]> git.ipfire.org Git - thirdparty/gcc.git/commitdiff
Unify UNSPEC_MASKED_EQ/GT to the form of UNSPEC_PCMP.
authorliuhongt <hongtao.liu@intel.com>
Mon, 30 Aug 2021 07:05:14 +0000 (15:05 +0800)
committerliuhongt <hongtao.liu@intel.com>
Tue, 31 Aug 2021 01:25:57 +0000 (09:25 +0800)
Currently for evex vpcmpeqb instruction, we have two forms of rtl
template representation, one is (unspec [op1 op2] UNSPEC_MASK_EQ), the
other is (unspec [op1, op2, const_int 0] UNSPEC_PCMP), which increases
the maintenance burden, such as optimization (not: vpcmpeqb)
to (vpcmpneqb) requires two define_insn_and_split to match the two
forms respectively, this patch removes UNSPEC_MASK_EQ/GT, unifying
them into the form of UNSPEC_PCMP.

gcc/ChangeLog:

* config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
define_split to define_insn_and_split.
(*avx2_eq<mode>3): Removed.
(<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
(<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
(*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
adjust pattern.
(*avx2_gt<mode>3): Removed.
(<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
define_insn to define_expand, and adjust pattern.
(UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.

gcc/testsuite/ChangeLog:

* gcc.target/i386/avx512bw-vpcmpeqb-1.c: Adjust testcase.
* gcc.target/i386/avx512bw-vpcmpeqw-1.c: Ditto.
* gcc.target/i386/avx512bw-vpcmpgtb-1.c: Ditto.
* gcc.target/i386/avx512bw-vpcmpgtw-1.c: Ditto.
* gcc.target/i386/avx512f-vpcmpeqd-1.c: Ditto.
* gcc.target/i386/avx512f-vpcmpeqq-1.c: Ditto.
* gcc.target/i386/avx512f-vpcmpgtd-1.c: Ditto.
* gcc.target/i386/avx512f-vpcmpgtq-1.c: Ditto.
* gcc.target/i386/avx512vl-vpcmpeqd-1.c: Ditto.
* gcc.target/i386/avx512vl-vpcmpeqq-1.c: Ditto.
* gcc.target/i386/avx512vl-vpcmpgtd-1.c: Ditto.
* gcc.target/i386/avx512vl-vpcmpgtq-1.c: Ditto.
* gcc.target/i386/bitwise_mask_op-1.c: Ditto.
* gcc.target/i386/bitwise_mask_op-2.c: Ditto.

15 files changed:
gcc/config/i386/sse.md
gcc/testsuite/gcc.target/i386/avx512bw-vpcmpeqb-1.c
gcc/testsuite/gcc.target/i386/avx512bw-vpcmpeqw-1.c
gcc/testsuite/gcc.target/i386/avx512bw-vpcmpgtb-1.c
gcc/testsuite/gcc.target/i386/avx512bw-vpcmpgtw-1.c
gcc/testsuite/gcc.target/i386/avx512f-vpcmpeqd-1.c
gcc/testsuite/gcc.target/i386/avx512f-vpcmpeqq-1.c
gcc/testsuite/gcc.target/i386/avx512f-vpcmpgtd-1.c
gcc/testsuite/gcc.target/i386/avx512f-vpcmpgtq-1.c
gcc/testsuite/gcc.target/i386/avx512vl-vpcmpeqd-1.c
gcc/testsuite/gcc.target/i386/avx512vl-vpcmpeqq-1.c
gcc/testsuite/gcc.target/i386/avx512vl-vpcmpgtd-1.c
gcc/testsuite/gcc.target/i386/avx512vl-vpcmpgtq-1.c
gcc/testsuite/gcc.target/i386/bitwise_mask_op-1.c
gcc/testsuite/gcc.target/i386/bitwise_mask_op-2.c

index ac0c46328f20736ceacb444cf76ecb2f92c0a151..5785e73241c47923b0d62301461d381468328cf4 100644 (file)
   UNSPEC_COMPRESS
   UNSPEC_COMPRESS_STORE
   UNSPEC_EXPAND
-  UNSPEC_MASKED_EQ
-  UNSPEC_MASKED_GT
 
   ;; Mask operations
   UNSPEC_MASKOP
 ;; For signed comparison, handle EQ 0: NEQ 4,
 ;; for unsigned comparison extra handle LE:2, NLE:6, equivalent to EQ and NEQ.
 
-(define_split
+(define_insn_and_split "*<avx512>_ucmp<mode>3_1"
   [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
        (unspec:<avx512fmaskmode>
          [(us_minus:VI12_AVX512VL
           (match_operand:VI12_AVX512VL 3 "const0_operand")
           (match_operand:SI 4 "const_0_to_7_operand")]
          UNSPEC_PCMP_ITER))]
-  "TARGET_AVX512BW
+  "TARGET_AVX512BW && ix86_pre_reload_split ()
   && ix86_binary_operator_ok (US_MINUS, <MODE>mode, operands)
   && (INTVAL (operands[4]) & <pcmp_signed_mask>) == 0"
+  "#"
+  "&& 1"
   [(const_int 0)]
   {
     /* LE: 2, NLT: 5, NLE: 6, LT: 1  */
    (set_attr "prefix" "vex")
    (set_attr "mode" "OI")])
 
-(define_insn_and_split "*avx2_eq<mode>3"
- [(set (match_operand:VI_128_256  0 "register_operand")
-       (vec_merge:VI_128_256
-         (match_operand:VI_128_256 1 "vector_all_ones_operand")
-         (match_operand:VI_128_256 2 "const0_operand")
-         (unspec:<avx512fmaskmode>
-           [(match_operand:VI_128_256 3 "nonimmediate_operand")
-            (match_operand:VI_128_256 4 "nonimmediate_operand")]
-            UNSPEC_MASKED_EQ)))]
-  "TARGET_AVX512VL && ix86_pre_reload_split ()
-  && !(MEM_P (operands[3]) && MEM_P (operands[4]))"
-  "#"
-  "&& 1"
-  [(set (match_dup 0)
-       (eq:VI_128_256
-          (match_dup 3)
-          (match_dup 4)))])
-
 (define_insn_and_split "*avx2_pcmp<mode>3_1"
  [(set (match_operand:VI_128_256  0 "register_operand")
        (vec_merge:VI_128_256
   [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
        (unspec:<avx512fmaskmode>
          [(match_operand:VI12_AVX512VL 1 "nonimmediate_operand")
-          (match_operand:VI12_AVX512VL 2 "nonimmediate_operand")]
-         UNSPEC_MASKED_EQ))]
+          (match_operand:VI12_AVX512VL 2 "nonimmediate_operand")
+          (const_int 0)]
+         UNSPEC_PCMP))]
   "TARGET_AVX512BW"
   "ix86_fixup_binary_operands_no_copy (EQ, <MODE>mode, operands);")
 
   [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
        (unspec:<avx512fmaskmode>
          [(match_operand:VI48_AVX512VL 1 "nonimmediate_operand")
-          (match_operand:VI48_AVX512VL 2 "nonimmediate_operand")]
-         UNSPEC_MASKED_EQ))]
+          (match_operand:VI48_AVX512VL 2 "nonimmediate_operand")
+          (const_int 0)]
+         UNSPEC_PCMP))]
   "TARGET_AVX512F"
   "ix86_fixup_binary_operands_no_copy (EQ, <MODE>mode, operands);")
 
-(define_insn "<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
+(define_insn "*<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
   [(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=k,k")
        (unspec:<avx512fmaskmode>
          [(match_operand:VI12_AVX512VL 1 "nonimm_or_0_operand" "%v,v")
-          (match_operand:VI12_AVX512VL 2 "nonimm_or_0_operand" "vm,C")]
-         UNSPEC_MASKED_EQ))]
+          (match_operand:VI12_AVX512VL 2 "nonimm_or_0_operand" "vm,C")
+          (const_int 0)]
+         UNSPEC_PCMP))]
   "TARGET_AVX512BW && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
   "@
    vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}
    (set_attr "prefix" "evex")
    (set_attr "mode" "<sseinsnmode>")])
 
-(define_insn "<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
+(define_insn "*<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
   [(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=k,k")
        (unspec:<avx512fmaskmode>
          [(match_operand:VI48_AVX512VL 1 "nonimm_or_0_operand" "%v,v")
-          (match_operand:VI48_AVX512VL 2 "nonimm_or_0_operand" "vm,C")]
-         UNSPEC_MASKED_EQ))]
+          (match_operand:VI48_AVX512VL 2 "nonimm_or_0_operand" "vm,C")
+          (const_int 0)]
+         UNSPEC_PCMP))]
   "TARGET_AVX512F && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
   "@
    vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}
    (set_attr "prefix" "vex")
    (set_attr "mode" "OI")])
 
-(define_insn_and_split "*avx2_gt<mode>3"
- [(set (match_operand:VI_128_256  0 "register_operand")
-       (vec_merge:VI_128_256
-         (match_operand:VI_128_256 1 "vector_all_ones_operand")
-         (match_operand:VI_128_256 2 "const0_operand")
-         (unspec:<avx512fmaskmode>
-           [(match_operand:VI_128_256 3 "register_operand")
-            (match_operand:VI_128_256 4 "nonimmediate_operand")]
-            UNSPEC_MASKED_GT)))]
-  "TARGET_AVX512VL
-  && ix86_pre_reload_split ()"
-  "#"
-  "&& 1"
-  [(set (match_dup 0)
-       (gt:VI_128_256
-          (match_dup 3)
-          (match_dup 4)))])
-
-(define_insn "<avx512>_gt<mode>3<mask_scalar_merge_name>"
-  [(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=k")
+(define_expand "<avx512>_gt<mode>3<mask_scalar_merge_name>"
+  [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
        (unspec:<avx512fmaskmode>
-         [(match_operand:VI48_AVX512VL 1 "register_operand" "v")
-          (match_operand:VI48_AVX512VL 2 "nonimmediate_operand" "vm")] UNSPEC_MASKED_GT))]
-  "TARGET_AVX512F"
-  "vpcmpgt<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}"
-  [(set_attr "type" "ssecmp")
-   (set_attr "prefix_extra" "1")
-   (set_attr "prefix" "evex")
-   (set_attr "mode" "<sseinsnmode>")])
+         [(match_operand:VI48_AVX512VL 1 "register_operand")
+          (match_operand:VI48_AVX512VL 2 "nonimmediate_operand")
+          (const_int 6)] UNSPEC_PCMP))]
+  "TARGET_AVX512F")
 
-(define_insn "<avx512>_gt<mode>3<mask_scalar_merge_name>"
-  [(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=k")
+(define_expand "<avx512>_gt<mode>3<mask_scalar_merge_name>"
+  [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
        (unspec:<avx512fmaskmode>
-         [(match_operand:VI12_AVX512VL 1 "register_operand" "v")
-          (match_operand:VI12_AVX512VL 2 "nonimmediate_operand" "vm")] UNSPEC_MASKED_GT))]
-  "TARGET_AVX512BW"
-  "vpcmpgt<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}"
-  [(set_attr "type" "ssecmp")
-   (set_attr "prefix_extra" "1")
-   (set_attr "prefix" "evex")
-   (set_attr "mode" "<sseinsnmode>")])
+         [(match_operand:VI12_AVX512VL 1 "register_operand")
+          (match_operand:VI12_AVX512VL 2 "nonimmediate_operand")
+          (const_int 6)] UNSPEC_PCMP))]
+  "TARGET_AVX512BW")
 
 (define_insn "*sse2_gt<mode>3"
   [(set (match_operand:VI124_128 0 "register_operand" "=x,x")
index 506fcf9dba0f52902e6a1c2c2229b2a7ee6e1b53..11a6007c609ade0a9259ba5ae5a0bae363d0cd33 100644 (file)
@@ -1,11 +1,11 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512bw -mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqb\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 1b2670b68044bb59333df25800a94124bcd0e2cb..5cf8198502d2334175b184fd9181cadae4eb3097 100644 (file)
@@ -1,11 +1,11 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512bw -mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index e4014b37847aec336118e3e1288e8d1e0e51b8fb..bc7e8b152ce2478f3e61795d1518cd8dcb57e3ea 100644 (file)
@@ -1,11 +1,11 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512bw -mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtb\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtb\[ \\t\]+|vpcmpb\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 37e8d32833602801531aaa6dcbb13123b6adf8d7..02999e00e361f01a78540804af74692943d755d1 100644 (file)
@@ -1,11 +1,11 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512bw -mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtw\[ \\t\]+|vpcmpw\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 46244427413a37984fcc2ad61dec5b8796dfab0d..33f4e73c8c4791a12299d68f40fa9a36bbb07263 100644 (file)
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512f -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 4a6505d9f97a0a71c73c340d108b23a56463b81a..595ba5527036a2a5ba5c2c37bfd61d846c41d0bc 100644 (file)
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512f -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 6b4a7ffe477eabd6383d681b0ea384297ac7968b..d038cfbe1f3c1ce1f763c482e25f508e6e1e3b3a 100644 (file)
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512f -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index d3647d43a9dbc776c6a0626e15f9fe243e5035e1..ce2e41ad259b4e32699d7983f603af5b8720f569 100644 (file)
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512f -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%zmm\[0-9\]+\[^\n^k\]*%k\[0-7\]\{%k\[0-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 863bad9bc57ef052753a43aaf32f9faeb68f054b..4322c5db74983c47921172d06f4f69eb73cdf7e8 100644 (file)
@@ -1,9 +1,9 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n^k\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 35ddeeb5f7c459ac65f97496dbfad2d9b67f493d..69b200a6be729640bd1e2fb3d6aabdc628d3e227 100644 (file)
@@ -1,9 +1,9 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpeqq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpeqq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$0)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 95ec53d93e0947273aaf49d754a542e7bb84eeb5..2278a27e2dd511f21273089c52aa6b25aa17cffc 100644 (file)
@@ -1,9 +1,9 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtd\[ \\t\]+|vpcmpd\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index a83a9011ecd1472857724045749758396fa0a48e..26cac3a1bad3bb67ec2ba4bf41078e8d5215f57f 100644 (file)
@@ -1,9 +1,9 @@
 /* { dg-do compile } */
 /* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
-/* { dg-final { scan-assembler-times "vpcmpgtq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\](?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "(?:vpcmpgtq\[ \\t\]+|vpcmpq\[ \\t\]+\\\$6)\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*%k\[0-7\]\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
 
 #include <immintrin.h>
 
index 61f71ab8b23c616dabaa6d1219ecf2729e0211ff..826ed6921d44e5156643c5b42fea7cc8981a5d36 100644 (file)
@@ -150,8 +150,6 @@ foo_notq (__m512i a, __m512i b, __m512i c, __m512i d)
   return _mm512_mask_add_epi8 (c, ~m1, a, d);
 }
 
-/* { dg-final { scan-assembler-times "knotq" "2" { target { ! ia32 } } } }  */
-
 __m512i
 foo_notd (__m512i a, __m512i b, __m512i c, __m512i d)
 {
@@ -159,8 +157,6 @@ foo_notd (__m512i a, __m512i b, __m512i c, __m512i d)
   return _mm512_mask_add_epi16 (c, ~m1, a, d);
 }
 
-/* { dg-final { scan-assembler-times "knotd" "2" { target { ! ia32 } } } }  */
-
 __m512i
 foo_notw (__m512i a, __m512i b, __m512i c, __m512i d)
 {
@@ -174,5 +170,3 @@ foo_notb (__m512i a, __m512i b, __m512i c, __m512i d)
   __mmask8 m1 = _mm512_cmpeq_epi64_mask (a, b);
   return _mm512_mask_add_epi64 (c, ~m1, a, d);
 }
-
-/* { dg-final { scan-assembler-times "knotw" "4" } }  */
index 850f0b42652c4870b099610c698be523c88d6364..e93269bcd6c75b50d6bd6ded5aaee07edf850f9e 100644 (file)
@@ -1,7 +1,6 @@
 /* PR target/88808  */
 /* { dg-do compile } */
 /* { dg-options "-mavx512bw -mavx512dq -O2" } */
-/* { dg-final { scan-assembler-times "knotb" "2" } }  */
 /* { dg-final { scan-assembler-times "korb" "1" } }  */
 /* { dg-final { scan-assembler-times "kxorb" "1" } }  */
 #include "bitwise_mask_op-1.c"