float64(e1 >> 16, true, s_std); + float64 e2r = float16_to_float64(e2 & 0xffff, true, s_std); + float64 e2c = float16_to_float64(e2 >> 16, true, s_std); + float64 t64; + float32 t32; + + /* + * The ARM pseudocode function FPDot performs both multiplies + * and the add with a single rounding operation. Emulate this + * by performing the first multiply in round-to-odd, then doing + * the second multiply as fused multiply-add, and rounding to + * float32 all in one step. + */ + t64 = float64_mul(e1r, e2r, s_odd); + t64 = float64r32_muladd(e1c, e2c, t64, 0, s_std); + + /* This conversion is exact, because we've already rounded. */ + t32 = float64_to_float32(t64, s_std); + + /* The final accumulation step is not fused. */ + return float32_add(sum, t32, s_std); +} + +void HELPER(sme_fmopa_h)(void *vza, void *vzn, void *vzm, void *vpn, + void *vpm, void *vst, uint32_t desc) +{ + intptr_t row, col, oprsz = simd_maxsz(desc); + uint32_t neg = simd_data(desc) * 0x80008000u; + uint16_t *pn = vpn, *pm = vpm; + float_status fpst_odd, fpst_std; + + /* + * Make a copy of float_status because this operation does not + * update the cumulative fp exception status. It also produces + * default nans. Make a second copy with round-to-odd -- see above. + */ + fpst_std = *(float_status *)vst; + set_default_nan_mode(true, &fpst_std); + fpst_odd = fpst_std; + set_float_rounding_mode(float_round_to_odd, &fpst_odd); + + for (row = 0; row < oprsz; ) { + uint16_t prow = pn[H2(row >> 4)]; + do { + void *vza_row = vza + tile_vslice_offset(row); + uint32_t n = *(uint32_t *)(vzn + H1_4(row)); + + n = f16mop_adj_pair(n, prow, neg); + + for (col = 0; col < oprsz; ) { + uint16_t pcol = pm[H2(col >> 4)]; + do { + if (prow & pcol & 0b0101) { + uint32_t *a = vza_row + H1_4(col); + uint32_t m = *(uint32_t *)(vzm + H1_4(col)); + + m = f16mop_adj_pair(m, pcol, 0); + *a = f16_dotadd(*a, n, m, &fpst_std, &fpst_odd); + + col += 4; + pcol >>= 4; + } + } while (col & 15); + } + row += 4; + prow >>= 4; + } while (row & 15); + } +} + void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, void *vpm, uint32_t desc) { diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ static bool do_outprod_fpst(DisasContext *s, arg_op *a, MemOp esz, return true; } +TRANS_FEAT(FMOPA_h, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_h) TRANS_FEAT(FMOPA_s, aa64_sme, do_outprod_fpst, a, MO_32, gen_helper_sme_fmopa_s) TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_fmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is SMOPA, SUMOPA, USMOPA_s, UMOPA, for both Int8 and Int16. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-28-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sme.h | 16 ++++++++ target/arm/sme.decode | 10 +++++ target/arm/sme_helper.c | 82 ++++++++++++++++++++++++++++++++++++++ target/arm/translate-sme.c | 10 +++++ 4 files changed, 118 insertions(+) diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sme.h +++ b/target/arm/helper-sme.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_7(sme_fmopa_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_6(sme_bfmopa, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_smopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_umopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_sumopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_usmopa_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_smopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_umopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_sumopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_6(sme_usmopa_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, ptr, i32) diff --git a/target/arm/sme.decode b/target/arm/sme.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme.decode +++ b/target/arm/sme.decode @@ -XXX,XX +XXX,XX @@ FMOPA_d 10000000 110 ..... ... ... ..... . 0 ... @op_64 BFMOPA 10000001 100 ..... ... ... ..... . 00 .. @op_32 FMOPA_h 10000001 101 ..... ... ... ..... . 00 .. @op_32 + +SMOPA_s 1010000 0 10 0 ..... ... ... ..... . 00 .. @op_32 +SUMOPA_s 1010000 0 10 1 ..... ... ... ..... . 00 .. @op_32 +USMOPA_s 1010000 1 10 0 ..... ... ... ..... . 00 .. @op_32 +UMOPA_s 1010000 1 10 1 ..... ... ... ..... . 00 .. @op_32 + +SMOPA_d 1010000 0 11 0 ..... ... ... ..... . 0 ... @op_64 +SUMOPA_d 1010000 0 11 1 ..... ... ... ..... . 0 ... @op_64 +USMOPA_d 1010000 1 11 0 ..... ... ... ..... . 0 ... @op_64 +UMOPA_d 1010000 1 11 1 ..... ... ... ..... . 0 ... @op_64 diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sme_helper.c +++ b/target/arm/sme_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(sme_bfmopa)(void *vza, void *vzn, void *vzm, void *vpn, } while (row & 15); } } + +typedef uint64_t IMOPFn(uint64_t, uint64_t, uint64_t, uint8_t, bool); + +static inline void do_imopa(uint64_t *za, uint64_t *zn, uint64_t *zm, + uint8_t *pn, uint8_t *pm, + uint32_t desc, IMOPFn *fn) +{ + intptr_t row, col, oprsz = simd_oprsz(desc) / 8; + bool neg = simd_data(desc); + + for (row = 0; row < oprsz; ++row) { + uint8_t pa = pn[H1(row)]; + uint64_t *za_row = &za[tile_vslice_index(row)]; + uint64_t n = zn[row]; + + for (col = 0; col < oprsz; ++col) { + uint8_t pb = pm[H1(col)]; + uint64_t *a = &za_row[col]; + + *a = fn(n, zm[col], *a, pa & pb, neg); + } + } +} + +#define DEF_IMOP_32(NAME, NTYPE, MTYPE) \ +static uint64_t NAME(uint64_t n, uint64_t m, uint64_t a, uint8_t p, bool neg) \ +{ \ + uint32_t sum0 = 0, sum1 = 0; \ + /* Apply P to N as a mask, making the inactive elements 0. */ \ + n &= expand_pred_b(p); \ + sum0 += (NTYPE)(n >> 0) * (MTYPE)(m >> 0); \ + sum0 += (NTYPE)(n >> 8) * (MTYPE)(m >> 8); \ + sum0 += (NTYPE)(n >> 16) * (MTYPE)(m >> 16); \ + sum0 += (NTYPE)(n >> 24) * (MTYPE)(m >> 24); \ + sum1 += (NTYPE)(n >> 32) * (MTYPE)(m >> 32); \ + sum1 += (NTYPE)(n >> 40) * (MTYPE)(m >> 40); \ + sum1 += (NTYPE)(n >> 48) * (MTYPE)(m >> 48); \ + sum1 += (NTYPE)(n >> 56) * (MTYPE)(m >> 56); \ + if (neg) { \ + sum0 = (uint32_t)a - sum0, sum1 = (uint32_t)(a >> 32) - sum1; \ + } else { \ + sum0 = (uint32_t)a + sum0, sum1 = (uint32_t)(a >> 32) + sum1; \ + } \ + return ((uint64_t)sum1 << 32) | sum0; \ +} + +#define DEF_IMOP_64(NAME, NTYPE, MTYPE) \ +static uint64_t NAME(uint64_t n, uint64_t m, uint64_t a, uint8_t p, bool neg) \ +{ \ + uint64_t sum = 0; \ + /* Apply P to N as a mask, making the inactive elements 0. */ \ + n &= expand_pred_h(p); \ + sum += (NTYPE)(n >> 0) * (MTYPE)(m >> 0); \ + sum += (NTYPE)(n >> 16) * (MTYPE)(m >> 16); \ + sum += (NTYPE)(n >> 32) * (MTYPE)(m >> 32); \ + sum += (NTYPE)(n >> 48) * (MTYPE)(m >> 48); \ + return neg ? a - sum : a + sum; \ +} + +DEF_IMOP_32(smopa_s, int8_t, int8_t) +DEF_IMOP_32(umopa_s, uint8_t, uint8_t) +DEF_IMOP_32(sumopa_s, int8_t, uint8_t) +DEF_IMOP_32(usmopa_s, uint8_t, int8_t) + +DEF_IMOP_64(smopa_d, int16_t, int16_t) +DEF_IMOP_64(umopa_d, uint16_t, uint16_t) +DEF_IMOP_64(sumopa_d, int16_t, uint16_t) +DEF_IMOP_64(usmopa_d, uint16_t, int16_t) + +#define DEF_IMOPH(NAME) \ + void HELPER(sme_##NAME)(void *vza, void *vzn, void *vzm, void *vpn, \ + void *vpm, uint32_t desc) \ + { do_imopa(vza, vzn, vzm, vpn, vpm, desc, NAME); } + +DEF_IMOPH(smopa_s) +DEF_IMOPH(umopa_s) +DEF_IMOPH(sumopa_s) +DEF_IMOPH(usmopa_s) +DEF_IMOPH(smopa_d) +DEF_IMOPH(umopa_d) +DEF_IMOPH(sumopa_d) +DEF_IMOPH(usmopa_d) diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sme.c +++ b/target/arm/translate-sme.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(FMOPA_d, aa64_sme_f64f64, do_outprod_fpst, a, MO_64, gen_helper_sme_f /* TODO: FEAT_EBF16 */ TRANS_FEAT(BFMOPA, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_bfmopa) + +TRANS_FEAT(SMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_smopa_s) +TRANS_FEAT(UMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_umopa_s) +TRANS_FEAT(SUMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_sumopa_s) +TRANS_FEAT(USMOPA_s, aa64_sme, do_outprod, a, MO_32, gen_helper_sme_usmopa_s) + +TRANS_FEAT(SMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_smopa_d) +TRANS_FEAT(UMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_umopa_d) +TRANS_FEAT(SUMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_sumopa_d) +TRANS_FEAT(USMOPA_d, aa64_sme_i16i64, do_outprod, a, MO_64, gen_helper_sme_usmopa_d) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-29-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/sve.decode | 20 +++++++++++++ target/arm/translate-sve.c | 57 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 77 insertions(+) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ BFMLALT_zzxw 01100100 11 1 ..... 0100.1 ..... ..... @rrxr_3a esz=2 ### SVE2 floating-point bfloat16 dot-product (indexed) BFDOT_zzxz 01100100 01 1 ..... 010000 ..... ..... @rrxr_2 esz=2 + +### SVE broadcast predicate element + +&psel esz pd pn pm rv imm +%psel_rv 16:2 !function=plus_12 +%psel_imm_b 22:2 19:2 +%psel_imm_h 22:2 20:1 +%psel_imm_s 22:2 +%psel_imm_d 23:1 +@psel ........ .. . ... .. .. pn:4 . pm:4 . pd:4 \ + &psel rv=%psel_rv + +PSEL 00100101 .. 1 ..1 .. 01 .... 0 .... 0 .... \ + @psel esz=0 imm=%psel_imm_b +PSEL 00100101 .. 1 .10 .. 01 .... 0 .... 0 .... \ + @psel esz=1 imm=%psel_imm_h +PSEL 00100101 .. 1 100 .. 01 .... 0 .... 0 .... \ + @psel esz=2 imm=%psel_imm_s +PSEL 00100101 .1 1 000 .. 01 .... 0 .... 0 .... \ + @psel esz=3 imm=%psel_imm_d diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool do_BFMLAL_zzxw(DisasContext *s, arg_rrxr_esz *a, bool sel) TRANS_FEAT(BFMLALB_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, false) TRANS_FEAT(BFMLALT_zzxw, aa64_sve_bf16, do_BFMLAL_zzxw, a, true) + +static bool trans_PSEL(DisasContext *s, arg_psel *a) +{ + int vl = vec_full_reg_size(s); + int pl = pred_gvec_reg_size(s); + int elements = vl >> a->esz; + TCGv_i64 tmp, didx, dbit; + TCGv_ptr ptr; + + if (!dc_isar_feature(aa64_sme, s)) { + return false; + } + if (!sve_access_check(s)) { + return true; + } + + tmp = tcg_temp_new_i64(); + dbit = tcg_temp_new_i64(); + didx = tcg_temp_new_i64(); + ptr = tcg_temp_new_ptr(); + + /* Compute the predicate element. */ + tcg_gen_addi_i64(tmp, cpu_reg(s, a->rv), a->imm); + if (is_power_of_2(elements)) { + tcg_gen_andi_i64(tmp, tmp, elements - 1); + } else { + tcg_gen_remu_i64(tmp, tmp, tcg_constant_i64(elements)); + } + + /* Extract the predicate byte and bit indices. */ + tcg_gen_shli_i64(tmp, tmp, a->esz); + tcg_gen_andi_i64(dbit, tmp, 7); + tcg_gen_shri_i64(didx, tmp, 3); + if (HOST_BIG_ENDIAN) { + tcg_gen_xori_i64(didx, didx, 7); + } + + /* Load the predicate word. */ + tcg_gen_trunc_i64_ptr(ptr, didx); + tcg_gen_add_ptr(ptr, ptr, cpu_env); + tcg_gen_ld8u_i64(tmp, ptr, pred_full_reg_offset(s, a->pm)); + + /* Extract the predicate bit and replicate to MO_64. */ + tcg_gen_shr_i64(tmp, tmp, dbit); + tcg_gen_andi_i64(tmp, tmp, 1); + tcg_gen_neg_i64(tmp, tmp); + + /* Apply to either copy the source, or write zeros. */ + tcg_gen_gvec_ands(MO_64, pred_full_reg_offset(s, a->pd), + pred_full_reg_offset(s, a->pn), tmp, pl, pl); + + tcg_temp_free_i64(tmp); + tcg_temp_free_i64(dbit); + tcg_temp_free_i64(didx); + tcg_temp_free_ptr(ptr); + return true; +} -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-30-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper-sve.h | 2 ++ target/arm/sve.decode | 1 + target/arm/sve_helper.c | 16 ++++++++++++++++ target/arm/translate-sve.c | 2 ++ 4 files changed, 21 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(sve_revh_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_revw_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(sme_revd_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(sve_rbit_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(sve_rbit_s, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ REVB 00000101 .. 1001 00 100 ... ..... ..... @rd_pg_rn REVH 00000101 .. 1001 01 100 ... ..... ..... @rd_pg_rn REVW 00000101 .. 1001 10 100 ... ..... ..... @rd_pg_rn RBIT 00000101 .. 1001 11 100 ... ..... ..... @rd_pg_rn +REVD 00000101 00 1011 10 100 ... ..... ..... @rd_pg_rn_e0 # SVE vector splice (predicated, destructive) SPLICE 00000101 .. 101 100 100 ... ..... ..... @rdn_pg_rm diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -XXX,XX +XXX,XX @@ DO_ZPZ_D(sve_revh_d, uint64_t, hswap64) DO_ZPZ_D(sve_revw_d, uint64_t, wswap64) +void HELPER(sme_revd_q)(void *vd, void *vn, void *vg, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc) / 8; + uint64_t *d = vd, *n = vn; + uint8_t *pg = vg; + + for (i = 0; i < opr_sz; i += 2) { + if (pg[H1(i)] & 1) { + uint64_t n0 = n[i + 0]; + uint64_t n1 = n[i + 1]; + d[i + 0] = n1; + d[i + 1] = n0; + } + } +} + DO_ZPZ(sve_rbit_b, uint8_t, H1, revbit8) DO_ZPZ(sve_rbit_h, uint16_t, H1_2, revbit16) DO_ZPZ(sve_rbit_s, uint32_t, H1_4, revbit32) diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ TRANS_FEAT(REVH, aa64_sve, gen_gvec_ool_arg_zpz, revh_fns[a->esz], a, 0) TRANS_FEAT(REVW, aa64_sve, gen_gvec_ool_arg_zpz, a->esz == 3 ? gen_helper_sve_revw_d : NULL, a, 0) +TRANS_FEAT(REVD, aa64_sme, gen_gvec_ool_arg_zpz, gen_helper_sme_revd_q, a, 0) + TRANS_FEAT(SPLICE, aa64_sve, gen_gvec_ool_arg_zpzz, gen_helper_sve_splice, a, a->esz) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> This is an SVE instruction that operates using the SVE vector length but that it is present only if SME is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-31-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.h | 18 +++++++ target/arm/sve.decode | 5 ++ target/arm/translate-sve.c | 102 +++++++++++++++++++++++++++++++++++++ target/arm/vec_helper.c | 24 +++++++++ 4 files changed, 149 insertions(+) diff --git a/target/arm/helper.h b/target/arm/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_6(gvec_bfmlal, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_6(gvec_bfmlal_idx, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_b, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_sclamp_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_5(gvec_uclamp_b, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_h, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_uclamp_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + #ifdef TARGET_AARCH64 #include "helper-a64.h" #include "helper-sve.h" diff --git a/target/arm/sve.decode b/target/arm/sve.decode index XXXXXXX..XXXXXXX 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -XXX,XX +XXX,XX @@ PSEL 00100101 .. 1 100 .. 01 .... 0 .... 0 .... \ @psel esz=2 imm=%psel_imm_s PSEL 00100101 .1 1 000 .. 01 .... 0 .... 0 .... \ @psel esz=3 imm=%psel_imm_d + +### SVE clamp + +SCLAMP 01000100 .. 0 ..... 110000 ..... ..... @rda_rn_rm +UCLAMP 01000100 .. 0 ..... 110001 ..... ..... @rda_rn_rm diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -XXX,XX +XXX,XX @@ static bool trans_PSEL(DisasContext *s, arg_psel *a) tcg_temp_free_ptr(ptr); return true; } + +static void gen_sclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a) +{ + tcg_gen_smax_i32(d, a, n); + tcg_gen_smin_i32(d, d, m); +} + +static void gen_sclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a) +{ + tcg_gen_smax_i64(d, a, n); + tcg_gen_smin_i64(d, d, m); +} + +static void gen_sclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n, + TCGv_vec m, TCGv_vec a) +{ + tcg_gen_smax_vec(vece, d, a, n); + tcg_gen_smin_vec(vece, d, d, m); +} + +static void gen_sclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m, + uint32_t a, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop[] = { + INDEX_op_smin_vec, INDEX_op_smax_vec, 0 + }; + static const GVecGen4 ops[4] = { + { .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_b, + .opt_opc = vecop, + .vece = MO_8 }, + { .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_h, + .opt_opc = vecop, + .vece = MO_16 }, + { .fni4 = gen_sclamp_i32, + .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_s, + .opt_opc = vecop, + .vece = MO_32 }, + { .fni8 = gen_sclamp_i64, + .fniv = gen_sclamp_vec, + .fno = gen_helper_gvec_sclamp_d, + .opt_opc = vecop, + .vece = MO_64, + .prefer_i64 = TCG_TARGET_REG_BITS == 64 } + }; + tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]); +} + +TRANS_FEAT(SCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_sclamp, a) + +static void gen_uclamp_i32(TCGv_i32 d, TCGv_i32 n, TCGv_i32 m, TCGv_i32 a) +{ + tcg_gen_umax_i32(d, a, n); + tcg_gen_umin_i32(d, d, m); +} + +static void gen_uclamp_i64(TCGv_i64 d, TCGv_i64 n, TCGv_i64 m, TCGv_i64 a) +{ + tcg_gen_umax_i64(d, a, n); + tcg_gen_umin_i64(d, d, m); +} + +static void gen_uclamp_vec(unsigned vece, TCGv_vec d, TCGv_vec n, + TCGv_vec m, TCGv_vec a) +{ + tcg_gen_umax_vec(vece, d, a, n); + tcg_gen_umin_vec(vece, d, d, m); +} + +static void gen_uclamp(unsigned vece, uint32_t d, uint32_t n, uint32_t m, + uint32_t a, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop[] = { + INDEX_op_umin_vec, INDEX_op_umax_vec, 0 + }; + static const GVecGen4 ops[4] = { + { .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_b, + .opt_opc = vecop, + .vece = MO_8 }, + { .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_h, + .opt_opc = vecop, + .vece = MO_16 }, + { .fni4 = gen_uclamp_i32, + .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_s, + .opt_opc = vecop, + .vece = MO_32 }, + { .fni8 = gen_uclamp_i64, + .fniv = gen_uclamp_vec, + .fno = gen_helper_gvec_uclamp_d, + .opt_opc = vecop, + .vece = MO_64, + .prefer_i64 = TCG_TARGET_REG_BITS == 64 } + }; + tcg_gen_gvec_4(d, n, m, a, oprsz, maxsz, &ops[vece]); +} + +TRANS_FEAT(UCLAMP, aa64_sme, gen_gvec_fn_arg_zzzz, gen_uclamp, a) diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_bfmlal_idx)(void *vd, void *vn, void *vm, } clear_tail(d, opr_sz, simd_maxsz(desc)); } + +#define DO_CLAMP(NAME, TYPE) \ +void HELPER(NAME)(void *d, void *n, void *m, void *a, uint32_t desc) \ +{ \ + intptr_t i, opr_sz = simd_oprsz(desc); \ + for (i = 0; i < opr_sz; i += sizeof(TYPE)) { \ + TYPE aa = *(TYPE *)(a + i); \ + TYPE nn = *(TYPE *)(n + i); \ + TYPE mm = *(TYPE *)(m + i); \ + TYPE dd = MIN(MAX(aa, nn), mm); \ + *(TYPE *)(d + i) = dd; \ + } \ + clear_tail(d, opr_sz, simd_maxsz(desc)); \ +} + +DO_CLAMP(gvec_sclamp_b, int8_t) +DO_CLAMP(gvec_sclamp_h, int16_t) +DO_CLAMP(gvec_sclamp_s, int32_t) +DO_CLAMP(gvec_sclamp_d, int64_t) + +DO_CLAMP(gvec_uclamp_b, uint8_t) +DO_CLAMP(gvec_uclamp_h, uint16_t) +DO_CLAMP(gvec_uclamp_s, uint32_t) +DO_CLAMP(gvec_uclamp_d, uint64_t) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> We can handle both exception entry and exception return by hooking into aarch64_sve_change_el. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-32-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/helper.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) diff --git a/target/arm/helper.c b/target/arm/helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/helper.c +++ b/target/arm/helper.c @@ -XXX,XX +XXX,XX @@ void aarch64_sve_change_el(CPUARMState *env, int old_el, return; } + old_a64 = old_el ? arm_el_is_aa64(env, old_el) : el0_a64; + new_a64 = new_el ? arm_el_is_aa64(env, new_el) : el0_a64; + + /* + * Both AArch64.TakeException and AArch64.ExceptionReturn + * invoke ResetSVEState when taking an exception from, or + * returning to, AArch32 state when PSTATE.SM is enabled. + */ + if (old_a64 != new_a64 && FIELD_EX64(env->svcr, SVCR, SM)) { + arm_reset_sve_state(env); + return; + } + /* * DDI0584A.d sec 3.2: "If SVE instructions are disabled or trapped * at ELx, or not available because the EL is in AArch32 state, then @@ -XXX,XX +XXX,XX @@ void aarch64_sve_change_el(CPUARMState *env, int old_el, * we already have the correct register contents when encountering the * vq0->vq0 transition between EL0->EL1. */ - old_a64 = old_el ? arm_el_is_aa64(env, old_el) : el0_a64; old_len = (old_a64 && !sve_exception_el(env, old_el) ? sve_vqm1_for_el(env, old_el) : 0); - new_a64 = new_el ? arm_el_is_aa64(env, new_el) : el0_a64; new_len = (new_a64 && !sve_exception_el(env, new_el) ? sve_vqm1_for_el(env, new_el) : 0); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Note that SME remains effectively disabled for user-only, because we do not yet set CPACR_EL1.SMEN. This needs to wait until the kernel ABI is implemented. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-33-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- docs/system/arm/emulation.rst | 4 ++++ target/arm/cpu64.c | 11 +++++++++++ 2 files changed, 15 insertions(+) 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_SHA512 (Advanced SIMD SHA512 instructions) - FEAT_SM3 (Advanced SIMD SM3 instructions) - FEAT_SM4 (Advanced SIMD SM4 instructions) +- FEAT_SME (Scalable Matrix Extension) +- FEAT_SME_FA64 (Full A64 instruction set in Streaming SVE mode) +- FEAT_SME_F64F64 (Double-precision floating-point outer product instructions) +- FEAT_SME_I16I64 (16-bit to 64-bit integer widening outer product instructions) - FEAT_SPECRES (Speculation restriction instructions) - FEAT_SSBS (Speculative Store Bypass Safe) - FEAT_TLBIOS (TLB invalidate instructions in Outer Shareable domain) diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu64.c +++ b/target/arm/cpu64.c @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) */ t = FIELD_DP64(t, ID_AA64PFR1, MTE, 3); /* FEAT_MTE3 */ t = FIELD_DP64(t, ID_AA64PFR1, RAS_FRAC, 0); /* FEAT_RASv1p1 + FEAT_DoubleFault */ + t = FIELD_DP64(t, ID_AA64PFR1, SME, 1); /* FEAT_SME */ t = FIELD_DP64(t, ID_AA64PFR1, CSV2_FRAC, 0); /* FEAT_CSV2_2 */ cpu->isar.id_aa64pfr1 = t; @@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj) t = FIELD_DP64(t, ID_AA64DFR0, PMUVER, 5); /* FEAT_PMUv3p4 */ cpu->isar.id_aa64dfr0 = t; + t = cpu->isar.id_aa64smfr0; + t = FIELD_DP64(t, ID_AA64SMFR0, F32F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, B16F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, F16F32, 1); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, I8I32, 0xf); /* FEAT_SME */ + t = FIELD_DP64(t, ID_AA64SMFR0, F64F64, 1); /* FEAT_SME_F64F64 */ + t = FIELD_DP64(t, ID_AA64SMFR0, I16I64, 0xf); /* FEAT_SME_I16I64 */ + t = FIELD_DP64(t, ID_AA64SMFR0, FA64, 1); /* FEAT_SME_FA64 */ + cpu->isar.id_aa64smfr0 = t; + /* Replicate the same data to the 32-bit id registers. */ aa32_max_features(cpu); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-34-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_cpu.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/target_cpu.h b/linux-user/aarch64/target_cpu.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_cpu.h +++ b/linux-user/aarch64/target_cpu.h @@ -XXX,XX +XXX,XX @@ static inline void cpu_clone_regs_parent(CPUARMState *env, unsigned flags) static inline void cpu_set_tls(CPUARMState *env, target_ulong newtls) { - /* Note that AArch64 Linux keeps the TLS pointer in TPIDR; this is + /* + * Note that AArch64 Linux keeps the TLS pointer in TPIDR; this is * different from AArch32 Linux, which uses TPIDRRO. */ env->cp15.tpidr_el[0] = newtls; + /* TPIDR2_EL0 is cleared with CLONE_SETTLS. */ + env->cp15.tpidr2_el0 = 0; } static inline abi_ulong get_sp_from_cpustate(CPUARMState *state) -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-35-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/cpu_loop.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/linux-user/aarch64/cpu_loop.c b/linux-user/aarch64/cpu_loop.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/cpu_loop.c +++ b/linux-user/aarch64/cpu_loop.c @@ -XXX,XX +XXX,XX @@ void cpu_loop(CPUARMState *env) switch (trapnr) { case EXCP_SWI: + /* + * On syscall, PSTATE.ZA is preserved, along with the ZA matrix. + * PSTATE.SM is cleared, per SMSTOP, which does ResetSVEState. + */ + if (FIELD_EX64(env->svcr, SVCR, SM)) { + env->svcr = FIELD_DP64(env->svcr, SVCR, SM, 0); + arm_rebuild_hflags(env); + arm_reset_sve_state(env); + } ret = do_syscall(env, env->xregs[8], env->xregs[0], -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Make sure to zero the currently reserved fields. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-36-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ struct target_extra_context { struct target_sve_context { struct target_aarch64_ctx head; uint16_t vl; - uint16_t reserved[3]; + uint16_t flags; + uint16_t reserved[2]; /* The actual SVE data immediately follows. It is laid out * according to TARGET_SVE_SIG_{Z,P}REG_OFFSET, based off of * the original struct pointer. @@ -XXX,XX +XXX,XX @@ struct target_sve_context { #define TARGET_SVE_SIG_CONTEXT_SIZE(VQ) \ (TARGET_SVE_SIG_PREG_OFFSET(VQ, 17)) +#define TARGET_SVE_SIG_FLAG_SM 1 + struct target_rt_sigframe { struct target_siginfo info; struct target_ucontext uc; @@ -XXX,XX +XXX,XX @@ static void target_setup_sve_record(struct target_sve_context *sve, { int i, j; + memset(sve, 0, sizeof(*sve)); __put_user(TARGET_SVE_MAGIC, &sve->head.magic); __put_user(size, &sve->head.size); __put_user(vq * TARGET_SVE_VQ_BYTES, &sve->vl); + if (FIELD_EX64(env->svcr, SVCR, SM)) { + __put_user(TARGET_SVE_SIG_FLAG_SM, &sve->flags); + } /* Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian store -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Fold the return value setting into the goto, so each point of failure need not do both. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-37-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 26 +++++++++++--------------- 1 file changed, 11 insertions(+), 15 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, struct target_sve_context *sve = NULL; uint64_t extra_datap = 0; bool used_extra = false; - bool err = false; int vq = 0, sve_size = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, switch (magic) { case 0: if (size != 0) { - err = true; - goto exit; + goto err; } if (used_extra) { ctx = NULL; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, case TARGET_FPSIMD_MAGIC: if (fpsimd || size != sizeof(struct target_fpsimd_context)) { - err = true; - goto exit; + goto err; } fpsimd = (struct target_fpsimd_context *)ctx; break; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, break; } } - err = true; - goto exit; + goto err; case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { - err = true; - goto exit; + goto err; } __get_user(extra_datap, &((struct target_extra_context *)ctx)->datap); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, /* Unknown record -- we certainly didn't generate it. * Did we in fact get out of sync? */ - err = true; - goto exit; + goto err; } ctx = (void *)ctx + size; } @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, if (fpsimd) { target_restore_fpsimd_record(env, fpsimd); } else { - err = true; + goto err; } /* SVE data, if present, overwrites FPSIMD data. */ if (sve) { target_restore_sve_record(env, sve, vq); } - - exit: unlock_user(extra, extra_datap, 0); - return err; + return 0; + + err: + unlock_user(extra, extra_datap, 0); + return 1; } static abi_ulong get_sigframe(struct target_sigaction *ka, -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> In parse_user_sigframe, the kernel rejects duplicate sve records, or records that are smaller than the header. We were silently allowing these cases to pass, dropping the record. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-38-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, break; case TARGET_SVE_MAGIC: + if (sve || size < sizeof(struct target_sve_context)) { + goto err; + } if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { vq = sve_vq(env); sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); - if (!sve && size == sve_size) { + if (size == sve_size) { sve = (struct target_sve_context *)ctx; break; } -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-39-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, __get_user(extra_size, &((struct target_extra_context *)ctx)->size); extra = lock_user(VERIFY_READ, extra_datap, extra_size, 0); + if (!extra) { + return 1; + } break; default: -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Move the checks out of the parsing loop and into the restore function. This more closely mirrors the code structure in the kernel, and is slightly clearer. Reject rather than silently skip incorrect VL and SVE record sizes, bringing our checks in to line with those the kernel does. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-40-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 51 +++++++++++++++++++++++++------------ 1 file changed, 35 insertions(+), 16 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ static void target_restore_fpsimd_record(CPUARMState *env, } } -static void target_restore_sve_record(CPUARMState *env, - struct target_sve_context *sve, int vq) +static bool target_restore_sve_record(CPUARMState *env, + struct target_sve_context *sve, + int size) { - int i, j; + int i, j, vl, vq; - /* Note that SVE regs are stored as a byte stream, with each byte element + if (!cpu_isar_feature(aa64_sve, env_archcpu(env))) { + return false; + } + + __get_user(vl, &sve->vl); + vq = sve_vq(env); + + /* Reject mismatched VL. */ + if (vl != vq * TARGET_SVE_VQ_BYTES) { + return false; + } + + /* Accept empty record -- used to clear PSTATE.SM. */ + if (size <= sizeof(*sve)) { + return true; + } + + /* Reject non-empty but incomplete record. */ + if (size < TARGET_SVE_SIG_CONTEXT_SIZE(vq)) { + return false; + } + + /* + * Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian load * of our 64-bit hunks. */ @@ -XXX,XX +XXX,XX @@ static void target_restore_sve_record(CPUARMState *env, } } } + return true; } static int target_restore_sigframe(CPUARMState *env, @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, struct target_sve_context *sve = NULL; uint64_t extra_datap = 0; bool used_extra = false; - int vq = 0, sve_size = 0; + int sve_size = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, if (sve || size < sizeof(struct target_sve_context)) { goto err; } - if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { - vq = sve_vq(env); - sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); - if (size == sve_size) { - sve = (struct target_sve_context *)ctx; - break; - } - } - goto err; + sve = (struct target_sve_context *)ctx; + sve_size = size; + break; case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, } /* SVE data, if present, overwrites FPSIMD data. */ - if (sve) { - target_restore_sve_record(env, sve, vq); + if (sve && !target_restore_sve_record(env, sve, sve_size)) { + goto err; } unlock_user(extra, extra_datap, 0); return 0; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Set the SM bit in the SVE record on signal delivery, create the ZA record. Restore SM and ZA state according to the records present on return. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-41-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/signal.c | 167 +++++++++++++++++++++++++++++++++--- 1 file changed, 154 insertions(+), 13 deletions(-) diff --git a/linux-user/aarch64/signal.c b/linux-user/aarch64/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/signal.c +++ b/linux-user/aarch64/signal.c @@ -XXX,XX +XXX,XX @@ struct target_sve_context { #define TARGET_SVE_SIG_FLAG_SM 1 +#define TARGET_ZA_MAGIC 0x54366345 + +struct target_za_context { + struct target_aarch64_ctx head; + uint16_t vl; + uint16_t reserved[3]; + /* The actual ZA data immediately follows. */ +}; + +#define TARGET_ZA_SIG_REGS_OFFSET \ + QEMU_ALIGN_UP(sizeof(struct target_za_context), TARGET_SVE_VQ_BYTES) +#define TARGET_ZA_SIG_ZAV_OFFSET(VQ, N) \ + (TARGET_ZA_SIG_REGS_OFFSET + (VQ) * TARGET_SVE_VQ_BYTES * (N)) +#define TARGET_ZA_SIG_CONTEXT_SIZE(VQ) \ + TARGET_ZA_SIG_ZAV_OFFSET(VQ, VQ * TARGET_SVE_VQ_BYTES) + struct target_rt_sigframe { struct target_siginfo info; struct target_ucontext uc; @@ -XXX,XX +XXX,XX @@ static void target_setup_end_record(struct target_aarch64_ctx *end) } static void target_setup_sve_record(struct target_sve_context *sve, - CPUARMState *env, int vq, int size) + CPUARMState *env, int size) { - int i, j; + int i, j, vq = sve_vq(env); memset(sve, 0, sizeof(*sve)); __put_user(TARGET_SVE_MAGIC, &sve->head.magic); @@ -XXX,XX +XXX,XX @@ static void target_setup_sve_record(struct target_sve_context *sve, } } +static void target_setup_za_record(struct target_za_context *za, + CPUARMState *env, int size) +{ + int vq = sme_vq(env); + int vl = vq * TARGET_SVE_VQ_BYTES; + int i, j; + + memset(za, 0, sizeof(*za)); + __put_user(TARGET_ZA_MAGIC, &za->head.magic); + __put_user(size, &za->head.size); + __put_user(vl, &za->vl); + + if (size == TARGET_ZA_SIG_CONTEXT_SIZE(0)) { + return; + } + assert(size == TARGET_ZA_SIG_CONTEXT_SIZE(vq)); + + /* + * Note that ZA vectors are stored as a byte stream, + * with each byte element at a subsequent address. + */ + for (i = 0; i < vl; ++i) { + uint64_t *z = (void *)za + TARGET_ZA_SIG_ZAV_OFFSET(vq, i); + for (j = 0; j < vq * 2; ++j) { + __put_user_e(env->zarray[i].d[j], z + j, le); + } + } +} + static void target_restore_general_frame(CPUARMState *env, struct target_rt_sigframe *sf) { @@ -XXX,XX +XXX,XX @@ static void target_restore_fpsimd_record(CPUARMState *env, static bool target_restore_sve_record(CPUARMState *env, struct target_sve_context *sve, - int size) + int size, int *svcr) { - int i, j, vl, vq; + int i, j, vl, vq, flags; + bool sm; - if (!cpu_isar_feature(aa64_sve, env_archcpu(env))) { + __get_user(vl, &sve->vl); + __get_user(flags, &sve->flags); + + sm = flags & TARGET_SVE_SIG_FLAG_SM; + + /* The cpu must support Streaming or Non-streaming SVE. */ + if (sm + ? !cpu_isar_feature(aa64_sme, env_archcpu(env)) + : !cpu_isar_feature(aa64_sve, env_archcpu(env))) { return false; } - __get_user(vl, &sve->vl); - vq = sve_vq(env); + /* + * Note that we cannot use sve_vq() because that depends on the + * current setting of PSTATE.SM, not the state to be restored. + */ + vq = sve_vqm1_for_el_sm(env, 0, sm) + 1; /* Reject mismatched VL. */ if (vl != vq * TARGET_SVE_VQ_BYTES) { @@ -XXX,XX +XXX,XX @@ static bool target_restore_sve_record(CPUARMState *env, return false; } + *svcr = FIELD_DP64(*svcr, SVCR, SM, sm); + /* * Note that SVE regs are stored as a byte stream, with each byte element * at a subsequent address. This corresponds to a little-endian load @@ -XXX,XX +XXX,XX @@ static bool target_restore_sve_record(CPUARMState *env, return true; } +static bool target_restore_za_record(CPUARMState *env, + struct target_za_context *za, + int size, int *svcr) +{ + int i, j, vl, vq; + + if (!cpu_isar_feature(aa64_sme, env_archcpu(env))) { + return false; + } + + __get_user(vl, &za->vl); + vq = sme_vq(env); + + /* Reject mismatched VL. */ + if (vl != vq * TARGET_SVE_VQ_BYTES) { + return false; + } + + /* Accept empty record -- used to clear PSTATE.ZA. */ + if (size <= TARGET_ZA_SIG_CONTEXT_SIZE(0)) { + return true; + } + + /* Reject non-empty but incomplete record. */ + if (size < TARGET_ZA_SIG_CONTEXT_SIZE(vq)) { + return false; + } + + *svcr = FIELD_DP64(*svcr, SVCR, ZA, 1); + + for (i = 0; i < vl; ++i) { + uint64_t *z = (void *)za + TARGET_ZA_SIG_ZAV_OFFSET(vq, i); + for (j = 0; j < vq * 2; ++j) { + __get_user_e(env->zarray[i].d[j], z + j, le); + } + } + return true; +} + static int target_restore_sigframe(CPUARMState *env, struct target_rt_sigframe *sf) { struct target_aarch64_ctx *ctx, *extra = NULL; struct target_fpsimd_context *fpsimd = NULL; struct target_sve_context *sve = NULL; + struct target_za_context *za = NULL; uint64_t extra_datap = 0; bool used_extra = false; int sve_size = 0; + int za_size = 0; + int svcr = 0; target_restore_general_frame(env, sf); @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, sve_size = size; break; + case TARGET_ZA_MAGIC: + if (za || size < sizeof(struct target_za_context)) { + goto err; + } + za = (struct target_za_context *)ctx; + za_size = size; + break; + case TARGET_EXTRA_MAGIC: if (extra || size != sizeof(struct target_extra_context)) { goto err; @@ -XXX,XX +XXX,XX @@ static int target_restore_sigframe(CPUARMState *env, } /* SVE data, if present, overwrites FPSIMD data. */ - if (sve && !target_restore_sve_record(env, sve, sve_size)) { + if (sve && !target_restore_sve_record(env, sve, sve_size, &svcr)) { goto err; } + if (za && !target_restore_za_record(env, za, za_size, &svcr)) { + goto err; + } + if (env->svcr != svcr) { + env->svcr = svcr; + arm_rebuild_hflags(env); + } unlock_user(extra, extra_datap, 0); return 0; @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, .total_size = offsetof(struct target_rt_sigframe, uc.tuc_mcontext.__reserved), }; - int fpsimd_ofs, fr_ofs, sve_ofs = 0, vq = 0, sve_size = 0; + int fpsimd_ofs, fr_ofs, sve_ofs = 0, za_ofs = 0; + int sve_size = 0, za_size = 0; struct target_rt_sigframe *frame; struct target_rt_frame_record *fr; abi_ulong frame_addr, return_addr; @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, &layout); /* SVE state needs saving only if it exists. */ - if (cpu_isar_feature(aa64_sve, env_archcpu(env))) { - vq = sve_vq(env); - sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(vq), 16); + if (cpu_isar_feature(aa64_sve, env_archcpu(env)) || + cpu_isar_feature(aa64_sme, env_archcpu(env))) { + sve_size = QEMU_ALIGN_UP(TARGET_SVE_SIG_CONTEXT_SIZE(sve_vq(env)), 16); sve_ofs = alloc_sigframe_space(sve_size, &layout); } + if (cpu_isar_feature(aa64_sme, env_archcpu(env))) { + /* ZA state needs saving only if it is enabled. */ + if (FIELD_EX64(env->svcr, SVCR, ZA)) { + za_size = TARGET_ZA_SIG_CONTEXT_SIZE(sme_vq(env)); + } else { + za_size = TARGET_ZA_SIG_CONTEXT_SIZE(0); + } + za_ofs = alloc_sigframe_space(za_size, &layout); + } if (layout.extra_ofs) { /* Reserve space for the extra end marker. The standard end marker @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, target_setup_end_record((void *)frame + layout.extra_end_ofs); } if (sve_ofs) { - target_setup_sve_record((void *)frame + sve_ofs, env, vq, sve_size); + target_setup_sve_record((void *)frame + sve_ofs, env, sve_size); + } + if (za_ofs) { + target_setup_za_record((void *)frame + za_ofs, env, za_size); } /* Set up the stack frame for unwinding. */ @@ -XXX,XX +XXX,XX @@ static void target_setup_frame(int usig, struct target_sigaction *ka, env->btype = 2; } + /* + * Invoke the signal handler with both SM and ZA disabled. + * When clearing SM, ResetSVEState, per SMSTOP. + */ + if (FIELD_EX64(env->svcr, SVCR, SM)) { + arm_reset_sve_state(env); + } + if (env->svcr) { + env->svcr = 0; + arm_rebuild_hflags(env); + } + if (info) { tswap_siginfo(&frame->info, info); env->xregs[1] = frame_addr + offsetof(struct target_rt_sigframe, info); -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Add "sve" to the sve prctl functions, to distinguish them from the coming "sme" prctls with similar names. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-42-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_prctl.h | 8 ++++---- linux-user/syscall.c | 12 ++++++------ 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/linux-user/aarch64/target_prctl.h b/linux-user/aarch64/target_prctl.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_prctl.h +++ b/linux-user/aarch64/target_prctl.h @@ -XXX,XX +XXX,XX @@ #ifndef AARCH64_TARGET_PRCTL_H #define AARCH64_TARGET_PRCTL_H -static abi_long do_prctl_get_vl(CPUArchState *env) +static abi_long do_prctl_sve_get_vl(CPUArchState *env) { ARMCPU *cpu = env_archcpu(env); if (cpu_isar_feature(aa64_sve, cpu)) { @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_get_vl(CPUArchState *env) } return -TARGET_EINVAL; } -#define do_prctl_get_vl do_prctl_get_vl +#define do_prctl_sve_get_vl do_prctl_sve_get_vl -static abi_long do_prctl_set_vl(CPUArchState *env, abi_long arg2) +static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) { /* * We cannot support either PR_SVE_SET_VL_ONEXEC or PR_SVE_VL_INHERIT. @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_set_vl(CPUArchState *env, abi_long arg2) } return -TARGET_EINVAL; } -#define do_prctl_set_vl do_prctl_set_vl +#define do_prctl_sve_set_vl do_prctl_sve_set_vl static abi_long do_prctl_reset_keys(CPUArchState *env, abi_long arg2) { diff --git a/linux-user/syscall.c b/linux-user/syscall.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_inval1(CPUArchState *env, abi_long arg2) #ifndef do_prctl_set_fp_mode #define do_prctl_set_fp_mode do_prctl_inval1 #endif -#ifndef do_prctl_get_vl -#define do_prctl_get_vl do_prctl_inval0 +#ifndef do_prctl_sve_get_vl +#define do_prctl_sve_get_vl do_prctl_inval0 #endif -#ifndef do_prctl_set_vl -#define do_prctl_set_vl do_prctl_inval1 +#ifndef do_prctl_sve_set_vl +#define do_prctl_sve_set_vl do_prctl_inval1 #endif #ifndef do_prctl_reset_keys #define do_prctl_reset_keys do_prctl_inval1 @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, case PR_SET_FP_MODE: return do_prctl_set_fp_mode(env, arg2); case PR_SVE_GET_VL: - return do_prctl_get_vl(env); + return do_prctl_sve_get_vl(env); case PR_SVE_SET_VL: - return do_prctl_set_vl(env, arg2); + return do_prctl_sve_set_vl(env, arg2); case PR_PAC_RESET_KEYS: if (arg3 || arg4 || arg5) { return -TARGET_EINVAL; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> These prctl set the Streaming SVE vector length, which may be completely different from the Normal SVE vector length. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-43-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/aarch64/target_prctl.h | 54 +++++++++++++++++++++++++++++++ linux-user/syscall.c | 16 +++++++++ 2 files changed, 70 insertions(+) diff --git a/linux-user/aarch64/target_prctl.h b/linux-user/aarch64/target_prctl.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/aarch64/target_prctl.h +++ b/linux-user/aarch64/target_prctl.h @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_get_vl(CPUArchState *env) { ARMCPU *cpu = env_archcpu(env); if (cpu_isar_feature(aa64_sve, cpu)) { + /* PSTATE.SM is always unset on syscall entry. */ return sve_vq(env) * 16; } return -TARGET_EINVAL; @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) && arg2 >= 0 && arg2 <= 512 * 16 && !(arg2 & 15)) { uint32_t vq, old_vq; + /* PSTATE.SM is always unset on syscall entry. */ old_vq = sve_vq(env); /* @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_sve_set_vl(CPUArchState *env, abi_long arg2) } #define do_prctl_sve_set_vl do_prctl_sve_set_vl +static abi_long do_prctl_sme_get_vl(CPUArchState *env) +{ + ARMCPU *cpu = env_archcpu(env); + if (cpu_isar_feature(aa64_sme, cpu)) { + return sme_vq(env) * 16; + } + return -TARGET_EINVAL; +} +#define do_prctl_sme_get_vl do_prctl_sme_get_vl + +static abi_long do_prctl_sme_set_vl(CPUArchState *env, abi_long arg2) +{ + /* + * We cannot support either PR_SME_SET_VL_ONEXEC or PR_SME_VL_INHERIT. + * Note the kernel definition of sve_vl_valid allows for VQ=512, + * i.e. VL=8192, even though the architectural maximum is VQ=16. + */ + if (cpu_isar_feature(aa64_sme, env_archcpu(env)) + && arg2 >= 0 && arg2 <= 512 * 16 && !(arg2 & 15)) { + int vq, old_vq; + + old_vq = sme_vq(env); + + /* + * Bound the value of vq, so that we know that it fits into + * the 4-bit field in SMCR_EL1. Because PSTATE.SM is cleared + * on syscall entry, we are not modifying the current SVE + * vector length. + */ + vq = MAX(arg2 / 16, 1); + vq = MIN(vq, 16); + env->vfp.smcr_el[1] = + FIELD_DP64(env->vfp.smcr_el[1], SMCR, LEN, vq - 1); + + /* Delay rebuilding hflags until we know if ZA must change. */ + vq = sve_vqm1_for_el_sm(env, 0, true) + 1; + + if (vq != old_vq) { + /* + * PSTATE.ZA state is cleared on any change to SVL. + * We need not call arm_rebuild_hflags because PSTATE.SM was + * cleared on syscall entry, so this hasn't changed VL. + */ + env->svcr = FIELD_DP64(env->svcr, SVCR, ZA, 0); + arm_rebuild_hflags(env); + } + return vq * 16; + } + return -TARGET_EINVAL; +} +#define do_prctl_sme_set_vl do_prctl_sme_set_vl + static abi_long do_prctl_reset_keys(CPUArchState *env, abi_long arg2) { ARMCPU *cpu = env_archcpu(env); diff --git a/linux-user/syscall.c b/linux-user/syscall.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -XXX,XX +XXX,XX @@ abi_long do_arch_prctl(CPUX86State *env, int code, abi_ulong addr) #ifndef PR_SET_SYSCALL_USER_DISPATCH # define PR_SET_SYSCALL_USER_DISPATCH 59 #endif +#ifndef PR_SME_SET_VL +# define PR_SME_SET_VL 63 +# define PR_SME_GET_VL 64 +# define PR_SME_VL_LEN_MASK 0xffff +# define PR_SME_VL_INHERIT (1 << 17) +#endif #include "target_prctl.h" @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl_inval1(CPUArchState *env, abi_long arg2) #ifndef do_prctl_set_unalign #define do_prctl_set_unalign do_prctl_inval1 #endif +#ifndef do_prctl_sme_get_vl +#define do_prctl_sme_get_vl do_prctl_inval0 +#endif +#ifndef do_prctl_sme_set_vl +#define do_prctl_sme_set_vl do_prctl_inval1 +#endif static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, abi_long arg3, abi_long arg4, abi_long arg5) @@ -XXX,XX +XXX,XX @@ static abi_long do_prctl(CPUArchState *env, abi_long option, abi_long arg2, return do_prctl_sve_get_vl(env); case PR_SVE_SET_VL: return do_prctl_sve_set_vl(env, arg2); + case PR_SME_GET_VL: + return do_prctl_sme_get_vl(env); + case PR_SME_SET_VL: + return do_prctl_sme_set_vl(env, arg2); case PR_PAC_RESET_KEYS: if (arg3 || arg4 || arg5) { return -TARGET_EINVAL; -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> There's no reason to set CPACR_EL1.ZEN if SVE disabled. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-44-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_reset(DeviceState *dev) /* and to the FP/Neon instructions */ env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, CPACR_EL1, FPEN, 3); - /* and to the SVE instructions */ - env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, - CPACR_EL1, ZEN, 3); - /* with reasonable vector length */ + /* and to the SVE instructions, with default vector length */ if (cpu_isar_feature(aa64_sve, cpu)) { + env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, + CPACR_EL1, ZEN, 3); env->vfp.zcr_el[1] = cpu->sve_default_vq - 1; } /* -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Enable SME, TPIDR2_EL0, and FA64 if supported by the cpu. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-45-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- target/arm/cpu.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static void arm_cpu_reset(DeviceState *dev) CPACR_EL1, ZEN, 3); env->vfp.zcr_el[1] = cpu->sve_default_vq - 1; } + /* and for SME instructions, with default vector length, and TPIDR2 */ + if (cpu_isar_feature(aa64_sme, cpu)) { + env->cp15.sctlr_el[1] |= SCTLR_EnTP2; + env->cp15.cpacr_el1 = FIELD_DP64(env->cp15.cpacr_el1, + CPACR_EL1, SMEN, 3); + env->vfp.smcr_el[1] = cpu->sme_default_vq - 1; + if (cpu_isar_feature(aa64_sme_fa64, cpu)) { + env->vfp.smcr_el[1] = FIELD_DP64(env->vfp.smcr_el[1], + SMCR, FA64, 1); + } + } /* * Enable 48-bit address space (TODO: take reserved_va into account). * Enable TBI0 but not TBI1. -- 2.25.1 From: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-id: 20220708151540.18136-46-richard.henderson@linaro.org Signed-off-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/elfload.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/linux-user/elfload.c b/linux-user/elfload.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/elfload.c +++ b/linux-user/elfload.c @@ -XXX,XX +XXX,XX @@ enum { ARM_HWCAP2_A64_RNG = 1 << 16, ARM_HWCAP2_A64_BTI = 1 << 17, ARM_HWCAP2_A64_MTE = 1 << 18, + ARM_HWCAP2_A64_ECV = 1 << 19, + ARM_HWCAP2_A64_AFP = 1 << 20, + ARM_HWCAP2_A64_RPRES = 1 << 21, + ARM_HWCAP2_A64_MTE3 = 1 << 22, + ARM_HWCAP2_A64_SME = 1 << 23, + ARM_HWCAP2_A64_SME_I16I64 = 1 << 24, + ARM_HWCAP2_A64_SME_F64F64 = 1 << 25, + ARM_HWCAP2_A64_SME_I8I32 = 1 << 26, + ARM_HWCAP2_A64_SME_F16F32 = 1 << 27, + ARM_HWCAP2_A64_SME_B16F32 = 1 << 28, + ARM_HWCAP2_A64_SME_F32F32 = 1 << 29, + ARM_HWCAP2_A64_SME_FA64 = 1 << 30, }; #define ELF_HWCAP get_elf_hwcap() @@ -XXX,XX +XXX,XX @@ static uint32_t get_elf_hwcap2(void) GET_FEATURE_ID(aa64_rndr, ARM_HWCAP2_A64_RNG); GET_FEATURE_ID(aa64_bti, ARM_HWCAP2_A64_BTI); GET_FEATURE_ID(aa64_mte, ARM_HWCAP2_A64_MTE); + GET_FEATURE_ID(aa64_sme, (ARM_HWCAP2_A64_SME | + ARM_HWCAP2_A64_SME_F32F32 | + ARM_HWCAP2_A64_SME_B16F32 | + ARM_HWCAP2_A64_SME_F16F32 | + ARM_HWCAP2_A64_SME_I8I32)); + GET_FEATURE_ID(aa64_sme_f64f64, ARM_HWCAP2_A64_SME_F64F64); + GET_FEATURE_ID(aa64_sme_i16i64, ARM_HWCAP2_A64_SME_I16I64); + GET_FEATURE_ID(aa64_sme_fa64, ARM_HWCAP2_A64_SME_FA64); return hwcaps; } -- 2.25.1