From f56ea76ecbc6b5bc9f6e11252cd1e8b4a566c66d Mon Sep 17 00:00:00 2001 From: Uros Bizjak Date: Thu, 20 Oct 2016 19:53:10 +0200 Subject: [PATCH] re PR target/78037 (Incorrect code generated at optimization level -O2 for tzcnt and binary and) PR target/78037 * config/i386/bmiintrin.h (__tzcnt_u16): Call __builtin_ia32_tzcnt_u16. (__tzcnt_u32, _tzcnt_u32): Call __builtin_ia32_tzcnt_u32. (__tzcnt_u64, _tzcnt_u64): Call __builtin_ia32_tzcnt_u64. * config/i386/lzcntintrin.h (__lzcnt_u16): Call __builtin_ia32_lzcnt_u16. (__lzcnt_u32, _lzcnt_u32): Call __builtin_ia32_lzcnt_u32. (__lzcnt_u64, _lzcnt_u64): Call __builtin_ia32_lzcnt_u64. * config/i386/i386.md (UNSPEC_LZCNT, UNSPEC_TZCNT): New unspecs. (ctz2, *ctz2): Use SWI48 mode iterator. (bmi_tzcnt_): New expander. (*bmi_tzcnt__falsedep_1): New define_insn_and_split pattern. (*bmi_tzcnt__falsedep, *bmi_tzcnt_): New insn patterns. (clz2_lzcnt, *clz2_lzcnt): Use SWI48 mode iterator. (lzcnt_): New expander. (*lzcnt__falsedep_1): New define_insn_and_split pattern. (*lzcnt__falsedep, *lzcnt_): New insn patterns. * config/i386/i386-builtin-types.def (UINT_FTYPE_UINT): New. (UINT64_FTYPE_UINT64): New. * config/i386/i386-builtin.def (__builtin_clzs): Remove description. (__builtin_ia32_lzcnt_u16): New description. (__builtin_ia32_lzcnt_u32): Ditto. (__builtin_ia32_lzcnt_u64): Ditto. (__builtin_ctzs): Remove description. (__builtin_ia32_tzcnt_u16): New description. (__builtin_ia32_tzcnt_u32): Ditto. (__builtin_ia32_tzcnt_u64): Ditto. * config/i386/i386.c (ix86_expand_args_builtin): Handle UINT_FTYPE_UINT and UINT64_FTYPE_UINT64. testsuite/ChangeLog: PR target/78037 * gcc.target/i386/pr78037.c: New test. From-SVN: r241381 --- gcc/ChangeLog | 32 +++++ gcc/config/i386/bmiintrin.h | 10 +- gcc/config/i386/i386-builtin-types.def | 4 +- gcc/config/i386/i386-builtin.def | 10 +- gcc/config/i386/i386.c | 4 +- gcc/config/i386/i386.md | 158 +++++++++++++++++++++--- gcc/config/i386/lzcntintrin.h | 10 +- gcc/testsuite/ChangeLog | 5 + gcc/testsuite/gcc.target/i386/pr78037.c | 21 ++++ 9 files changed, 226 insertions(+), 28 deletions(-) create mode 100644 gcc/testsuite/gcc.target/i386/pr78037.c diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 1d06d98cfcb..7eb9476f52f 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,35 @@ +2016-10-20 Uros Bizjak + + PR target/78037 + * config/i386/bmiintrin.h (__tzcnt_u16): Call __builtin_ia32_tzcnt_u16. + (__tzcnt_u32, _tzcnt_u32): Call __builtin_ia32_tzcnt_u32. + (__tzcnt_u64, _tzcnt_u64): Call __builtin_ia32_tzcnt_u64. + * config/i386/lzcntintrin.h (__lzcnt_u16): Call + __builtin_ia32_lzcnt_u16. + (__lzcnt_u32, _lzcnt_u32): Call __builtin_ia32_lzcnt_u32. + (__lzcnt_u64, _lzcnt_u64): Call __builtin_ia32_lzcnt_u64. + * config/i386/i386.md (UNSPEC_LZCNT, UNSPEC_TZCNT): New unspecs. + (ctz2, *ctz2): Use SWI48 mode iterator. + (bmi_tzcnt_): New expander. + (*bmi_tzcnt__falsedep_1): New define_insn_and_split pattern. + (*bmi_tzcnt__falsedep, *bmi_tzcnt_): New insn patterns. + (clz2_lzcnt, *clz2_lzcnt): Use SWI48 mode iterator. + (lzcnt_): New expander. + (*lzcnt__falsedep_1): New define_insn_and_split pattern. + (*lzcnt__falsedep, *lzcnt_): New insn patterns. + * config/i386/i386-builtin-types.def (UINT_FTYPE_UINT): New. + (UINT64_FTYPE_UINT64): New. + * config/i386/i386-builtin.def (__builtin_clzs): Remove description. + (__builtin_ia32_lzcnt_u16): New description. + (__builtin_ia32_lzcnt_u32): Ditto. + (__builtin_ia32_lzcnt_u64): Ditto. + (__builtin_ctzs): Remove description. + (__builtin_ia32_tzcnt_u16): New description. + (__builtin_ia32_tzcnt_u32): Ditto. + (__builtin_ia32_tzcnt_u64): Ditto. + * config/i386/i386.c (ix86_expand_args_builtin): Handle + UINT_FTYPE_UINT and UINT64_FTYPE_UINT64. + 2016-10-20 Martin Liska PR lto/78049 diff --git a/gcc/config/i386/bmiintrin.h b/gcc/config/i386/bmiintrin.h index a838a927be9..3ebe6d7d508 100644 --- a/gcc/config/i386/bmiintrin.h +++ b/gcc/config/i386/bmiintrin.h @@ -37,7 +37,7 @@ extern __inline unsigned short __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __tzcnt_u16 (unsigned short __X) { - return __builtin_ctzs (__X); + return __builtin_ia32_tzcnt_u16 (__X); } extern __inline unsigned int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) @@ -97,13 +97,13 @@ _blsr_u32 (unsigned int __X) extern __inline unsigned int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __tzcnt_u32 (unsigned int __X) { - return __builtin_ctz (__X); + return __builtin_ia32_tzcnt_u32 (__X); } extern __inline unsigned int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) _tzcnt_u32 (unsigned int __X) { - return __builtin_ctz (__X); + return __builtin_ia32_tzcnt_u32 (__X); } @@ -165,13 +165,13 @@ _blsr_u64 (unsigned long long __X) extern __inline unsigned long long __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __tzcnt_u64 (unsigned long long __X) { - return __builtin_ctzll (__X); + return __builtin_ia32_tzcnt_u64 (__X); } extern __inline unsigned long long __attribute__((__gnu_inline__, __always_inline__, __artificial__)) _tzcnt_u64 (unsigned long long __X) { - return __builtin_ctzll (__X); + return __builtin_ia32_tzcnt_u64 (__X); } #endif /* __x86_64__ */ diff --git a/gcc/config/i386/i386-builtin-types.def b/gcc/config/i386/i386-builtin-types.def index 7eb6fc96e66..b34cfdabd4d 100644 --- a/gcc/config/i386/i386-builtin-types.def +++ b/gcc/config/i386/i386-builtin-types.def @@ -205,9 +205,11 @@ DEF_FUNCTION_TYPE (INT, PCCHAR) DEF_FUNCTION_TYPE (INT64, INT64) DEF_FUNCTION_TYPE (INT64, V2DF) DEF_FUNCTION_TYPE (INT64, V4SF) -DEF_FUNCTION_TYPE (UINT64, INT) +DEF_FUNCTION_TYPE (UINT, UINT) DEF_FUNCTION_TYPE (UINT16, UINT16) +DEF_FUNCTION_TYPE (UINT64, INT) DEF_FUNCTION_TYPE (UINT64, PUNSIGNED) +DEF_FUNCTION_TYPE (UINT64, UINT64) DEF_FUNCTION_TYPE (V16QI, PCCHAR) DEF_FUNCTION_TYPE (V16QI, V16QI) DEF_FUNCTION_TYPE (V2DF, PCDOUBLE) diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def index 8518b7445de..227526bcd72 100644 --- a/gcc/config/i386/i386-builtin.def +++ b/gcc/config/i386/i386-builtin.def @@ -1186,12 +1186,18 @@ BDESC (OPTION_MASK_ISA_AVX2, CODE_FOR_avx2_lshrvv2di, "__builtin_ia32_psrlv2di", BDESC (OPTION_MASK_ISA_AVX2, CODE_FOR_avx2_lshrvv8si, "__builtin_ia32_psrlv8si", IX86_BUILTIN_PSRLVV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI) BDESC (OPTION_MASK_ISA_AVX2, CODE_FOR_avx2_lshrvv4si, "__builtin_ia32_psrlv4si", IX86_BUILTIN_PSRLVV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI) -BDESC (OPTION_MASK_ISA_LZCNT, CODE_FOR_clzhi2_lzcnt, "__builtin_clzs", IX86_BUILTIN_CLZS, UNKNOWN, (int) UINT16_FTYPE_UINT16) +/* LZCNT */ +BDESC (OPTION_MASK_ISA_LZCNT, CODE_FOR_lzcnt_hi, "__builtin_ia32_lzcnt_u16", IX86_BUILTIN_LZCNT16, UNKNOWN, (int) UINT16_FTYPE_UINT16) +BDESC (OPTION_MASK_ISA_LZCNT, CODE_FOR_lzcnt_si, "__builtin_ia32_lzcnt_u32", IX86_BUILTIN_LZCNT32, UNKNOWN, (int) UINT_FTYPE_UINT) +BDESC (OPTION_MASK_ISA_LZCNT | OPTION_MASK_ISA_64BIT, CODE_FOR_lzcnt_di, "__builtin_ia32_lzcnt_u64", IX86_BUILTIN_LZCNT64, UNKNOWN, (int) UINT64_FTYPE_UINT64) /* BMI */ BDESC (OPTION_MASK_ISA_BMI, CODE_FOR_bmi_bextr_si, "__builtin_ia32_bextr_u32", IX86_BUILTIN_BEXTR32, UNKNOWN, (int) UINT_FTYPE_UINT_UINT) BDESC (OPTION_MASK_ISA_BMI | OPTION_MASK_ISA_64BIT, CODE_FOR_bmi_bextr_di, "__builtin_ia32_bextr_u64", IX86_BUILTIN_BEXTR64, UNKNOWN, (int) UINT64_FTYPE_UINT64_UINT64) -BDESC (OPTION_MASK_ISA_BMI, CODE_FOR_ctzhi2, "__builtin_ctzs", IX86_BUILTIN_CTZS, UNKNOWN, (int) UINT16_FTYPE_UINT16) + +BDESC (OPTION_MASK_ISA_BMI, CODE_FOR_bmi_tzcnt_hi, "__builtin_ia32_tzcnt_u16", IX86_BUILTIN_TZCNT16, UNKNOWN, (int) UINT16_FTYPE_UINT16) +BDESC (OPTION_MASK_ISA_BMI, CODE_FOR_bmi_tzcnt_si, "__builtin_ia32_tzcnt_u32", IX86_BUILTIN_TZCNT32, UNKNOWN, (int) UINT_FTYPE_UINT) +BDESC (OPTION_MASK_ISA_BMI | OPTION_MASK_ISA_64BIT, CODE_FOR_bmi_tzcnt_di, "__builtin_ia32_tzcnt_u64", IX86_BUILTIN_TZCNT64, UNKNOWN, (int) UINT64_FTYPE_UINT64) /* TBM */ BDESC (OPTION_MASK_ISA_TBM, CODE_FOR_tbm_bextri_si, "__builtin_ia32_bextri_u32", IX86_BUILTIN_BEXTRI32, UNKNOWN, (int) UINT_FTYPE_UINT_UINT) diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index 5cac21d4141..3e6f8fde127 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -34317,8 +34317,10 @@ ix86_expand_args_builtin (const struct builtin_description *d, case FLOAT128_FTYPE_FLOAT128: case FLOAT_FTYPE_FLOAT: case INT_FTYPE_INT: - case UINT64_FTYPE_INT: + case UINT_FTYPE_UINT: case UINT16_FTYPE_UINT16: + case UINT64_FTYPE_INT: + case UINT64_FTYPE_UINT64: case INT64_FTYPE_INT64: case INT64_FTYPE_V4SF: case INT64_FTYPE_V2DF: diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md index 32794ac30ed..6e0348de7bc 100644 --- a/gcc/config/i386/i386.md +++ b/gcc/config/i386/i386.md @@ -175,7 +175,11 @@ ;; For CRC32 support UNSPEC_CRC32 + ;; For LZCNT suppoprt + UNSPEC_LZCNT + ;; For BMI support + UNSPEC_TZCNT UNSPEC_BEXTR ;; For BMI2 support @@ -12850,9 +12854,9 @@ (define_expand "ctz2" [(parallel - [(set (match_operand:SWI248 0 "register_operand") - (ctz:SWI248 - (match_operand:SWI248 1 "nonimmediate_operand"))) + [(set (match_operand:SWI48 0 "register_operand") + (ctz:SWI48 + (match_operand:SWI48 1 "nonimmediate_operand"))) (clobber (reg:CC FLAGS_REG))])]) ; False dependency happens when destination is only updated by tzcnt, @@ -12900,8 +12904,8 @@ (set_attr "mode" "")]) (define_insn "*ctz2" - [(set (match_operand:SWI248 0 "register_operand" "=r") - (ctz:SWI248 (match_operand:SWI248 1 "nonimmediate_operand" "rm"))) + [(set (match_operand:SWI48 0 "register_operand" "=r") + (ctz:SWI48 (match_operand:SWI48 1 "nonimmediate_operand" "rm"))) (clobber (reg:CC FLAGS_REG))] "" { @@ -12926,15 +12930,78 @@ (const_string "0"))) (set_attr "mode" "")]) +;; Version of tzcnt that is expanded from intrinsics. This version provides +;; operand size as output when source operand is zero. + +(define_expand "bmi_tzcnt_" + [(parallel + [(set (match_operand:SWI248 0 "register_operand") + (unspec:SWI248 + [(match_operand:SWI248 1 "nonimmediate_operand")] + UNSPEC_TZCNT)) + (clobber (reg:CC FLAGS_REG))])] + "TARGET_BMI") + +; False dependency happens when destination is only updated by tzcnt, +; lzcnt or popcnt. There is no false dependency when destination is +; also used in source. +(define_insn_and_split "*bmi_tzcnt__falsedep_1" + [(set (match_operand:SWI48 0 "register_operand" "=r") + (unspec:SWI48 + [(match_operand:SWI48 1 "nonimmediate_operand" "rm")] + UNSPEC_TZCNT)) + (clobber (reg:CC FLAGS_REG))] + "TARGET_BMI + && TARGET_AVOID_FALSE_DEP_FOR_BMI && optimize_function_for_speed_p (cfun)" + "#" + "&& reload_completed" + [(parallel + [(set (match_dup 0) + (unspec:SWI48 [(match_dup 1)] UNSPEC_TZCNT)) + (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP) + (clobber (reg:CC FLAGS_REG))])] +{ + if (!reg_mentioned_p (operands[0], operands[1])) + ix86_expand_clear (operands[0]); +}) + +(define_insn "*bmi_tzcnt__falsedep" + [(set (match_operand:SWI48 0 "register_operand" "=r") + (unspec:SWI48 + [(match_operand:SWI48 1 "nonimmediate_operand" "rm")] + UNSPEC_TZCNT)) + (unspec [(match_operand:SWI48 2 "register_operand" "0")] + UNSPEC_INSN_FALSE_DEP) + (clobber (reg:CC FLAGS_REG))] + "TARGET_BMI" + "tzcnt{}\t{%1, %0|%0, %1}" + [(set_attr "type" "alu1") + (set_attr "prefix_0f" "1") + (set_attr "prefix_rep" "1") + (set_attr "mode" "")]) + +(define_insn "*bmi_tzcnt_" + [(set (match_operand:SWI248 0 "register_operand" "=r") + (unspec:SWI248 + [(match_operand:SWI248 1 "nonimmediate_operand" "rm")] + UNSPEC_TZCNT)) + (clobber (reg:CC FLAGS_REG))] + "TARGET_BMI" + "tzcnt{}\t{%1, %0|%0, %1}" + [(set_attr "type" "alu1") + (set_attr "prefix_0f" "1") + (set_attr "prefix_rep" "1") + (set_attr "mode" "")]) + (define_expand "clz2" [(parallel - [(set (match_operand:SWI248 0 "register_operand") - (minus:SWI248 + [(set (match_operand:SWI48 0 "register_operand") + (minus:SWI48 (match_dup 2) - (clz:SWI248 (match_operand:SWI248 1 "nonimmediate_operand")))) + (clz:SWI48 (match_operand:SWI48 1 "nonimmediate_operand")))) (clobber (reg:CC FLAGS_REG))]) (parallel - [(set (match_dup 0) (xor:SWI248 (match_dup 0) (match_dup 2))) + [(set (match_dup 0) (xor:SWI48 (match_dup 0) (match_dup 2))) (clobber (reg:CC FLAGS_REG))])] "" { @@ -12948,9 +13015,9 @@ (define_expand "clz2_lzcnt" [(parallel - [(set (match_operand:SWI248 0 "register_operand") - (clz:SWI248 - (match_operand:SWI248 1 "nonimmediate_operand"))) + [(set (match_operand:SWI48 0 "register_operand") + (clz:SWI48 + (match_operand:SWI48 1 "nonimmediate_operand"))) (clobber (reg:CC FLAGS_REG))])] "TARGET_LZCNT") @@ -12987,8 +13054,8 @@ (set_attr "mode" "")]) (define_insn "*clz2_lzcnt" - [(set (match_operand:SWI248 0 "register_operand" "=r") - (clz:SWI248 (match_operand:SWI248 1 "nonimmediate_operand" "rm"))) + [(set (match_operand:SWI48 0 "register_operand" "=r") + (clz:SWI48 (match_operand:SWI48 1 "nonimmediate_operand" "rm"))) (clobber (reg:CC FLAGS_REG))] "TARGET_LZCNT" "lzcnt{}\t{%1, %0|%0, %1}" @@ -12996,6 +13063,69 @@ (set_attr "type" "bitmanip") (set_attr "mode" "")]) +;; Version of lzcnt that is expanded from intrinsics. This version provides +;; operand size as output when source operand is zero. + +(define_expand "lzcnt_" + [(parallel + [(set (match_operand:SWI248 0 "register_operand") + (unspec:SWI248 + [(match_operand:SWI248 1 "nonimmediate_operand")] + UNSPEC_LZCNT)) + (clobber (reg:CC FLAGS_REG))])] + "TARGET_LZCNT") + +; False dependency happens when destination is only updated by tzcnt, +; lzcnt or popcnt. There is no false dependency when destination is +; also used in source. +(define_insn_and_split "*lzcnt__falsedep_1" + [(set (match_operand:SWI48 0 "register_operand" "=r") + (unspec:SWI48 + [(match_operand:SWI48 1 "nonimmediate_operand" "rm")] + UNSPEC_LZCNT)) + (clobber (reg:CC FLAGS_REG))] + "TARGET_LZCNT + && TARGET_AVOID_FALSE_DEP_FOR_BMI && optimize_function_for_speed_p (cfun)" + "#" + "&& reload_completed" + [(parallel + [(set (match_dup 0) + (unspec:SWI48 [(match_dup 1)] UNSPEC_LZCNT)) + (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP) + (clobber (reg:CC FLAGS_REG))])] +{ + if (!reg_mentioned_p (operands[0], operands[1])) + ix86_expand_clear (operands[0]); +}) + +(define_insn "*lzcnt__falsedep" + [(set (match_operand:SWI48 0 "register_operand" "=r") + (unspec:SWI48 + [(match_operand:SWI48 1 "nonimmediate_operand" "rm")] + UNSPEC_LZCNT)) + (unspec [(match_operand:SWI48 2 "register_operand" "0")] + UNSPEC_INSN_FALSE_DEP) + (clobber (reg:CC FLAGS_REG))] + "TARGET_LZCNT" + "lzcnt{}\t{%1, %0|%0, %1}" + [(set_attr "type" "alu1") + (set_attr "prefix_0f" "1") + (set_attr "prefix_rep" "1") + (set_attr "mode" "")]) + +(define_insn "*lzcnt_" + [(set (match_operand:SWI248 0 "register_operand" "=r") + (unspec:SWI248 + [(match_operand:SWI248 1 "nonimmediate_operand" "rm")] + UNSPEC_LZCNT)) + (clobber (reg:CC FLAGS_REG))] + "TARGET_LZCNT" + "lzcnt{}\t{%1, %0|%0, %1}" + [(set_attr "type" "alu1") + (set_attr "prefix_0f" "1") + (set_attr "prefix_rep" "1") + (set_attr "mode" "")]) + ;; BMI instructions. (define_insn "*bmi_andn_" [(set (match_operand:SWI48 0 "register_operand" "=r,r") diff --git a/gcc/config/i386/lzcntintrin.h b/gcc/config/i386/lzcntintrin.h index 413267ac21c..6d9d3b2c046 100644 --- a/gcc/config/i386/lzcntintrin.h +++ b/gcc/config/i386/lzcntintrin.h @@ -38,32 +38,32 @@ extern __inline unsigned short __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __lzcnt16 (unsigned short __X) { - return __builtin_clzs (__X); + return __builtin_ia32_lzcnt_u16 (__X); } extern __inline unsigned int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __lzcnt32 (unsigned int __X) { - return __builtin_clz (__X); + return __builtin_ia32_lzcnt_u32 (__X); } extern __inline unsigned int __attribute__((__gnu_inline__, __always_inline__, __artificial__)) _lzcnt_u32 (unsigned int __X) { - return __builtin_clz (__X); + return __builtin_ia32_lzcnt_u32 (__X); } #ifdef __x86_64__ extern __inline unsigned long long __attribute__((__gnu_inline__, __always_inline__, __artificial__)) __lzcnt64 (unsigned long long __X) { - return __builtin_clzll (__X); + return __builtin_ia32_lzcnt_u64 (__X); } extern __inline unsigned long long __attribute__((__gnu_inline__, __always_inline__, __artificial__)) _lzcnt_u64 (unsigned long long __X) { - return __builtin_clzll (__X); + return __builtin_ia32_lzcnt_u64 (__X); } #endif diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 9805c78dfa0..eeeda8d8d97 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,8 @@ +2016-10-20 Uros Bizjak + + PR target/78037 + * gcc.target/i386/pr78037.c: New test. + 2016-10-20 Eric Botcazou * gcc.target/visium/overflow8.c: New. diff --git a/gcc/testsuite/gcc.target/i386/pr78037.c b/gcc/testsuite/gcc.target/i386/pr78037.c new file mode 100644 index 00000000000..d2e66ced7bc --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr78037.c @@ -0,0 +1,21 @@ +/* { dg-do run } */ +/* { dg-require-effective-target bmi } */ +/* { dg-options "-O2 -mbmi" } */ + +#include + +#include "bmi-check.h" + +int +__attribute__((noinline, noclone)) +foo (int x) +{ + return __tzcnt_u32 (x) & 0x1f; +} + +static void +bmi_test () +{ + if (foo (0) != 0) + abort (); +} -- 2.30.2