From bf9a1a07adb0594b8060bbe23e2cde3475885fdc Mon Sep 17 00:00:00 2001 From: Ilya Verbin Date: Wed, 25 May 2016 15:08:14 +0000 Subject: [PATCH] AVX-512F: Add vectorizer support builtins gcc/ * config/i386/i386-builtin-types.def: Add V16SI_FTYPE_V16SF, V8DF_FTYPE_V8DF_ROUND, V16SF_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF_ROUND. * config/i386/i386.c (enum ix86_builtins): Add IX86_BUILTIN_CVTPS2DQ512_MASK, IX86_BUILTIN_FLOORPS512, IX86_BUILTIN_FLOORPD512, IX86_BUILTIN_CEILPS512, IX86_BUILTIN_CEILPD512, IX86_BUILTIN_TRUNCPS512, IX86_BUILTIN_TRUNCPD512, IX86_BUILTIN_CVTPS2DQ512, IX86_BUILTIN_VEC_PACK_SFIX512, IX86_BUILTIN_FLOORPS_SFIX512, IX86_BUILTIN_CEILPS_SFIX512, IX86_BUILTIN_ROUNDPS_AZ_SFIX512. (builtin_description bdesc_args): Add __builtin_ia32_floorps512, __builtin_ia32_ceilps512, __builtin_ia32_truncps512, __builtin_ia32_floorpd512, __builtin_ia32_ceilpd512, __builtin_ia32_truncpd512, __builtin_ia32_cvtps2dq512, __builtin_ia32_vec_pack_sfix512, __builtin_ia32_roundps_az_sfix512, __builtin_ia32_floorps_sfix512, __builtin_ia32_ceilps_sfix512. Change IX86_BUILTIN_CVTPS2DQ512 to IX86_BUILTIN_CVTPS2DQ512_MASK for __builtin_ia32_cvtps2dq512_mask. (ix86_expand_args_builtin): Handle V8DF_FTYPE_V8DF_ROUND, V16SF_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF. (ix86_builtin_vectorized_function): Handle builtins mentioned above. * config/i386/sse.md (avx512f_fix_notruncv16sfv16si): Rename to ... (avx512f_fix_notruncv16sfv16si): ... this. (avx512f_cvtpd2dq512): Rename to ... (avx512f_cvtpd2dq512): ... this. (avx512f_vec_pack_sfix_v8df): New define_expand. (avx512f_roundpd512): Rename to ... (avx512f_round512): ... this. Change iterator. (avx512f_roundps512_sfix): New define_expand. (round2_sfix): Change iterator. gcc/testsuite/ * gcc.target/i386/avx512f-ceil-vec-1.c: New test. * gcc.target/i386/avx512f-ceil-vec-2.c: New test. * gcc.target/i386/avx512f-ceilf-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-ceilf-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-ceilf-vec-1.c: New test. * gcc.target/i386/avx512f-ceilf-vec-2.c: New test. * gcc.target/i386/avx512f-floor-vec-1.c: New test. * gcc.target/i386/avx512f-floor-vec-2.c: New test. * gcc.target/i386/avx512f-floorf-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-floorf-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-floorf-vec-1.c: New test. * gcc.target/i386/avx512f-floorf-vec-2.c: New test. * gcc.target/i386/avx512f-rint-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-rint-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-rintf-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-rintf-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-round-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-round-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-roundf-sfix-vec-1.c: New test. * gcc.target/i386/avx512f-roundf-sfix-vec-2.c: New test. * gcc.target/i386/avx512f-trunc-vec-1.c: New test. * gcc.target/i386/avx512f-trunc-vec-2.c: New test. * gcc.target/i386/avx512f-truncf-vec-1.c: New test. * gcc.target/i386/avx512f-truncf-vec-2.c: New test. From-SVN: r236709 --- gcc/ChangeLog | 35 ++++++++++++ gcc/config/i386/i386-builtin-types.def | 4 ++ gcc/config/i386/i386.c | 54 +++++++++++++++++-- gcc/config/i386/sse.md | 43 ++++++++++++--- gcc/testsuite/ChangeLog | 27 ++++++++++ .../gcc.target/i386/avx512f-ceil-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-ceil-vec-2.c | 6 +++ .../i386/avx512f-ceilf-sfix-vec-1.c | 52 ++++++++++++++++++ .../i386/avx512f-ceilf-sfix-vec-2.c | 7 +++ .../gcc.target/i386/avx512f-ceilf-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-ceilf-vec-2.c | 6 +++ .../gcc.target/i386/avx512f-floor-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-floor-vec-2.c | 6 +++ .../i386/avx512f-floorf-sfix-vec-1.c | 52 ++++++++++++++++++ .../i386/avx512f-floorf-sfix-vec-2.c | 7 +++ .../gcc.target/i386/avx512f-floorf-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-floorf-vec-2.c | 6 +++ .../gcc.target/i386/avx512f-rint-sfix-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-rint-sfix-vec-2.c | 7 +++ .../i386/avx512f-rintf-sfix-vec-1.c | 52 ++++++++++++++++++ .../i386/avx512f-rintf-sfix-vec-2.c | 6 +++ .../i386/avx512f-round-sfix-vec-1.c | 52 ++++++++++++++++++ .../i386/avx512f-round-sfix-vec-2.c | 7 +++ .../i386/avx512f-roundf-sfix-vec-1.c | 52 ++++++++++++++++++ .../i386/avx512f-roundf-sfix-vec-2.c | 7 +++ .../gcc.target/i386/avx512f-trunc-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-trunc-vec-2.c | 6 +++ .../gcc.target/i386/avx512f-truncf-vec-1.c | 52 ++++++++++++++++++ .../gcc.target/i386/avx512f-truncf-vec-2.c | 6 +++ 29 files changed, 854 insertions(+), 10 deletions(-) create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floor-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floor-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-2.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-1.c create mode 100644 gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-2.c diff --git a/gcc/ChangeLog b/gcc/ChangeLog index d7eb4333d1d..4af0254ca98 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,38 @@ +2016-05-25 Ilya Verbin + + * config/i386/i386-builtin-types.def: Add V16SI_FTYPE_V16SF, + V8DF_FTYPE_V8DF_ROUND, V16SF_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF_ROUND. + * config/i386/i386.c (enum ix86_builtins): Add + IX86_BUILTIN_CVTPS2DQ512_MASK, IX86_BUILTIN_FLOORPS512, + IX86_BUILTIN_FLOORPD512, IX86_BUILTIN_CEILPS512, IX86_BUILTIN_CEILPD512, + IX86_BUILTIN_TRUNCPS512, IX86_BUILTIN_TRUNCPD512, + IX86_BUILTIN_CVTPS2DQ512, IX86_BUILTIN_VEC_PACK_SFIX512, + IX86_BUILTIN_FLOORPS_SFIX512, IX86_BUILTIN_CEILPS_SFIX512, + IX86_BUILTIN_ROUNDPS_AZ_SFIX512. + (builtin_description bdesc_args): Add __builtin_ia32_floorps512, + __builtin_ia32_ceilps512, __builtin_ia32_truncps512, + __builtin_ia32_floorpd512, __builtin_ia32_ceilpd512, + __builtin_ia32_truncpd512, __builtin_ia32_cvtps2dq512, + __builtin_ia32_vec_pack_sfix512, __builtin_ia32_roundps_az_sfix512, + __builtin_ia32_floorps_sfix512, __builtin_ia32_ceilps_sfix512. + Change IX86_BUILTIN_CVTPS2DQ512 to IX86_BUILTIN_CVTPS2DQ512_MASK for + __builtin_ia32_cvtps2dq512_mask. + (ix86_expand_args_builtin): Handle V8DF_FTYPE_V8DF_ROUND, + V16SF_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF_ROUND, V16SI_FTYPE_V16SF. + (ix86_builtin_vectorized_function): Handle builtins mentioned above. + * config/i386/sse.md + (avx512f_fix_notruncv16sfv16si): + Rename to ... + (avx512f_fix_notruncv16sfv16si): ... this. + (avx512f_cvtpd2dq512): Rename + to ... + (avx512f_cvtpd2dq512): ... this. + (avx512f_vec_pack_sfix_v8df): New define_expand. + (avx512f_roundpd512): Rename to ... + (avx512f_round512): ... this. Change iterator. + (avx512f_roundps512_sfix): New define_expand. + (round2_sfix): Change iterator. + 2016-05-25 Nick Clifton * config/msp430/msp430.c (msp430_attr): Produce an error if a diff --git a/gcc/config/i386/i386-builtin-types.def b/gcc/config/i386/i386-builtin-types.def index 75d57d968c7..c66f65108f8 100644 --- a/gcc/config/i386/i386-builtin-types.def +++ b/gcc/config/i386/i386-builtin-types.def @@ -294,6 +294,7 @@ DEF_FUNCTION_TYPE (V8DF, V4DF) DEF_FUNCTION_TYPE (V8DF, V2DF) DEF_FUNCTION_TYPE (V16SI, V4SI) DEF_FUNCTION_TYPE (V16SI, V8SI) +DEF_FUNCTION_TYPE (V16SI, V16SF) DEF_FUNCTION_TYPE (V16SI, V16SI, V16SI, UHI) DEF_FUNCTION_TYPE (V8DI, V8DI, V8DI, UQI) DEF_FUNCTION_TYPE (V8DI, PV8DI) @@ -1061,14 +1062,17 @@ DEF_FUNCTION_TYPE (VOID, QI, V8DI, PCINT, INT, INT) DEF_FUNCTION_TYPE_ALIAS (V2DF_FTYPE_V2DF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V4DF_FTYPE_V4DF, ROUND) +DEF_FUNCTION_TYPE_ALIAS (V8DF_FTYPE_V8DF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V4SF_FTYPE_V4SF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V8SF_FTYPE_V8SF, ROUND) +DEF_FUNCTION_TYPE_ALIAS (V16SF_FTYPE_V16SF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V4SI_FTYPE_V2DF_V2DF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V8SI_FTYPE_V4DF_V4DF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V16SI_FTYPE_V8DF_V8DF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V4SI_FTYPE_V4SF, ROUND) DEF_FUNCTION_TYPE_ALIAS (V8SI_FTYPE_V8SF, ROUND) +DEF_FUNCTION_TYPE_ALIAS (V16SI_FTYPE_V16SF, ROUND) DEF_FUNCTION_TYPE_ALIAS (INT_FTYPE_V2DF_V2DF, PTEST) DEF_FUNCTION_TYPE_ALIAS (INT_FTYPE_V2DI_V2DI, PTEST) diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index 51364f3f50f..9a75bed92d0 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -30935,7 +30935,7 @@ enum ix86_builtins IX86_BUILTIN_CVTPD2PS512, IX86_BUILTIN_CVTPD2UDQ512, IX86_BUILTIN_CVTPH2PS512, - IX86_BUILTIN_CVTPS2DQ512, + IX86_BUILTIN_CVTPS2DQ512_MASK, IX86_BUILTIN_CVTPS2PD512, IX86_BUILTIN_CVTPS2PH512, IX86_BUILTIN_CVTPS2UDQ512, @@ -32375,14 +32375,25 @@ enum ix86_builtins IX86_BUILTIN_COPYSIGNQ, /* Vectorizer support builtins. */ - IX86_BUILTIN_CEILPD_VEC_PACK_SFIX512, IX86_BUILTIN_CPYSGNPS, IX86_BUILTIN_CPYSGNPD, IX86_BUILTIN_CPYSGNPS256, IX86_BUILTIN_CPYSGNPS512, IX86_BUILTIN_CPYSGNPD256, IX86_BUILTIN_CPYSGNPD512, + IX86_BUILTIN_FLOORPS512, + IX86_BUILTIN_FLOORPD512, + IX86_BUILTIN_CEILPS512, + IX86_BUILTIN_CEILPD512, + IX86_BUILTIN_TRUNCPS512, + IX86_BUILTIN_TRUNCPD512, + IX86_BUILTIN_CVTPS2DQ512, + IX86_BUILTIN_VEC_PACK_SFIX512, + IX86_BUILTIN_FLOORPS_SFIX512, IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX512, + IX86_BUILTIN_CEILPS_SFIX512, + IX86_BUILTIN_CEILPD_VEC_PACK_SFIX512, + IX86_BUILTIN_ROUNDPS_AZ_SFIX512, IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX512, @@ -34197,6 +34208,17 @@ static const struct builtin_description bdesc_args[] = { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_sqrtv8df2, "__builtin_ia32_sqrtpd512", IX86_BUILTIN_SQRTPD512, UNKNOWN, (int) V8DF_FTYPE_V8DF }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_sqrtv16sf2, "__builtin_ia32_sqrtps512", IX86_BUILTIN_SQRTPS_NR512, UNKNOWN, (int) V16SF_FTYPE_V16SF }, { OPTION_MASK_ISA_AVX512ER, CODE_FOR_avx512er_exp2v16sf, "__builtin_ia32_exp2ps", IX86_BUILTIN_EXP2PS, UNKNOWN, (int) V16SF_FTYPE_V16SF }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundps512, "__builtin_ia32_floorps512", IX86_BUILTIN_FLOORPS512, (enum rtx_code) ROUND_FLOOR, (int) V16SF_FTYPE_V16SF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundps512, "__builtin_ia32_ceilps512", IX86_BUILTIN_CEILPS512, (enum rtx_code) ROUND_CEIL, (int) V16SF_FTYPE_V16SF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundps512, "__builtin_ia32_truncps512", IX86_BUILTIN_TRUNCPS512, (enum rtx_code) ROUND_TRUNC, (int) V16SF_FTYPE_V16SF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundpd512, "__builtin_ia32_floorpd512", IX86_BUILTIN_FLOORPD512, (enum rtx_code) ROUND_FLOOR, (int) V8DF_FTYPE_V8DF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundpd512, "__builtin_ia32_ceilpd512", IX86_BUILTIN_CEILPD512, (enum rtx_code) ROUND_CEIL, (int) V8DF_FTYPE_V8DF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundpd512, "__builtin_ia32_truncpd512", IX86_BUILTIN_TRUNCPD512, (enum rtx_code) ROUND_TRUNC, (int) V8DF_FTYPE_V8DF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fix_notruncv16sfv16si, "__builtin_ia32_cvtps2dq512", IX86_BUILTIN_CVTPS2DQ512, UNKNOWN, (int) V16SI_FTYPE_V16SF }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_vec_pack_sfix_v8df, "__builtin_ia32_vec_pack_sfix512", IX86_BUILTIN_VEC_PACK_SFIX512, UNKNOWN, (int) V16SI_FTYPE_V8DF_V8DF }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_roundv16sf2_sfix, "__builtin_ia32_roundps_az_sfix512", IX86_BUILTIN_ROUNDPS_AZ_SFIX512, UNKNOWN, (int) V16SI_FTYPE_V16SF }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundps512_sfix, "__builtin_ia32_floorps_sfix512", IX86_BUILTIN_FLOORPS_SFIX512, (enum rtx_code) ROUND_FLOOR, (int) V16SI_FTYPE_V16SF_ROUND }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundps512_sfix, "__builtin_ia32_ceilps_sfix512", IX86_BUILTIN_CEILPS_SFIX512, (enum rtx_code) ROUND_CEIL, (int) V16SI_FTYPE_V16SF_ROUND }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_roundv8df2_vec_pack_sfix, "__builtin_ia32_roundpd_az_vec_pack_sfix512", IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX512, UNKNOWN, (int) V16SI_FTYPE_V8DF_V8DF }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundpd_vec_pack_sfix512, "__builtin_ia32_floorpd_vec_pack_sfix512", IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX512, (enum rtx_code) ROUND_FLOOR, (int) V16SI_FTYPE_V8DF_V8DF_ROUND }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_roundpd_vec_pack_sfix512, "__builtin_ia32_ceilpd_vec_pack_sfix512", IX86_BUILTIN_CEILPD_VEC_PACK_SFIX512, (enum rtx_code) ROUND_CEIL, (int) V16SI_FTYPE_V8DF_V8DF_ROUND }, @@ -35113,7 +35135,7 @@ static const struct builtin_description bdesc_round_args[] = { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_cvtpd2ps512_mask_round, "__builtin_ia32_cvtpd2ps512_mask", IX86_BUILTIN_CVTPD2PS512, UNKNOWN, (int) V8SF_FTYPE_V8DF_V8SF_QI_INT }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_ufix_notruncv8dfv8si2_mask_round, "__builtin_ia32_cvtpd2udq512_mask", IX86_BUILTIN_CVTPD2UDQ512, UNKNOWN, (int) V8SI_FTYPE_V8DF_V8SI_QI_INT }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_vcvtph2ps512_mask_round, "__builtin_ia32_vcvtph2ps512_mask", IX86_BUILTIN_CVTPH2PS512, UNKNOWN, (int) V16SF_FTYPE_V16HI_V16SF_HI_INT }, - { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fix_notruncv16sfv16si_mask_round, "__builtin_ia32_cvtps2dq512_mask", IX86_BUILTIN_CVTPS2DQ512, UNKNOWN, (int) V16SI_FTYPE_V16SF_V16SI_HI_INT }, + { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fix_notruncv16sfv16si_mask_round, "__builtin_ia32_cvtps2dq512_mask", IX86_BUILTIN_CVTPS2DQ512_MASK, UNKNOWN, (int) V16SI_FTYPE_V16SF_V16SI_HI_INT }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_cvtps2pd512_mask_round, "__builtin_ia32_cvtps2pd512_mask", IX86_BUILTIN_CVTPS2PD512, UNKNOWN, (int) V8DF_FTYPE_V8SF_V8DF_QI_INT }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_ufix_notruncv16sfv16si_mask_round, "__builtin_ia32_cvtps2udq512_mask", IX86_BUILTIN_CVTPS2UDQ512, UNKNOWN, (int) V16SI_FTYPE_V16SF_V16SI_HI_INT }, { OPTION_MASK_ISA_AVX512F, CODE_FOR_sse2_cvtsd2ss_round, "__builtin_ia32_cvtsd2ss_round", IX86_BUILTIN_CVTSD2SS_ROUND, UNKNOWN, (int) V4SF_FTYPE_V4SF_V2DF_INT }, @@ -38690,10 +38712,13 @@ ix86_expand_args_builtin (const struct builtin_description *d, { case V2DF_FTYPE_V2DF_ROUND: case V4DF_FTYPE_V4DF_ROUND: + case V8DF_FTYPE_V8DF_ROUND: case V4SF_FTYPE_V4SF_ROUND: case V8SF_FTYPE_V8SF_ROUND: + case V16SF_FTYPE_V16SF_ROUND: case V4SI_FTYPE_V4SF_ROUND: case V8SI_FTYPE_V8SF_ROUND: + case V16SI_FTYPE_V16SF_ROUND: return ix86_expand_sse_round (d, exp, target); case V4SI_FTYPE_V2DF_V2DF_ROUND: case V8SI_FTYPE_V4DF_V4DF_ROUND: @@ -38807,6 +38832,7 @@ ix86_expand_args_builtin (const struct builtin_description *d, case V16SI_FTYPE_V8SI: case V16SF_FTYPE_V4SF: case V16SI_FTYPE_V4SI: + case V16SI_FTYPE_V16SF: case V16SF_FTYPE_V16SF: case V8DI_FTYPE_UQI: case V8DF_FTYPE_V4DF: @@ -42536,6 +42562,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_FLOORPS_SFIX); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_FLOORPS_SFIX256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_FLOORPS_SFIX512); } break; @@ -42561,6 +42589,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_CEILPS_SFIX); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_CEILPS_SFIX256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_CEILPS_SFIX512); } break; @@ -42573,6 +42603,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_VEC_PACK_SFIX); else if (out_n == 8 && in_n == 4) return ix86_get_builtin (IX86_BUILTIN_VEC_PACK_SFIX256); + else if (out_n == 16 && in_n == 8) + return ix86_get_builtin (IX86_BUILTIN_VEC_PACK_SFIX512); } if (out_mode == SImode && in_mode == SFmode) { @@ -42580,6 +42612,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_CVTPS2DQ); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_CVTPS2DQ256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_CVTPS2DQ512); } break; @@ -42605,6 +42639,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ_SFIX); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ_SFIX256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ_SFIX512); } break; @@ -42619,6 +42655,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_FLOORPD); else if (out_n == 4 && in_n == 4) return ix86_get_builtin (IX86_BUILTIN_FLOORPD256); + else if (out_n == 8 && in_n == 8) + return ix86_get_builtin (IX86_BUILTIN_FLOORPD512); } if (out_mode == SFmode && in_mode == SFmode) { @@ -42626,6 +42664,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_FLOORPS); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_FLOORPS256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_FLOORPS512); } break; @@ -42640,6 +42680,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_CEILPD); else if (out_n == 4 && in_n == 4) return ix86_get_builtin (IX86_BUILTIN_CEILPD256); + else if (out_n == 8 && in_n == 8) + return ix86_get_builtin (IX86_BUILTIN_CEILPD512); } if (out_mode == SFmode && in_mode == SFmode) { @@ -42647,6 +42689,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_CEILPS); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_CEILPS256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_CEILPS512); } break; @@ -42661,6 +42705,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_TRUNCPD); else if (out_n == 4 && in_n == 4) return ix86_get_builtin (IX86_BUILTIN_TRUNCPD256); + else if (out_n == 8 && in_n == 8) + return ix86_get_builtin (IX86_BUILTIN_TRUNCPD512); } if (out_mode == SFmode && in_mode == SFmode) { @@ -42668,6 +42714,8 @@ ix86_builtin_vectorized_function (unsigned int fn, tree type_out, return ix86_get_builtin (IX86_BUILTIN_TRUNCPS); else if (out_n == 8 && in_n == 8) return ix86_get_builtin (IX86_BUILTIN_TRUNCPS256); + else if (out_n == 16 && in_n == 16) + return ix86_get_builtin (IX86_BUILTIN_TRUNCPS512); } break; diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 3c6099808be..2297ca2fd77 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -4242,7 +4242,7 @@ (set_attr "prefix" "maybe_vex") (set_attr "mode" "")]) -(define_insn "avx512f_fix_notruncv16sfv16si" +(define_insn "avx512f_fix_notruncv16sfv16si" [(set (match_operand:V16SI 0 "register_operand" "=v") (unspec:V16SI [(match_operand:V16SF 1 "" "")] @@ -4799,7 +4799,7 @@ (set_attr "prefix" "maybe_vex") (set_attr "mode" "V2DF")]) -(define_insn "avx512f_cvtpd2dq512" +(define_insn "avx512f_cvtpd2dq512" [(set (match_operand:V8SI 0 "register_operand" "=v") (unspec:V8SI [(match_operand:V8DF 1 "" "")] @@ -5802,6 +5802,23 @@ DONE; }) +(define_expand "avx512f_vec_pack_sfix_v8df" + [(match_operand:V16SI 0 "register_operand") + (match_operand:V8DF 1 "nonimmediate_operand") + (match_operand:V8DF 2 "nonimmediate_operand")] + "TARGET_AVX512F" +{ + rtx r1, r2; + + r1 = gen_reg_rtx (V8SImode); + r2 = gen_reg_rtx (V8SImode); + + emit_insn (gen_avx512f_cvtpd2dq512 (r1, operands[1])); + emit_insn (gen_avx512f_cvtpd2dq512 (r2, operands[2])); + emit_insn (gen_avx_vec_concatv16si (operands[0], r1, r2)); + DONE; +}) + (define_expand "vec_pack_sfix_v4df" [(match_operand:V8SI 0 "register_operand") (match_operand:V4DF 1 "nonimmediate_operand") @@ -15027,13 +15044,25 @@ DONE; }) -(define_expand "avx512f_roundpd512" - [(match_operand:V8DF 0 "register_operand") - (match_operand:V8DF 1 "nonimmediate_operand") +(define_expand "avx512f_round512" + [(match_operand:VF_512 0 "register_operand") + (match_operand:VF_512 1 "nonimmediate_operand") + (match_operand:SI 2 "const_0_to_15_operand")] + "TARGET_AVX512F" +{ + emit_insn (gen_avx512f_rndscale (operands[0], operands[1], operands[2])); + DONE; +}) + +(define_expand "avx512f_roundps512_sfix" + [(match_operand:V16SI 0 "register_operand") + (match_operand:V16SF 1 "nonimmediate_operand") (match_operand:SI 2 "const_0_to_15_operand")] "TARGET_AVX512F" { - emit_insn (gen_avx512f_rndscalev8df (operands[0], operands[1], operands[2])); + rtx tmp = gen_reg_rtx (V16SFmode); + emit_insn (gen_avx512f_rndscalev16sf (tmp, operands[1], operands[2])); + emit_insn (gen_fix_truncv16sfv16si2 (operands[0], tmp)); DONE; }) @@ -15134,7 +15163,7 @@ (define_expand "round2_sfix" [(match_operand: 0 "register_operand") - (match_operand:VF1_128_256 1 "register_operand")] + (match_operand:VF1 1 "register_operand")] "TARGET_ROUND && !flag_trapping_math" { rtx tmp = gen_reg_rtx (mode); diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 1b1b97e2c73..8070adb9df0 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,30 @@ +2016-05-25 Ilya Verbin + + * gcc.target/i386/avx512f-ceil-vec-1.c: New test. + * gcc.target/i386/avx512f-ceil-vec-2.c: New test. + * gcc.target/i386/avx512f-ceilf-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-ceilf-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-ceilf-vec-1.c: New test. + * gcc.target/i386/avx512f-ceilf-vec-2.c: New test. + * gcc.target/i386/avx512f-floor-vec-1.c: New test. + * gcc.target/i386/avx512f-floor-vec-2.c: New test. + * gcc.target/i386/avx512f-floorf-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-floorf-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-floorf-vec-1.c: New test. + * gcc.target/i386/avx512f-floorf-vec-2.c: New test. + * gcc.target/i386/avx512f-rint-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-rint-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-rintf-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-rintf-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-round-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-round-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-roundf-sfix-vec-1.c: New test. + * gcc.target/i386/avx512f-roundf-sfix-vec-2.c: New test. + * gcc.target/i386/avx512f-trunc-vec-1.c: New test. + * gcc.target/i386/avx512f-trunc-vec-2.c: New test. + * gcc.target/i386/avx512f-truncf-vec-1.c: New test. + * gcc.target/i386/avx512f-truncf-vec-2.c: New test. + 2016-05-25 Marek Polacek PR c/71265 diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-1.c new file mode 100644 index 00000000000..fc48b1572b9 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (double *src) +{ + int i, sign = 1; + double f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + double a[NUM]; + double r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = ceil (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != ceil (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-2.c new file mode 100644 index 00000000000..bf8af064cfe --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceil-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-ceil-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscalepd\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-1.c new file mode 100644 index 00000000000..c6d53d89fc6 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) ceilf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) ceilf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-2.c new file mode 100644 index 00000000000..80e594dbfa4 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-sfix-vec-2.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-ceilf-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vcvttps2dq\[^\n\]+zmm\[0-9\].{7}(?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-1.c new file mode 100644 index 00000000000..4788825fc3d --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + float r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = ceilf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != ceilf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-2.c new file mode 100644 index 00000000000..95a79e29d12 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-ceilf-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-ceilf-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-1.c new file mode 100644 index 00000000000..b7cbed005dd --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (double *src) +{ + int i, sign = 1; + double f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + double a[NUM]; + double r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = floor (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != floor (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-2.c new file mode 100644 index 00000000000..0d401f78d63 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floor-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-floor-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscalepd\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-1.c new file mode 100644 index 00000000000..6a25f438a8e --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) floorf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) floorf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-2.c new file mode 100644 index 00000000000..f4bfec5385e --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floorf-sfix-vec-2.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-floorf-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vcvttps2dq\[^\n\]+zmm\[0-9\].{7}(?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-1.c new file mode 100644 index 00000000000..69fc73d78ef --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + float r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = floorf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != floorf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-2.c new file mode 100644 index 00000000000..90c6c0fade5 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-floorf-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-floorf-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-1.c new file mode 100644 index 00000000000..8e1745aa13a --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (double *src) +{ + int i, sign = 1; + double f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + double a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) rint (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) rint (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-2.c new file mode 100644 index 00000000000..c3f78ac3f25 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-rint-sfix-vec-2.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-rint-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vcvtpd2dq\[^\n\]+ymm\[0-9\](?:\n|\[ \\t\]+#)" 2 } } */ +/* { dg-final { scan-assembler-times "vinserti64x4\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-1.c new file mode 100644 index 00000000000..ac3e9a25973 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) rintf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) rintf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-2.c new file mode 100644 index 00000000000..c172e61f84a --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-rintf-sfix-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-rintf-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vcvtps2dq\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-1.c new file mode 100644 index 00000000000..61bea578e18 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (double *src) +{ + int i, sign = 1; + double f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + double a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) round (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) round (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-2.c new file mode 100644 index 00000000000..5982c65d1e1 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-round-sfix-vec-2.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-round-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscalepd\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 2 } } */ +/* { dg-final { scan-assembler-times "vcvttpd2dq\[^\n\]+zmm\[0-9\].{7}(?:\n|\[ \\t\]+#)" 2 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-1.c new file mode 100644 index 00000000000..c5ec9e7ec00 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + int r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = (int) roundf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != (int) roundf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-2.c new file mode 100644 index 00000000000..0d8abb892d1 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-roundf-sfix-vec-2.c @@ -0,0 +1,7 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-roundf-sfix-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vcvttps2dq\[^\n\]+zmm\[0-9\].{7}(?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-1.c new file mode 100644 index 00000000000..dfb93d72324 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (double *src) +{ + int i, sign = 1; + double f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + double a[NUM]; + double r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = trunc (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != trunc (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-2.c new file mode 100644 index 00000000000..e8ec0227653 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-trunc-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-trunc-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscalepd\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-1.c b/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-1.c new file mode 100644 index 00000000000..db13e712829 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-1.c @@ -0,0 +1,52 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ +/* { dg-require-effective-target avx512f } */ +/* { dg-skip-if "no M_PI" { vxworks_kernel } } */ + +#define __NO_MATH_INLINES +#include +#include "avx512f-check.h" + +#define NUM 64 + +static void +__attribute__((__target__("fpmath=sse"))) +init_src (float *src) +{ + int i, sign = 1; + float f = rand (); + + for (i = 0; i < NUM; i++) + { + src[i] = (i + 1) * f * M_PI * sign; + if (i < (NUM / 2)) + { + if ((i % 6) == 0) + f = f * src[i]; + } + else if (i == (NUM / 2)) + f = rand (); + else if ((i % 6) == 0) + f = 1 / (f * (i + 1) * src[i] * M_PI * sign); + sign = -sign; + } +} + +static void +__attribute__((__target__("fpmath=387"))) +avx512f_test (void) +{ + float a[NUM]; + float r[NUM]; + int i; + + init_src (a); + + for (i = 0; i < NUM; i++) + r[i] = truncf (a[i]); + + /* check results: */ + for (i = 0; i < NUM; i++) + if (r[i] != truncf (a[i])) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-2.c b/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-2.c new file mode 100644 index 00000000000..ae542d8276b --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avx512f-truncf-vec-2.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -ffast-math -ftree-vectorize -mavx512f" } */ + +#include "avx512f-truncf-vec-1.c" + +/* { dg-final { scan-assembler-times "vrndscaleps\[^\n\]+zmm\[0-9\](?:\n|\[ \\t\]+#)" 1 } } */ -- 2.30.2