Many AVX512 vector operations can broadcast from a scalar memory source.
This patch enables memory broadcast for INT logic operations.
gcc/
PR target/72782
* config/i386/sse.md (*<code><mode>3_bcst): New.
gcc/testsuite/
PR target/72782
* gcc.target/i386/avx512f-and-di-zmm-1.c: New test.
* gcc.target/i386/avx512f-and-si-zmm-1.c: Likewise.
* gcc.target/i386/avx512f-and-si-zmm-2.c: Likewise.
* gcc.target/i386/avx512f-and-si-zmm-3.c: Likewise.
* gcc.target/i386/avx512f-and-si-zmm-4.c: Likewise.
* gcc.target/i386/avx512f-and-si-zmm-5.c: Likewise.
* gcc.target/i386/avx512f-and-si-zmm-6.c: Likewise.
* gcc.target/i386/avx512f-or-di-zmm-1.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-1.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-2.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-3.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-4.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-5.c: Likewise.
* gcc.target/i386/avx512f-or-si-zmm-6.c: Likewise.
* gcc.target/i386/avx512f-xor-di-zmm-1.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-1.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-2.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-3.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-4.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-5.c: Likewise.
* gcc.target/i386/avx512f-xor-si-zmm-6.c: Likewise.
* gcc.target/i386/avx512vl-and-si-xmm-1.c: Likewise.
* gcc.target/i386/avx512vl-and-si-ymm-1.c: Likewise.
* gcc.target/i386/avx512vl-or-si-xmm-1.c: Likewise.
* gcc.target/i386/avx512vl-or-si-ymm-1.c: Likewise.
* gcc.target/i386/avx512vl-xor-si-xmm-1.c: Likewise.
* gcc.target/i386/avx512vl-xor-si-ymm-1.c: Likewise.
From-SVN: r265369
+2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
+
+ PR target/72782
+ * config/i386/sse.md (*<code><mode>3_bcst): New.
+
2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
PR target/72782
]
(const_string "<sseinsnmode>")))])
+(define_insn "*<code><mode>3_bcst"
+ [(set (match_operand:VI48_AVX512VL 0 "register_operand" "=v")
+ (any_logic:VI48_AVX512VL
+ (vec_duplicate:VI48_AVX512VL
+ (match_operand:<ssescalarmode> 1 "memory_operand" "m"))
+ (match_operand:VI48_AVX512VL 2 "register_operand" "v")))]
+ "TARGET_AVX512F && <mask_avx512vl_condition>"
+ "vp<logic><ssemodesuffix>\t{%1<avx512bcst>, %2, %0<mask_operand3>|%0<mask_operand3>, %2, %1<avx512bcst>}"
+ [(set_attr "type" "sseiadd")
+ (set_attr "prefix" "evex")
+ (set_attr "mode" "<sseinsnmode>")])
+
(define_insn "<avx512>_testm<mode>3<mask_scalar_merge_name>"
[(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=Yk")
(unspec:<avx512fmaskmode>
+2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
+
+ PR target/72782
+ * gcc.target/i386/avx512f-and-di-zmm-1.c: New test.
+ * gcc.target/i386/avx512f-and-si-zmm-1.c: Likewise.
+ * gcc.target/i386/avx512f-and-si-zmm-2.c: Likewise.
+ * gcc.target/i386/avx512f-and-si-zmm-3.c: Likewise.
+ * gcc.target/i386/avx512f-and-si-zmm-4.c: Likewise.
+ * gcc.target/i386/avx512f-and-si-zmm-5.c: Likewise.
+ * gcc.target/i386/avx512f-and-si-zmm-6.c: Likewise.
+ * gcc.target/i386/avx512f-or-di-zmm-1.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-1.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-2.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-3.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-4.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-5.c: Likewise.
+ * gcc.target/i386/avx512f-or-si-zmm-6.c: Likewise.
+ * gcc.target/i386/avx512f-xor-di-zmm-1.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-1.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-2.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-3.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-4.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-5.c: Likewise.
+ * gcc.target/i386/avx512f-xor-si-zmm-6.c: Likewise.
+ * gcc.target/i386/avx512vl-and-si-xmm-1.c: Likewise.
+ * gcc.target/i386/avx512vl-and-si-ymm-1.c: Likewise.
+ * gcc.target/i386/avx512vl-or-si-xmm-1.c: Likewise.
+ * gcc.target/i386/avx512vl-or-si-ymm-1.c: Likewise.
+ * gcc.target/i386/avx512vl-xor-si-xmm-1.c: Likewise.
+ * gcc.target/i386/avx512vl-xor-si-ymm-1.c: Likewise.
+
2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
PR target/72782
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpandq\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastq\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi64
+#define SCALAR long long
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-2.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpandd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-3.h"
--- /dev/null
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+%zmm\[0-9\]+, %zmm\[0-9\]+, %zmm0" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-4.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-5.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op and
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-6.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vporq\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastq\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi64
+#define SCALAR long long
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-2.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpord\[^\n\]*%zmm\[0-9\]+" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-3.h"
--- /dev/null
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+%zmm\[0-9\]+, %zmm\[0-9\]+, %zmm0" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-4.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-5.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-6.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpxorq\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastq\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi64
+#define SCALAR long long
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-2.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpxord\[^\n\]*%zmm\[0-9\]+" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-3.h"
--- /dev/null
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" 1 } } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+%zmm\[0-9\]+, %zmm\[0-9\]+, %zmm0" 1 } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-4.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-5.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\[^\n\]*\\\{1to\[1-8\]+\\\}, %zmm\[0-9\]+, %zmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%zmm\[0-9\]+" } } */
+
+#define type __m512i
+#define vec 512
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-6.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %xmm\[0-9\]+, %xmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%xmm\[0-9\]+" } } */
+
+#include <immintrin.h>
+
+__m128i
+foo (__m128i x, int *f)
+{
+ return (__m128i) ((__v4su) x & (__v4su) _mm_set1_epi32 (*f));
+}
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpandd\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %ymm\[0-9\]+, %ymm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%ymm\[0-9\]+" } } */
+
+#include <immintrin.h>
+
+__m256i
+foo (__m256i x, int *f)
+{
+ return (__m256i) ((__v8su) x & (__v8su) _mm256_set1_epi32 (*f));
+}
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %xmm\[0-9\]+, %xmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%xmm\[0-9\]+" } } */
+
+#define type __m128i
+#define vec
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %ymm\[0-9\]+, %ymm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%ymm\[0-9\]+" } } */
+
+#define type __m256i
+#define vec 256
+#define op or
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %xmm\[0-9\]+, %xmm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%xmm\[0-9\]+" } } */
+
+#define type __m128i
+#define vec
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vpxord\[ \\t\]+\\(%(?:eax|rdi|edi)\\)\\\{1to\[1-8\]+\\\}, %ymm\[0-9\]+, %ymm0" 1 } } */
+/* { dg-final { scan-assembler-not "vpbroadcastd\[^\n\]*%ymm\[0-9\]+" } } */
+
+#define type __m256i
+#define vec 256
+#define op xor
+#define suffix epi32
+#define SCALAR int
+
+#include "avx512-binop-1.h"