[PULL,05/23] tcg: Add opcodes for vector minmax arithmetic

Message ID	20190128155907.20607-6-richard.henderson@linaro.org
State	New
Headers	show Delivered-To: patch@linaro.org Received-SPF: pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; From: Richard Henderson <richard.henderson@linaro.org> To: qemu-devel@nongnu.org Date: Mon, 28 Jan 2019 07:58:49 -0800 Message-Id: <20190128155907.20607-6-richard.henderson@linaro.org> In-Reply-To: <20190128155907.20607-1-richard.henderson@linaro.org> References: <20190128155907.20607-1-richard.henderson@linaro.org> Subject: [Qemu-devel] [PULL 05/23] tcg: Add opcodes for vector minmax arithmetic Precedence: list Cc: peter.maydell@linaro.org Errors-To: qemu-devel-bounces+patch=linaro.org@nongnu.org Sender: "Qemu-devel" <qemu-devel-bounces+patch=linaro.org@nongnu.org>
Series	tcg queued patches \| expand [PULL,00/23] tcg queued patches [PULL,01/23] tcg: Add logical simplifications during gvec expand [PULL,02/23] tcg: Add gvec expanders for nand, nor, eqv [PULL,03/23] tcg: Add write_aofs to GVecGen4 [PULL,04/23] tcg: Add opcodes for vector saturated arithmetic [PULL,05/23] tcg: Add opcodes for vector minmax arithmetic [PULL,06/23] tcg/i386: Split subroutines out of tcg_expand_vec_op [PULL,07/23] tcg/i386: Implement vector saturating arithmetic [PULL,08/23] tcg/i386: Implement vector minmax arithmetic [PULL,09/23] tcg/aarch64: Implement vector saturating arithmetic [PULL,10/23] tcg/aarch64: Implement vector minmax arithmetic [PULL,11/23] cputlb: do not evict empty entries to the vtlb [PULL,12/23] tcg: introduce dynamic TLB sizing [PULL,13/23] tcg/i386: enable dynamic TLB sizing [PULL,14/23] tcg/aarch64: enable dynamic TLB sizing [PULL,15/23] tcg/ppc: enable dynamic TLB sizing [PULL,16/23] tcg/sparc: enable dynamic TLB sizing [PULL,17/23] tcg/s390: enable dynamic TLB sizing [PULL,18/23] tcg/riscv: enable dynamic TLB sizing [PULL,19/23] tcg/arm: enable dynamic TLB sizing [PULL,20/23] tcg/mips: Fix tcg_out_qemu_ld_slow_path [PULL,21/23] tcg/mips: enable dynamic TLB sizing [PULL,22/23] tcg/tci: enable dynamic TLB sizing [PULL,23/23] cputlb: Remove static tlb sizing

diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h index 835ddfebb2..dfe325625c 100644 --- a/accel/tcg/tcg-runtime.h +++ b/accel/tcg/tcg-runtime.h @@ -200,6 +200,26 @@ DEF_HELPER_FLAGS_4(gvec_ussub16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_ussub32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_ussub64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smin8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smin16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smin32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smin64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(gvec_smax8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smax16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smax32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_smax64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(gvec_umin8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umin16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umin32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umin64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(gvec_umax8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umax16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umax32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_umax64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_3(gvec_neg8, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_neg16, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_neg32, TCG_CALL_NO_RWG, void, ptr, ptr, i32) diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index 98556bcf22..545a6eec75 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -136,6 +136,7 @@ typedef enum { #define TCG_TARGET_HAS_cmp_vec 1 #define TCG_TARGET_HAS_mul_vec 1 #define TCG_TARGET_HAS_sat_vec 0 +#define TCG_TARGET_HAS_minmax_vec 0 #define TCG_TARGET_DEFAULT_MO (0) #define TCG_TARGET_HAS_MEMORY_BSWAP 1 diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index 44381062e6..7bd7eae672 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -186,6 +186,7 @@ extern bool have_avx2; #define TCG_TARGET_HAS_cmp_vec 1 #define TCG_TARGET_HAS_mul_vec 1 #define TCG_TARGET_HAS_sat_vec 0 +#define TCG_TARGET_HAS_minmax_vec 0 #define TCG_TARGET_deposit_i32_valid(ofs, len) \ (((ofs) == 0 && (len) == 8) || ((ofs) == 8 && (len) == 8) || \ diff --git a/tcg/tcg-op-gvec.h b/tcg/tcg-op-gvec.h index 2cb447112e..4734eef7de 100644 --- a/tcg/tcg-op-gvec.h +++ b/tcg/tcg-op-gvec.h @@ -234,6 +234,16 @@ void tcg_gen_gvec_usadd(unsigned vece, uint32_t dofs, uint32_t aofs, void tcg_gen_gvec_ussub(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +/* Min/max. */ +void tcg_gen_gvec_smin(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_umin(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_smax(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_umax(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz); + void tcg_gen_gvec_and(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_or(unsigned vece, uint32_t dofs, uint32_t aofs, diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h index 4a93d730e8..2d98868d8f 100644 --- a/tcg/tcg-op.h +++ b/tcg/tcg-op.h @@ -971,6 +971,10 @@ void tcg_gen_ssadd_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_usadd_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_sssub_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_ussub_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); +void tcg_gen_smin_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); +void tcg_gen_umin_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); +void tcg_gen_smax_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); +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); diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h index 94b2ed80af..4e0238ad1a 100644 --- a/tcg/tcg-opc.h +++ b/tcg/tcg-opc.h @@ -226,6 +226,10 @@ DEF(ssadd_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_sat_vec)) DEF(usadd_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_sat_vec)) DEF(sssub_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_sat_vec)) DEF(ussub_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_sat_vec)) +DEF(smin_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_minmax_vec)) +DEF(umin_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_minmax_vec)) +DEF(smax_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_minmax_vec)) +DEF(umax_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_minmax_vec)) DEF(and_vec, 1, 2, 0, IMPLVEC) DEF(or_vec, 1, 2, 0, IMPLVEC) diff --git a/tcg/tcg.h b/tcg/tcg.h index 5590dc512a..045c24a357 100644 --- a/tcg/tcg.h +++ b/tcg/tcg.h @@ -184,6 +184,7 @@ typedef uint64_t TCGRegSet; #define TCG_TARGET_HAS_shv_vec 0 #define TCG_TARGET_HAS_mul_vec 0 #define TCG_TARGET_HAS_sat_vec 0 +#define TCG_TARGET_HAS_minmax_vec 0 #else #define TCG_TARGET_MAYBE_vec 1 #endif diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c index d1802467d5..9358749741 100644 --- a/accel/tcg/tcg-runtime-gvec.c +++ b/accel/tcg/tcg-runtime-gvec.c @@ -1028,3 +1028,227 @@ void HELPER(gvec_ussub64)(void *d, void *a, void *b, uint32_t desc) } clear_high(d, oprsz, desc); } + +void HELPER(gvec_smin8)(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(int8_t)) { + int8_t aa = *(int8_t *)(a + i); + int8_t bb = *(int8_t *)(b + i); + int8_t dd = aa < bb ? aa : bb; + *(int8_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smin16)(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(int16_t)) { + int16_t aa = *(int16_t *)(a + i); + int16_t bb = *(int16_t *)(b + i); + int16_t dd = aa < bb ? aa : bb; + *(int16_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smin32)(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(int32_t)) { + int32_t aa = *(int32_t *)(a + i); + int32_t bb = *(int32_t *)(b + i); + int32_t dd = aa < bb ? aa : bb; + *(int32_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smin64)(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(int64_t)) { + int64_t aa = *(int64_t *)(a + i); + int64_t bb = *(int64_t *)(b + i); + int64_t dd = aa < bb ? aa : bb; + *(int64_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smax8)(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(int8_t)) { + int8_t aa = *(int8_t *)(a + i); + int8_t bb = *(int8_t *)(b + i); + int8_t dd = aa > bb ? aa : bb; + *(int8_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smax16)(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(int16_t)) { + int16_t aa = *(int16_t *)(a + i); + int16_t bb = *(int16_t *)(b + i); + int16_t dd = aa > bb ? aa : bb; + *(int16_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smax32)(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(int32_t)) { + int32_t aa = *(int32_t *)(a + i); + int32_t bb = *(int32_t *)(b + i); + int32_t dd = aa > bb ? aa : bb; + *(int32_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_smax64)(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(int64_t)) { + int64_t aa = *(int64_t *)(a + i); + int64_t bb = *(int64_t *)(b + i); + int64_t dd = aa > bb ? aa : bb; + *(int64_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umin8)(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 aa = *(uint8_t *)(a + i); + uint8_t bb = *(uint8_t *)(b + i); + uint8_t dd = aa < bb ? aa : bb; + *(uint8_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umin16)(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)) { + uint16_t aa = *(uint16_t *)(a + i); + uint16_t bb = *(uint16_t *)(b + i); + uint16_t dd = aa < bb ? aa : bb; + *(uint16_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umin32)(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)) { + uint32_t aa = *(uint32_t *)(a + i); + uint32_t bb = *(uint32_t *)(b + i); + uint32_t dd = aa < bb ? aa : bb; + *(uint32_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umin64)(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)) { + uint64_t aa = *(uint64_t *)(a + i); + uint64_t bb = *(uint64_t *)(b + i); + uint64_t dd = aa < bb ? aa : bb; + *(uint64_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umax8)(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 aa = *(uint8_t *)(a + i); + uint8_t bb = *(uint8_t *)(b + i); + uint8_t dd = aa > bb ? aa : bb; + *(uint8_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umax16)(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)) { + uint16_t aa = *(uint16_t *)(a + i); + uint16_t bb = *(uint16_t *)(b + i); + uint16_t dd = aa > bb ? aa : bb; + *(uint16_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umax32)(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)) { + uint32_t aa = *(uint32_t *)(a + i); + uint32_t bb = *(uint32_t *)(b + i); + uint32_t dd = aa > bb ? aa : bb; + *(uint32_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_umax64)(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)) { + uint64_t aa = *(uint64_t *)(a + i); + uint64_t bb = *(uint64_t *)(b + i); + uint64_t dd = aa > bb ? aa : bb; + *(uint64_t *)(d + i) = dd; + } + clear_high(d, oprsz, desc); +} diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index 0a33f51065..3ee44fcb75 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -1810,6 +1810,114 @@ void tcg_gen_gvec_ussub(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); } +void tcg_gen_gvec_smin(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_smin_vec, + .fno = gen_helper_gvec_smin8, + .opc = INDEX_op_smin_vec, + .vece = MO_8 }, + { .fniv = tcg_gen_smin_vec, + .fno = gen_helper_gvec_smin16, + .opc = INDEX_op_smin_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_smin_i32, + .fniv = tcg_gen_smin_vec, + .fno = gen_helper_gvec_smin32, + .opc = INDEX_op_smin_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_smin_i64, + .fniv = tcg_gen_smin_vec, + .fno = gen_helper_gvec_smin64, + .opc = INDEX_op_smin_vec, + .vece = MO_64 } + }; + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + +void tcg_gen_gvec_umin(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_umin_vec, + .fno = gen_helper_gvec_umin8, + .opc = INDEX_op_umin_vec, + .vece = MO_8 }, + { .fniv = tcg_gen_umin_vec, + .fno = gen_helper_gvec_umin16, + .opc = INDEX_op_umin_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_umin_i32, + .fniv = tcg_gen_umin_vec, + .fno = gen_helper_gvec_umin32, + .opc = INDEX_op_umin_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_umin_i64, + .fniv = tcg_gen_umin_vec, + .fno = gen_helper_gvec_umin64, + .opc = INDEX_op_umin_vec, + .vece = MO_64 } + }; + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + +void tcg_gen_gvec_smax(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_smax_vec, + .fno = gen_helper_gvec_smax8, + .opc = INDEX_op_smax_vec, + .vece = MO_8 }, + { .fniv = tcg_gen_smax_vec, + .fno = gen_helper_gvec_smax16, + .opc = INDEX_op_smax_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_smax_i32, + .fniv = tcg_gen_smax_vec, + .fno = gen_helper_gvec_smax32, + .opc = INDEX_op_smax_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_smax_i64, + .fniv = tcg_gen_smax_vec, + .fno = gen_helper_gvec_smax64, + .opc = INDEX_op_smax_vec, + .vece = MO_64 } + }; + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + +void tcg_gen_gvec_umax(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen3 g[4] = { + { .fniv = tcg_gen_umax_vec, + .fno = gen_helper_gvec_umax8, + .opc = INDEX_op_umax_vec, + .vece = MO_8 }, + { .fniv = tcg_gen_umax_vec, + .fno = gen_helper_gvec_umax16, + .opc = INDEX_op_umax_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_umax_i32, + .fniv = tcg_gen_umax_vec, + .fno = gen_helper_gvec_umax32, + .opc = INDEX_op_umax_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_umax_i64, + .fniv = tcg_gen_umax_vec, + .fno = gen_helper_gvec_umax64, + .opc = INDEX_op_umax_vec, + .vece = MO_64 } + }; + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); +} + /* Perform a vector negation using normal negation and a mask. Compare gen_subv_mask above. */ static void gen_negv_mask(TCGv_i64 d, TCGv_i64 b, TCGv_i64 m) diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index 675aa09258..36f35022ac 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -433,3 +433,23 @@ void tcg_gen_ussub_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) { do_op3(vece, r, a, b, INDEX_op_ussub_vec); } + +void tcg_gen_smin_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3(vece, r, a, b, INDEX_op_smin_vec); +} + +void tcg_gen_umin_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3(vece, r, a, b, INDEX_op_umin_vec); +} + +void tcg_gen_smax_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3(vece, r, a, b, INDEX_op_smax_vec); +} + +void tcg_gen_umax_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) +{ + do_op3(vece, r, a, b, INDEX_op_umax_vec); +} diff --git a/tcg/tcg.c b/tcg/tcg.c index 6363f0cb29..20a5d8f315 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -1612,6 +1612,11 @@ bool tcg_op_supported(TCGOpcode op) case INDEX_op_sssub_vec: case INDEX_op_ussub_vec: return have_vec && TCG_TARGET_HAS_sat_vec; + case INDEX_op_smin_vec: + case INDEX_op_umin_vec: + case INDEX_op_smax_vec: + case INDEX_op_umax_vec: + return have_vec && TCG_TARGET_HAS_minmax_vec; default: tcg_debug_assert(op > INDEX_op_last_generic && op < NB_OPS); diff --git a/tcg/README b/tcg/README index dd9fd9c86c..603f4df659 100644 --- a/tcg/README +++ b/tcg/README @@ -554,6 +554,16 @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32. Similarly, v0 = -v1. +* smin_vec: +* umin_vec: + + Similarly, v0 = MIN(v1, v2), for signed and unsigned element types. + +* smax_vec: +* umax_vec: + + Similarly, v0 = MAX(v1, v2), for signed and unsigned element types. + * ssadd_vec: * sssub_vec: * usadd_vec:

[PULL,05/23] tcg: Add opcodes for vector minmax arithmetic

Commit Message

Patch