Committed, thanks Kito and Juzhe. Pan
From: Kito Cheng <kito.ch...@gmail.com> Sent: Monday, June 12, 2023 11:33 AM To: 钟居哲 <juzhe.zh...@rivai.ai> Cc: Li, Pan2 <pan2...@intel.com>; gcc-patches <gcc-patches@gcc.gnu.org>; rdapp.gcc <rdapp....@gmail.com>; Jeff Law <jeffreya...@gmail.com>; Wang, Yanzhang <yanzhang.w...@intel.com> Subject: Re: [PATCH v1] RISC-V: Support RVV FP16 MISC vlmul ext intrinsic API Lgtm too :) 钟居哲 <juzhe.zh...@rivai.ai<mailto:juzhe.zh...@rivai.ai>> 於 2023年6月12日 週一 05:48 寫道： LGTM juzhe.zh...@rivai.ai<mailto:juzhe.zh...@rivai.ai> From: pan2.li<http://pan2.li> Date: 2023-06-11 08:33 To: gcc-patches CC: juzhe.zhong; rdapp.gcc; jeffreyalaw; pan2.li<http://pan2.li>; yanzhang.wang; kito.cheng Subject: [PATCH v1] RISC-V: Support RVV FP16 MISC vlmul ext intrinsic API From: Pan Li <pan2...@intel.com<mailto:pan2...@intel.com>> This patch support the intrinsic API of FP16 ZVFHMIN vlmul ext. Aka: vfloat16*_t <==> vfloat16*_t. From the user's perspective, it is reasonable to do some type convert between vfloat16*_t and vfloat16*_t when only ZVFHMIN is enabled. Signed-off-by: Pan Li <pan2...@intel.com<mailto:pan2...@intel.com>> gcc/ChangeLog: * config/riscv/riscv-vector-builtins-types.def (vfloat16mf4_t): Add type to X2/X4/X8/X16/X32 vlmul ext ops. (vfloat16mf2_t): Ditto. (vfloat16m1_t): Ditto. (vfloat16m2_t): Ditto. (vfloat16m4_t): Ditto. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/zvfh-over-zvfhmin.c: Add new test cases. * gcc.target/riscv/rvv/base/zvfhmin-intrinsic.c: Add new test cases. --- .../riscv/riscv-vector-builtins-types.def | 15 ++++++ .../riscv/rvv/base/zvfh-over-zvfhmin.c | 18 +++++-- .../riscv/rvv/base/zvfhmin-intrinsic.c | 54 +++++++++++++++++-- 3 files changed, 79 insertions(+), 8 deletions(-) diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 589ea532727..db8e61fea6a 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -978,6 +978,11 @@ DEF_RVV_X2_VLMUL_EXT_OPS (vuint32m4_t, 0) DEF_RVV_X2_VLMUL_EXT_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) DEF_RVV_X2_VLMUL_EXT_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_X2_VLMUL_EXT_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_X2_VLMUL_EXT_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X2_VLMUL_EXT_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X2_VLMUL_EXT_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X2_VLMUL_EXT_OPS (vfloat16m2_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X2_VLMUL_EXT_OPS (vfloat16m4_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_X2_VLMUL_EXT_OPS (vfloat32mf2_t, RVV_REQUIRE_ELEN_FP_32 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X2_VLMUL_EXT_OPS (vfloat32m1_t, RVV_REQUIRE_ELEN_FP_32) DEF_RVV_X2_VLMUL_EXT_OPS (vfloat32m2_t, RVV_REQUIRE_ELEN_FP_32) @@ -1014,6 +1019,10 @@ DEF_RVV_X4_VLMUL_EXT_OPS (vuint32m1_t, 0) DEF_RVV_X4_VLMUL_EXT_OPS (vuint32m2_t, 0) DEF_RVV_X4_VLMUL_EXT_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) DEF_RVV_X4_VLMUL_EXT_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_X4_VLMUL_EXT_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X4_VLMUL_EXT_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X4_VLMUL_EXT_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X4_VLMUL_EXT_OPS (vfloat16m2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_X4_VLMUL_EXT_OPS (vfloat32mf2_t, RVV_REQUIRE_ELEN_FP_32 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X4_VLMUL_EXT_OPS (vfloat32m1_t, RVV_REQUIRE_ELEN_FP_32) DEF_RVV_X4_VLMUL_EXT_OPS (vfloat32m2_t, RVV_REQUIRE_ELEN_FP_32) @@ -1040,6 +1049,9 @@ DEF_RVV_X8_VLMUL_EXT_OPS (vuint16m1_t, 0) DEF_RVV_X8_VLMUL_EXT_OPS (vuint32mf2_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X8_VLMUL_EXT_OPS (vuint32m1_t, 0) DEF_RVV_X8_VLMUL_EXT_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_X8_VLMUL_EXT_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X8_VLMUL_EXT_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) +DEF_RVV_X8_VLMUL_EXT_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_X8_VLMUL_EXT_OPS (vfloat32mf2_t, RVV_REQUIRE_ELEN_FP_32 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X8_VLMUL_EXT_OPS (vfloat32m1_t, RVV_REQUIRE_ELEN_FP_32) DEF_RVV_X8_VLMUL_EXT_OPS (vfloat64m1_t, RVV_REQUIRE_ELEN_FP_64) @@ -1056,6 +1068,8 @@ DEF_RVV_X16_VLMUL_EXT_OPS (vuint8mf2_t, 0) DEF_RVV_X16_VLMUL_EXT_OPS (vuint16mf4_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X16_VLMUL_EXT_OPS (vuint16mf2_t, 0) DEF_RVV_X16_VLMUL_EXT_OPS (vuint32mf2_t, RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X16_VLMUL_EXT_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X16_VLMUL_EXT_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_X16_VLMUL_EXT_OPS (vfloat32mf2_t, RVV_REQUIRE_ELEN_FP_32 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X32_VLMUL_EXT_OPS (vint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) @@ -1064,6 +1078,7 @@ DEF_RVV_X32_VLMUL_EXT_OPS (vint16mf4_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X32_VLMUL_EXT_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X32_VLMUL_EXT_OPS (vuint8mf4_t, 0) DEF_RVV_X32_VLMUL_EXT_OPS (vuint16mf4_t, RVV_REQUIRE_MIN_VLEN_64) +DEF_RVV_X32_VLMUL_EXT_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X64_VLMUL_EXT_OPS (vint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_X64_VLMUL_EXT_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/zvfh-over-zvfhmin.c b/gcc/testsuite/gcc.target/riscv/rvv/base/zvfh-over-zvfhmin.c index d5bcdd5156a..ff9e0156a68 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/zvfh-over-zvfhmin.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/zvfh-over-zvfhmin.c @@ -37,13 +37,23 @@ vuint16m8_t test_vreinterpret_v_f16m8_u16m8(vfloat16m8_t src) { return __riscv_vreinterpret_v_f16m8_u16m8(src); } +vfloat16mf2_t test_vlmul_ext_v_f16mf4_f16mf2(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16mf2(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16mf4_f16m8(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16m8(op1); +} + /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 3 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m4,\s*t[au],\s*m[au]} 2 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m8,\s*t[au],\s*m[au]} 1 } } */ -/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 4 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 6 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf2,\s*t[au],\s*m[au]} 1 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*m8,\s*t[au],\s*m[au]} 1 } } */ /* { dg-final { scan-assembler-times {vfwcvt\.f\.f\.v\s+v[0-9]+,\s*v[0-9]+} 2 } } */ /* { dg-final { scan-assembler-times {vfncvt\.f\.f\.w\s+v[0-9]+,\s*v[0-9]+} 2 } } */ -/* { dg-final { scan-assembler-times {vle16\.v\s+v[0-9]+,\s*0\([0-9ax]+\)} 4 } } */ -/* { dg-final { scan-assembler-times {vse16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 3 } } */ +/* { dg-final { scan-assembler-times {vle16\.v\s+v[0-9]+,\s*0\([0-9ax]+\)} 6 } } */ +/* { dg-final { scan-assembler-times {vse16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 4 } } */ /* { dg-final { scan-assembler-times {vl8re16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 1 } } */ -/* { dg-final { scan-assembler-times {vs8r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 3 } } */ +/* { dg-final { scan-assembler-times {vs8r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 4 } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/zvfhmin-intrinsic.c b/gcc/testsuite/gcc.target/riscv/rvv/base/zvfhmin-intrinsic.c index e56b2751d4c..68720e64926 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/zvfhmin-intrinsic.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/zvfhmin-intrinsic.c @@ -85,16 +85,62 @@ vuint16m8_t test_vreinterpret_v_f16m8_u16m8(vfloat16m8_t src) { return __riscv_vreinterpret_v_f16m8_u16m8(src); } +vfloat16mf2_t test_vlmul_ext_v_f16mf4_f16mf2(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16mf2(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16m4_f16m8(vfloat16m4_t op1) { + return __riscv_vlmul_ext_v_f16m4_f16m8(op1); +} + +vfloat16m1_t test_vlmul_ext_v_f16mf4_f16m1(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16m1(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16m2_f16m8(vfloat16m2_t op1) { + return __riscv_vlmul_ext_v_f16m2_f16m8(op1); +} + +vfloat16m2_t test_vlmul_ext_v_f16mf4_f16m2(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16m2(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16m1_f16m8(vfloat16m1_t op1) { + return __riscv_vlmul_ext_v_f16m1_f16m8(op1); +} + +vfloat16m4_t test_vlmul_ext_v_f16mf4_f16m4(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16m4(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16mf2_f16m8(vfloat16mf2_t op1) { + return __riscv_vlmul_ext_v_f16mf2_f16m8(op1); +} + +vfloat16m8_t test_vlmul_ext_v_f16mf4_f16m8(vfloat16mf4_t op1) { + return __riscv_vlmul_ext_v_f16mf4_f16m8(op1); +} + /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 3 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*mf2,\s*t[au],\s*m[au]} 2 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m1,\s*t[au],\s*m[au]} 2 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m2,\s*t[au],\s*m[au]} 2 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m4,\s*t[au],\s*m[au]} 2 } } */ /* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e16,\s*m8,\s*t[au],\s*m[au]} 1 } } */ -/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 7 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf4,\s*t[au],\s*m[au]} 12 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*mf2,\s*t[au],\s*m[au]} 4 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*m2,\s*t[au],\s*m[au]} 1 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*m4,\s*t[au],\s*m[au]} 1 } } */ +/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e16,\s*m8,\s*t[au],\s*m[au]} 5 } } */ /* { dg-final { scan-assembler-times {vfwcvt\.f\.f\.v\s+v[0-9]+,\s*v[0-9]+} 5 } } */ /* { dg-final { scan-assembler-times {vfncvt\.f\.f\.w\s+v[0-9]+,\s*v[0-9]+} 5 } } */ -/* { dg-final { scan-assembler-times {vle16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 8 } } */ -/* { dg-final { scan-assembler-times {vse16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 7 } } */ +/* { dg-final { scan-assembler-times {vle16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 14 } } */ +/* { dg-final { scan-assembler-times {vse16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 8 } } */ +/* { dg-final { scan-assembler-times {vl1re16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 2 } } */ +/* { dg-final { scan-assembler-times {vl2re16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 2 } } */ /* { dg-final { scan-assembler-times {vl8re16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 4 } } */ -/* { dg-final { scan-assembler-times {vs8r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 6 } } */ +/* { dg-final { scan-assembler-times {vl4re16\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 2 } } */ +/* { dg-final { scan-assembler-times {vs1r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 3 } } */ +/* { dg-final { scan-assembler-times {vs2r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 3 } } */ +/* { dg-final { scan-assembler-times {vs4r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 3 } } */ +/* { dg-final { scan-assembler-times {vs8r\.v\s+v[0-9]+,\s*0\([a-x][0-9]+\)} 11 } } */ -- 2.34.1
