I have not been able to prod reviews of all of the rotate patches in 4 weeks, but let's not let that block ARM work forever. r~ The following changes since commit cccdd8c7971896c339d59c9c5d4647d4ffd9568a: Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging (2020-06-02 10:25:55 +0100) are available in the Git repository at: https://github.com/rth7680/qemu.git tags/pull-tcg-20200602 for you to fetch changes up to 71b04329c4f7d5824a289ca5225e1883a278cf3b: accel/tcg: Provide a NetBSD specific aarch64 cpu_signal_handler (2020-06-02 08:42:37 -0700) ---------------------------------------------------------------- Vector rotate support Signal handling support for NetBSD arm/aarch64 ---------------------------------------------------------------- Nick Hudson (2): accel/tcg: Adjust cpu_signal_handler for NetBSD/arm accel/tcg: Provide a NetBSD specific aarch64 cpu_signal_handler Richard Henderson (10): tcg: Implement gvec support for rotate by immediate tcg: Implement gvec support for rotate by vector tcg: Remove expansion to shift by vector from do_shifts tcg: Implement gvec support for rotate by scalar tcg/i386: Implement INDEX_op_rotl{i,s,v}_vec tcg/aarch64: Implement INDEX_op_rotl{i,v}_vec tcg/ppc: Implement INDEX_op_rot[lr]v_vec target/ppc: Use tcg_gen_gvec_rotlv target/s390x: Use tcg_gen_gvec_rotl{i,s,v} tcg: Improve move ops in liveness_pass_2 accel/tcg/tcg-runtime.h | 15 +++ include/tcg/tcg-op-gvec.h | 12 ++ include/tcg/tcg-op.h | 5 + include/tcg/tcg-opc.h | 4 + include/tcg/tcg.h | 3 + target/ppc/helper.h | 4 - target/s390x/helper.h | 4 - tcg/aarch64/tcg-target.h | 3 + tcg/aarch64/tcg-target.opc.h | 1 + tcg/i386/tcg-target.h | 3 + tcg/ppc/tcg-target.h | 3 + tcg/ppc/tcg-target.opc.h | 1 - accel/tcg/tcg-runtime-gvec.c | 144 ++++++++++++++++++++++++ accel/tcg/user-exec.c | 43 +++++++- target/ppc/int_helper.c | 17 --- target/ppc/translate/vmx-impl.inc.c | 8 +- target/s390x/translate_vx.inc.c | 66 ++--------- target/s390x/vec_int_helper.c | 31 ------ tcg/aarch64/tcg-target.inc.c | 53 ++++++++- tcg/i386/tcg-target.inc.c | 116 +++++++++++++++++--- tcg/ppc/tcg-target.inc.c | 23 +++- tcg/tcg-op-gvec.c | 212 ++++++++++++++++++++++++++++++++++++ tcg/tcg-op-vec.c | 62 +++++++---- tcg/tcg.c | 85 +++++++++++---- target/s390x/insn-data.def | 4 +- tcg/README | 7 +- 26 files changed, 736 insertions(+), 193 deletions(-) No host backend support yet, but the interfaces for rotli are in place. Canonicalize immediate rotate to the left, based on a survey of architectures, but provide both left and right shift interfaces to the translators. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/tcg-runtime.h | 5 +++ include/tcg/tcg-op-gvec.h | 6 ++++ include/tcg/tcg-op.h | 2 ++ include/tcg/tcg-opc.h | 1 + include/tcg/tcg.h | 1 + tcg/aarch64/tcg-target.h | 1 + tcg/i386/tcg-target.h | 1 + tcg/ppc/tcg-target.h | 1 + accel/tcg/tcg-runtime-gvec.c | 48 +++++++++++++++++++++++++ tcg/tcg-op-gvec.c | 68 ++++++++++++++++++++++++++++++++++++ tcg/tcg-op-vec.c | 12 +++++++ tcg/tcg.c | 2 ++ tcg/README | 3 +- 13 files changed, 150 insertions(+), 1 deletion(-) diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-runtime.h +++ b/accel/tcg/tcg-runtime.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(gvec_sar16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_sar32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_sar64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_rotl8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_rotl16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_rotl32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_rotl64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(gvec_shl8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_shl16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_shl32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_shri(unsigned vece, uint32_t dofs, uint32_t aofs, int64_t shift, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs, int64_t shift, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_rotli(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t shift, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_rotri(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t shift, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_shls(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz); @@ -XXX,XX +XXX,XX @@ void tcg_gen_vec_shr8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); void tcg_gen_vec_shr16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); void tcg_gen_vec_sar8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); void tcg_gen_vec_sar16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_rotl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c); +void tcg_gen_vec_rotl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c); #endif diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op.h +++ b/include/tcg/tcg-op.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_umax_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_shli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); void tcg_gen_shri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); +void tcg_gen_rotli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); +void tcg_gen_rotri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); void tcg_gen_shls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); void tcg_gen_shrs_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-opc.h +++ b/include/tcg/tcg-opc.h @@ -XXX,XX +XXX,XX @@ DEF(not_vec, 1, 1, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_not_vec)) DEF(shli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) DEF(shri_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) DEF(sari_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) +DEF(rotli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_roti_vec)) DEF(shls_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) DEF(shrs_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ typedef uint64_t TCGRegSet; #define TCG_TARGET_HAS_not_vec 0 #define TCG_TARGET_HAS_andc_vec 0 #define TCG_TARGET_HAS_orc_vec 0 +#define TCG_TARGET_HAS_roti_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 0 diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_not_vec 1 #define TCG_TARGET_HAS_neg_vec 1 #define TCG_TARGET_HAS_abs_vec 1 +#define TCG_TARGET_HAS_roti_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 1 diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_avx2; #define TCG_TARGET_HAS_not_vec 0 #define TCG_TARGET_HAS_neg_vec 0 #define TCG_TARGET_HAS_abs_vec 1 +#define TCG_TARGET_HAS_roti_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 1 #define TCG_TARGET_HAS_shv_vec have_avx2 diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_vsx; #define TCG_TARGET_HAS_not_vec 1 #define TCG_TARGET_HAS_neg_vec have_isa_3_00 #define TCG_TARGET_HAS_abs_vec 0 +#define TCG_TARGET_HAS_roti_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 1 diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-runtime-gvec.c +++ b/accel/tcg/tcg-runtime-gvec.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_sar64i)(void *d, void *a, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_rotl8i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint8_t)) { + *(uint8_t *)(d + i) = rol8(*(uint8_t *)(a + i), shift); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl16i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint16_t)) { + *(uint16_t *)(d + i) = rol16(*(uint16_t *)(a + i), shift); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl32i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint32_t)) { + *(uint32_t *)(d + i) = rol32(*(uint32_t *)(a + i), shift); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl64i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint64_t)) { + *(uint64_t *)(d + i) = rol64(*(uint64_t *)(a + i), shift); + } + clear_high(d, oprsz, desc); +} + void HELPER(gvec_shl8v)(void *d, void *a, void *b, uint32_t desc) { intptr_t oprsz = simd_oprsz(desc); diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs, } } +void tcg_gen_vec_rotl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = dup_const(MO_8, 0xff << c); + + tcg_gen_shli_i64(d, a, c); + tcg_gen_shri_i64(a, a, 8 - c); + tcg_gen_andi_i64(d, d, mask); + tcg_gen_andi_i64(a, a, ~mask); + tcg_gen_or_i64(d, d, a); +} + +void tcg_gen_vec_rotl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = dup_const(MO_16, 0xffff << c); + + tcg_gen_shli_i64(d, a, c); + tcg_gen_shri_i64(a, a, 16 - c); + tcg_gen_andi_i64(d, d, mask); + tcg_gen_andi_i64(a, a, ~mask); + tcg_gen_or_i64(d, d, a); +} + +void tcg_gen_gvec_rotli(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t shift, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { INDEX_op_rotli_vec, 0 }; + static const GVecGen2i g[4] = { + { .fni8 = tcg_gen_vec_rotl8i_i64, + .fniv = tcg_gen_rotli_vec, + .fno = gen_helper_gvec_rotl8i, + .opt_opc = vecop_list, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_rotl16i_i64, + .fniv = tcg_gen_rotli_vec, + .fno = gen_helper_gvec_rotl16i, + .opt_opc = vecop_list, + .vece = MO_16 }, + { .fni4 = tcg_gen_rotli_i32, + .fniv = tcg_gen_rotli_vec, + .fno = gen_helper_gvec_rotl32i, + .opt_opc = vecop_list, + .vece = MO_32 }, + { .fni8 = tcg_gen_rotli_i64, + .fniv = tcg_gen_rotli_vec, + .fno = gen_helper_gvec_rotl64i, + .opt_opc = vecop_list, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_debug_assert(shift >= 0 && shift < (8 << vece)); + if (shift == 0) { + tcg_gen_gvec_mov(vece, dofs, aofs, oprsz, maxsz); + } else { + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, shift, &g[vece]); + } +} + +void tcg_gen_gvec_rotri(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t shift, uint32_t oprsz, uint32_t maxsz) +{ + tcg_debug_assert(vece <= MO_64); + tcg_debug_assert(shift >= 0 && shift < (8 << vece)); + tcg_gen_gvec_rotli(vece, dofs, aofs, -shift & ((8 << vece) - 1), + oprsz, maxsz); +} + /* * Specialized generation vector shifts by a non-constant scalar. */ diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) do_shifti(INDEX_op_sari_vec, vece, r, a, i); } +void tcg_gen_rotli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) +{ + do_shifti(INDEX_op_rotli_vec, vece, r, a, i); +} + +void tcg_gen_rotri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) +{ + int bits = 8 << vece; + tcg_debug_assert(i >= 0 && i < bits); + do_shifti(INDEX_op_rotli_vec, vece, r, a, -i & (bits - 1)); +} + void tcg_gen_cmp_vec(TCGCond cond, unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) { diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ bool tcg_op_supported(TCGOpcode op) case INDEX_op_shrv_vec: case INDEX_op_sarv_vec: return have_vec && TCG_TARGET_HAS_shv_vec; + case INDEX_op_rotli_vec: + return have_vec && TCG_TARGET_HAS_roti_vec; case INDEX_op_ssadd_vec: case INDEX_op_usadd_vec: case INDEX_op_sssub_vec: diff --git a/tcg/README b/tcg/README index XXXXXXX..XXXXXXX 100644 --- a/tcg/README +++ b/tcg/README @@ -XXX,XX +XXX,XX @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32. * shri_vec v0, v1, i2 * sari_vec v0, v1, i2 +* rotli_vec v0, v1, i2 * shrs_vec v0, v1, s2 * sars_vec v0, v1, s2 - Similarly for logical and arithmetic right shift. + Similarly for logical and arithmetic right shift, and left rotate. * shlv_vec v0, v1, v2 -- 2.25.1 No host backend support yet, but the interfaces for rotlv and rotrv are in place. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- v3: Drop the generic expansion from rot to shift; we can do better for each backend, and then this code becomes unused. --- accel/tcg/tcg-runtime.h | 10 +++ include/tcg/tcg-op-gvec.h | 4 ++ include/tcg/tcg-op.h | 2 + include/tcg/tcg-opc.h | 2 + include/tcg/tcg.h | 1 + tcg/aarch64/tcg-target.h | 1 + tcg/i386/tcg-target.h | 1 + tcg/ppc/tcg-target.h | 1 + accel/tcg/tcg-runtime-gvec.c | 96 +++++++++++++++++++++++++++ tcg/tcg-op-gvec.c | 122 +++++++++++++++++++++++++++++++++++ tcg/tcg-op-vec.c | 10 +++ tcg/tcg.c | 3 + tcg/README | 4 +- 13 files changed, 256 insertions(+), 1 deletion(-) diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-runtime.h +++ b/accel/tcg/tcg-runtime.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(gvec_sar16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sar32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sar64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotl8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotl16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotl32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotl64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(gvec_rotr8v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotr16v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotr32v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_rotr64v, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(gvec_eq8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_eq16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_eq32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_shrv(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_sarv(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_rotlv(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_rotrv(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_cmp(TCGCond cond, unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op.h +++ b/include/tcg/tcg-op.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_sars_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); void tcg_gen_shlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); void tcg_gen_shrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); void tcg_gen_sarv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); +void tcg_gen_rotlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); +void tcg_gen_rotrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); void tcg_gen_cmp_vec(TCGCond cond, unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-opc.h +++ b/include/tcg/tcg-opc.h @@ -XXX,XX +XXX,XX @@ DEF(sars_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) DEF(shlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) DEF(shrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) DEF(sarv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) +DEF(rotlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_rotv_vec)) +DEF(rotrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_rotv_vec)) DEF(cmp_vec, 1, 2, 1, IMPLVEC) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ typedef uint64_t TCGRegSet; #define TCG_TARGET_HAS_andc_vec 0 #define TCG_TARGET_HAS_orc_vec 0 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 0 diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_neg_vec 1 #define TCG_TARGET_HAS_abs_vec 1 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 1 diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_avx2; #define TCG_TARGET_HAS_neg_vec 0 #define TCG_TARGET_HAS_abs_vec 1 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 1 #define TCG_TARGET_HAS_shv_vec have_avx2 diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_vsx; #define TCG_TARGET_HAS_neg_vec have_isa_3_00 #define TCG_TARGET_HAS_abs_vec 0 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 1 diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-runtime-gvec.c +++ b/accel/tcg/tcg-runtime-gvec.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_sar64v)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_rotl8v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint8_t)) { + uint8_t sh = *(uint8_t *)(b + i) & 7; + *(uint8_t *)(d + i) = rol8(*(uint8_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl16v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint16_t)) { + uint8_t sh = *(uint16_t *)(b + i) & 15; + *(uint16_t *)(d + i) = rol16(*(uint16_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl32v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint32_t)) { + uint8_t sh = *(uint32_t *)(b + i) & 31; + *(uint32_t *)(d + i) = rol32(*(uint32_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotl64v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint64_t)) { + uint8_t sh = *(uint64_t *)(b + i) & 63; + *(uint64_t *)(d + i) = rol64(*(uint64_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotr8v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint8_t)) { + uint8_t sh = *(uint8_t *)(b + i) & 7; + *(uint8_t *)(d + i) = ror8(*(uint8_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotr16v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint16_t)) { + uint8_t sh = *(uint16_t *)(b + i) & 15; + *(uint16_t *)(d + i) = ror16(*(uint16_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotr32v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint32_t)) { + uint8_t sh = *(uint32_t *)(b + i) & 31; + *(uint32_t *)(d + i) = ror32(*(uint32_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_rotr64v)(void *d, void *a, void *b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(uint64_t)) { + uint8_t sh = *(uint64_t *)(b + i) & 63; + *(uint64_t *)(d + i) = ror64(*(uint64_t *)(a + i), sh); + } + clear_high(d, oprsz, desc); +} + #define DO_CMP1(NAME, TYPE, OP) \ void HELPER(NAME)(void *d, void *a, void *b, uint32_t desc) \ { \ diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_sarv(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); } +/* + * Similarly for rotates. + */ + +static void tcg_gen_rotlv_mod_vec(unsigned vece, TCGv_vec d, + TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + + tcg_gen_dupi_vec(vece, t, (8 << vece) - 1); + tcg_gen_and_vec(vece, t, t, b); + tcg_gen_rotlv_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +static void tcg_gen_rotl_mod_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t = tcg_temp_new_i32(); + + tcg_gen_andi_i32(t, b, 31); + tcg_gen_rotl_i32(d, a, t); + tcg_temp_free_i32(t); +} + +static void tcg_gen_rotl_mod_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t = tcg_temp_new_i64(); + + tcg_gen_andi_i64(t, b, 63); + tcg_gen_rotl_i64(d, a, t); + tcg_temp_free_i64(t); +} + +void tcg_gen_gvec_rotlv(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { INDEX_op_rotlv_vec, 0 }; + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_rotlv_mod_vec, + .fno = gen_helper_gvec_rotl8v, + .opt_opc = vecop_list, + .vece = MO_8 }, + { .fniv = tcg_gen_rotlv_mod_vec, + .fno = gen_helper_gvec_rotl16v, + .opt_opc = vecop_list, + .vece = MO_16 }, + { .fni4 = tcg_gen_rotl_mod_i32, + .fniv = tcg_gen_rotlv_mod_vec, + .fno = gen_helper_gvec_rotl32v, + .opt_opc = vecop_list, + .vece = MO_32 }, + { .fni8 = tcg_gen_rotl_mod_i64, + .fniv = tcg_gen_rotlv_mod_vec, + .fno = gen_helper_gvec_rotl64v, + .opt_opc = vecop_list, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + +static void tcg_gen_rotrv_mod_vec(unsigned vece, TCGv_vec d, + TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + + tcg_gen_dupi_vec(vece, t, (8 << vece) - 1); + tcg_gen_and_vec(vece, t, t, b); + tcg_gen_rotrv_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +static void tcg_gen_rotr_mod_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t = tcg_temp_new_i32(); + + tcg_gen_andi_i32(t, b, 31); + tcg_gen_rotr_i32(d, a, t); + tcg_temp_free_i32(t); +} + +static void tcg_gen_rotr_mod_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t = tcg_temp_new_i64(); + + tcg_gen_andi_i64(t, b, 63); + tcg_gen_rotr_i64(d, a, t); + tcg_temp_free_i64(t); +} + +void tcg_gen_gvec_rotrv(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { INDEX_op_rotrv_vec, 0 }; + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_rotrv_mod_vec, + .fno = gen_helper_gvec_rotr8v, + .opt_opc = vecop_list, + .vece = MO_8 }, + { .fniv = tcg_gen_rotrv_mod_vec, + .fno = gen_helper_gvec_rotr16v, + .opt_opc = vecop_list, + .vece = MO_16 }, + { .fni4 = tcg_gen_rotr_mod_i32, + .fniv = tcg_gen_rotrv_mod_vec, + .fno = gen_helper_gvec_rotr32v, + .opt_opc = vecop_list, + .vece = MO_32 }, + { .fni8 = tcg_gen_rotr_mod_i64, + .fniv = tcg_gen_rotrv_mod_vec, + .fno = gen_helper_gvec_rotr64v, + .opt_opc = vecop_list, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + /* Expand OPSZ bytes worth of three-operand operations using i32 elements. */ static void expand_cmp_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, TCGCond cond) diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_sarv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) do_op3_nofail(vece, r, a, b, INDEX_op_sarv_vec); } +void tcg_gen_rotlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3_nofail(vece, r, a, b, INDEX_op_rotlv_vec); +} + +void tcg_gen_rotrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3_nofail(vece, r, a, b, INDEX_op_rotrv_vec); +} + static void do_shifts(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s, TCGOpcode opc_s, TCGOpcode opc_v) { diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ bool tcg_op_supported(TCGOpcode op) return have_vec && TCG_TARGET_HAS_shv_vec; case INDEX_op_rotli_vec: return have_vec && TCG_TARGET_HAS_roti_vec; + case INDEX_op_rotlv_vec: + case INDEX_op_rotrv_vec: + return have_vec && TCG_TARGET_HAS_rotv_vec; case INDEX_op_ssadd_vec: case INDEX_op_usadd_vec: case INDEX_op_sssub_vec: diff --git a/tcg/README b/tcg/README index XXXXXXX..XXXXXXX 100644 --- a/tcg/README +++ b/tcg/README @@ -XXX,XX +XXX,XX @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32. * shrv_vec v0, v1, v2 * sarv_vec v0, v1, v2 +* rotlv_vec v0, v1, v2 +* rotrv_vec v0, v1, v2 - Similarly for logical and arithmetic right shift. + Similarly for logical and arithmetic right shift, and rotates. * cmp_vec v0, v1, v2, cond -- 2.25.1 We do not reflect this expansion in tcg_can_emit_vecop_list, so it is unused and unusable. However, we actually perform the same expansion in do_gvec_shifts, so it is also unneeded. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg-op-vec.c | 35 +++++++++++------------------------ 1 file changed, 11 insertions(+), 24 deletions(-) diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_rotrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) } static void do_shifts(unsigned vece, TCGv_vec r, TCGv_vec a, - TCGv_i32 s, TCGOpcode opc_s, TCGOpcode opc_v) + TCGv_i32 s, TCGOpcode opc) { TCGTemp *rt = tcgv_vec_temp(r); TCGTemp *at = tcgv_vec_temp(a); @@ -XXX,XX +XXX,XX @@ static void do_shifts(unsigned vece, TCGv_vec r, TCGv_vec a, TCGArg ai = temp_arg(at); TCGArg si = temp_arg(st); TCGType type = rt->base_type; - const TCGOpcode *hold_list; int can; tcg_debug_assert(at->base_type >= type); - tcg_assert_listed_vecop(opc_s); - hold_list = tcg_swap_vecop_list(NULL); - - can = tcg_can_emit_vec_op(opc_s, type, vece); + tcg_assert_listed_vecop(opc); + can = tcg_can_emit_vec_op(opc, type, vece); if (can > 0) { - vec_gen_3(opc_s, type, vece, ri, ai, si); + vec_gen_3(opc, type, vece, ri, ai, si); } else if (can < 0) { - tcg_expand_vec_op(opc_s, type, vece, ri, ai, si); + const TCGOpcode *hold_list = tcg_swap_vecop_list(NULL); + tcg_expand_vec_op(opc, type, vece, ri, ai, si); + tcg_swap_vecop_list(hold_list); } else { - TCGv_vec vec_s = tcg_temp_new_vec(type); - - if (vece == MO_64) { - TCGv_i64 s64 = tcg_temp_new_i64(); - tcg_gen_extu_i32_i64(s64, s); - tcg_gen_dup_i64_vec(MO_64, vec_s, s64); - tcg_temp_free_i64(s64); - } else { - tcg_gen_dup_i32_vec(vece, vec_s, s); - } - do_op3_nofail(vece, r, a, vec_s, opc_v); - tcg_temp_free_vec(vec_s); + g_assert_not_reached(); } - tcg_swap_vecop_list(hold_list); } void tcg_gen_shls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 b) { - do_shifts(vece, r, a, b, INDEX_op_shls_vec, INDEX_op_shlv_vec); + do_shifts(vece, r, a, b, INDEX_op_shls_vec); } void tcg_gen_shrs_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 b) { - do_shifts(vece, r, a, b, INDEX_op_shrs_vec, INDEX_op_shrv_vec); + do_shifts(vece, r, a, b, INDEX_op_shrs_vec); } void tcg_gen_sars_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 b) { - do_shifts(vece, r, a, b, INDEX_op_sars_vec, INDEX_op_sarv_vec); + do_shifts(vece, r, a, b, INDEX_op_sars_vec); } void tcg_gen_bitsel_vec(unsigned vece, TCGv_vec r, TCGv_vec a, -- 2.25.1 No host backend support yet, but the interfaces for rotls are in place. Only implement left-rotate for now, as the only known use of vector rotate by scalar is s390x, so any right-rotate would be unused and untestable. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg-op-gvec.h | 2 ++ include/tcg/tcg-op.h | 1 + include/tcg/tcg-opc.h | 1 + include/tcg/tcg.h | 1 + tcg/aarch64/tcg-target.h | 1 + tcg/i386/tcg-target.h | 1 + tcg/ppc/tcg-target.h | 1 + tcg/tcg-op-gvec.c | 22 ++++++++++++++++++++++ tcg/tcg-op-vec.c | 5 +++++ tcg/tcg.c | 2 ++ 10 files changed, 37 insertions(+) diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_shrs(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_sars(unsigned vece, uint32_t dofs, uint32_t aofs, TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_rotls(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz); /* * Perform vector shift by vector element, modulo the element size. diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-op.h +++ b/include/tcg/tcg-op.h @@ -XXX,XX +XXX,XX @@ void tcg_gen_rotri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); void tcg_gen_shls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); void tcg_gen_shrs_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); void tcg_gen_sars_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); +void tcg_gen_rotls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s); void tcg_gen_shlv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); void tcg_gen_shrv_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec s); diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-opc.h +++ b/include/tcg/tcg-opc.h @@ -XXX,XX +XXX,XX @@ DEF(rotli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_roti_vec)) DEF(shls_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) DEF(shrs_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) DEF(sars_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) +DEF(rotls_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_rots_vec)) DEF(shlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) DEF(shrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ typedef uint64_t TCGRegSet; #define TCG_TARGET_HAS_andc_vec 0 #define TCG_TARGET_HAS_orc_vec 0 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rots_vec 0 #define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_neg_vec 1 #define TCG_TARGET_HAS_abs_vec 1 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rots_vec 0 #define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 0 diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_avx2; #define TCG_TARGET_HAS_neg_vec 0 #define TCG_TARGET_HAS_abs_vec 1 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rots_vec 0 #define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 1 #define TCG_TARGET_HAS_shs_vec 1 diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_vsx; #define TCG_TARGET_HAS_neg_vec have_isa_3_00 #define TCG_TARGET_HAS_abs_vec 0 #define TCG_TARGET_HAS_roti_vec 0 +#define TCG_TARGET_HAS_rots_vec 0 #define TCG_TARGET_HAS_rotv_vec 0 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_gvec_sars(unsigned vece, uint32_t dofs, uint32_t aofs, do_gvec_shifts(vece, dofs, aofs, shift, oprsz, maxsz, &g); } +void tcg_gen_gvec_rotls(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i32 shift, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2sh g = { + .fni4 = tcg_gen_rotl_i32, + .fni8 = tcg_gen_rotl_i64, + .fniv_s = tcg_gen_rotls_vec, + .fniv_v = tcg_gen_rotlv_vec, + .fno = { + gen_helper_gvec_rotl8i, + gen_helper_gvec_rotl16i, + gen_helper_gvec_rotl32i, + gen_helper_gvec_rotl64i, + }, + .s_list = { INDEX_op_rotls_vec, 0 }, + .v_list = { INDEX_op_rotlv_vec, 0 }, + }; + + tcg_debug_assert(vece <= MO_64); + do_gvec_shifts(vece, dofs, aofs, shift, oprsz, maxsz, &g); +} + /* * Expand D = A << (B % element bits) * diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -XXX,XX +XXX,XX @@ void tcg_gen_sars_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 b) do_shifts(vece, r, a, b, INDEX_op_sars_vec); } +void tcg_gen_rotls_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_i32 s) +{ + do_shifts(vece, r, a, s, INDEX_op_rotls_vec); +} + void tcg_gen_bitsel_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b, TCGv_vec c) { diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ bool tcg_op_supported(TCGOpcode op) return have_vec && TCG_TARGET_HAS_shv_vec; case INDEX_op_rotli_vec: return have_vec && TCG_TARGET_HAS_roti_vec; + case INDEX_op_rotls_vec: + return have_vec && TCG_TARGET_HAS_rots_vec; case INDEX_op_rotlv_vec: case INDEX_op_rotrv_vec: return have_vec && TCG_TARGET_HAS_rotv_vec; -- 2.25.1 For immediates, we must continue the special casing of 8-bit elements. The other element sizes and shift types are trivially implemented with shifts. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.inc.c | 116 ++++++++++++++++++++++++++++++++------ 1 file changed, 100 insertions(+), 16 deletions(-) diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.inc.c +++ b/tcg/i386/tcg-target.inc.c @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) case INDEX_op_shls_vec: case INDEX_op_shrs_vec: case INDEX_op_sars_vec: + case INDEX_op_rotls_vec: case INDEX_op_cmp_vec: case INDEX_op_x86_shufps_vec: case INDEX_op_x86_blend_vec: @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) case INDEX_op_xor_vec: case INDEX_op_andc_vec: return 1; + case INDEX_op_rotli_vec: case INDEX_op_cmp_vec: case INDEX_op_cmpsel_vec: return -1; @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) return vece >= MO_16; case INDEX_op_sars_vec: return vece >= MO_16 && vece <= MO_32; + case INDEX_op_rotls_vec: + return vece >= MO_16 ? -1 : 0; case INDEX_op_shlv_vec: case INDEX_op_shrv_vec: return have_avx2 && vece >= MO_32; case INDEX_op_sarv_vec: return have_avx2 && vece == MO_32; + case INDEX_op_rotlv_vec: + case INDEX_op_rotrv_vec: + return have_avx2 && vece >= MO_32 ? -1 : 0; case INDEX_op_mul_vec: if (vece == MO_8) { @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) } } -static void expand_vec_shi(TCGType type, unsigned vece, bool shr, +static void expand_vec_shi(TCGType type, unsigned vece, TCGOpcode opc, TCGv_vec v0, TCGv_vec v1, TCGArg imm) { TCGv_vec t1, t2; @@ -XXX,XX +XXX,XX @@ static void expand_vec_shi(TCGType type, unsigned vece, bool shr, t1 = tcg_temp_new_vec(type); t2 = tcg_temp_new_vec(type); - /* Unpack to W, shift, and repack. Tricky bits: - (1) Use punpck*bw x,x to produce DDCCBBAA, - i.e. duplicate in other half of the 16-bit lane. - (2) For right-shift, add 8 so that the high half of - the lane becomes zero. For left-shift, we must - shift up and down again. - (3) Step 2 leaves high half zero such that PACKUSWB - (pack with unsigned saturation) does not modify - the quantity. */ + /* + * Unpack to W, shift, and repack. Tricky bits: + * (1) Use punpck*bw x,x to produce DDCCBBAA, + * i.e. duplicate in other half of the 16-bit lane. + * (2) For right-shift, add 8 so that the high half of the lane + * becomes zero. For left-shift, and left-rotate, we must + * shift up and down again. + * (3) Step 2 leaves high half zero such that PACKUSWB + * (pack with unsigned saturation) does not modify + * the quantity. + */ vec_gen_3(INDEX_op_x86_punpckl_vec, type, MO_8, tcgv_vec_arg(t1), tcgv_vec_arg(v1), tcgv_vec_arg(v1)); vec_gen_3(INDEX_op_x86_punpckh_vec, type, MO_8, tcgv_vec_arg(t2), tcgv_vec_arg(v1), tcgv_vec_arg(v1)); - if (shr) { - tcg_gen_shri_vec(MO_16, t1, t1, imm + 8); - tcg_gen_shri_vec(MO_16, t2, t2, imm + 8); + if (opc != INDEX_op_rotli_vec) { + imm += 8; + } + if (opc == INDEX_op_shri_vec) { + tcg_gen_shri_vec(MO_16, t1, t1, imm); + tcg_gen_shri_vec(MO_16, t2, t2, imm); } else { - tcg_gen_shli_vec(MO_16, t1, t1, imm + 8); - tcg_gen_shli_vec(MO_16, t2, t2, imm + 8); + tcg_gen_shli_vec(MO_16, t1, t1, imm); + tcg_gen_shli_vec(MO_16, t2, t2, imm); tcg_gen_shri_vec(MO_16, t1, t1, 8); tcg_gen_shri_vec(MO_16, t2, t2, 8); } @@ -XXX,XX +XXX,XX @@ static void expand_vec_sari(TCGType type, unsigned vece, } } +static void expand_vec_rotli(TCGType type, unsigned vece, + TCGv_vec v0, TCGv_vec v1, TCGArg imm) +{ + TCGv_vec t; + + if (vece == MO_8) { + expand_vec_shi(type, vece, INDEX_op_rotli_vec, v0, v1, imm); + return; + } + + t = tcg_temp_new_vec(type); + tcg_gen_shli_vec(vece, t, v1, imm); + tcg_gen_shri_vec(vece, v0, v1, (8 << vece) - imm); + tcg_gen_or_vec(vece, v0, v0, t); + tcg_temp_free_vec(t); +} + +static void expand_vec_rotls(TCGType type, unsigned vece, + TCGv_vec v0, TCGv_vec v1, TCGv_i32 lsh) +{ + TCGv_i32 rsh; + TCGv_vec t; + + tcg_debug_assert(vece != MO_8); + + t = tcg_temp_new_vec(type); + rsh = tcg_temp_new_i32(); + + tcg_gen_neg_i32(rsh, lsh); + tcg_gen_andi_i32(rsh, rsh, (8 << vece) - 1); + tcg_gen_shls_vec(vece, t, v1, lsh); + tcg_gen_shrs_vec(vece, v0, v1, rsh); + tcg_gen_or_vec(vece, v0, v0, t); + tcg_temp_free_vec(t); + tcg_temp_free_i32(rsh); +} + +static void expand_vec_rotv(TCGType type, unsigned vece, TCGv_vec v0, + TCGv_vec v1, TCGv_vec sh, bool right) +{ + TCGv_vec t = tcg_temp_new_vec(type); + + tcg_gen_dupi_vec(vece, t, 8 << vece); + tcg_gen_sub_vec(vece, t, t, sh); + if (right) { + tcg_gen_shlv_vec(vece, t, v1, t); + tcg_gen_shrv_vec(vece, v0, v1, sh); + } else { + tcg_gen_shrv_vec(vece, t, v1, t); + tcg_gen_shlv_vec(vece, v0, v1, sh); + } + tcg_gen_or_vec(vece, v0, v0, t); + tcg_temp_free_vec(t); +} + static void expand_vec_mul(TCGType type, unsigned vece, TCGv_vec v0, TCGv_vec v1, TCGv_vec v2) { @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, switch (opc) { case INDEX_op_shli_vec: case INDEX_op_shri_vec: - expand_vec_shi(type, vece, opc == INDEX_op_shri_vec, v0, v1, a2); + expand_vec_shi(type, vece, opc, v0, v1, a2); break; case INDEX_op_sari_vec: expand_vec_sari(type, vece, v0, v1, a2); break; + case INDEX_op_rotli_vec: + expand_vec_rotli(type, vece, v0, v1, a2); + break; + + case INDEX_op_rotls_vec: + expand_vec_rotls(type, vece, v0, v1, temp_tcgv_i32(arg_temp(a2))); + break; + + case INDEX_op_rotlv_vec: + v2 = temp_tcgv_vec(arg_temp(a2)); + expand_vec_rotv(type, vece, v0, v1, v2, false); + break; + case INDEX_op_rotrv_vec: + v2 = temp_tcgv_vec(arg_temp(a2)); + expand_vec_rotv(type, vece, v0, v1, v2, true); + break; + case INDEX_op_mul_vec: v2 = temp_tcgv_vec(arg_temp(a2)); expand_vec_mul(type, vece, v0, v1, v2); -- 2.25.1 For immediate rotate , we can implement this in two instructions, using SLI. For variable rotate, the oddness of aarch64 right-shift- as-negative-left-shift means a backend-specific expansion works best. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.opc.h | 1 + tcg/aarch64/tcg-target.inc.c | 53 ++++++++++++++++++++++++++++++++++-- 2 files changed, 52 insertions(+), 2 deletions(-) diff --git a/tcg/aarch64/tcg-target.opc.h b/tcg/aarch64/tcg-target.opc.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.opc.h +++ b/tcg/aarch64/tcg-target.opc.h @@ -XXX,XX +XXX,XX @@ */ DEF(aa64_sshl_vec, 1, 2, 0, IMPLVEC) +DEF(aa64_sli_vec, 1, 2, 1, IMPLVEC) diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.inc.c +++ b/tcg/aarch64/tcg-target.inc.c @@ -XXX,XX +XXX,XX @@ typedef enum { I3614_SSHR = 0x0f000400, I3614_SSRA = 0x0f001400, I3614_SHL = 0x0f005400, + I3614_SLI = 0x2f005400, I3614_USHR = 0x2f000400, I3614_USRA = 0x2f001400, @@ -XXX,XX +XXX,XX @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc, case INDEX_op_sari_vec: tcg_out_insn(s, 3614, SSHR, is_q, a0, a1, (16 << vece) - a2); break; + case INDEX_op_aa64_sli_vec: + tcg_out_insn(s, 3614, SLI, is_q, a0, a2, args[3] + (8 << vece)); + break; case INDEX_op_shlv_vec: tcg_out_insn(s, 3616, USHL, is_q, vece, a0, a1, a2); break; @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) case INDEX_op_shlv_vec: case INDEX_op_bitsel_vec: return 1; + case INDEX_op_rotli_vec: case INDEX_op_shrv_vec: case INDEX_op_sarv_vec: + case INDEX_op_rotlv_vec: + case INDEX_op_rotrv_vec: return -1; case INDEX_op_mul_vec: case INDEX_op_smax_vec: @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, TCGArg a0, ...) { va_list va; - TCGv_vec v0, v1, v2, t1; + TCGv_vec v0, v1, v2, t1, t2; + TCGArg a2; va_start(va, a0); v0 = temp_tcgv_vec(arg_temp(a0)); v1 = temp_tcgv_vec(arg_temp(va_arg(va, TCGArg))); - v2 = temp_tcgv_vec(arg_temp(va_arg(va, TCGArg))); + a2 = va_arg(va, TCGArg); + v2 = temp_tcgv_vec(arg_temp(a2)); switch (opc) { + case INDEX_op_rotli_vec: + t1 = tcg_temp_new_vec(type); + tcg_gen_shri_vec(vece, t1, v1, -a2 & ((8 << vece) - 1)); + vec_gen_4(INDEX_op_aa64_sli_vec, type, vece, + tcgv_vec_arg(v0), tcgv_vec_arg(t1), tcgv_vec_arg(v1), a2); + tcg_temp_free_vec(t1); + break; + case INDEX_op_shrv_vec: case INDEX_op_sarv_vec: /* Right shifts are negative left shifts for AArch64. */ @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, tcg_temp_free_vec(t1); break; + case INDEX_op_rotlv_vec: + t1 = tcg_temp_new_vec(type); + tcg_gen_dupi_vec(vece, t1, 8 << vece); + tcg_gen_sub_vec(vece, t1, v2, t1); + /* Right shifts are negative left shifts for AArch64. */ + vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1), + tcgv_vec_arg(v1), tcgv_vec_arg(t1)); + vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(v0), + tcgv_vec_arg(v1), tcgv_vec_arg(v2)); + tcg_gen_or_vec(vece, v0, v0, t1); + tcg_temp_free_vec(t1); + break; + + case INDEX_op_rotrv_vec: + t1 = tcg_temp_new_vec(type); + t2 = tcg_temp_new_vec(type); + tcg_gen_neg_vec(vece, t1, v2); + tcg_gen_dupi_vec(vece, t2, 8 << vece); + tcg_gen_add_vec(vece, t2, t1, t2); + /* Right shifts are negative left shifts for AArch64. */ + vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1), + tcgv_vec_arg(v1), tcgv_vec_arg(t1)); + vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t2), + tcgv_vec_arg(v1), tcgv_vec_arg(t2)); + tcg_gen_or_vec(vece, v0, t1, t2); + tcg_temp_free_vec(t1); + tcg_temp_free_vec(t2); + break; + default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) static const TCGTargetOpDef lZ_l = { .args_ct_str = { "lZ", "l" } }; static const TCGTargetOpDef r_r_r = { .args_ct_str = { "r", "r", "r" } }; static const TCGTargetOpDef w_w_w = { .args_ct_str = { "w", "w", "w" } }; + static const TCGTargetOpDef w_0_w = { .args_ct_str = { "w", "0", "w" } }; static const TCGTargetOpDef w_w_wO = { .args_ct_str = { "w", "w", "wO" } }; static const TCGTargetOpDef w_w_wN = { .args_ct_str = { "w", "w", "wN" } }; static const TCGTargetOpDef w_w_wZ = { .args_ct_str = { "w", "w", "wZ" } }; @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) return &w_w_wZ; case INDEX_op_bitsel_vec: return &w_w_w_w; + case INDEX_op_aa64_sli_vec: + return &w_0_w; default: return NULL; -- 2.25.1 We already had support for rotlv, using a target-specific opcode; convert to use the generic opcode. Handle rotrv via simple negation. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.h | 2 +- tcg/ppc/tcg-target.opc.h | 1 - tcg/ppc/tcg-target.inc.c | 23 +++++++++++++++++++---- 3 files changed, 20 insertions(+), 6 deletions(-) diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_vsx; #define TCG_TARGET_HAS_abs_vec 0 #define TCG_TARGET_HAS_roti_vec 0 #define TCG_TARGET_HAS_rots_vec 0 -#define TCG_TARGET_HAS_rotv_vec 0 +#define TCG_TARGET_HAS_rotv_vec 1 #define TCG_TARGET_HAS_shi_vec 0 #define TCG_TARGET_HAS_shs_vec 0 #define TCG_TARGET_HAS_shv_vec 1 diff --git a/tcg/ppc/tcg-target.opc.h b/tcg/ppc/tcg-target.opc.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.opc.h +++ b/tcg/ppc/tcg-target.opc.h @@ -XXX,XX +XXX,XX @@ DEF(ppc_msum_vec, 1, 3, 0, IMPLVEC) DEF(ppc_muleu_vec, 1, 2, 0, IMPLVEC) DEF(ppc_mulou_vec, 1, 2, 0, IMPLVEC) DEF(ppc_pkum_vec, 1, 2, 0, IMPLVEC) -DEF(ppc_rotl_vec, 1, 2, 0, IMPLVEC) diff --git a/tcg/ppc/tcg-target.inc.c b/tcg/ppc/tcg-target.inc.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.inc.c +++ b/tcg/ppc/tcg-target.inc.c @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) case INDEX_op_shlv_vec: case INDEX_op_shrv_vec: case INDEX_op_sarv_vec: + case INDEX_op_rotlv_vec: return vece <= MO_32 || have_isa_2_07; case INDEX_op_ssadd_vec: case INDEX_op_sssub_vec: @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) case INDEX_op_shli_vec: case INDEX_op_shri_vec: case INDEX_op_sari_vec: + case INDEX_op_rotli_vec: return vece <= MO_32 || have_isa_2_07 ? -1 : 0; case INDEX_op_neg_vec: return vece >= MO_32 && have_isa_3_00; @@ -XXX,XX +XXX,XX @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece) return 0; case INDEX_op_bitsel_vec: return have_vsx; + case INDEX_op_rotrv_vec: + return -1; default: return 0; } @@ -XXX,XX +XXX,XX @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc, case INDEX_op_ppc_pkum_vec: insn = pkum_op[vece]; break; - case INDEX_op_ppc_rotl_vec: + case INDEX_op_rotlv_vec: insn = rotl_op[vece]; break; case INDEX_op_ppc_msum_vec: @@ -XXX,XX +XXX,XX @@ static void expand_vec_mul(TCGType type, unsigned vece, TCGv_vec v0, t3 = tcg_temp_new_vec(type); t4 = tcg_temp_new_vec(type); tcg_gen_dupi_vec(MO_8, t4, -16); - vec_gen_3(INDEX_op_ppc_rotl_vec, type, MO_32, tcgv_vec_arg(t1), + vec_gen_3(INDEX_op_rotlv_vec, type, MO_32, tcgv_vec_arg(t1), tcgv_vec_arg(v2), tcgv_vec_arg(t4)); vec_gen_3(INDEX_op_ppc_mulou_vec, type, MO_16, tcgv_vec_arg(t2), tcgv_vec_arg(v1), tcgv_vec_arg(v2)); @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, TCGArg a0, ...) { va_list va; - TCGv_vec v0, v1, v2; + TCGv_vec v0, v1, v2, t0; TCGArg a2; va_start(va, a0); @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, case INDEX_op_sari_vec: expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_sarv_vec); break; + case INDEX_op_rotli_vec: + expand_vec_shi(type, vece, v0, v1, a2, INDEX_op_rotlv_vec); + break; case INDEX_op_cmp_vec: v2 = temp_tcgv_vec(arg_temp(a2)); expand_vec_cmp(type, vece, v0, v1, v2, va_arg(va, TCGArg)); @@ -XXX,XX +XXX,XX @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece, v2 = temp_tcgv_vec(arg_temp(a2)); expand_vec_mul(type, vece, v0, v1, v2); break; + case INDEX_op_rotlv_vec: + v2 = temp_tcgv_vec(arg_temp(a2)); + t0 = tcg_temp_new_vec(type); + tcg_gen_neg_vec(vece, t0, v2); + tcg_gen_rotlv_vec(vece, v0, v1, t0); + tcg_temp_free_vec(t0); + break; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op) case INDEX_op_shlv_vec: case INDEX_op_shrv_vec: case INDEX_op_sarv_vec: + case INDEX_op_rotlv_vec: + case INDEX_op_rotrv_vec: case INDEX_op_ppc_mrgh_vec: case INDEX_op_ppc_mrgl_vec: case INDEX_op_ppc_muleu_vec: case INDEX_op_ppc_mulou_vec: case INDEX_op_ppc_pkum_vec: - case INDEX_op_ppc_rotl_vec: case INDEX_op_dup2_vec: return &v_v_v; case INDEX_op_not_vec: -- 2.25.1 Acked-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/ppc/helper.h | 4 ---- target/ppc/int_helper.c | 17 ----------------- target/ppc/translate/vmx-impl.inc.c | 8 ++++---- 3 files changed, 4 insertions(+), 25 deletions(-) diff --git a/target/ppc/helper.h b/target/ppc/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/helper.h +++ b/target/ppc/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_3(vsubuqm, void, avr, avr, avr) DEF_HELPER_4(vsubecuq, void, avr, avr, avr, avr) DEF_HELPER_4(vsubeuqm, void, avr, avr, avr, avr) DEF_HELPER_3(vsubcuq, void, avr, avr, avr) -DEF_HELPER_3(vrlb, void, avr, avr, avr) -DEF_HELPER_3(vrlh, void, avr, avr, avr) -DEF_HELPER_3(vrlw, void, avr, avr, avr) -DEF_HELPER_3(vrld, void, avr, avr, avr) DEF_HELPER_4(vsldoi, void, avr, avr, avr, i32) DEF_HELPER_3(vextractub, void, avr, avr, i32) DEF_HELPER_3(vextractuh, void, avr, avr, i32) diff --git a/target/ppc/int_helper.c b/target/ppc/int_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/int_helper.c +++ b/target/ppc/int_helper.c @@ -XXX,XX +XXX,XX @@ VRFI(p, float_round_up) VRFI(z, float_round_to_zero) #undef VRFI -#define VROTATE(suffix, element, mask) \ - void helper_vrl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ - { \ - int i; \ - \ - for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ - unsigned int shift = b->element[i] & mask; \ - r->element[i] = (a->element[i] << shift) | \ - (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \ - } \ - } -VROTATE(b, u8, 0x7) -VROTATE(h, u16, 0xF) -VROTATE(w, u32, 0x1F) -VROTATE(d, u64, 0x3F) -#undef VROTATE - void helper_vrsqrtefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b) { int i; diff --git a/target/ppc/translate/vmx-impl.inc.c b/target/ppc/translate/vmx-impl.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/translate/vmx-impl.inc.c +++ b/target/ppc/translate/vmx-impl.inc.c @@ -XXX,XX +XXX,XX @@ GEN_VXFORM3(vsubeuqm, 31, 0); GEN_VXFORM3(vsubecuq, 31, 0); GEN_VXFORM_DUAL(vsubeuqm, PPC_NONE, PPC2_ALTIVEC_207, \ vsubecuq, PPC_NONE, PPC2_ALTIVEC_207) -GEN_VXFORM(vrlb, 2, 0); -GEN_VXFORM(vrlh, 2, 1); -GEN_VXFORM(vrlw, 2, 2); +GEN_VXFORM_V(vrlb, MO_8, tcg_gen_gvec_rotlv, 2, 0); +GEN_VXFORM_V(vrlh, MO_16, tcg_gen_gvec_rotlv, 2, 1); +GEN_VXFORM_V(vrlw, MO_32, tcg_gen_gvec_rotlv, 2, 2); GEN_VXFORM(vrlwmi, 2, 2); GEN_VXFORM_DUAL(vrlw, PPC_ALTIVEC, PPC_NONE, \ vrlwmi, PPC_NONE, PPC2_ISA300) -GEN_VXFORM(vrld, 2, 3); +GEN_VXFORM_V(vrld, MO_64, tcg_gen_gvec_rotlv, 2, 3); GEN_VXFORM(vrldmi, 2, 3); GEN_VXFORM_DUAL(vrld, PPC_NONE, PPC2_ALTIVEC_207, \ vrldmi, PPC_NONE, PPC2_ISA300) -- 2.25.1 Merge VERLL and VERLLV into op_vesv and op_ves, alongside all of the other vector shift operations. Reviewed-by: David Hildenbrand <david@redhat.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/s390x/helper.h | 4 -- target/s390x/translate_vx.inc.c | 66 +++++---------------------------- target/s390x/vec_int_helper.c | 31 ---------------- target/s390x/insn-data.def | 4 +- 4 files changed, 11 insertions(+), 94 deletions(-) diff --git a/target/s390x/helper.h b/target/s390x/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/helper.h +++ b/target/s390x/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_4(gvec_vmlo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vmlo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vpopct8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vpopct16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) -DEF_HELPER_FLAGS_4(gvec_verllv8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) -DEF_HELPER_FLAGS_4(gvec_verllv16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) -DEF_HELPER_FLAGS_4(gvec_verll8, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) -DEF_HELPER_FLAGS_4(gvec_verll16, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_verim8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_verim16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vsl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) diff --git a/target/s390x/translate_vx.inc.c b/target/s390x/translate_vx.inc.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/translate_vx.inc.c +++ b/target/s390x/translate_vx.inc.c @@ -XXX,XX +XXX,XX @@ static DisasJumpType op_vpopct(DisasContext *s, DisasOps *o) return DISAS_NEXT; } -static void gen_rll_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b) -{ - TCGv_i32 t0 = tcg_temp_new_i32(); - - tcg_gen_andi_i32(t0, b, 31); - tcg_gen_rotl_i32(d, a, t0); - tcg_temp_free_i32(t0); -} - -static void gen_rll_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b) -{ - TCGv_i64 t0 = tcg_temp_new_i64(); - - tcg_gen_andi_i64(t0, b, 63); - tcg_gen_rotl_i64(d, a, t0); - tcg_temp_free_i64(t0); -} - -static DisasJumpType op_verllv(DisasContext *s, DisasOps *o) -{ - const uint8_t es = get_field(s, m4); - static const GVecGen3 g[4] = { - { .fno = gen_helper_gvec_verllv8, }, - { .fno = gen_helper_gvec_verllv16, }, - { .fni4 = gen_rll_i32, }, - { .fni8 = gen_rll_i64, }, - }; - - if (es > ES_64) { - gen_program_exception(s, PGM_SPECIFICATION); - return DISAS_NORETURN; - } - - gen_gvec_3(get_field(s, v1), get_field(s, v2), - get_field(s, v3), &g[es]); - return DISAS_NEXT; -} - -static DisasJumpType op_verll(DisasContext *s, DisasOps *o) -{ - const uint8_t es = get_field(s, m4); - static const GVecGen2s g[4] = { - { .fno = gen_helper_gvec_verll8, }, - { .fno = gen_helper_gvec_verll16, }, - { .fni4 = gen_rll_i32, }, - { .fni8 = gen_rll_i64, }, - }; - - if (es > ES_64) { - gen_program_exception(s, PGM_SPECIFICATION); - return DISAS_NORETURN; - } - gen_gvec_2s(get_field(s, v1), get_field(s, v3), o->addr1, - &g[es]); - return DISAS_NEXT; -} - static void gen_rim_i32(TCGv_i32 d, TCGv_i32 a, TCGv_i32 b, int32_t c) { TCGv_i32 t = tcg_temp_new_i32(); @@ -XXX,XX +XXX,XX @@ static DisasJumpType op_vesv(DisasContext *s, DisasOps *o) case 0x70: gen_gvec_fn_3(shlv, es, v1, v2, v3); break; + case 0x73: + gen_gvec_fn_3(rotlv, es, v1, v2, v3); + break; case 0x7a: gen_gvec_fn_3(sarv, es, v1, v2, v3); break; @@ -XXX,XX +XXX,XX @@ static DisasJumpType op_ves(DisasContext *s, DisasOps *o) case 0x30: gen_gvec_fn_2i(shli, es, v1, v3, d2); break; + case 0x33: + gen_gvec_fn_2i(rotli, es, v1, v3, d2); + break; case 0x3a: gen_gvec_fn_2i(sari, es, v1, v3, d2); break; @@ -XXX,XX +XXX,XX @@ static DisasJumpType op_ves(DisasContext *s, DisasOps *o) case 0x30: gen_gvec_fn_2s(shls, es, v1, v3, shift); break; + case 0x33: + gen_gvec_fn_2s(rotls, es, v1, v3, shift); + break; case 0x3a: gen_gvec_fn_2s(sars, es, v1, v3, shift); break; diff --git a/target/s390x/vec_int_helper.c b/target/s390x/vec_int_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/vec_int_helper.c +++ b/target/s390x/vec_int_helper.c @@ -XXX,XX +XXX,XX @@ void HELPER(gvec_vpopct##BITS)(void *v1, const void *v2, uint32_t desc) \ DEF_VPOPCT(8) DEF_VPOPCT(16) -#define DEF_VERLLV(BITS) \ -void HELPER(gvec_verllv##BITS)(void *v1, const void *v2, const void *v3, \ - uint32_t desc) \ -{ \ - int i; \ - \ - for (i = 0; i < (128 / BITS); i++) { \ - const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ - const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ - \ - s390_vec_write_element##BITS(v1, i, rol##BITS(a, b)); \ - } \ -} -DEF_VERLLV(8) -DEF_VERLLV(16) - -#define DEF_VERLL(BITS) \ -void HELPER(gvec_verll##BITS)(void *v1, const void *v2, uint64_t count, \ - uint32_t desc) \ -{ \ - int i; \ - \ - for (i = 0; i < (128 / BITS); i++) { \ - const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ - \ - s390_vec_write_element##BITS(v1, i, rol##BITS(a, count)); \ - } \ -} -DEF_VERLL(8) -DEF_VERLL(16) - #define DEF_VERIM(BITS) \ void HELPER(gvec_verim##BITS)(void *v1, const void *v2, const void *v3, \ uint32_t desc) \ diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/insn-data.def +++ b/target/s390x/insn-data.def @@ -XXX,XX +XXX,XX @@ /* VECTOR POPULATION COUNT */ F(0xe750, VPOPCT, VRR_a, V, 0, 0, 0, 0, vpopct, 0, IF_VEC) /* VECTOR ELEMENT ROTATE LEFT LOGICAL */ - F(0xe773, VERLLV, VRR_c, V, 0, 0, 0, 0, verllv, 0, IF_VEC) - F(0xe733, VERLL, VRS_a, V, la2, 0, 0, 0, verll, 0, IF_VEC) + F(0xe773, VERLLV, VRR_c, V, 0, 0, 0, 0, vesv, 0, IF_VEC) + F(0xe733, VERLL, VRS_a, V, la2, 0, 0, 0, ves, 0, IF_VEC) /* VECTOR ELEMENT ROTATE AND INSERT UNDER MASK */ F(0xe772, VERIM, VRI_d, V, 0, 0, 0, 0, verim, 0, IF_VEC) /* VECTOR ELEMENT SHIFT LEFT */ -- 2.25.1 If the output of the move is dead, then the last use is in the store. If we propagate the input to the store, then we can remove the move opcode entirely. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg.c | 78 +++++++++++++++++++++++++++++++++++++++---------------- 1 file changed, 56 insertions(+), 22 deletions(-) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static bool liveness_pass_2(TCGContext *s) } /* Outputs become available. */ - for (i = 0; i < nb_oargs; i++) { - arg_ts = arg_temp(op->args[i]); + if (opc == INDEX_op_mov_i32 || opc == INDEX_op_mov_i64) { + arg_ts = arg_temp(op->args[0]); dir_ts = arg_ts->state_ptr; - if (!dir_ts) { - continue; + if (dir_ts) { + op->args[0] = temp_arg(dir_ts); + changes = true; + + /* The output is now live and modified. */ + arg_ts->state = 0; + + if (NEED_SYNC_ARG(0)) { + TCGOpcode sopc = (arg_ts->type == TCG_TYPE_I32 + ? INDEX_op_st_i32 + : INDEX_op_st_i64); + TCGOp *sop = tcg_op_insert_after(s, op, sopc); + TCGTemp *out_ts = dir_ts; + + if (IS_DEAD_ARG(0)) { + out_ts = arg_temp(op->args[1]); + arg_ts->state = TS_DEAD; + tcg_op_remove(s, op); + } else { + arg_ts->state = TS_MEM; + } + + sop->args[0] = temp_arg(out_ts); + sop->args[1] = temp_arg(arg_ts->mem_base); + sop->args[2] = arg_ts->mem_offset; + } else { + tcg_debug_assert(!IS_DEAD_ARG(0)); + } } - op->args[i] = temp_arg(dir_ts); - changes = true; + } else { + for (i = 0; i < nb_oargs; i++) { + arg_ts = arg_temp(op->args[i]); + dir_ts = arg_ts->state_ptr; + if (!dir_ts) { + continue; + } + op->args[i] = temp_arg(dir_ts); + changes = true; - /* The output is now live and modified. */ - arg_ts->state = 0; + /* The output is now live and modified. */ + arg_ts->state = 0; - /* Sync outputs upon their last write. */ - if (NEED_SYNC_ARG(i)) { - TCGOpcode sopc = (arg_ts->type == TCG_TYPE_I32 - ? INDEX_op_st_i32 - : INDEX_op_st_i64); - TCGOp *sop = tcg_op_insert_after(s, op, sopc); + /* Sync outputs upon their last write. */ + if (NEED_SYNC_ARG(i)) { + TCGOpcode sopc = (arg_ts->type == TCG_TYPE_I32 + ? INDEX_op_st_i32 + : INDEX_op_st_i64); + TCGOp *sop = tcg_op_insert_after(s, op, sopc); - sop->args[0] = temp_arg(dir_ts); - sop->args[1] = temp_arg(arg_ts->mem_base); - sop->args[2] = arg_ts->mem_offset; + sop->args[0] = temp_arg(dir_ts); + sop->args[1] = temp_arg(arg_ts->mem_base); + sop->args[2] = arg_ts->mem_offset; - arg_ts->state = TS_MEM; - } - /* Drop outputs that are dead. */ - if (IS_DEAD_ARG(i)) { - arg_ts->state = TS_DEAD; + arg_ts->state = TS_MEM; + } + /* Drop outputs that are dead. */ + if (IS_DEAD_ARG(i)) { + arg_ts->state = TS_DEAD; + } } } } -- 2.25.1 From: Nick Hudson <skrll@netbsd.org> Fix building on NetBSD/arm by extracting the FSR value from the correct siginfo_t field. Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Nick Hudson <skrll@netbsd.org> Message-Id: <20200516154147.24842-1-skrll@netbsd.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/user-exec.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/user-exec.c +++ b/accel/tcg/user-exec.c @@ -XXX,XX +XXX,XX @@ int cpu_signal_handler(int host_signum, void *pinfo, #if defined(__NetBSD__) #include <ucontext.h> +#include <sys/siginfo.h> #endif int cpu_signal_handler(int host_signum, void *pinfo, @@ -XXX,XX +XXX,XX @@ int cpu_signal_handler(int host_signum, void *pinfo, siginfo_t *info = pinfo; #if defined(__NetBSD__) ucontext_t *uc = puc; + siginfo_t *si = pinfo; #else ucontext_t *uc = puc; #endif unsigned long pc; + uint32_t fsr; int is_write; #if defined(__NetBSD__) @@ -XXX,XX +XXX,XX @@ int cpu_signal_handler(int host_signum, void *pinfo, pc = uc->uc_mcontext.arm_pc; #endif - /* error_code is the FSR value, in which bit 11 is WnR (assuming a v6 or - * later processor; on v5 we will always report this as a read). +#ifdef __NetBSD__ + fsr = si->si_trap; +#else + fsr = uc->uc_mcontext.error_code; +#endif + /* + * In the FSR, bit 11 is WnR, assuming a v6 or + * later processor. On v5 we will always report + * this as a read, which will fail later. */ - is_write = extract32(uc->uc_mcontext.error_code, 11, 1); + is_write = extract32(fsr, 11, 1); return handle_cpu_signal(pc, info, is_write, &uc->uc_sigmask); } -- 2.25.1 From: Nick Hudson <skrll@netbsd.org> Fix qemu build on NetBSD/evbarm-aarch64 by providing a NetBSD specific cpu_signal_handler. Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Nick Hudson <skrll@netbsd.org> Message-Id: <20200517101529.5367-1-skrll@netbsd.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/user-exec.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/user-exec.c +++ b/accel/tcg/user-exec.c @@ -XXX,XX +XXX,XX @@ int cpu_signal_handler(int host_signum, void *pinfo, #elif defined(__aarch64__) +#if defined(__NetBSD__) + +#include <ucontext.h> +#include <sys/siginfo.h> + +int cpu_signal_handler(int host_signum, void *pinfo, void *puc) +{ + ucontext_t *uc = puc; + siginfo_t *si = pinfo; + unsigned long pc; + int is_write; + uint32_t esr; + + pc = uc->uc_mcontext.__gregs[_REG_PC]; + esr = si->si_trap; + + /* + * siginfo_t::si_trap is the ESR value, for data aborts ESR.EC + * is 0b10010x: then bit 6 is the WnR bit + */ + is_write = extract32(esr, 27, 5) == 0x12 && extract32(esr, 6, 1) == 1; + return handle_cpu_signal(pc, si, is_write, &uc->uc_sigmask); +} + +#else + #ifndef ESR_MAGIC /* Pre-3.16 kernel headers don't have these, so provide fallback definitions */ #define ESR_MAGIC 0x45535201 @@ -XXX,XX +XXX,XX @@ int cpu_signal_handler(int host_signum, void *pinfo, void *puc) } return handle_cpu_signal(pc, info, is_write, &uc->uc_sigmask); } +#endif #elif defined(__s390__) -- 2.25.1 The following changes since commit aa3a285b5bc56a4208b3b57d4a55291e9c260107: Merge tag 'mem-2024-12-21' of https://github.com/davidhildenbrand/qemu into staging (2024-12-22 14:33:27 -0500) are available in the Git repository at: https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20241224 for you to fetch changes up to e4a8e093dc74be049f4829831dce76e5edab0003: accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core (2024-12-24 08:32:15 -0800) ---------------------------------------------------------------- tcg/optimize: Remove in-flight mask data from OptContext fpu: Add float*_muladd_scalbn fpu: Remove float_muladd_halve_result fpu: Add float_round_nearest_even_max fpu: Add float_muladd_suppress_add_product_zero target/hexagon: Use float32_muladd accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core ---------------------------------------------------------------- Ilya Leoshkevich (1): tests/tcg: Do not use inttypes.h in multiarch/system/memory.c Pierrick Bouvier (1): plugins: optimize cpu_index code generation Richard Henderson (70): tcg/optimize: Split out finish_bb, finish_ebb tcg/optimize: Split out fold_affected_mask tcg/optimize: Copy mask writeback to fold_masks tcg/optimize: Split out fold_masks_zs tcg/optimize: Augment s_mask from z_mask in fold_masks_zs tcg/optimize: Change representation of s_mask tcg/optimize: Use finish_folding in fold_add, fold_add_vec, fold_addsub2 tcg/optimize: Introduce const value accessors for TempOptInfo tcg/optimize: Use fold_masks_zs in fold_and tcg/optimize: Use fold_masks_zs in fold_andc tcg/optimize: Use fold_masks_zs in fold_bswap tcg/optimize: Use fold_masks_zs in fold_count_zeros tcg/optimize: Use fold_masks_z in fold_ctpop tcg/optimize: Use fold_and and fold_masks_z in fold_deposit tcg/optimize: Compute sign mask in fold_deposit tcg/optimize: Use finish_folding in fold_divide tcg/optimize: Use finish_folding in fold_dup, fold_dup2 tcg/optimize: Use fold_masks_s in fold_eqv tcg/optimize: Use fold_masks_z in fold_extract tcg/optimize: Use finish_folding in fold_extract2 tcg/optimize: Use fold_masks_zs in fold_exts tcg/optimize: Use fold_masks_z in fold_extu tcg/optimize: Use fold_masks_zs in fold_movcond tcg/optimize: Use finish_folding in fold_mul* tcg/optimize: Use fold_masks_s in fold_nand tcg/optimize: Use fold_masks_z in fold_neg_no_const tcg/optimize: Use fold_masks_s in fold_nor tcg/optimize: Use fold_masks_s in fold_not tcg/optimize: Use fold_masks_zs in fold_or tcg/optimize: Use fold_masks_zs in fold_orc tcg/optimize: Use fold_masks_zs in fold_qemu_ld tcg/optimize: Return true from fold_qemu_st, fold_tcg_st tcg/optimize: Use finish_folding in fold_remainder tcg/optimize: Distinguish simplification in fold_setcond_zmask tcg/optimize: Use fold_masks_z in fold_setcond tcg/optimize: Use fold_masks_s in fold_negsetcond tcg/optimize: Use fold_masks_z in fold_setcond2 tcg/optimize: Use finish_folding in fold_cmp_vec tcg/optimize: Use finish_folding in fold_cmpsel_vec tcg/optimize: Use fold_masks_zs in fold_sextract tcg/optimize: Use fold_masks_zs, fold_masks_s in fold_shift tcg/optimize: Simplify sign bit test in fold_shift tcg/optimize: Use finish_folding in fold_sub, fold_sub_vec tcg/optimize: Use fold_masks_zs in fold_tcg_ld tcg/optimize: Use finish_folding in fold_tcg_ld_memcopy tcg/optimize: Use fold_masks_zs in fold_xor tcg/optimize: Use finish_folding in fold_bitsel_vec tcg/optimize: Use finish_folding as default in tcg_optimize tcg/optimize: Remove z_mask, s_mask from OptContext tcg/optimize: Re-enable sign-mask optimizations tcg/optimize: Move fold_bitsel_vec into alphabetic sort tcg/optimize: Move fold_cmp_vec, fold_cmpsel_vec into alphabetic sort softfloat: Add float{16,32,64}_muladd_scalbn target/arm: Use float*_muladd_scalbn target/sparc: Use float*_muladd_scalbn softfloat: Remove float_muladd_halve_result softfloat: Add float_round_nearest_even_max softfloat: Add float_muladd_suppress_add_product_zero target/hexagon: Use float32_mul in helper_sfmpy target/hexagon: Use float32_muladd for helper_sffma target/hexagon: Use float32_muladd for helper_sffms target/hexagon: Use float32_muladd_scalbn for helper_sffma_sc target/hexagon: Use float32_muladd for helper_sffm[as]_lib target/hexagon: Remove internal_fmafx target/hexagon: Expand GEN_XF_ROUND target/hexagon: Remove Float target/hexagon: Remove Double target/hexagon: Use mulu64 for int128_mul_6464 target/hexagon: Simplify internal_mpyhh setup accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core include/exec/translator.h | 14 - include/fpu/softfloat-types.h | 2 + include/fpu/softfloat.h | 14 +- include/hw/core/tcg-cpu-ops.h | 13 + target/alpha/cpu.h | 2 + target/arm/internals.h | 2 + target/avr/cpu.h | 2 + target/hexagon/cpu.h | 2 + target/hexagon/fma_emu.h | 3 - target/hppa/cpu.h | 2 + target/i386/tcg/helper-tcg.h | 2 + target/loongarch/internals.h | 2 + target/m68k/cpu.h | 2 + target/microblaze/cpu.h | 2 + target/mips/tcg/tcg-internal.h | 2 + target/openrisc/cpu.h | 2 + target/ppc/cpu.h | 2 + target/riscv/cpu.h | 3 + target/rx/cpu.h | 2 + target/s390x/s390x-internal.h | 2 + target/sh4/cpu.h | 2 + target/sparc/cpu.h | 2 + target/sparc/helper.h | 4 +- target/tricore/cpu.h | 2 + target/xtensa/cpu.h | 2 + accel/tcg/cpu-exec.c | 8 +- accel/tcg/plugin-gen.c | 9 + accel/tcg/translate-all.c | 8 +- fpu/softfloat.c | 63 +-- target/alpha/cpu.c | 1 + target/alpha/translate.c | 4 +- target/arm/cpu.c | 1 + target/arm/tcg/cpu-v7m.c | 1 + target/arm/tcg/helper-a64.c | 6 +- target/arm/tcg/translate.c | 5 +- target/avr/cpu.c | 1 + target/avr/translate.c | 6 +- target/hexagon/cpu.c | 1 + target/hexagon/fma_emu.c | 496 ++++++--------------- target/hexagon/op_helper.c | 125 ++---- target/hexagon/translate.c | 4 +- target/hppa/cpu.c | 1 + target/hppa/translate.c | 4 +- target/i386/tcg/tcg-cpu.c | 1 + target/i386/tcg/translate.c | 5 +- target/loongarch/cpu.c | 1 + target/loongarch/tcg/translate.c | 4 +- target/m68k/cpu.c | 1 + target/m68k/translate.c | 4 +- target/microblaze/cpu.c | 1 + target/microblaze/translate.c | 4 +- target/mips/cpu.c | 1 + target/mips/tcg/translate.c | 4 +- target/openrisc/cpu.c | 1 + target/openrisc/translate.c | 4 +- target/ppc/cpu_init.c | 1 + target/ppc/translate.c | 4 +- target/riscv/tcg/tcg-cpu.c | 1 + target/riscv/translate.c | 4 +- target/rx/cpu.c | 1 + target/rx/translate.c | 4 +- target/s390x/cpu.c | 1 + target/s390x/tcg/translate.c | 4 +- target/sh4/cpu.c | 1 + target/sh4/translate.c | 4 +- target/sparc/cpu.c | 1 + target/sparc/fop_helper.c | 8 +- target/sparc/translate.c | 84 ++-- target/tricore/cpu.c | 1 + target/tricore/translate.c | 5 +- target/xtensa/cpu.c | 1 + target/xtensa/translate.c | 4 +- tcg/optimize.c | 857 +++++++++++++++++++----------------- tests/tcg/multiarch/system/memory.c | 9 +- fpu/softfloat-parts.c.inc | 16 +- 75 files changed, 866 insertions(+), 1009 deletions(-) From: Ilya Leoshkevich <iii@linux.ibm.com> make check-tcg fails on Fedora with the following error message: alpha-linux-gnu-gcc [...] qemu/tests/tcg/multiarch/system/memory.c -o memory [...] qemu/tests/tcg/multiarch/system/memory.c:17:10: fatal error: inttypes.h: No such file or directory 17 | #include <inttypes.h> | ^~~~~~~~~~~~ compilation terminated. The reason is that Fedora has cross-compilers, but no cross-glibc headers. Fix by hardcoding the format specifiers and dropping the include. An alternative fix would be to introduce a configure check for inttypes.h. But this would make it impossible to use Fedora cross-compilers for softmmu tests, which used to work so far. Fixes: ecbcc9ead2f8 ("tests/tcg: add a system test to check memory instrumentation") Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-ID: <20241010085906.226249-1-iii@linux.ibm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tests/tcg/multiarch/system/memory.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/tests/tcg/multiarch/system/memory.c b/tests/tcg/multiarch/system/memory.c index XXXXXXX..XXXXXXX 100644 --- a/tests/tcg/multiarch/system/memory.c +++ b/tests/tcg/multiarch/system/memory.c @@ -XXX,XX +XXX,XX @@ #include <stdint.h> #include <stdbool.h> -#include <inttypes.h> #include <minilib.h> #ifndef CHECK_UNALIGNED @@ -XXX,XX +XXX,XX @@ int main(void) int i; bool ok = true; - ml_printf("Test data start: 0x%"PRIxPTR"\n", &test_data[0]); - ml_printf("Test data end: 0x%"PRIxPTR"\n", &test_data[TEST_SIZE]); + ml_printf("Test data start: 0x%lx\n", (unsigned long)&test_data[0]); + ml_printf("Test data end: 0x%lx\n", (unsigned long)&test_data[TEST_SIZE]); /* Run through the unsigned tests first */ for (i = 0; i < ARRAY_SIZE(init_ufns) && ok; i++) { @@ -XXX,XX +XXX,XX @@ int main(void) ok = do_signed_reads(true); } - ml_printf("Test data read: %"PRId32"\n", test_read_count); - ml_printf("Test data write: %"PRId32"\n", test_write_count); + ml_printf("Test data read: %lu\n", (unsigned long)test_read_count); + ml_printf("Test data write: %lu\n", (unsigned long)test_write_count); ml_printf("Test complete: %s\n", ok ? "PASSED" : "FAILED"); return ok ? 0 : -1; } -- 2.43.0 From: Pierrick Bouvier <pierrick.bouvier@linaro.org> When running with a single vcpu, we can return a constant instead of a load when accessing cpu_index. A side effect is that all tcg operations using it are optimized, most notably scoreboard access. When running a simple loop in user-mode, the speedup is around 20%. Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-ID: <20241128213843.1023080-1-pierrick.bouvier@linaro.org> --- accel/tcg/plugin-gen.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/accel/tcg/plugin-gen.c b/accel/tcg/plugin-gen.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/plugin-gen.c +++ b/accel/tcg/plugin-gen.c @@ -XXX,XX +XXX,XX @@ static void gen_disable_mem_helper(void) static TCGv_i32 gen_cpu_index(void) { + /* + * Optimize when we run with a single vcpu. All values using cpu_index, + * including scoreboard index, will be optimized out. + * User-mode calls tb_flush when setting this flag. In system-mode, all + * vcpus are created before generating code. + */ + if (!tcg_cflags_has(current_cpu, CF_PARALLEL)) { + return tcg_constant_i32(current_cpu->cpu_index); + } TCGv_i32 cpu_index = tcg_temp_ebb_new_i32(); tcg_gen_ld_i32(cpu_index, tcg_env, -offsetof(ArchCPU, env) + offsetof(CPUState, cpu_index)); -- 2.43.0 Call them directly from the opcode switch statement in tcg_optimize, rather than in finish_folding based on opcode flags. Adjust folding of conditional branches to match. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 47 +++++++++++++++++++++++++++++++---------------- 1 file changed, 31 insertions(+), 16 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static void copy_propagate(OptContext *ctx, TCGOp *op, } } +static void finish_bb(OptContext *ctx) +{ + /* We only optimize memory barriers across basic blocks. */ + ctx->prev_mb = NULL; +} + +static void finish_ebb(OptContext *ctx) +{ + finish_bb(ctx); + /* We only optimize across extended basic blocks. */ + memset(&ctx->temps_used, 0, sizeof(ctx->temps_used)); + remove_mem_copy_all(ctx); +} + static void finish_folding(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; int i, nb_oargs; - /* - * We only optimize extended basic blocks. If the opcode ends a BB - * and is not a conditional branch, reset all temp data. - */ - if (def->flags & TCG_OPF_BB_END) { - ctx->prev_mb = NULL; - if (!(def->flags & TCG_OPF_COND_BRANCH)) { - memset(&ctx->temps_used, 0, sizeof(ctx->temps_used)); - remove_mem_copy_all(ctx); - } - return; - } - nb_oargs = def->nb_oargs; for (i = 0; i < nb_oargs; i++) { TCGTemp *ts = arg_temp(op->args[i]); @@ -XXX,XX +XXX,XX @@ static bool fold_brcond(OptContext *ctx, TCGOp *op) if (i > 0) { op->opc = INDEX_op_br; op->args[0] = op->args[3]; + finish_ebb(ctx); + } else { + finish_bb(ctx); } - return false; + return true; } static bool fold_brcond2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) } op->opc = INDEX_op_br; op->args[0] = label; - break; + finish_ebb(ctx); + return true; } - return false; + + finish_bb(ctx); + return true; } static bool fold_bswap(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) CASE_OP_32_64_VEC(xor): done = fold_xor(&ctx, op); break; + case INDEX_op_set_label: + case INDEX_op_br: + case INDEX_op_exit_tb: + case INDEX_op_goto_tb: + case INDEX_op_goto_ptr: + finish_ebb(&ctx); + done = true; + break; default: break; } -- 2.43.0 There are only a few logical operations which can compute an "affected" mask. Split out handling of this optimization to a separate function, only to be called when applicable. Remove the a_mask field from OptContext, as the mask is no longer stored anywhere. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 42 +++++++++++++++++++++++++++--------------- 1 file changed, 27 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { QSIMPLEQ_HEAD(, MemCopyInfo) mem_free; /* In flight values from optimization. */ - uint64_t a_mask; /* mask bit is 0 iff value identical to first input */ uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ uint64_t s_mask; /* mask of clrsb(value) bits */ TCGType type; @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) static bool fold_masks(OptContext *ctx, TCGOp *op) { - uint64_t a_mask = ctx->a_mask; uint64_t z_mask = ctx->z_mask; uint64_t s_mask = ctx->s_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) * type changing opcodes. */ if (ctx->type == TCG_TYPE_I32) { - a_mask = (int32_t)a_mask; z_mask = (int32_t)z_mask; s_mask |= MAKE_64BIT_MASK(32, 32); ctx->z_mask = z_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) if (z_mask == 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], 0); } + return false; +} + +/* + * An "affected" mask bit is 0 if and only if the result is identical + * to the first input. Thus if the entire mask is 0, the operation + * is equivalent to a copy. + */ +static bool fold_affected_mask(OptContext *ctx, TCGOp *op, uint64_t a_mask) +{ + if (ctx->type == TCG_TYPE_I32) { + a_mask = (uint32_t)a_mask; + } if (a_mask == 0) { return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); } @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer affected bits from it. */ - if (arg_is_const(op->args[2])) { - ctx->a_mask = z1 & ~z2; + if (arg_is_const(op->args[2]) && + fold_affected_mask(ctx, op, z1 & ~z2)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) */ if (arg_is_const(op->args[2])) { uint64_t z2 = ~arg_info(op->args[2])->z_mask; - ctx->a_mask = z1 & ~z2; + if (fold_affected_mask(ctx, op, z1 & ~z2)) { + return true; + } z1 &= z2; } ctx->z_mask = z1; @@ -XXX,XX +XXX,XX @@ static bool fold_extract(OptContext *ctx, TCGOp *op) z_mask_old = arg_info(op->args[1])->z_mask; z_mask = extract64(z_mask_old, pos, len); - if (pos == 0) { - ctx->a_mask = z_mask_old ^ z_mask; + if (pos == 0 && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { + return true; } ctx->z_mask = z_mask; ctx->s_mask = smask_from_zmask(z_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = s_mask; - if (!type_change) { - ctx->a_mask = s_mask & ~s_mask_old; + if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = smask_from_zmask(z_mask); - if (!type_change) { - ctx->a_mask = z_mask_old ^ z_mask; + if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { + return true; } return fold_masks(ctx, op); } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask |= MAKE_64BIT_MASK(len, 64 - len); ctx->s_mask = s_mask; - if (pos == 0) { - ctx->a_mask = s_mask & ~s_mask_old; + if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) } /* Assume all bits affected, no bits known zero, no sign reps. */ - ctx.a_mask = -1; ctx.z_mask = -1; ctx.s_mask = 0; -- 2.43.0 Use of fold_masks should be restricted to those opcodes that can reliably make use of it -- those with a single output, and from higher-level folders that set up the masks. Prepare for conversion of each folder in turn. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) { uint64_t z_mask = ctx->z_mask; uint64_t s_mask = ctx->s_mask; + const TCGOpDef *def = &tcg_op_defs[op->opc]; + TCGTemp *ts; + TempOptInfo *ti; + + /* Only single-output opcodes are supported here. */ + tcg_debug_assert(def->nb_oargs == 1); /* * 32-bit ops generate 32-bit results, which for the purpose of @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) if (ctx->type == TCG_TYPE_I32) { z_mask = (int32_t)z_mask; s_mask |= MAKE_64BIT_MASK(32, 32); - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; } if (z_mask == 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], 0); } - return false; + + ts = arg_temp(op->args[0]); + reset_ts(ctx, ts); + + ti = ts_info(ts); + ti->z_mask = z_mask; + ti->s_mask = s_mask; + return true; } /* -- 2.43.0 Add a routine to which masks can be passed directly, rather than storing them into OptContext. To be used in upcoming patches. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) return fold_const2(ctx, op); } -static bool fold_masks(OptContext *ctx, TCGOp *op) +/* + * Record "zero" and "sign" masks for the single output of @op. + * See TempOptInfo definition of z_mask and s_mask. + * If z_mask allows, fold the output to constant zero. + */ +static bool fold_masks_zs(OptContext *ctx, TCGOp *op, + uint64_t z_mask, uint64_t s_mask) { - uint64_t z_mask = ctx->z_mask; - uint64_t s_mask = ctx->s_mask; const TCGOpDef *def = &tcg_op_defs[op->opc]; TCGTemp *ts; TempOptInfo *ti; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) return true; } +static bool fold_masks(OptContext *ctx, TCGOp *op) +{ + return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); +} + /* * An "affected" mask bit is 0 if and only if the result is identical * to the first input. Thus if the entire mask is 0, the operation -- 2.43.0 Consider the passed s_mask to be a minimum deduced from either existing s_mask or from a sign-extension operation. We may be able to deduce more from the set of known zeros. Remove identical logic from several opcode folders. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 ++++++--------------- 1 file changed, 6 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) * Record "zero" and "sign" masks for the single output of @op. * See TempOptInfo definition of z_mask and s_mask. * If z_mask allows, fold the output to constant zero. + * The passed s_mask may be augmented by z_mask. */ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, uint64_t z_mask, uint64_t s_mask) @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, ti = ts_info(ts); ti->z_mask = z_mask; - ti->s_mask = s_mask; + ti->s_mask = s_mask | smask_from_zmask(z_mask); return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - s_mask = smask_from_zmask(z_mask); + s_mask = 0; switch (op->args[2] & (TCG_BSWAP_OZ | TCG_BSWAP_OS)) { case TCG_BSWAP_OZ: break; @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) default: /* The high bits are undefined: force all bits above the sign to 1. */ z_mask |= sign << 1; - s_mask = 0; break; } ctx->z_mask = z_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } ctx->z_mask = arg_info(op->args[2])->z_mask | z_mask; - ctx->s_mask = smask_from_zmask(ctx->z_mask); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_ctpop(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - ctx->s_mask = smask_from_zmask(ctx->z_mask); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_extract(OptContext *ctx, TCGOp *op) return true; } ctx->z_mask = z_mask; - ctx->s_mask = smask_from_zmask(z_mask); return fold_masks(ctx, op); } @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) } ctx->z_mask = z_mask; - ctx->s_mask = smask_from_zmask(z_mask); if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) int width = 8 * memop_size(mop); if (width < 64) { - ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); - if (!(mop & MO_SIGN)) { + if (mop & MO_SIGN) { + ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); + } else { ctx->z_mask = MAKE_64BIT_MASK(0, width); - ctx->s_mask <<= 1; } } @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) fold_setcond_tst_pow2(ctx, op, false); ctx->z_mask = 1; - ctx->s_mask = smask_from_zmask(1); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) } ctx->z_mask = 1; - ctx->s_mask = smask_from_zmask(1); return false; do_setcond_const: @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) break; CASE_OP_32_64(ld8u): ctx->z_mask = MAKE_64BIT_MASK(0, 8); - ctx->s_mask = MAKE_64BIT_MASK(9, 55); break; CASE_OP_32_64(ld16s): ctx->s_mask = MAKE_64BIT_MASK(16, 48); break; CASE_OP_32_64(ld16u): ctx->z_mask = MAKE_64BIT_MASK(0, 16); - ctx->s_mask = MAKE_64BIT_MASK(17, 47); break; case INDEX_op_ld32s_i64: ctx->s_mask = MAKE_64BIT_MASK(32, 32); break; case INDEX_op_ld32u_i64: ctx->z_mask = MAKE_64BIT_MASK(0, 32); - ctx->s_mask = MAKE_64BIT_MASK(33, 31); break; default: g_assert_not_reached(); -- 2.43.0 Change the representation from sign bit repetitions to all bits equal to the sign bit, including the sign bit itself. The previous format has a problem in that it is difficult to recreate a valid sign mask after a shift operation: the "repetitions" part of the previous format meant that applying the same shift as for the value lead to an off-by-one value. The new format, including the sign bit itself, means that the sign mask can be manipulated in exactly the same way as the value, canonicalization is easier. Canonicalize the s_mask in fold_masks_zs, rather than requiring callers to do so. Treat 0 as a non-canonical but typeless input for no sign information, which will be reset as appropriate for the data type. We can easily fold in the data from z_mask while canonicalizing. Temporarily disable optimizations using s_mask while each operation is converted to use fold_masks_zs and to the new form. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 64 ++++++++++++-------------------------------------- 1 file changed, 15 insertions(+), 49 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct TempOptInfo { QSIMPLEQ_HEAD(, MemCopyInfo) mem_copy; uint64_t val; uint64_t z_mask; /* mask bit is 0 if and only if value bit is 0 */ - uint64_t s_mask; /* a left-aligned mask of clrsb(value) bits. */ + uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ } TempOptInfo; typedef struct OptContext { @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { /* In flight values from optimization. */ uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ - uint64_t s_mask; /* mask of clrsb(value) bits */ + uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ TCGType type; } OptContext; -/* Calculate the smask for a specific value. */ -static uint64_t smask_from_value(uint64_t value) -{ - int rep = clrsb64(value); - return ~(~0ull >> rep); -} - -/* - * Calculate the smask for a given set of known-zeros. - * If there are lots of zeros on the left, we can consider the remainder - * an unsigned field, and thus the corresponding signed field is one bit - * larger. - */ -static uint64_t smask_from_zmask(uint64_t zmask) -{ - /* - * Only the 0 bits are significant for zmask, thus the msb itself - * must be zero, else we have no sign information. - */ - int rep = clz64(zmask); - if (rep == 0) { - return 0; - } - rep -= 1; - return ~(~0ull >> rep); -} - -/* - * Recreate a properly left-aligned smask after manipulation. - * Some bit-shuffling, particularly shifts and rotates, may - * retain sign bits on the left, but may scatter disconnected - * sign bits on the right. Retain only what remains to the left. - */ -static uint64_t smask_from_smask(int64_t smask) -{ - /* Only the 1 bits are significant for smask */ - return smask_from_zmask(~smask); -} - static inline TempOptInfo *ts_info(TCGTemp *ts) { return ts->state_ptr; @@ -XXX,XX +XXX,XX @@ static void init_ts_info(OptContext *ctx, TCGTemp *ts) ti->is_const = true; ti->val = ts->val; ti->z_mask = ts->val; - ti->s_mask = smask_from_value(ts->val); + ti->s_mask = INT64_MIN >> clrsb64(ts->val); } else { ti->is_const = false; ti->z_mask = -1; @@ -XXX,XX +XXX,XX @@ static void finish_folding(OptContext *ctx, TCGOp *op) */ if (i == 0) { ts_info(ts)->z_mask = ctx->z_mask; - ts_info(ts)->s_mask = ctx->s_mask; } } } @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) * The passed s_mask may be augmented by z_mask. */ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, - uint64_t z_mask, uint64_t s_mask) + uint64_t z_mask, int64_t s_mask) { const TCGOpDef *def = &tcg_op_defs[op->opc]; TCGTemp *ts; TempOptInfo *ti; + int rep; /* Only single-output opcodes are supported here. */ tcg_debug_assert(def->nb_oargs == 1); @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, */ if (ctx->type == TCG_TYPE_I32) { z_mask = (int32_t)z_mask; - s_mask |= MAKE_64BIT_MASK(32, 32); + s_mask |= INT32_MIN; } if (z_mask == 0) { @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, ti = ts_info(ts); ti->z_mask = z_mask; - ti->s_mask = s_mask | smask_from_zmask(z_mask); + + /* Canonicalize s_mask and incorporate data from z_mask. */ + rep = clz64(~s_mask); + rep = MAX(rep, clz64(z_mask)); + rep = MAX(rep - 1, 0); + ti->s_mask = INT64_MIN >> rep; + return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = s_mask; - if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask |= MAKE_64BIT_MASK(len, 64 - len); ctx->s_mask = s_mask; - if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) ctx->z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); s_mask = do_constant_folding(op->opc, ctx->type, s_mask, sh); - ctx->s_mask = smask_from_smask(s_mask); return fold_masks(ctx, op); } -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static void finish_ebb(OptContext *ctx) remove_mem_copy_all(ctx); } -static void finish_folding(OptContext *ctx, TCGOp *op) +static bool finish_folding(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; int i, nb_oargs; @@ -XXX,XX +XXX,XX @@ static void finish_folding(OptContext *ctx, TCGOp *op) ts_info(ts)->z_mask = ctx->z_mask; } } + return true; } /* @@ -XXX,XX +XXX,XX @@ static bool fold_add(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } /* We cannot as yet do_constant_folding with vectors. */ @@ -XXX,XX +XXX,XX @@ static bool fold_add_vec(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_addsub2(OptContext *ctx, TCGOp *op, bool add) @@ -XXX,XX +XXX,XX @@ static bool fold_addsub2(OptContext *ctx, TCGOp *op, bool add) op->args[4] = arg_new_constant(ctx, bl); op->args[5] = arg_new_constant(ctx, bh); } - return false; + return finish_folding(ctx, op); } static bool fold_add2(OptContext *ctx, TCGOp *op) -- 2.43.0 Introduce ti_is_const, ti_const_val, ti_is_const_val. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static inline TempOptInfo *arg_info(TCGArg arg) return ts_info(arg_temp(arg)); } +static inline bool ti_is_const(TempOptInfo *ti) +{ + return ti->is_const; +} + +static inline uint64_t ti_const_val(TempOptInfo *ti) +{ + return ti->val; +} + +static inline bool ti_is_const_val(TempOptInfo *ti, uint64_t val) +{ + return ti_is_const(ti) && ti_const_val(ti) == val; +} + static inline bool ts_is_const(TCGTemp *ts) { - return ts_info(ts)->is_const; + return ti_is_const(ts_info(ts)); } static inline bool ts_is_const_val(TCGTemp *ts, uint64_t val) { - TempOptInfo *ti = ts_info(ts); - return ti->is_const && ti->val == val; + return ti_is_const_val(ts_info(ts), val); } static inline bool arg_is_const(TCGArg arg) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Sink mask computation below fold_affected_mask early exit. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_add2(OptContext *ctx, TCGOp *op) static bool fold_and(OptContext *ctx, TCGOp *op) { - uint64_t z1, z2; + uint64_t z1, z2, z_mask, s_mask; + TempOptInfo *t1, *t2; if (fold_const2_commutative(ctx, op) || fold_xi_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) return true; } - z1 = arg_info(op->args[1])->z_mask; - z2 = arg_info(op->args[2])->z_mask; - ctx->z_mask = z1 & z2; - - /* - * Sign repetitions are perforce all identical, whether they are 1 or 0. - * Bitwise operations preserve the relative quantity of the repetitions. - */ - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z1 = t1->z_mask; + z2 = t2->z_mask; /* * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer affected bits from it. */ - if (arg_is_const(op->args[2]) && - fold_affected_mask(ctx, op, z1 & ~z2)) { + if (ti_is_const(t2) && fold_affected_mask(ctx, op, z1 & ~z2)) { return true; } - return fold_masks(ctx, op); + z_mask = z1 & z2; + + /* + * Sign repetitions are perforce all identical, whether they are 1 or 0. + * Bitwise operations preserve the relative quantity of the repetitions. + */ + s_mask = t1->s_mask & t2->s_mask; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_andc(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Avoid double inversion of the value of second const operand. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) static bool fold_andc(OptContext *ctx, TCGOp *op) { - uint64_t z1; + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || fold_xx_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) return true; } - z1 = arg_info(op->args[1])->z_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask; /* * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer anything from it. */ - if (arg_is_const(op->args[2])) { - uint64_t z2 = ~arg_info(op->args[2])->z_mask; - if (fold_affected_mask(ctx, op, z1 & ~z2)) { + if (ti_is_const(t2)) { + uint64_t v2 = ti_const_val(t2); + if (fold_affected_mask(ctx, op, z_mask & v2)) { return true; } - z1 &= z2; + z_mask &= ~v2; } - ctx->z_mask = z1; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_brcond(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Always set s_mask along the BSWAP_OS path, since the result is being explicitly sign-extended. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) static bool fold_bswap(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, sign; + TempOptInfo *t1 = arg_info(op->args[1]); - if (arg_is_const(op->args[1])) { - uint64_t t = arg_info(op->args[1])->val; - - t = do_constant_folding(op->opc, ctx->type, t, op->args[2]); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + do_constant_folding(op->opc, ctx->type, + ti_const_val(t1), + op->args[2])); } - z_mask = arg_info(op->args[1])->z_mask; - + z_mask = t1->z_mask; switch (op->opc) { case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) /* If the sign bit may be 1, force all the bits above to 1. */ if (z_mask & sign) { z_mask |= sign; - s_mask = sign << 1; } + /* The value and therefore s_mask is explicitly sign-extended. */ + s_mask = sign; break; default: /* The high bits are undefined: force all bits above the sign to 1. */ z_mask |= sign << 1; break; } - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_call(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Compute s_mask from the union of the maximum count and the op2 fallback for op1 being zero. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_call(OptContext *ctx, TCGOp *op) static bool fold_count_zeros(OptContext *ctx, TCGOp *op) { - uint64_t z_mask; + uint64_t z_mask, s_mask; + TempOptInfo *t1 = arg_info(op->args[1]); + TempOptInfo *t2 = arg_info(op->args[2]); - if (arg_is_const(op->args[1])) { - uint64_t t = arg_info(op->args[1])->val; + if (ti_is_const(t1)) { + uint64_t t = ti_const_val(t1); if (t != 0) { t = do_constant_folding(op->opc, ctx->type, t, 0); @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - ctx->z_mask = arg_info(op->args[2])->z_mask | z_mask; - return false; + s_mask = ~z_mask; + z_mask |= t2->z_mask; + s_mask &= t2->s_mask; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_ctpop(OptContext *ctx, TCGOp *op) -- 2.43.0 Add fold_masks_z as a trivial wrapper around fold_masks_zs. Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, return true; } +static bool fold_masks_z(OptContext *ctx, TCGOp *op, uint64_t z_mask) +{ + return fold_masks_zs(ctx, op, z_mask, 0); +} + static bool fold_masks(OptContext *ctx, TCGOp *op) { return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) static bool fold_ctpop(OptContext *ctx, TCGOp *op) { + uint64_t z_mask; + if (fold_const1(ctx, op)) { return true; } switch (ctx->type) { case TCG_TYPE_I32: - ctx->z_mask = 32 | 31; + z_mask = 32 | 31; break; case TCG_TYPE_I64: - ctx->z_mask = 64 | 63; + z_mask = 64 | 63; break; default: g_assert_not_reached(); } - return false; + return fold_masks_z(ctx, op, z_mask); } static bool fold_deposit(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. When we fold to and, use fold_and. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 35 +++++++++++++++++------------------ 1 file changed, 17 insertions(+), 18 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_ctpop(OptContext *ctx, TCGOp *op) static bool fold_deposit(OptContext *ctx, TCGOp *op) { + TempOptInfo *t1 = arg_info(op->args[1]); + TempOptInfo *t2 = arg_info(op->args[2]); + int ofs = op->args[3]; + int len = op->args[4]; TCGOpcode and_opc; + uint64_t z_mask; - if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { - uint64_t t1 = arg_info(op->args[1])->val; - uint64_t t2 = arg_info(op->args[2])->val; - - t1 = deposit64(t1, op->args[3], op->args[4], t2); - return tcg_opt_gen_movi(ctx, op, op->args[0], t1); + if (ti_is_const(t1) && ti_is_const(t2)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + deposit64(ti_const_val(t1), ofs, len, + ti_const_val(t2))); } switch (ctx->type) { @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) } /* Inserting a value into zero at offset 0. */ - if (arg_is_const_val(op->args[1], 0) && op->args[3] == 0) { - uint64_t mask = MAKE_64BIT_MASK(0, op->args[4]); + if (ti_is_const_val(t1, 0) && ofs == 0) { + uint64_t mask = MAKE_64BIT_MASK(0, len); op->opc = and_opc; op->args[1] = op->args[2]; op->args[2] = arg_new_constant(ctx, mask); - ctx->z_mask = mask & arg_info(op->args[1])->z_mask; - return false; + return fold_and(ctx, op); } /* Inserting zero into a value. */ - if (arg_is_const_val(op->args[2], 0)) { - uint64_t mask = deposit64(-1, op->args[3], op->args[4], 0); + if (ti_is_const_val(t2, 0)) { + uint64_t mask = deposit64(-1, ofs, len, 0); op->opc = and_opc; op->args[2] = arg_new_constant(ctx, mask); - ctx->z_mask = mask & arg_info(op->args[1])->z_mask; - return false; + return fold_and(ctx, op); } - ctx->z_mask = deposit64(arg_info(op->args[1])->z_mask, - op->args[3], op->args[4], - arg_info(op->args[2])->z_mask); - return false; + z_mask = deposit64(t1->z_mask, ofs, len, t2->z_mask); + return fold_masks_z(ctx, op, z_mask); } static bool fold_divide(OptContext *ctx, TCGOp *op) -- 2.43.0 The input which overlaps the sign bit of the output can have its input s_mask propagated to the output s_mask. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) TempOptInfo *t2 = arg_info(op->args[2]); int ofs = op->args[3]; int len = op->args[4]; + int width; TCGOpcode and_opc; - uint64_t z_mask; + uint64_t z_mask, s_mask; if (ti_is_const(t1) && ti_is_const(t2)) { return tcg_opt_gen_movi(ctx, op, op->args[0], @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) switch (ctx->type) { case TCG_TYPE_I32: and_opc = INDEX_op_and_i32; + width = 32; break; case TCG_TYPE_I64: and_opc = INDEX_op_and_i64; + width = 64; break; default: g_assert_not_reached(); @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) return fold_and(ctx, op); } + /* The s_mask from the top portion of the deposit is still valid. */ + if (ofs + len == width) { + s_mask = t2->s_mask << ofs; + } else { + s_mask = t1->s_mask & ~MAKE_64BIT_MASK(0, ofs + len); + } + z_mask = deposit64(t1->z_mask, ofs, len, t2->z_mask); - return fold_masks_z(ctx, op, z_mask); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_divide(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_divide(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 1)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_dup(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_dup(OptContext *ctx, TCGOp *op) t = dup_const(TCGOP_VECE(op), t); return tcg_opt_gen_movi(ctx, op, op->args[0], t); } - return false; + return finish_folding(ctx, op); } static bool fold_dup2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_dup2(OptContext *ctx, TCGOp *op) op->opc = INDEX_op_dup_vec; TCGOP_VECE(op) = MO_32; } - return false; + return finish_folding(ctx, op); } static bool fold_eqv(OptContext *ctx, TCGOp *op) -- 2.43.0 Add fold_masks_s as a trivial wrapper around fold_masks_zs. Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_z(OptContext *ctx, TCGOp *op, uint64_t z_mask) return fold_masks_zs(ctx, op, z_mask, 0); } +static bool fold_masks_s(OptContext *ctx, TCGOp *op, uint64_t s_mask) +{ + return fold_masks_zs(ctx, op, -1, s_mask); +} + static bool fold_masks(OptContext *ctx, TCGOp *op) { return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_dup2(OptContext *ctx, TCGOp *op) static bool fold_eqv(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_x(ctx, op, -1) || fold_xi_to_not(ctx, op, 0)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_extract(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_eqv(OptContext *ctx, TCGOp *op) static bool fold_extract(OptContext *ctx, TCGOp *op) { uint64_t z_mask_old, z_mask; + TempOptInfo *t1 = arg_info(op->args[1]); int pos = op->args[2]; int len = op->args[3]; - if (arg_is_const(op->args[1])) { - uint64_t t; - - t = arg_info(op->args[1])->val; - t = extract64(t, pos, len); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + extract64(ti_const_val(t1), pos, len)); } - z_mask_old = arg_info(op->args[1])->z_mask; + z_mask_old = t1->z_mask; z_mask = extract64(z_mask_old, pos, len); if (pos == 0 && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } - ctx->z_mask = z_mask; - return fold_masks(ctx, op); + return fold_masks_z(ctx, op, z_mask); } static bool fold_extract2(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extract2(OptContext *ctx, TCGOp *op) } return tcg_opt_gen_movi(ctx, op, op->args[0], v1 | v2); } - return false; + return finish_folding(ctx, op); } static bool fold_exts(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Explicitly sign-extend z_mask instead of doing that manually. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 29 ++++++++++++----------------- 1 file changed, 12 insertions(+), 17 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extract2(OptContext *ctx, TCGOp *op) static bool fold_exts(OptContext *ctx, TCGOp *op) { - uint64_t s_mask_old, s_mask, z_mask, sign; + uint64_t s_mask_old, s_mask, z_mask; bool type_change = false; + TempOptInfo *t1; if (fold_const1(ctx, op)) { return true; } - z_mask = arg_info(op->args[1])->z_mask; - s_mask = arg_info(op->args[1])->s_mask; + t1 = arg_info(op->args[1]); + z_mask = t1->z_mask; + s_mask = t1->s_mask; s_mask_old = s_mask; switch (op->opc) { CASE_OP_32_64(ext8s): - sign = INT8_MIN; - z_mask = (uint8_t)z_mask; + s_mask |= INT8_MIN; + z_mask = (int8_t)z_mask; break; CASE_OP_32_64(ext16s): - sign = INT16_MIN; - z_mask = (uint16_t)z_mask; + s_mask |= INT16_MIN; + z_mask = (int16_t)z_mask; break; case INDEX_op_ext_i32_i64: type_change = true; QEMU_FALLTHROUGH; case INDEX_op_ext32s_i64: - sign = INT32_MIN; - z_mask = (uint32_t)z_mask; + s_mask |= INT32_MIN; + z_mask = (int32_t)z_mask; break; default: g_assert_not_reached(); } - if (z_mask & sign) { - z_mask |= sign; - } - s_mask |= sign << 1; - - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_extu(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } - ctx->z_mask = z_mask; if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } - return fold_masks(ctx, op); + + return fold_masks_z(ctx, op, z_mask); } static bool fold_mb(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_mov(OptContext *ctx, TCGOp *op) static bool fold_movcond(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *tt, *ft; int i; /* If true and false values are the same, eliminate the cmp. */ @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[4 - i]); } - ctx->z_mask = arg_info(op->args[3])->z_mask - | arg_info(op->args[4])->z_mask; - ctx->s_mask = arg_info(op->args[3])->s_mask - & arg_info(op->args[4])->s_mask; + tt = arg_info(op->args[3]); + ft = arg_info(op->args[4]); + z_mask = tt->z_mask | ft->z_mask; + s_mask = tt->s_mask & ft->s_mask; - if (arg_is_const(op->args[3]) && arg_is_const(op->args[4])) { - uint64_t tv = arg_info(op->args[3])->val; - uint64_t fv = arg_info(op->args[4])->val; + if (ti_is_const(tt) && ti_is_const(ft)) { + uint64_t tv = ti_const_val(tt); + uint64_t fv = ti_const_val(ft); TCGOpcode opc, negopc = 0; TCGCond cond = op->args[5]; @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) } } } - return false; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_mul(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_mul(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 1)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_mul_highpart(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_mul_highpart(OptContext *ctx, TCGOp *op) fold_xi_to_i(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_multiply2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_multiply2(OptContext *ctx, TCGOp *op) tcg_opt_gen_movi(ctx, op2, rh, h); return true; } - return false; + return finish_folding(ctx, op); } static bool fold_nand(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_multiply2(OptContext *ctx, TCGOp *op) static bool fold_nand(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_not(ctx, op, -1)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_neg_no_const(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_neg_no_const(OptContext *ctx, TCGOp *op) { /* Set to 1 all bits to the left of the rightmost. */ uint64_t z_mask = arg_info(op->args[1])->z_mask; - ctx->z_mask = -(z_mask & -z_mask); + z_mask = -(z_mask & -z_mask); - /* - * Because of fold_sub_to_neg, we want to always return true, - * via finish_folding. - */ - finish_folding(ctx, op); - return true; + return fold_masks_z(ctx, op, z_mask); } static bool fold_neg(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_neg(OptContext *ctx, TCGOp *op) static bool fold_nor(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_not(ctx, op, 0)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_not(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_not(OptContext *ctx, TCGOp *op) if (fold_const1(ctx, op)) { return true; } - - ctx->s_mask = arg_info(op->args[1])->s_mask; - - /* Because of fold_to_not, we want to always return true, via finish. */ - finish_folding(ctx, op); - return true; + return fold_masks_s(ctx, op, arg_info(op->args[1])->s_mask); } static bool fold_or(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_not(OptContext *ctx, TCGOp *op) static bool fold_or(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; + if (fold_const2_commutative(ctx, op) || fold_xi_to_x(ctx, op, 0) || fold_xx_to_x(ctx, op)) { return true; } - ctx->z_mask = arg_info(op->args[1])->z_mask - | arg_info(op->args[2])->z_mask; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask | t2->z_mask; + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_orc(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_or(OptContext *ctx, TCGOp *op) static bool fold_orc(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2(ctx, op) || fold_xx_to_i(ctx, op, -1) || fold_xi_to_x(ctx, op, -1) || @@ -XXX,XX +XXX,XX @@ static bool fold_orc(OptContext *ctx, TCGOp *op) return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Be careful not to call fold_masks_zs when the memory operation is wide enough to require multiple outputs, so split into two functions: fold_qemu_ld_1reg and fold_qemu_ld_2reg. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 26 +++++++++++++++++++++----- 1 file changed, 21 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_orc(OptContext *ctx, TCGOp *op) return fold_masks_s(ctx, op, s_mask); } -static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) +static bool fold_qemu_ld_1reg(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; MemOpIdx oi = op->args[def->nb_oargs + def->nb_iargs]; MemOp mop = get_memop(oi); int width = 8 * memop_size(mop); + uint64_t z_mask = -1, s_mask = 0; if (width < 64) { if (mop & MO_SIGN) { - ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); + s_mask = MAKE_64BIT_MASK(width - 1, 64 - (width - 1)); } else { - ctx->z_mask = MAKE_64BIT_MASK(0, width); + z_mask = MAKE_64BIT_MASK(0, width); } } /* Opcodes that touch guest memory stop the mb optimization. */ ctx->prev_mb = NULL; - return false; + + return fold_masks_zs(ctx, op, z_mask, s_mask); +} + +static bool fold_qemu_ld_2reg(OptContext *ctx, TCGOp *op) +{ + /* Opcodes that touch guest memory stop the mb optimization. */ + ctx->prev_mb = NULL; + return finish_folding(ctx, op); } static bool fold_qemu_st(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) break; case INDEX_op_qemu_ld_a32_i32: case INDEX_op_qemu_ld_a64_i32: + done = fold_qemu_ld_1reg(&ctx, op); + break; case INDEX_op_qemu_ld_a32_i64: case INDEX_op_qemu_ld_a64_i64: + if (TCG_TARGET_REG_BITS == 64) { + done = fold_qemu_ld_1reg(&ctx, op); + break; + } + QEMU_FALLTHROUGH; case INDEX_op_qemu_ld_a32_i128: case INDEX_op_qemu_ld_a64_i128: - done = fold_qemu_ld(&ctx, op); + done = fold_qemu_ld_2reg(&ctx, op); break; case INDEX_op_qemu_st8_a32_i32: case INDEX_op_qemu_st8_a64_i32: -- 2.43.0 Stores have no output operands, and so need no further work. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_qemu_st(OptContext *ctx, TCGOp *op) { /* Opcodes that touch guest memory stop the mb optimization. */ ctx->prev_mb = NULL; - return false; + return true; } static bool fold_remainder(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st(OptContext *ctx, TCGOp *op) if (op->args[1] != tcgv_ptr_arg(tcg_env)) { remove_mem_copy_all(ctx); - return false; + return true; } switch (op->opc) { @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } remove_mem_copy_in(ctx, ofs, ofs + lm1); - return false; + return true; } static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) TCGType type; if (op->args[1] != tcgv_ptr_arg(tcg_env)) { - fold_tcg_st(ctx, op); - return false; + return fold_tcg_st(ctx, op); } src = arg_temp(op->args[0]); @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) last = ofs + tcg_type_size(type) - 1; remove_mem_copy_in(ctx, ofs, last); record_mem_copy(ctx, type, src, ofs, last); - return false; + return true; } static bool fold_xor(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) fold_xx_to_i(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) -- 2.43.0 Change return from bool to int; distinguish between complete folding, simplification, and no change. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) return finish_folding(ctx, op); } -static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) +/* Return 1 if finished, -1 if simplified, 0 if unchanged. */ +static int fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) { uint64_t a_zmask, b_val; TCGCond cond; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) op->opc = xor_opc; op->args[2] = arg_new_constant(ctx, 1); } - return false; + return -1; } } - - return false; + return 0; } static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], i); } - if (fold_setcond_zmask(ctx, op, false)) { + i = fold_setcond_zmask(ctx, op, false); + if (i > 0) { return true; } - fold_setcond_tst_pow2(ctx, op, false); + if (i == 0) { + fold_setcond_tst_pow2(ctx, op, false); + } ctx->z_mask = 1; return false; @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], -i); } - if (fold_setcond_zmask(ctx, op, true)) { + i = fold_setcond_zmask(ctx, op, true); + if (i > 0) { return true; } - fold_setcond_tst_pow2(ctx, op, true); + if (i == 0) { + fold_setcond_tst_pow2(ctx, op, true); + } /* Value is {0,-1} so all bits are repetitions of the sign. */ ctx->s_mask = -1; -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) fold_setcond_tst_pow2(ctx, op, false); } - ctx->z_mask = 1; - return false; + return fold_masks_z(ctx, op, 1); } static bool fold_negsetcond(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) } /* Value is {0,-1} so all bits are repetitions of the sign. */ - ctx->s_mask = -1; - return false; + return fold_masks_s(ctx, op, -1); } static bool fold_setcond2(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) return fold_setcond(ctx, op); } - ctx->z_mask = 1; - return false; + return fold_masks_z(ctx, op, 1); do_setcond_const: return tcg_opt_gen_movi(ctx, op, op->args[0], i); -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { op->args[3] = tcg_swap_cond(op->args[3]); } - return false; + return finish_folding(ctx, op); } static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { op->args[5] = tcg_invert_cond(op->args[5]); } - return false; + return finish_folding(ctx, op); } static bool fold_sextract(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) static bool fold_sextract(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, s_mask_old; + TempOptInfo *t1 = arg_info(op->args[1]); int pos = op->args[2]; int len = op->args[3]; - if (arg_is_const(op->args[1])) { - uint64_t t; - - t = arg_info(op->args[1])->val; - t = sextract64(t, pos, len); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + sextract64(ti_const_val(t1), pos, len)); } - z_mask = arg_info(op->args[1])->z_mask; - z_mask = sextract64(z_mask, pos, len); - ctx->z_mask = z_mask; - - s_mask_old = arg_info(op->args[1])->s_mask; - s_mask = sextract64(s_mask_old, pos, len); - s_mask |= MAKE_64BIT_MASK(len, 64 - len); - ctx->s_mask = s_mask; + s_mask_old = t1->s_mask; + s_mask = s_mask_old >> pos; + s_mask |= -1ull << (len - 1); if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } - return fold_masks(ctx, op); + z_mask = sextract64(t1->z_mask, pos, len); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_shift(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 27 ++++++++++++++------------- 1 file changed, 14 insertions(+), 13 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) static bool fold_shift(OptContext *ctx, TCGOp *op) { uint64_t s_mask, z_mask, sign; + TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || fold_ix_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) return true; } - s_mask = arg_info(op->args[1])->s_mask; - z_mask = arg_info(op->args[1])->z_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + s_mask = t1->s_mask; + z_mask = t1->z_mask; - if (arg_is_const(op->args[2])) { - int sh = arg_info(op->args[2])->val; - - ctx->z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); + if (ti_is_const(t2)) { + int sh = ti_const_val(t2); + z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); s_mask = do_constant_folding(op->opc, ctx->type, s_mask, sh); - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } switch (op->opc) { @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) * Arithmetic right shift will not reduce the number of * input sign repetitions. */ - ctx->s_mask = s_mask; - break; + return fold_masks_s(ctx, op, s_mask); CASE_OP_32_64(shr): /* * If the sign bit is known zero, then logical right shift - * will not reduced the number of input sign repetitions. + * will not reduce the number of input sign repetitions. */ - sign = (s_mask & -s_mask) >> 1; + sign = -s_mask; if (sign && !(z_mask & sign)) { - ctx->s_mask = s_mask; + return fold_masks_s(ctx, op, s_mask); } break; default: break; } - return false; + return finish_folding(ctx, op); } static bool fold_sub_to_neg(OptContext *ctx, TCGOp *op) -- 2.43.0 Merge the two conditions, sign != 0 && !(z_mask & sign), by testing ~z_mask & sign. If sign == 0, the logical and will produce false. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) static bool fold_shift(OptContext *ctx, TCGOp *op) { - uint64_t s_mask, z_mask, sign; + uint64_t s_mask, z_mask; TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) * If the sign bit is known zero, then logical right shift * will not reduce the number of input sign repetitions. */ - sign = -s_mask; - if (sign && !(z_mask & sign)) { + if (~z_mask & -s_mask) { return fold_masks_s(ctx, op, s_mask); } break; -- 2.43.0 Duplicate fold_sub_vec into fold_sub instead of calling it, now that fold_sub_vec always returns true. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sub_vec(OptContext *ctx, TCGOp *op) fold_sub_to_neg(ctx, op)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_sub(OptContext *ctx, TCGOp *op) { - if (fold_const2(ctx, op) || fold_sub_vec(ctx, op)) { + if (fold_const2(ctx, op) || + fold_xx_to_i(ctx, op, 0) || + fold_xi_to_x(ctx, op, 0) || + fold_sub_to_neg(ctx, op)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sub(OptContext *ctx, TCGOp *op) ? INDEX_op_add_i32 : INDEX_op_add_i64); op->args[2] = arg_new_constant(ctx, -val); } - return false; + return finish_folding(ctx, op); } static bool fold_sub2(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sub2(OptContext *ctx, TCGOp *op) static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) { + uint64_t z_mask = -1, s_mask = 0; + /* We can't do any folding with a load, but we can record bits. */ switch (op->opc) { CASE_OP_32_64(ld8s): - ctx->s_mask = MAKE_64BIT_MASK(8, 56); + s_mask = INT8_MIN; break; CASE_OP_32_64(ld8u): - ctx->z_mask = MAKE_64BIT_MASK(0, 8); + z_mask = MAKE_64BIT_MASK(0, 8); break; CASE_OP_32_64(ld16s): - ctx->s_mask = MAKE_64BIT_MASK(16, 48); + s_mask = INT16_MIN; break; CASE_OP_32_64(ld16u): - ctx->z_mask = MAKE_64BIT_MASK(0, 16); + z_mask = MAKE_64BIT_MASK(0, 16); break; case INDEX_op_ld32s_i64: - ctx->s_mask = MAKE_64BIT_MASK(32, 32); + s_mask = INT32_MIN; break; case INDEX_op_ld32u_i64: - ctx->z_mask = MAKE_64BIT_MASK(0, 32); + z_mask = MAKE_64BIT_MASK(0, 32); break; default: g_assert_not_reached(); } - return false; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_tcg_ld_memcopy(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_ld_memcopy(OptContext *ctx, TCGOp *op) TCGType type; if (op->args[1] != tcgv_ptr_arg(tcg_env)) { - return false; + return finish_folding(ctx, op); } type = ctx->type; -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Remove fold_masks as the function becomes unused. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_s(OptContext *ctx, TCGOp *op, uint64_t s_mask) return fold_masks_zs(ctx, op, -1, s_mask); } -static bool fold_masks(OptContext *ctx, TCGOp *op) -{ - return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); -} - /* * An "affected" mask bit is 0 if and only if the result is identical * to the first input. Thus if the entire mask is 0, the operation @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) static bool fold_xor(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; + if (fold_const2_commutative(ctx, op) || fold_xx_to_i(ctx, op, 0) || fold_xi_to_x(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_xor(OptContext *ctx, TCGOp *op) return true; } - ctx->z_mask = arg_info(op->args[1])->z_mask - | arg_info(op->args[2])->z_mask; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask | t2->z_mask; + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) return fold_orc(ctx, op); } } - return false; + return finish_folding(ctx, op); } /* Propagate constants and copies, fold constant expressions. */ -- 2.43.0 All non-default cases now finish folding within each function. Do the same with the default case and assert it is done after. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) done = true; break; default: + done = finish_folding(&ctx, op); break; } - - if (!done) { - finish_folding(&ctx, op); - } + tcg_debug_assert(done); } } -- 2.43.0 All mask setting is now done with parameters via fold_masks_*. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ------------- 1 file changed, 13 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { QSIMPLEQ_HEAD(, MemCopyInfo) mem_free; /* In flight values from optimization. */ - uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ - uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ TCGType type; } OptContext; @@ -XXX,XX +XXX,XX @@ static bool finish_folding(OptContext *ctx, TCGOp *op) for (i = 0; i < nb_oargs; i++) { TCGTemp *ts = arg_temp(op->args[i]); reset_ts(ctx, ts); - /* - * Save the corresponding known-zero/sign bits mask for the - * first output argument (only one supported so far). - */ - if (i == 0) { - ts_info(ts)->z_mask = ctx->z_mask; - } } return true; } @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) ctx.type = TCG_TYPE_I32; } - /* Assume all bits affected, no bits known zero, no sign reps. */ - ctx.z_mask = -1; - ctx.s_mask = 0; - /* * Process each opcode. * Sorted alphabetically by opcode as much as possible. -- 2.43.0 All instances of s_mask have been converted to the new representation. We can now re-enable usage. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } - if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask = s_mask_old >> pos; s_mask |= -1ull << (len - 1); - if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } -- 2.43.0 The big comment just above says functions should be sorted. Add forward declarations as needed. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 114 +++++++++++++++++++++++++------------------------ 1 file changed, 59 insertions(+), 55 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_xx_to_x(OptContext *ctx, TCGOp *op) * 3) those that produce information about the result value. */ +static bool fold_or(OptContext *ctx, TCGOp *op); +static bool fold_orc(OptContext *ctx, TCGOp *op); +static bool fold_xor(OptContext *ctx, TCGOp *op); + static bool fold_add(OptContext *ctx, TCGOp *op) { if (fold_const2_commutative(ctx, op) || @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) return fold_masks_zs(ctx, op, z_mask, s_mask); } +static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) +{ + /* If true and false values are the same, eliminate the cmp. */ + if (args_are_copies(op->args[2], op->args[3])) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[2]); + } + + if (arg_is_const(op->args[2]) && arg_is_const(op->args[3])) { + uint64_t tv = arg_info(op->args[2])->val; + uint64_t fv = arg_info(op->args[3])->val; + + if (tv == -1 && fv == 0) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); + } + if (tv == 0 && fv == -1) { + if (TCG_TARGET_HAS_not_vec) { + op->opc = INDEX_op_not_vec; + return fold_not(ctx, op); + } else { + op->opc = INDEX_op_xor_vec; + op->args[2] = arg_new_constant(ctx, -1); + return fold_xor(ctx, op); + } + } + } + if (arg_is_const(op->args[2])) { + uint64_t tv = arg_info(op->args[2])->val; + if (tv == -1) { + op->opc = INDEX_op_or_vec; + op->args[2] = op->args[3]; + return fold_or(ctx, op); + } + if (tv == 0 && TCG_TARGET_HAS_andc_vec) { + op->opc = INDEX_op_andc_vec; + op->args[2] = op->args[1]; + op->args[1] = op->args[3]; + return fold_andc(ctx, op); + } + } + if (arg_is_const(op->args[3])) { + uint64_t fv = arg_info(op->args[3])->val; + if (fv == 0) { + op->opc = INDEX_op_and_vec; + return fold_and(ctx, op); + } + if (fv == -1 && TCG_TARGET_HAS_orc_vec) { + op->opc = INDEX_op_orc_vec; + op->args[2] = op->args[1]; + op->args[1] = op->args[3]; + return fold_orc(ctx, op); + } + } + return finish_folding(ctx, op); +} + static bool fold_brcond(OptContext *ctx, TCGOp *op) { int i = do_constant_folding_cond1(ctx, op, NO_DEST, &op->args[0], @@ -XXX,XX +XXX,XX @@ static bool fold_xor(OptContext *ctx, TCGOp *op) return fold_masks_zs(ctx, op, z_mask, s_mask); } -static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) -{ - /* If true and false values are the same, eliminate the cmp. */ - if (args_are_copies(op->args[2], op->args[3])) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[2]); - } - - if (arg_is_const(op->args[2]) && arg_is_const(op->args[3])) { - uint64_t tv = arg_info(op->args[2])->val; - uint64_t fv = arg_info(op->args[3])->val; - - if (tv == -1 && fv == 0) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); - } - if (tv == 0 && fv == -1) { - if (TCG_TARGET_HAS_not_vec) { - op->opc = INDEX_op_not_vec; - return fold_not(ctx, op); - } else { - op->opc = INDEX_op_xor_vec; - op->args[2] = arg_new_constant(ctx, -1); - return fold_xor(ctx, op); - } - } - } - if (arg_is_const(op->args[2])) { - uint64_t tv = arg_info(op->args[2])->val; - if (tv == -1) { - op->opc = INDEX_op_or_vec; - op->args[2] = op->args[3]; - return fold_or(ctx, op); - } - if (tv == 0 && TCG_TARGET_HAS_andc_vec) { - op->opc = INDEX_op_andc_vec; - op->args[2] = op->args[1]; - op->args[1] = op->args[3]; - return fold_andc(ctx, op); - } - } - if (arg_is_const(op->args[3])) { - uint64_t fv = arg_info(op->args[3])->val; - if (fv == 0) { - op->opc = INDEX_op_and_vec; - return fold_and(ctx, op); - } - if (fv == -1 && TCG_TARGET_HAS_orc_vec) { - op->opc = INDEX_op_orc_vec; - op->args[2] = op->args[1]; - op->args[1] = op->args[3]; - return fold_orc(ctx, op); - } - } - return finish_folding(ctx, op); -} - /* Propagate constants and copies, fold constant expressions. */ void tcg_optimize(TCGContext *s) { -- 2.43.0 The big comment just above says functions should be sorted. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 60 +++++++++++++++++++++++++------------------------- 1 file changed, 30 insertions(+), 30 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_call(OptContext *ctx, TCGOp *op) return true; } +static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) +{ + /* Canonicalize the comparison to put immediate second. */ + if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { + op->args[3] = tcg_swap_cond(op->args[3]); + } + return finish_folding(ctx, op); +} + +static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) +{ + /* If true and false values are the same, eliminate the cmp. */ + if (args_are_copies(op->args[3], op->args[4])) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[3]); + } + + /* Canonicalize the comparison to put immediate second. */ + if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { + op->args[5] = tcg_swap_cond(op->args[5]); + } + /* + * Canonicalize the "false" input reg to match the destination, + * so that the tcg backend can implement "move if true". + */ + if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { + op->args[5] = tcg_invert_cond(op->args[5]); + } + return finish_folding(ctx, op); +} + static bool fold_count_zeros(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], i); } -static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) -{ - /* Canonicalize the comparison to put immediate second. */ - if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { - op->args[3] = tcg_swap_cond(op->args[3]); - } - return finish_folding(ctx, op); -} - -static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) -{ - /* If true and false values are the same, eliminate the cmp. */ - if (args_are_copies(op->args[3], op->args[4])) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[3]); - } - - /* Canonicalize the comparison to put immediate second. */ - if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { - op->args[5] = tcg_swap_cond(op->args[5]); - } - /* - * Canonicalize the "false" input reg to match the destination, - * so that the tcg backend can implement "move if true". - */ - if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { - op->args[5] = tcg_invert_cond(op->args[5]); - } - return finish_folding(ctx, op); -} - static bool fold_sextract(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, s_mask_old; -- 2.43.0 We currently have a flag, float_muladd_halve_result, to scale the result by 2**-1. Extend this to handle arbitrary scaling. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 6 ++++ fpu/softfloat.c | 58 ++++++++++++++++++++++----------------- fpu/softfloat-parts.c.inc | 7 +++-- 3 files changed, 44 insertions(+), 27 deletions(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ float16 float16_add(float16, float16, float_status *status); float16 float16_sub(float16, float16, float_status *status); float16 float16_mul(float16, float16, float_status *status); float16 float16_muladd(float16, float16, float16, int, float_status *status); +float16 float16_muladd_scalbn(float16, float16, float16, + int, int, float_status *status); float16 float16_div(float16, float16, float_status *status); float16 float16_scalbn(float16, int, float_status *status); float16 float16_min(float16, float16, float_status *status); @@ -XXX,XX +XXX,XX @@ float32 float32_mul(float32, float32, float_status *status); float32 float32_div(float32, float32, float_status *status); float32 float32_rem(float32, float32, float_status *status); float32 float32_muladd(float32, float32, float32, int, float_status *status); +float32 float32_muladd_scalbn(float32, float32, float32, + int, int, float_status *status); float32 float32_sqrt(float32, float_status *status); float32 float32_exp2(float32, float_status *status); float32 float32_log2(float32, float_status *status); @@ -XXX,XX +XXX,XX @@ float64 float64_mul(float64, float64, float_status *status); float64 float64_div(float64, float64, float_status *status); float64 float64_rem(float64, float64, float_status *status); float64 float64_muladd(float64, float64, float64, int, float_status *status); +float64 float64_muladd_scalbn(float64, float64, float64, + int, int, float_status *status); float64 float64_sqrt(float64, float_status *status); float64 float64_log2(float64, float_status *status); FloatRelation float64_compare(float64, float64, float_status *status); diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ static FloatParts128 *parts128_mul(FloatParts128 *a, FloatParts128 *b, #define parts_mul(A, B, S) \ PARTS_GENERIC_64_128(mul, A)(A, B, S) -static FloatParts64 *parts64_muladd(FloatParts64 *a, FloatParts64 *b, - FloatParts64 *c, int flags, - float_status *s); -static FloatParts128 *parts128_muladd(FloatParts128 *a, FloatParts128 *b, - FloatParts128 *c, int flags, - float_status *s); +static FloatParts64 *parts64_muladd_scalbn(FloatParts64 *a, FloatParts64 *b, + FloatParts64 *c, int scale, + int flags, float_status *s); +static FloatParts128 *parts128_muladd_scalbn(FloatParts128 *a, FloatParts128 *b, + FloatParts128 *c, int scale, + int flags, float_status *s); -#define parts_muladd(A, B, C, Z, S) \ - PARTS_GENERIC_64_128(muladd, A)(A, B, C, Z, S) +#define parts_muladd_scalbn(A, B, C, Z, Y, S) \ + PARTS_GENERIC_64_128(muladd_scalbn, A)(A, B, C, Z, Y, S) static FloatParts64 *parts64_div(FloatParts64 *a, FloatParts64 *b, float_status *s); @@ -XXX,XX +XXX,XX @@ floatx80_mul(floatx80 a, floatx80 b, float_status *status) * Fused multiply-add */ -float16 QEMU_FLATTEN float16_muladd(float16 a, float16 b, float16 c, - int flags, float_status *status) +float16 QEMU_FLATTEN +float16_muladd_scalbn(float16 a, float16 b, float16 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float16_unpack_canonical(&pa, a, status); float16_unpack_canonical(&pb, b, status); float16_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float16_round_pack_canonical(pr, status); } -static float32 QEMU_SOFTFLOAT_ATTR -soft_f32_muladd(float32 a, float32 b, float32 c, int flags, - float_status *status) +float16 float16_muladd(float16 a, float16 b, float16 c, + int flags, float_status *status) +{ + return float16_muladd_scalbn(a, b, c, 0, flags, status); +} + +float32 QEMU_SOFTFLOAT_ATTR +float32_muladd_scalbn(float32 a, float32 b, float32 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float32_unpack_canonical(&pa, a, status); float32_unpack_canonical(&pb, b, status); float32_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float32_round_pack_canonical(pr, status); } -static float64 QEMU_SOFTFLOAT_ATTR -soft_f64_muladd(float64 a, float64 b, float64 c, int flags, - float_status *status) +float64 QEMU_SOFTFLOAT_ATTR +float64_muladd_scalbn(float64 a, float64 b, float64 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float64_unpack_canonical(&pa, a, status); float64_unpack_canonical(&pb, b, status); float64_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float64_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) return ur.s; soft: - return soft_f32_muladd(ua.s, ub.s, uc.s, flags, s); + return float32_muladd_scalbn(ua.s, ub.s, uc.s, 0, flags, s); } float64 QEMU_FLATTEN @@ -XXX,XX +XXX,XX @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s) return ur.s; soft: - return soft_f64_muladd(ua.s, ub.s, uc.s, flags, s); + return float64_muladd_scalbn(ua.s, ub.s, uc.s, 0, flags, s); } float64 float64r32_muladd(float64 a, float64 b, float64 c, @@ -XXX,XX +XXX,XX @@ float64 float64r32_muladd(float64 a, float64 b, float64 c, float64_unpack_canonical(&pa, a, status); float64_unpack_canonical(&pb, b, status); float64_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return float64r32_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ bfloat16 QEMU_FLATTEN bfloat16_muladd(bfloat16 a, bfloat16 b, bfloat16 c, bfloat16_unpack_canonical(&pa, a, status); bfloat16_unpack_canonical(&pb, b, status); bfloat16_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return bfloat16_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float128 QEMU_FLATTEN float128_muladd(float128 a, float128 b, float128 c, float128_unpack_canonical(&pa, a, status); float128_unpack_canonical(&pb, b, status); float128_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return float128_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float32 float32_exp2(float32 a, float_status *status) float64_unpack_canonical(&rp, float64_one, status); for (i = 0 ; i < 15 ; i++) { + float64_unpack_canonical(&tp, float32_exp2_coefficients[i], status); - rp = *parts_muladd(&tp, &xnp, &rp, 0, status); + rp = *parts_muladd_scalbn(&tp, &xnp, &rp, 0, 0, status); xnp = *parts_mul(&xnp, &xp, status); } diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(mul)(FloatPartsN *a, FloatPartsN *b, * Requires A and C extracted into a double-sized structure to provide the * extra space for the widening multiply. */ -static FloatPartsN *partsN(muladd)(FloatPartsN *a, FloatPartsN *b, - FloatPartsN *c, int flags, float_status *s) +static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, + FloatPartsN *c, int scale, + int flags, float_status *s) { int ab_mask, abc_mask; FloatPartsW p_widen, c_widen; @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd)(FloatPartsN *a, FloatPartsN *b, a->exp = p_widen.exp; return_normal: + /* TODO: Replace all use of float_muladd_halve_result with scale. */ if (flags & float_muladd_halve_result) { a->exp -= 1; } + a->exp += scale; finish_sign: if (flags & float_muladd_negate_result) { a->sign ^= 1; -- 2.43.0 Use the scalbn interface instead of float_muladd_halve_result. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/arm/tcg/helper-a64.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/helper-a64.c +++ b/target/arm/tcg/helper-a64.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(rsqrtsf_f16)(uint32_t a, uint32_t b, float_status *fpst) (float16_is_infinity(b) && float16_is_zero(a))) { return float16_one_point_five; } - return float16_muladd(a, b, float16_three, float_muladd_halve_result, fpst); + return float16_muladd_scalbn(a, b, float16_three, -1, 0, fpst); } float32 HELPER(rsqrtsf_f32)(float32 a, float32 b, float_status *fpst) @@ -XXX,XX +XXX,XX @@ float32 HELPER(rsqrtsf_f32)(float32 a, float32 b, float_status *fpst) (float32_is_infinity(b) && float32_is_zero(a))) { return float32_one_point_five; } - return float32_muladd(a, b, float32_three, float_muladd_halve_result, fpst); + return float32_muladd_scalbn(a, b, float32_three, -1, 0, fpst); } float64 HELPER(rsqrtsf_f64)(float64 a, float64 b, float_status *fpst) @@ -XXX,XX +XXX,XX @@ float64 HELPER(rsqrtsf_f64)(float64 a, float64 b, float_status *fpst) (float64_is_infinity(b) && float64_is_zero(a))) { return float64_one_point_five; } - return float64_muladd(a, b, float64_three, float_muladd_halve_result, fpst); + return float64_muladd_scalbn(a, b, float64_three, -1, 0, fpst); } /* Floating-point reciprocal exponent - see FPRecpX in ARM ARM */ -- 2.43.0 Use the scalbn interface instead of float_muladd_halve_result. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/sparc/helper.h | 4 +- target/sparc/fop_helper.c | 8 ++-- target/sparc/translate.c | 80 +++++++++++++++++++++++---------------- 3 files changed, 54 insertions(+), 38 deletions(-) diff --git a/target/sparc/helper.h b/target/sparc/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/helper.h +++ b/target/sparc/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(faddd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fsubd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fmuld, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fdivd, TCG_CALL_NO_WG, f64, env, f64, f64) -DEF_HELPER_FLAGS_5(fmaddd, TCG_CALL_NO_WG, f64, env, f64, f64, f64, i32) +DEF_HELPER_FLAGS_6(fmaddd, TCG_CALL_NO_WG, f64, env, f64, f64, f64, s32, i32) DEF_HELPER_FLAGS_3(fnaddd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fnmuld, TCG_CALL_NO_WG, f64, env, f64, f64) @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(fadds, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fsubs, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fmuls, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fdivs, TCG_CALL_NO_WG, f32, env, f32, f32) -DEF_HELPER_FLAGS_5(fmadds, TCG_CALL_NO_WG, f32, env, f32, f32, f32, i32) +DEF_HELPER_FLAGS_6(fmadds, TCG_CALL_NO_WG, f32, env, f32, f32, f32, s32, i32) DEF_HELPER_FLAGS_3(fnadds, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fnmuls, TCG_CALL_NO_WG, f32, env, f32, f32) diff --git a/target/sparc/fop_helper.c b/target/sparc/fop_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/fop_helper.c +++ b/target/sparc/fop_helper.c @@ -XXX,XX +XXX,XX @@ Int128 helper_fsqrtq(CPUSPARCState *env, Int128 src) } float32 helper_fmadds(CPUSPARCState *env, float32 s1, - float32 s2, float32 s3, uint32_t op) + float32 s2, float32 s3, int32_t sc, uint32_t op) { - float32 ret = float32_muladd(s1, s2, s3, op, &env->fp_status); + float32 ret = float32_muladd_scalbn(s1, s2, s3, sc, op, &env->fp_status); check_ieee_exceptions(env, GETPC()); return ret; } float64 helper_fmaddd(CPUSPARCState *env, float64 s1, - float64 s2, float64 s3, uint32_t op) + float64 s2, float64 s3, int32_t sc, uint32_t op) { - float64 ret = float64_muladd(s1, s2, s3, op, &env->fp_status); + float64 ret = float64_muladd_scalbn(s1, s2, s3, sc, op, &env->fp_status); check_ieee_exceptions(env, GETPC()); return ret; } diff --git a/target/sparc/translate.c b/target/sparc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/translate.c +++ b/target/sparc/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_op_fabsq(TCGv_i128 dst, TCGv_i128 src) static void gen_op_fmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(0)); + TCGv_i32 z = tcg_constant_i32(0); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, z); } static void gen_op_fmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(0)); + TCGv_i32 z = tcg_constant_i32(0); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, z); } static void gen_op_fmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_c; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_c; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_c | float_muladd_negate_result; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c | + float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_c | float_muladd_negate_result; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c | + float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_result; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_result; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } /* Use muladd to compute (1 * src1) + src2 / 2 with one rounding. */ static void gen_op_fhadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(0); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fhaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(0); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } /* Use muladd to compute (1 * src1) - src2 / 2 with one rounding. */ static void gen_op_fhsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_negate_c | float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fhsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_negate_c | float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } /* Use muladd to compute -((1 * src1) + src2 / 2) with one rounding. */ static void gen_op_fnhadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_negate_result | float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fnhaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_negate_result | float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fpexception_im(DisasContext *dc, int ftt) -- 2.43.0 All uses have been convered to float*_muladd_scalbn. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 3 --- fpu/softfloat.c | 6 ------ fpu/softfloat-parts.c.inc | 4 ---- 3 files changed, 13 deletions(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ bfloat16 bfloat16_squash_input_denormal(bfloat16 a, float_status *status); | Using these differs from negating an input or output before calling | the muladd function in that this means that a NaN doesn't have its | sign bit inverted before it is propagated. -| We also support halving the result before rounding, as a special -| case to support the ARM fused-sqrt-step instruction FRSQRTS. *----------------------------------------------------------------------------*/ enum { float_muladd_negate_c = 1, float_muladd_negate_product = 2, float_muladd_negate_result = 4, - float_muladd_halve_result = 8, }; /*---------------------------------------------------------------------------- diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } - if (unlikely(flags & float_muladd_halve_result)) { - goto soft; - } float32_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f32_is_zon3(ua, ub, uc))) { @@ -XXX,XX +XXX,XX @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } - if (unlikely(flags & float_muladd_halve_result)) { - goto soft; - } float64_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f64_is_zon3(ua, ub, uc))) { diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, a->exp = p_widen.exp; return_normal: - /* TODO: Replace all use of float_muladd_halve_result with scale. */ - if (flags & float_muladd_halve_result) { - a->exp -= 1; - } a->exp += scale; finish_sign: if (flags & float_muladd_negate_result) { -- 2.43.0 This rounding mode is used by Hexagon. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat-types.h | 2 ++ fpu/softfloat-parts.c.inc | 3 +++ 2 files changed, 5 insertions(+) diff --git a/include/fpu/softfloat-types.h b/include/fpu/softfloat-types.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat-types.h +++ b/include/fpu/softfloat-types.h @@ -XXX,XX +XXX,XX @@ typedef enum __attribute__((__packed__)) { float_round_to_odd = 5, /* Not an IEEE rounding mode: round to closest odd, overflow to inf */ float_round_to_odd_inf = 6, + /* Not an IEEE rounding mode: round to nearest even, overflow to max */ + float_round_nearest_even_max = 7, } FloatRoundMode; /* diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static void partsN(uncanon_normal)(FloatPartsN *p, float_status *s, int exp, flags = 0; switch (s->float_rounding_mode) { + case float_round_nearest_even_max: + overflow_norm = true; + /* fall through */ case float_round_nearest_even: if (N > 64 && frac_lsb == 0) { inc = ((p->frac_hi & 1) || (p->frac_lo & round_mask) != frac_lsbm1 -- 2.43.0 Certain Hexagon instructions suppress changes to the result when the product of fma() is a true zero. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 5 +++++ fpu/softfloat.c | 3 +++ fpu/softfloat-parts.c.inc | 4 +++- 3 files changed, 11 insertions(+), 1 deletion(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ bfloat16 bfloat16_squash_input_denormal(bfloat16 a, float_status *status); | Using these differs from negating an input or output before calling | the muladd function in that this means that a NaN doesn't have its | sign bit inverted before it is propagated. +| +| With float_muladd_suppress_add_product_zero, if A or B is zero +| such that the product is a true zero, then return C without addition. +| This preserves the sign of C when C is +/- 0. Used for Hexagon. *----------------------------------------------------------------------------*/ enum { float_muladd_negate_c = 1, float_muladd_negate_product = 2, float_muladd_negate_result = 4, + float_muladd_suppress_add_product_zero = 8, }; /*---------------------------------------------------------------------------- diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } + if (unlikely(flags & float_muladd_suppress_add_product_zero)) { + goto soft; + } float32_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f32_is_zon3(ua, ub, uc))) { diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, goto return_normal; } if (c->cls == float_class_zero) { - if (a->sign != c->sign) { + if (flags & float_muladd_suppress_add_product_zero) { + a->sign = c->sign; + } else if (a->sign != c->sign) { goto return_sub_zero; } goto return_zero; -- 2.43.0 There are no special cases for this instruction. Remove internal_mpyf as unused. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.h | 1 - target/hexagon/fma_emu.c | 8 -------- target/hexagon/op_helper.c | 2 +- 3 files changed, 1 insertion(+), 10 deletions(-) diff --git a/target/hexagon/fma_emu.h b/target/hexagon/fma_emu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.h +++ b/target/hexagon/fma_emu.h @@ -XXX,XX +XXX,XX @@ int32_t float32_getexp(float32 f32); float32 infinite_float32(uint8_t sign); float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, float_status *fp_status); -float32 internal_mpyf(float32 a, float32 b, float_status *fp_status); float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status); diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, return accum_round_float32(result, fp_status); } -float32 internal_mpyf(float32 a, float32 b, float_status *fp_status) -{ - if (float32_is_zero(a) || float32_is_zero(b)) { - return float32_mul(a, b, fp_status); - } - return internal_fmafx(a, b, float32_zero, 0, fp_status); -} - float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sfmpy)(CPUHexagonState *env, float32 RsV, float32 RtV) { float32 RdV; arch_fpop_start(env); - RdV = internal_mpyf(RsV, RtV, &env->fp_status); + RdV = float32_mul(RsV, RtV, &env->fp_status); arch_fpop_end(env); return RdV; } -- 2.43.0 There are no special cases for this instruction. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { arch_fpop_start(env); - RxV = internal_fmafx(RsV, RtV, RxV, 0, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, 0, &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 There are no special cases for this instruction. Since hexagon always uses default-nan mode, explicitly negating the first input is unnecessary. Use float_muladd_negate_product instead. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 HELPER(sffms)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - float32 neg_RsV; arch_fpop_start(env); - neg_RsV = float32_set_sign(RsV, float32_is_neg(RsV) ? 0 : 1); - RxV = internal_fmafx(neg_RsV, RtV, RxV, 0, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, float_muladd_negate_product, + &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 This instruction has a special case that 0 * x + c returns c without the normal sign folding that comes with 0 + -0. Use the new float_muladd_suppress_add_product_zero to describe this. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ static float32 check_nan(float32 dst, float32 x, float_status *fp_status) float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV, float32 PuV) { - size4s_t tmp; arch_fpop_start(env); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - tmp = internal_fmafx(RsV, RtV, RxV, fSXTN(8, 64, PuV), &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } + RxV = float32_muladd_scalbn(RsV, RtV, RxV, fSXTN(8, 64, PuV), + float_muladd_suppress_add_product_zero, + &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 There are multiple special cases for this instruction. (1) The saturate to normal maximum instead of overflow to infinity is handled by the new float_round_nearest_even_max rounding mode. (2) The 0 * n + c special case is handled by the new float_muladd_suppress_add_product_zero flag. (3) The Inf - Inf -> 0 special case can be detected after the fact by examining float_flag_invalid_isi. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 105 +++++++++---------------------------- 1 file changed, 26 insertions(+), 79 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma)(CPUHexagonState *env, float32 RxV, return RxV; } -static bool is_zero_prod(float32 a, float32 b) -{ - return ((float32_is_zero(a) && is_finite(b)) || - (float32_is_zero(b) && is_finite(a))); -} - -static float32 check_nan(float32 dst, float32 x, float_status *fp_status) -{ - float32 ret = dst; - if (float32_is_any_nan(x)) { - if (extract32(x, 22, 1) == 0) { - float_raise(float_flag_invalid, fp_status); - } - ret = make_float32(0xffffffff); /* nan */ - } - return ret; -} - float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV, float32 PuV) { @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffms)(CPUHexagonState *env, float32 RxV, return RxV; } -static bool is_inf_prod(int32_t a, int32_t b) +static float32 do_sffma_lib(CPUHexagonState *env, float32 RxV, + float32 RsV, float32 RtV, int negate) { - return (float32_is_infinity(a) && float32_is_infinity(b)) || - (float32_is_infinity(a) && is_finite(b) && !float32_is_zero(b)) || - (float32_is_infinity(b) && is_finite(a) && !float32_is_zero(a)); + int flags; + + arch_fpop_start(env); + + set_float_rounding_mode(float_round_nearest_even_max, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, + negate | float_muladd_suppress_add_product_zero, + &env->fp_status); + + flags = get_float_exception_flags(&env->fp_status); + if (flags) { + /* Flags are suppressed by this instruction. */ + set_float_exception_flags(0, &env->fp_status); + + /* Return 0 for Inf - Inf. */ + if (flags & float_flag_invalid_isi) { + RxV = 0; + } + } + + arch_fpop_end(env); + return RxV; } float32 HELPER(sffma_lib)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - bool infinp; - bool infminusinf; - float32 tmp; - - arch_fpop_start(env); - set_float_rounding_mode(float_round_nearest_even, &env->fp_status); - infminusinf = float32_is_infinity(RxV) && - is_inf_prod(RsV, RtV) && - (fGETBIT(31, RsV ^ RxV ^ RtV) != 0); - infinp = float32_is_infinity(RxV) || - float32_is_infinity(RtV) || - float32_is_infinity(RsV); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - tmp = internal_fmafx(RsV, RtV, RxV, 0, &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } - set_float_exception_flags(0, &env->fp_status); - if (float32_is_infinity(RxV) && !infinp) { - RxV = RxV - 1; - } - if (infminusinf) { - RxV = 0; - } - arch_fpop_end(env); - return RxV; + return do_sffma_lib(env, RxV, RsV, RtV, 0); } float32 HELPER(sffms_lib)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - bool infinp; - bool infminusinf; - float32 tmp; - - arch_fpop_start(env); - set_float_rounding_mode(float_round_nearest_even, &env->fp_status); - infminusinf = float32_is_infinity(RxV) && - is_inf_prod(RsV, RtV) && - (fGETBIT(31, RsV ^ RxV ^ RtV) == 0); - infinp = float32_is_infinity(RxV) || - float32_is_infinity(RtV) || - float32_is_infinity(RsV); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - float32 minus_RsV = float32_sub(float32_zero, RsV, &env->fp_status); - tmp = internal_fmafx(minus_RsV, RtV, RxV, 0, &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } - set_float_exception_flags(0, &env->fp_status); - if (float32_is_infinity(RxV) && !infinp) { - RxV = RxV - 1; - } - if (infminusinf) { - RxV = 0; - } - arch_fpop_end(env); - return RxV; + return do_sffma_lib(env, RxV, RsV, RtV, float_muladd_negate_product); } float64 HELPER(dfmpyfix)(CPUHexagonState *env, float64 RssV, float64 RttV) -- 2.43.0 The function is now unused. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.h | 2 - target/hexagon/fma_emu.c | 171 --------------------------------------- 2 files changed, 173 deletions(-) diff --git a/target/hexagon/fma_emu.h b/target/hexagon/fma_emu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.h +++ b/target/hexagon/fma_emu.h @@ -XXX,XX +XXX,XX @@ static inline uint32_t float32_getexp_raw(float32 f32) } int32_t float32_getexp(float32 f32); float32 infinite_float32(uint8_t sign); -float32 internal_fmafx(float32 a, float32 b, float32 c, - int scale, float_status *fp_status); float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status); diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ int32_t float64_getexp(float64 f64) return -1; } -static uint64_t float32_getmant(float32 f32) -{ - Float a = { .i = f32 }; - if (float32_is_normal(f32)) { - return a.mant | 1ULL << 23; - } - if (float32_is_zero(f32)) { - return 0; - } - if (float32_is_denormal(f32)) { - return a.mant; - } - return ~0ULL; -} - int32_t float32_getexp(float32 f32) { Float a = { .i = f32 }; @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) } /* Return a maximum finite value with the requested sign */ -static float32 maxfinite_float32(uint8_t sign) -{ - if (sign) { - return make_float32(SF_MINUS_MAXF); - } else { - return make_float32(SF_MAXF); - } -} - -/* Return a zero value with requested sign */ -static float32 zero_float32(uint8_t sign) -{ - if (sign) { - return make_float32(0x80000000); - } else { - return float32_zero; - } -} - #define GEN_XF_ROUND(SUFFIX, MANTBITS, INF_EXP, INTERNAL_TYPE) \ static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ { \ @@ -XXX,XX +XXX,XX @@ static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ } GEN_XF_ROUND(float64, DF_MANTBITS, DF_INF_EXP, Double) -GEN_XF_ROUND(float32, SF_MANTBITS, SF_INF_EXP, Float) - -static bool is_inf_prod(float64 a, float64 b) -{ - return ((float64_is_infinity(a) && float64_is_infinity(b)) || - (float64_is_infinity(a) && is_finite(b) && (!float64_is_zero(b))) || - (float64_is_infinity(b) && is_finite(a) && (!float64_is_zero(a)))); -} - -static float64 special_fma(float64 a, float64 b, float64 c, - float_status *fp_status) -{ - float64 ret = make_float64(0); - - /* - * If A multiplied by B is an exact infinity and C is also an infinity - * but with the opposite sign, FMA returns NaN and raises invalid. - */ - uint8_t a_sign = float64_is_neg(a); - uint8_t b_sign = float64_is_neg(b); - uint8_t c_sign = float64_is_neg(c); - if (is_inf_prod(a, b) && float64_is_infinity(c)) { - if ((a_sign ^ b_sign) != c_sign) { - ret = make_float64(DF_NAN); - float_raise(float_flag_invalid, fp_status); - return ret; - } - } - if ((float64_is_infinity(a) && float64_is_zero(b)) || - (float64_is_zero(a) && float64_is_infinity(b))) { - ret = make_float64(DF_NAN); - float_raise(float_flag_invalid, fp_status); - return ret; - } - /* - * If none of the above checks are true and C is a NaN, - * a NaN shall be returned - * If A or B are NaN, a NAN shall be returned. - */ - if (float64_is_any_nan(a) || - float64_is_any_nan(b) || - float64_is_any_nan(c)) { - if (float64_is_any_nan(a) && (fGETBIT(51, a) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - if (float64_is_any_nan(b) && (fGETBIT(51, b) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - if (float64_is_any_nan(c) && (fGETBIT(51, c) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - ret = make_float64(DF_NAN); - return ret; - } - /* - * We have checked for adding opposite-signed infinities. - * Other infinities return infinity with the correct sign - */ - if (float64_is_infinity(c)) { - ret = infinite_float64(c_sign); - return ret; - } - if (float64_is_infinity(a) || float64_is_infinity(b)) { - ret = infinite_float64(a_sign ^ b_sign); - return ret; - } - g_assert_not_reached(); -} - -static float32 special_fmaf(float32 a, float32 b, float32 c, - float_status *fp_status) -{ - float64 aa, bb, cc; - aa = float32_to_float64(a, fp_status); - bb = float32_to_float64(b, fp_status); - cc = float32_to_float64(c, fp_status); - return float64_to_float32(special_fma(aa, bb, cc, fp_status), fp_status); -} - -float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, - float_status *fp_status) -{ - Accum prod; - Accum acc; - Accum result; - accum_init(&prod); - accum_init(&acc); - accum_init(&result); - - uint8_t a_sign = float32_is_neg(a); - uint8_t b_sign = float32_is_neg(b); - uint8_t c_sign = float32_is_neg(c); - if (float32_is_infinity(a) || - float32_is_infinity(b) || - float32_is_infinity(c)) { - return special_fmaf(a, b, c, fp_status); - } - if (float32_is_any_nan(a) || - float32_is_any_nan(b) || - float32_is_any_nan(c)) { - return special_fmaf(a, b, c, fp_status); - } - if ((scale == 0) && (float32_is_zero(a) || float32_is_zero(b))) { - float32 tmp = float32_mul(a, b, fp_status); - tmp = float32_add(tmp, c, fp_status); - return tmp; - } - - /* (a * 2**b) * (c * 2**d) == a*c * 2**(b+d) */ - prod.mant = int128_mul_6464(float32_getmant(a), float32_getmant(b)); - - /* - * Note: extracting the mantissa into an int is multiplying by - * 2**23, so adjust here - */ - prod.exp = float32_getexp(a) + float32_getexp(b) - SF_BIAS - 23; - prod.sign = a_sign ^ b_sign; - if (float32_is_zero(a) || float32_is_zero(b)) { - prod.exp = -2 * WAY_BIG_EXP; - } - if ((scale > 0) && float32_is_denormal(c)) { - acc.mant = int128_mul_6464(0, 0); - acc.exp = -WAY_BIG_EXP; - acc.sign = c_sign; - acc.sticky = 1; - result = accum_add(prod, acc); - } else if (!float32_is_zero(c)) { - acc.mant = int128_mul_6464(float32_getmant(c), 1); - acc.exp = float32_getexp(c); - acc.sign = c_sign; - result = accum_add(prod, acc); - } else { - result = prod; - } - result.exp += scale; - return accum_round_float32(result, fp_status); -} float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, -- 2.43.0 This massive macro is now only used once. Expand it for use only by float64. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 255 +++++++++++++++++++-------------------- 1 file changed, 127 insertions(+), 128 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) } /* Return a maximum finite value with the requested sign */ -#define GEN_XF_ROUND(SUFFIX, MANTBITS, INF_EXP, INTERNAL_TYPE) \ -static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ -{ \ - if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) \ - && ((a.guard | a.round | a.sticky) == 0)) { \ - /* result zero */ \ - switch (fp_status->float_rounding_mode) { \ - case float_round_down: \ - return zero_##SUFFIX(1); \ - default: \ - return zero_##SUFFIX(0); \ - } \ - } \ - /* Normalize right */ \ - /* We want MANTBITS bits of mantissa plus the leading one. */ \ - /* That means that we want MANTBITS+1 bits, or 0x000000000000FF_FFFF */ \ - /* So we need to normalize right while the high word is non-zero and \ - * while the low word is nonzero when masked with 0xffe0_0000_0000_0000 */ \ - while ((int128_gethi(a.mant) != 0) || \ - ((int128_getlo(a.mant) >> (MANTBITS + 1)) != 0)) { \ - a = accum_norm_right(a, 1); \ - } \ - /* \ - * OK, now normalize left \ - * We want to normalize left until we have a leading one in bit 24 \ - * Theoretically, we only need to shift a maximum of one to the left if we \ - * shifted out lots of bits from B, or if we had no shift / 1 shift sticky \ - * should be 0 \ - */ \ - while ((int128_getlo(a.mant) & (1ULL << MANTBITS)) == 0) { \ - a = accum_norm_left(a); \ - } \ - /* \ - * OK, now we might need to denormalize because of potential underflow. \ - * We need to do this before rounding, and rounding might make us normal \ - * again \ - */ \ - while (a.exp <= 0) { \ - a = accum_norm_right(a, 1 - a.exp); \ - /* \ - * Do we have underflow? \ - * That's when we get an inexact answer because we ran out of bits \ - * in a denormal. \ - */ \ - if (a.guard || a.round || a.sticky) { \ - float_raise(float_flag_underflow, fp_status); \ - } \ - } \ - /* OK, we're relatively canonical... now we need to round */ \ - if (a.guard || a.round || a.sticky) { \ - float_raise(float_flag_inexact, fp_status); \ - switch (fp_status->float_rounding_mode) { \ - case float_round_to_zero: \ - /* Chop and we're done */ \ - break; \ - case float_round_up: \ - if (a.sign == 0) { \ - a.mant = int128_add(a.mant, int128_one()); \ - } \ - break; \ - case float_round_down: \ - if (a.sign != 0) { \ - a.mant = int128_add(a.mant, int128_one()); \ - } \ - break; \ - default: \ - if (a.round || a.sticky) { \ - /* round up if guard is 1, down if guard is zero */ \ - a.mant = int128_add(a.mant, int128_make64(a.guard)); \ - } else if (a.guard) { \ - /* exactly .5, round up if odd */ \ - a.mant = int128_add(a.mant, int128_and(a.mant, int128_one())); \ - } \ - break; \ - } \ - } \ - /* \ - * OK, now we might have carried all the way up. \ - * So we might need to shr once \ - * at least we know that the lsb should be zero if we rounded and \ - * got a carry out... \ - */ \ - if ((int128_getlo(a.mant) >> (MANTBITS + 1)) != 0) { \ - a = accum_norm_right(a, 1); \ - } \ - /* Overflow? */ \ - if (a.exp >= INF_EXP) { \ - /* Yep, inf result */ \ - float_raise(float_flag_overflow, fp_status); \ - float_raise(float_flag_inexact, fp_status); \ - switch (fp_status->float_rounding_mode) { \ - case float_round_to_zero: \ - return maxfinite_##SUFFIX(a.sign); \ - case float_round_up: \ - if (a.sign == 0) { \ - return infinite_##SUFFIX(a.sign); \ - } else { \ - return maxfinite_##SUFFIX(a.sign); \ - } \ - case float_round_down: \ - if (a.sign != 0) { \ - return infinite_##SUFFIX(a.sign); \ - } else { \ - return maxfinite_##SUFFIX(a.sign); \ - } \ - default: \ - return infinite_##SUFFIX(a.sign); \ - } \ - } \ - /* Underflow? */ \ - if (int128_getlo(a.mant) & (1ULL << MANTBITS)) { \ - /* Leading one means: No, we're normal. So, we should be done... */ \ - INTERNAL_TYPE ret; \ - ret.i = 0; \ - ret.sign = a.sign; \ - ret.exp = a.exp; \ - ret.mant = int128_getlo(a.mant); \ - return ret.i; \ - } \ - assert(a.exp == 1); \ - INTERNAL_TYPE ret; \ - ret.i = 0; \ - ret.sign = a.sign; \ - ret.exp = 0; \ - ret.mant = int128_getlo(a.mant); \ - return ret.i; \ +static float64 accum_round_float64(Accum a, float_status *fp_status) +{ + if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) + && ((a.guard | a.round | a.sticky) == 0)) { + /* result zero */ + switch (fp_status->float_rounding_mode) { + case float_round_down: + return zero_float64(1); + default: + return zero_float64(0); + } + } + /* + * Normalize right + * We want DF_MANTBITS bits of mantissa plus the leading one. + * That means that we want DF_MANTBITS+1 bits, or 0x000000000000FF_FFFF + * So we need to normalize right while the high word is non-zero and + * while the low word is nonzero when masked with 0xffe0_0000_0000_0000 + */ + while ((int128_gethi(a.mant) != 0) || + ((int128_getlo(a.mant) >> (DF_MANTBITS + 1)) != 0)) { + a = accum_norm_right(a, 1); + } + /* + * OK, now normalize left + * We want to normalize left until we have a leading one in bit 24 + * Theoretically, we only need to shift a maximum of one to the left if we + * shifted out lots of bits from B, or if we had no shift / 1 shift sticky + * should be 0 + */ + while ((int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) == 0) { + a = accum_norm_left(a); + } + /* + * OK, now we might need to denormalize because of potential underflow. + * We need to do this before rounding, and rounding might make us normal + * again + */ + while (a.exp <= 0) { + a = accum_norm_right(a, 1 - a.exp); + /* + * Do we have underflow? + * That's when we get an inexact answer because we ran out of bits + * in a denormal. + */ + if (a.guard || a.round || a.sticky) { + float_raise(float_flag_underflow, fp_status); + } + } + /* OK, we're relatively canonical... now we need to round */ + if (a.guard || a.round || a.sticky) { + float_raise(float_flag_inexact, fp_status); + switch (fp_status->float_rounding_mode) { + case float_round_to_zero: + /* Chop and we're done */ + break; + case float_round_up: + if (a.sign == 0) { + a.mant = int128_add(a.mant, int128_one()); + } + break; + case float_round_down: + if (a.sign != 0) { + a.mant = int128_add(a.mant, int128_one()); + } + break; + default: + if (a.round || a.sticky) { + /* round up if guard is 1, down if guard is zero */ + a.mant = int128_add(a.mant, int128_make64(a.guard)); + } else if (a.guard) { + /* exactly .5, round up if odd */ + a.mant = int128_add(a.mant, int128_and(a.mant, int128_one())); + } + break; + } + } + /* + * OK, now we might have carried all the way up. + * So we might need to shr once + * at least we know that the lsb should be zero if we rounded and + * got a carry out... + */ + if ((int128_getlo(a.mant) >> (DF_MANTBITS + 1)) != 0) { + a = accum_norm_right(a, 1); + } + /* Overflow? */ + if (a.exp >= DF_INF_EXP) { + /* Yep, inf result */ + float_raise(float_flag_overflow, fp_status); + float_raise(float_flag_inexact, fp_status); + switch (fp_status->float_rounding_mode) { + case float_round_to_zero: + return maxfinite_float64(a.sign); + case float_round_up: + if (a.sign == 0) { + return infinite_float64(a.sign); + } else { + return maxfinite_float64(a.sign); + } + case float_round_down: + if (a.sign != 0) { + return infinite_float64(a.sign); + } else { + return maxfinite_float64(a.sign); + } + default: + return infinite_float64(a.sign); + } + } + /* Underflow? */ + if (int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) { + /* Leading one means: No, we're normal. So, we should be done... */ + Double ret; + ret.i = 0; + ret.sign = a.sign; + ret.exp = a.exp; + ret.mant = int128_getlo(a.mant); + return ret.i; + } + assert(a.exp == 1); + Double ret; + ret.i = 0; + ret.sign = a.sign; + ret.exp = 0; + ret.mant = int128_getlo(a.mant); + return ret.i; } -GEN_XF_ROUND(float64, DF_MANTBITS, DF_INF_EXP, Double) - float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status) -- 2.43.0 This structure, with bitfields, is incorrect for big-endian. Use the existing float32_getexp_raw which uses extract32. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 16 +++------------- 1 file changed, 3 insertions(+), 13 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ typedef union { }; } Double; -typedef union { - float f; - uint32_t i; - struct { - uint32_t mant:23; - uint32_t exp:8; - uint32_t sign:1; - }; -} Float; - static uint64_t float64_getmant(float64 f64) { Double a = { .i = f64 }; @@ -XXX,XX +XXX,XX @@ int32_t float64_getexp(float64 f64) int32_t float32_getexp(float32 f32) { - Float a = { .i = f32 }; + int exp = float32_getexp_raw(f32); if (float32_is_normal(f32)) { - return a.exp; + return exp; } if (float32_is_denormal(f32)) { - return a.exp + 1; + return exp + 1; } return -1; } -- 2.43.0 This structure, with bitfields, is incorrect for big-endian. Use extract64 and deposit64 instead. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 46 ++++++++++++++-------------------------- 1 file changed, 16 insertions(+), 30 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ #define WAY_BIG_EXP 4096 -typedef union { - double f; - uint64_t i; - struct { - uint64_t mant:52; - uint64_t exp:11; - uint64_t sign:1; - }; -} Double; - static uint64_t float64_getmant(float64 f64) { - Double a = { .i = f64 }; + uint64_t mant = extract64(f64, 0, 52); if (float64_is_normal(f64)) { - return a.mant | 1ULL << 52; + return mant | 1ULL << 52; } if (float64_is_zero(f64)) { return 0; } if (float64_is_denormal(f64)) { - return a.mant; + return mant; } return ~0ULL; } int32_t float64_getexp(float64 f64) { - Double a = { .i = f64 }; + int exp = extract64(f64, 52, 11); if (float64_is_normal(f64)) { - return a.exp; + return exp; } if (float64_is_denormal(f64)) { - return a.exp + 1; + return exp + 1; } return -1; } @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) /* Return a maximum finite value with the requested sign */ static float64 accum_round_float64(Accum a, float_status *fp_status) { + uint64_t ret; + if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) && ((a.guard | a.round | a.sticky) == 0)) { /* result zero */ @@ -XXX,XX +XXX,XX @@ static float64 accum_round_float64(Accum a, float_status *fp_status) } } /* Underflow? */ - if (int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) { + ret = int128_getlo(a.mant); + if (ret & (1ULL << DF_MANTBITS)) { /* Leading one means: No, we're normal. So, we should be done... */ - Double ret; - ret.i = 0; - ret.sign = a.sign; - ret.exp = a.exp; - ret.mant = int128_getlo(a.mant); - return ret.i; + ret = deposit64(ret, 52, 11, a.exp); + } else { + assert(a.exp == 1); + ret = deposit64(ret, 52, 11, 0); } - assert(a.exp == 1); - Double ret; - ret.i = 0; - ret.sign = a.sign; - ret.exp = 0; - ret.mant = int128_getlo(a.mant); - return ret.i; + ret = deposit64(ret, 63, 1, a.sign); + return ret; } float64 internal_mpyhh(float64 a, float64 b, -- 2.43.0 No need to open-code 64x64->128-bit multiplication. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 32 +++----------------------------- 1 file changed, 3 insertions(+), 29 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ int32_t float32_getexp(float32 f32) return -1; } -static uint32_t int128_getw0(Int128 x) -{ - return int128_getlo(x); -} - -static uint32_t int128_getw1(Int128 x) -{ - return int128_getlo(x) >> 32; -} - static Int128 int128_mul_6464(uint64_t ai, uint64_t bi) { - Int128 a, b; - uint64_t pp0, pp1a, pp1b, pp1s, pp2; + uint64_t l, h; - a = int128_make64(ai); - b = int128_make64(bi); - pp0 = (uint64_t)int128_getw0(a) * (uint64_t)int128_getw0(b); - pp1a = (uint64_t)int128_getw1(a) * (uint64_t)int128_getw0(b); - pp1b = (uint64_t)int128_getw1(b) * (uint64_t)int128_getw0(a); - pp2 = (uint64_t)int128_getw1(a) * (uint64_t)int128_getw1(b); - - pp1s = pp1a + pp1b; - if ((pp1s < pp1a) || (pp1s < pp1b)) { - pp2 += (1ULL << 32); - } - uint64_t ret_low = pp0 + (pp1s << 32); - if ((ret_low < pp0) || (ret_low < (pp1s << 32))) { - pp2 += 1; - } - - return int128_make128(ret_low, pp2 + (pp1s >> 32)); + mulu64(&l, &h, ai, bi); + return int128_make128(l, h); } static Int128 int128_sub_borrow(Int128 a, Int128 b, int borrow) -- 2.43.0 Initialize x with accumulated via direct assignment, rather than multiplying by 1. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float64 internal_mpyhh(float64 a, float64 b, float64_is_infinity(b)) { return float64_mul(a, b, fp_status); } - x.mant = int128_mul_6464(accumulated, 1); + x.mant = int128_make64(accumulated); x.sticky = sticky; prod = fGETUWORD(1, float64_getmant(a)) * fGETUWORD(1, float64_getmant(b)); x.mant = int128_add(x.mant, int128_mul_6464(prod, 0x100000000ULL)); -- 2.43.0 Convert all targets simultaneously, as the gen_intermediate_code function disappears from the target. While there are possible workarounds, they're larger than simply performing the conversion. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/exec/translator.h | 14 -------------- include/hw/core/tcg-cpu-ops.h | 13 +++++++++++++ target/alpha/cpu.h | 2 ++ target/arm/internals.h | 2 ++ target/avr/cpu.h | 2 ++ target/hexagon/cpu.h | 2 ++ target/hppa/cpu.h | 2 ++ target/i386/tcg/helper-tcg.h | 2 ++ target/loongarch/internals.h | 2 ++ target/m68k/cpu.h | 2 ++ target/microblaze/cpu.h | 2 ++ target/mips/tcg/tcg-internal.h | 2 ++ target/openrisc/cpu.h | 2 ++ target/ppc/cpu.h | 2 ++ target/riscv/cpu.h | 3 +++ target/rx/cpu.h | 2 ++ target/s390x/s390x-internal.h | 2 ++ target/sh4/cpu.h | 2 ++ target/sparc/cpu.h | 2 ++ target/tricore/cpu.h | 2 ++ target/xtensa/cpu.h | 2 ++ accel/tcg/cpu-exec.c | 8 +++++--- accel/tcg/translate-all.c | 8 +++++--- target/alpha/cpu.c | 1 + target/alpha/translate.c | 4 ++-- target/arm/cpu.c | 1 + target/arm/tcg/cpu-v7m.c | 1 + target/arm/tcg/translate.c | 5 ++--- target/avr/cpu.c | 1 + target/avr/translate.c | 6 +++--- target/hexagon/cpu.c | 1 + target/hexagon/translate.c | 4 ++-- target/hppa/cpu.c | 1 + target/hppa/translate.c | 4 ++-- target/i386/tcg/tcg-cpu.c | 1 + target/i386/tcg/translate.c | 5 ++--- target/loongarch/cpu.c | 1 + target/loongarch/tcg/translate.c | 4 ++-- target/m68k/cpu.c | 1 + target/m68k/translate.c | 4 ++-- target/microblaze/cpu.c | 1 + target/microblaze/translate.c | 4 ++-- target/mips/cpu.c | 1 + target/mips/tcg/translate.c | 4 ++-- target/openrisc/cpu.c | 1 + target/openrisc/translate.c | 4 ++-- target/ppc/cpu_init.c | 1 + target/ppc/translate.c | 4 ++-- target/riscv/tcg/tcg-cpu.c | 1 + target/riscv/translate.c | 4 ++-- target/rx/cpu.c | 1 + target/rx/translate.c | 4 ++-- target/s390x/cpu.c | 1 + target/s390x/tcg/translate.c | 4 ++-- target/sh4/cpu.c | 1 + target/sh4/translate.c | 4 ++-- target/sparc/cpu.c | 1 + target/sparc/translate.c | 4 ++-- target/tricore/cpu.c | 1 + target/tricore/translate.c | 5 ++--- target/xtensa/cpu.c | 1 + target/xtensa/translate.c | 4 ++-- 62 files changed, 121 insertions(+), 62 deletions(-) diff --git a/include/exec/translator.h b/include/exec/translator.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/translator.h +++ b/include/exec/translator.h @@ -XXX,XX +XXX,XX @@ #include "qemu/bswap.h" #include "exec/vaddr.h" -/** - * gen_intermediate_code - * @cpu: cpu context - * @tb: translation block - * @max_insns: max number of instructions to translate - * @pc: guest virtual program counter address - * @host_pc: host physical program counter address - * - * This function must be provided by the target, which should create - * the target-specific DisasContext, and then invoke translator_loop. - */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc); - /** * DisasJumpType: * @DISAS_NEXT: Next instruction in program order. diff --git a/include/hw/core/tcg-cpu-ops.h b/include/hw/core/tcg-cpu-ops.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/core/tcg-cpu-ops.h +++ b/include/hw/core/tcg-cpu-ops.h @@ -XXX,XX +XXX,XX @@ struct TCGCPUOps { * Called when the first CPU is realized. */ void (*initialize)(void); + /** + * @translate_code: Translate guest instructions to TCGOps + * @cpu: cpu context + * @tb: translation block + * @max_insns: max number of instructions to translate + * @pc: guest virtual program counter address + * @host_pc: host physical program counter address + * + * This function must be provided by the target, which should create + * the target-specific DisasContext, and then invoke translator_loop. + */ + void (*translate_code)(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /** * @synchronize_from_tb: Synchronize state from a TCG #TranslationBlock * diff --git a/target/alpha/cpu.h b/target/alpha/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/cpu.h +++ b/target/alpha/cpu.h @@ -XXX,XX +XXX,XX @@ enum { }; void alpha_translate_init(void); +void alpha_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_ALPHA_CPU diff --git a/target/arm/internals.h b/target/arm/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/internals.h +++ b/target/arm/internals.h @@ -XXX,XX +XXX,XX @@ void init_cpreg_list(ARMCPU *cpu); void arm_cpu_register_gdb_regs_for_features(ARMCPU *cpu); void arm_translate_init(void); +void arm_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void arm_cpu_register_gdb_commands(ARMCPU *cpu); void aarch64_cpu_register_gdb_commands(ARMCPU *cpu, GString *, diff --git a/target/avr/cpu.h b/target/avr/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/avr/cpu.h +++ b/target/avr/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void set_avr_feature(CPUAVRState *env, int feature) } void avr_cpu_tcg_init(void); +void avr_cpu_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); int cpu_avr_exec(CPUState *cpu); diff --git a/target/hexagon/cpu.h b/target/hexagon/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/cpu.h +++ b/target/hexagon/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void cpu_get_tb_cpu_state(CPUHexagonState *env, vaddr *pc, typedef HexagonCPU ArchCPU; void hexagon_translate_init(void); +void hexagon_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #include "exec/cpu-all.h" diff --git a/target/hppa/cpu.h b/target/hppa/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/cpu.h +++ b/target/hppa/cpu.h @@ -XXX,XX +XXX,XX @@ static inline int HPPA_BTLB_ENTRIES(CPUHPPAState *env) } void hppa_translate_init(void); +void hppa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_HPPA_CPU diff --git a/target/i386/tcg/helper-tcg.h b/target/i386/tcg/helper-tcg.h index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/helper-tcg.h +++ b/target/i386/tcg/helper-tcg.h @@ -XXX,XX +XXX,XX @@ static inline target_long lshift(target_long x, int n) /* translate.c */ void tcg_x86_init(void); +void x86_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /* excp_helper.c */ G_NORETURN void raise_exception(CPUX86State *env, int exception_index); diff --git a/target/loongarch/internals.h b/target/loongarch/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/internals.h +++ b/target/loongarch/internals.h @@ -XXX,XX +XXX,XX @@ #define TARGET_VIRT_MASK MAKE_64BIT_MASK(0, TARGET_VIRT_ADDR_SPACE_BITS) void loongarch_translate_init(void); +void loongarch_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void G_NORETURN do_raise_exception(CPULoongArchState *env, uint32_t exception, diff --git a/target/m68k/cpu.h b/target/m68k/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/cpu.h +++ b/target/m68k/cpu.h @@ -XXX,XX +XXX,XX @@ int m68k_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int m68k_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void m68k_tcg_init(void); +void m68k_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void m68k_cpu_init_gdb(M68kCPU *cpu); uint32_t cpu_m68k_get_ccr(CPUM68KState *env); void cpu_m68k_set_ccr(CPUM68KState *env, uint32_t); diff --git a/target/microblaze/cpu.h b/target/microblaze/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/cpu.h +++ b/target/microblaze/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void mb_cpu_write_msr(CPUMBState *env, uint32_t val) } void mb_tcg_init(void); +void mb_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_MICROBLAZE_CPU diff --git a/target/mips/tcg/tcg-internal.h b/target/mips/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/target/mips/tcg/tcg-internal.h +++ b/target/mips/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ #include "cpu.h" void mips_tcg_init(void); +void mips_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void mips_cpu_synchronize_from_tb(CPUState *cs, const TranslationBlock *tb); G_NORETURN void mips_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, diff --git a/target/openrisc/cpu.h b/target/openrisc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/cpu.h +++ b/target/openrisc/cpu.h @@ -XXX,XX +XXX,XX @@ void openrisc_cpu_dump_state(CPUState *cpu, FILE *f, int flags); int openrisc_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int openrisc_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void openrisc_translate_init(void); +void openrisc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); int print_insn_or1k(bfd_vma addr, disassemble_info *info); #ifndef CONFIG_USER_ONLY diff --git a/target/ppc/cpu.h b/target/ppc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/cpu.h +++ b/target/ppc/cpu.h @@ -XXX,XX +XXX,XX @@ extern const VMStateDescription vmstate_ppc_cpu; /*****************************************************************************/ void ppc_translate_init(void); +void ppc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #if !defined(CONFIG_USER_ONLY) void ppc_store_sdr1(CPUPPCState *env, target_ulong value); diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/cpu.h +++ b/target/riscv/cpu.h @@ -XXX,XX +XXX,XX @@ RISCVException smstateen_acc_ok(CPURISCVState *env, int index, uint64_t bit); void riscv_cpu_set_mode(CPURISCVState *env, target_ulong newpriv, bool virt_en); void riscv_translate_init(void); +void riscv_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); + G_NORETURN void riscv_raise_exception(CPURISCVState *env, uint32_t exception, uintptr_t pc); diff --git a/target/rx/cpu.h b/target/rx/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/rx/cpu.h +++ b/target/rx/cpu.h @@ -XXX,XX +XXX,XX @@ int rx_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int rx_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void rx_translate_init(void); +void rx_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void rx_cpu_unpack_psw(CPURXState *env, uint32_t psw, int rte); #include "exec/cpu-all.h" diff --git a/target/s390x/s390x-internal.h b/target/s390x/s390x-internal.h index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/s390x-internal.h +++ b/target/s390x/s390x-internal.h @@ -XXX,XX +XXX,XX @@ void handle_diag_308(CPUS390XState *env, uint64_t r1, uint64_t r3, /* translate.c */ void s390x_translate_init(void); +void s390x_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void s390x_restore_state_to_opc(CPUState *cs, const TranslationBlock *tb, const uint64_t *data); diff --git a/target/sh4/cpu.h b/target/sh4/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/cpu.h +++ b/target/sh4/cpu.h @@ -XXX,XX +XXX,XX @@ G_NORETURN void superh_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, uintptr_t retaddr); void sh4_translate_init(void); +void sh4_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #if !defined(CONFIG_USER_ONLY) hwaddr superh_cpu_get_phys_page_debug(CPUState *cpu, vaddr addr); diff --git a/target/sparc/cpu.h b/target/sparc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/cpu.h +++ b/target/sparc/cpu.h @@ -XXX,XX +XXX,XX @@ int sparc_cpu_memory_rw_debug(CPUState *cpu, vaddr addr, /* translate.c */ void sparc_tcg_init(void); +void sparc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /* fop_helper.c */ target_ulong cpu_get_fsr(CPUSPARCState *); diff --git a/target/tricore/cpu.h b/target/tricore/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/cpu.h +++ b/target/tricore/cpu.h @@ -XXX,XX +XXX,XX @@ FIELD(TB_FLAGS, PRIV, 0, 2) void cpu_state_reset(CPUTriCoreState *s); void tricore_tcg_init(void); +void tricore_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); static inline void cpu_get_tb_cpu_state(CPUTriCoreState *env, vaddr *pc, uint64_t *cs_base, uint32_t *flags) diff --git a/target/xtensa/cpu.h b/target/xtensa/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/cpu.h +++ b/target/xtensa/cpu.h @@ -XXX,XX +XXX,XX @@ G_NORETURN void xtensa_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, void xtensa_collect_sr_names(const XtensaConfig *config); void xtensa_translate_init(void); +void xtensa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void **xtensa_get_regfile_by_name(const char *name, int entries, int bits); void xtensa_breakpoint_handler(CPUState *cs); void xtensa_register_core(XtensaConfigList *node); diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cpu-exec.c +++ b/accel/tcg/cpu-exec.c @@ -XXX,XX +XXX,XX @@ bool tcg_exec_realizefn(CPUState *cpu, Error **errp) if (!tcg_target_initialized) { /* Check mandatory TCGCPUOps handlers */ + const TCGCPUOps *tcg_ops = cpu->cc->tcg_ops; #ifndef CONFIG_USER_ONLY - assert(cpu->cc->tcg_ops->cpu_exec_halt); - assert(cpu->cc->tcg_ops->cpu_exec_interrupt); + assert(tcg_ops->cpu_exec_halt); + assert(tcg_ops->cpu_exec_interrupt); #endif /* !CONFIG_USER_ONLY */ - cpu->cc->tcg_ops->initialize(); + assert(tcg_ops->translate_code); + tcg_ops->initialize(); tcg_target_initialized = true; } diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/translate-all.c +++ b/accel/tcg/translate-all.c @@ -XXX,XX +XXX,XX @@ static int setjmp_gen_code(CPUArchState *env, TranslationBlock *tb, tcg_func_start(tcg_ctx); - tcg_ctx->cpu = env_cpu(env); - gen_intermediate_code(env_cpu(env), tb, max_insns, pc, host_pc); + CPUState *cs = env_cpu(env); + tcg_ctx->cpu = cs; + cs->cc->tcg_ops->translate_code(cs, tb, max_insns, pc, host_pc); + assert(tb->size != 0); tcg_ctx->cpu = NULL; *max_insns = tb->icount; @@ -XXX,XX +XXX,XX @@ TranslationBlock *tb_gen_code(CPUState *cpu, /* * Overflow of code_gen_buffer, or the current slice of it. * - * TODO: We don't need to re-do gen_intermediate_code, nor + * TODO: We don't need to re-do tcg_ops->translate_code, nor * should we re-do the tcg optimization currently hidden * inside tcg_gen_code. All that should be required is to * flush the TBs, allocate a new TB, re-initialize it per diff --git a/target/alpha/cpu.c b/target/alpha/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/cpu.c +++ b/target/alpha/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps alpha_sysemu_ops = { static const TCGCPUOps alpha_tcg_ops = { .initialize = alpha_translate_init, + .translate_code = alpha_translate_code, .synchronize_from_tb = alpha_cpu_synchronize_from_tb, .restore_state_to_opc = alpha_restore_state_to_opc, diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps alpha_tr_ops = { .tb_stop = alpha_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void alpha_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &alpha_tr_ops, &dc.base); diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps arm_sysemu_ops = { #ifdef CONFIG_TCG static const TCGCPUOps arm_tcg_ops = { .initialize = arm_translate_init, + .translate_code = arm_translate_code, .synchronize_from_tb = arm_cpu_synchronize_from_tb, .debug_excp_handler = arm_debug_excp_handler, .restore_state_to_opc = arm_restore_state_to_opc, diff --git a/target/arm/tcg/cpu-v7m.c b/target/arm/tcg/cpu-v7m.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/cpu-v7m.c +++ b/target/arm/tcg/cpu-v7m.c @@ -XXX,XX +XXX,XX @@ static void cortex_m55_initfn(Object *obj) static const TCGCPUOps arm_v7m_tcg_ops = { .initialize = arm_translate_init, + .translate_code = arm_translate_code, .synchronize_from_tb = arm_cpu_synchronize_from_tb, .debug_excp_handler = arm_debug_excp_handler, .restore_state_to_opc = arm_restore_state_to_opc, diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/translate.c +++ b/target/arm/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps thumb_translator_ops = { .tb_stop = arm_tr_tb_stop, }; -/* generate intermediate code for basic block 'tb'. */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void arm_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = { }; const TranslatorOps *ops = &arm_translator_ops; diff --git a/target/avr/cpu.c b/target/avr/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/avr/cpu.c +++ b/target/avr/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps avr_sysemu_ops = { static const TCGCPUOps avr_tcg_ops = { .initialize = avr_cpu_tcg_init, + .translate_code = avr_cpu_translate_code, .synchronize_from_tb = avr_cpu_synchronize_from_tb, .restore_state_to_opc = avr_restore_state_to_opc, .cpu_exec_interrupt = avr_cpu_exec_interrupt, diff --git a/target/avr/translate.c b/target/avr/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/avr/translate.c +++ b/target/avr/translate.c @@ -XXX,XX +XXX,XX @@ static bool trans_WDR(DisasContext *ctx, arg_WDR *a) * * - translate() * - canonicalize_skip() - * - gen_intermediate_code() + * - translate_code() * - restore_state_to_opc() * */ @@ -XXX,XX +XXX,XX @@ static const TranslatorOps avr_tr_ops = { .tb_stop = avr_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void avr_cpu_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = { }; translator_loop(cs, tb, max_insns, pc, host_pc, &avr_tr_ops, &dc.base); diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/cpu.c +++ b/target/hexagon/cpu.c @@ -XXX,XX +XXX,XX @@ static void hexagon_cpu_init(Object *obj) static const TCGCPUOps hexagon_tcg_ops = { .initialize = hexagon_translate_init, + .translate_code = hexagon_translate_code, .synchronize_from_tb = hexagon_cpu_synchronize_from_tb, .restore_state_to_opc = hexagon_restore_state_to_opc, }; diff --git a/target/hexagon/translate.c b/target/hexagon/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/translate.c +++ b/target/hexagon/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps hexagon_tr_ops = { .tb_stop = hexagon_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void hexagon_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/hppa/cpu.c b/target/hppa/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/cpu.c +++ b/target/hppa/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps hppa_sysemu_ops = { static const TCGCPUOps hppa_tcg_ops = { .initialize = hppa_translate_init, + .translate_code = hppa_translate_code, .synchronize_from_tb = hppa_cpu_synchronize_from_tb, .restore_state_to_opc = hppa_restore_state_to_opc, diff --git a/target/hppa/translate.c b/target/hppa/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/translate.c +++ b/target/hppa/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps hppa_tr_ops = { #endif }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void hppa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx = { }; translator_loop(cs, tb, max_insns, pc, host_pc, &hppa_tr_ops, &ctx.base); diff --git a/target/i386/tcg/tcg-cpu.c b/target/i386/tcg/tcg-cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/tcg-cpu.c +++ b/target/i386/tcg/tcg-cpu.c @@ -XXX,XX +XXX,XX @@ static bool x86_debug_check_breakpoint(CPUState *cs) static const TCGCPUOps x86_tcg_ops = { .initialize = tcg_x86_init, + .translate_code = x86_translate_code, .synchronize_from_tb = x86_cpu_synchronize_from_tb, .restore_state_to_opc = x86_restore_state_to_opc, .cpu_exec_enter = x86_cpu_exec_enter, diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/translate.c +++ b/target/i386/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps i386_tr_ops = { .tb_stop = i386_tr_tb_stop, }; -/* generate intermediate code for basic block 'tb'. */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void x86_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/loongarch/cpu.c b/target/loongarch/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/cpu.c +++ b/target/loongarch/cpu.c @@ -XXX,XX +XXX,XX @@ static void loongarch_cpu_dump_state(CPUState *cs, FILE *f, int flags) static const TCGCPUOps loongarch_tcg_ops = { .initialize = loongarch_translate_init, + .translate_code = loongarch_translate_code, .synchronize_from_tb = loongarch_cpu_synchronize_from_tb, .restore_state_to_opc = loongarch_restore_state_to_opc, diff --git a/target/loongarch/tcg/translate.c b/target/loongarch/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/tcg/translate.c +++ b/target/loongarch/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps loongarch_tr_ops = { .tb_stop = loongarch_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void loongarch_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/m68k/cpu.c b/target/m68k/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/cpu.c +++ b/target/m68k/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps m68k_sysemu_ops = { static const TCGCPUOps m68k_tcg_ops = { .initialize = m68k_tcg_init, + .translate_code = m68k_translate_code, .restore_state_to_opc = m68k_restore_state_to_opc, #ifndef CONFIG_USER_ONLY diff --git a/target/m68k/translate.c b/target/m68k/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/translate.c +++ b/target/m68k/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps m68k_tr_ops = { .tb_stop = m68k_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void m68k_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &m68k_tr_ops, &dc.base); diff --git a/target/microblaze/cpu.c b/target/microblaze/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/cpu.c +++ b/target/microblaze/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps mb_sysemu_ops = { static const TCGCPUOps mb_tcg_ops = { .initialize = mb_tcg_init, + .translate_code = mb_translate_code, .synchronize_from_tb = mb_cpu_synchronize_from_tb, .restore_state_to_opc = mb_restore_state_to_opc, diff --git a/target/microblaze/translate.c b/target/microblaze/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/translate.c +++ b/target/microblaze/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps mb_tr_ops = { .tb_stop = mb_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void mb_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &mb_tr_ops, &dc.base); diff --git a/target/mips/cpu.c b/target/mips/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/mips/cpu.c +++ b/target/mips/cpu.c @@ -XXX,XX +XXX,XX @@ static const Property mips_cpu_properties[] = { #include "hw/core/tcg-cpu-ops.h" static const TCGCPUOps mips_tcg_ops = { .initialize = mips_tcg_init, + .translate_code = mips_translate_code, .synchronize_from_tb = mips_cpu_synchronize_from_tb, .restore_state_to_opc = mips_restore_state_to_opc, diff --git a/target/mips/tcg/translate.c b/target/mips/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/mips/tcg/translate.c +++ b/target/mips/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps mips_tr_ops = { .tb_stop = mips_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void mips_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/openrisc/cpu.c b/target/openrisc/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/cpu.c +++ b/target/openrisc/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps openrisc_sysemu_ops = { static const TCGCPUOps openrisc_tcg_ops = { .initialize = openrisc_translate_init, + .translate_code = openrisc_translate_code, .synchronize_from_tb = openrisc_cpu_synchronize_from_tb, .restore_state_to_opc = openrisc_restore_state_to_opc, diff --git a/target/openrisc/translate.c b/target/openrisc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/translate.c +++ b/target/openrisc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps openrisc_tr_ops = { .tb_stop = openrisc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void openrisc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/cpu_init.c +++ b/target/ppc/cpu_init.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps ppc_sysemu_ops = { static const TCGCPUOps ppc_tcg_ops = { .initialize = ppc_translate_init, + .translate_code = ppc_translate_code, .restore_state_to_opc = ppc_restore_state_to_opc, #ifdef CONFIG_USER_ONLY diff --git a/target/ppc/translate.c b/target/ppc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/translate.c +++ b/target/ppc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps ppc_tr_ops = { .tb_stop = ppc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void ppc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/tcg/tcg-cpu.c +++ b/target/riscv/tcg/tcg-cpu.c @@ -XXX,XX +XXX,XX @@ static void riscv_restore_state_to_opc(CPUState *cs, static const TCGCPUOps riscv_tcg_ops = { .initialize = riscv_translate_init, + .translate_code = riscv_translate_code, .synchronize_from_tb = riscv_cpu_synchronize_from_tb, .restore_state_to_opc = riscv_restore_state_to_opc, diff --git a/target/riscv/translate.c b/target/riscv/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/translate.c +++ b/target/riscv/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps riscv_tr_ops = { .tb_stop = riscv_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void riscv_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/rx/cpu.c b/target/rx/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/rx/cpu.c +++ b/target/rx/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps rx_sysemu_ops = { static const TCGCPUOps rx_tcg_ops = { .initialize = rx_translate_init, + .translate_code = rx_translate_code, .synchronize_from_tb = rx_cpu_synchronize_from_tb, .restore_state_to_opc = rx_restore_state_to_opc, .tlb_fill = rx_cpu_tlb_fill, diff --git a/target/rx/translate.c b/target/rx/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/rx/translate.c +++ b/target/rx/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps rx_tr_ops = { .tb_stop = rx_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void rx_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/cpu.c +++ b/target/s390x/cpu.c @@ -XXX,XX +XXX,XX @@ void cpu_get_tb_cpu_state(CPUS390XState *env, vaddr *pc, static const TCGCPUOps s390_tcg_ops = { .initialize = s390x_translate_init, + .translate_code = s390x_translate_code, .restore_state_to_opc = s390x_restore_state_to_opc, #ifdef CONFIG_USER_ONLY diff --git a/target/s390x/tcg/translate.c b/target/s390x/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/tcg/translate.c +++ b/target/s390x/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps s390x_tr_ops = { .disas_log = s390x_tr_disas_log, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void s390x_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/sh4/cpu.c b/target/sh4/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/cpu.c +++ b/target/sh4/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps sh4_sysemu_ops = { static const TCGCPUOps superh_tcg_ops = { .initialize = sh4_translate_init, + .translate_code = sh4_translate_code, .synchronize_from_tb = superh_cpu_synchronize_from_tb, .restore_state_to_opc = superh_restore_state_to_opc, diff --git a/target/sh4/translate.c b/target/sh4/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/translate.c +++ b/target/sh4/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps sh4_tr_ops = { .tb_stop = sh4_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void sh4_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/sparc/cpu.c b/target/sparc/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/cpu.c +++ b/target/sparc/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps sparc_sysemu_ops = { static const TCGCPUOps sparc_tcg_ops = { .initialize = sparc_tcg_init, + .translate_code = sparc_translate_code, .synchronize_from_tb = sparc_cpu_synchronize_from_tb, .restore_state_to_opc = sparc_restore_state_to_opc, diff --git a/target/sparc/translate.c b/target/sparc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/translate.c +++ b/target/sparc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps sparc_tr_ops = { .tb_stop = sparc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void sparc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = {}; diff --git a/target/tricore/cpu.c b/target/tricore/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/cpu.c +++ b/target/tricore/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps tricore_sysemu_ops = { static const TCGCPUOps tricore_tcg_ops = { .initialize = tricore_tcg_init, + .translate_code = tricore_translate_code, .synchronize_from_tb = tricore_cpu_synchronize_from_tb, .restore_state_to_opc = tricore_restore_state_to_opc, .tlb_fill = tricore_cpu_tlb_fill, diff --git a/target/tricore/translate.c b/target/tricore/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/translate.c +++ b/target/tricore/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps tricore_tr_ops = { .tb_stop = tricore_tr_tb_stop, }; - -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void tricore_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; translator_loop(cs, tb, max_insns, pc, host_pc, diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/cpu.c +++ b/target/xtensa/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps xtensa_sysemu_ops = { static const TCGCPUOps xtensa_tcg_ops = { .initialize = xtensa_translate_init, + .translate_code = xtensa_translate_code, .debug_excp_handler = xtensa_breakpoint_handler, .restore_state_to_opc = xtensa_restore_state_to_opc, diff --git a/target/xtensa/translate.c b/target/xtensa/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/translate.c +++ b/target/xtensa/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps xtensa_translator_ops = { .tb_stop = xtensa_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void xtensa_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = {}; translator_loop(cpu, tb, max_insns, pc, host_pc, -- 2.43.0