On Sun, Oct 21, 2018 at 2:03 AM H.J. Lu <hjl.to...@gmail.com> wrote:
>
> gcc/
>
> PR target/87662
> * i386/avx512vlintrin.h (_mm256_or_epi32): New.
> (_mm_or_epi32): Likewise.
> (_mm256_xor_epi32): Likewise.
> (_mm_xor_epi32): Likewise.
> (_mm256_or_epi64): Likewise.
> (_mm_or_epi64): Likewise.
> (_mm256_xor_epi64): Likewise.
> (_mm_xor_epi64): Likewise.
>
> gcc/testsuite/
>
> PR target/87662
> * gcc.target/i386/pr87662.c
LGTM.
Thanks,
Uros.
> ---
> gcc/config/i386/avx512vlintrin.h | 48 ++++++++++++++++
> gcc/testsuite/gcc.target/i386/pr87662.c | 76 +++++++++++++++++++++++++
> 2 files changed, 124 insertions(+)
> create mode 100644 gcc/testsuite/gcc.target/i386/pr87662.c
>
> diff --git a/gcc/config/i386/avx512vlintrin.h
> b/gcc/config/i386/avx512vlintrin.h
> index 68b5537845b..a4fb0b0ac00 100644
> --- a/gcc/config/i386/avx512vlintrin.h
> +++ b/gcc/config/i386/avx512vlintrin.h
> @@ -4855,6 +4855,12 @@ _mm256_maskz_or_epi32 (__mmask8 __U, __m256i __A,
> __m256i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m256i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm256_or_epi32 (__m256i __A, __m256i __B)
> +{
> + return (__m256i) ((__v8su)__A | (__v8su)__B);
> +}
> +
> extern __inline __m128i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm_mask_or_epi32 (__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
> @@ -4876,6 +4882,12 @@ _mm_maskz_or_epi32 (__mmask8 __U, __m128i __A, __m128i
> __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm_or_epi32 (__m128i __A, __m128i __B)
> +{
> + return (__m128i) ((__v4su)__A | (__v4su)__B);
> +}
> +
> extern __inline __m256i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm256_mask_xor_epi32 (__m256i __W, __mmask8 __U, __m256i __A,
> @@ -4898,6 +4910,12 @@ _mm256_maskz_xor_epi32 (__mmask8 __U, __m256i __A,
> __m256i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m256i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm256_xor_epi32 (__m256i __A, __m256i __B)
> +{
> + return (__m256i) ((__v8su)__A ^ (__v8su)__B);
> +}
> +
> extern __inline __m128i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm_mask_xor_epi32 (__m128i __W, __mmask8 __U, __m128i __A,
> @@ -4920,6 +4938,12 @@ _mm_maskz_xor_epi32 (__mmask8 __U, __m128i __A,
> __m128i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm_xor_epi32 (__m128i __A, __m128i __B)
> +{
> + return (__m128i) ((__v4su)__A ^ (__v4su)__B);
> +}
> +
> extern __inline __m128
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm_mask_cvtpd_ps (__m128 __W, __mmask8 __U, __m128d __A)
> @@ -7340,6 +7364,12 @@ _mm256_maskz_or_epi64 (__mmask8 __U, __m256i __A,
> __m256i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m256i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm256_or_epi64 (__m256i __A, __m256i __B)
> +{
> + return (__m256i) ((__v4du)__A | (__v4du)__B);
> +}
> +
> extern __inline __m128i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm_mask_or_epi64 (__m128i __W, __mmask8 __U, __m128i __A, __m128i __B)
> @@ -7361,6 +7391,12 @@ _mm_maskz_or_epi64 (__mmask8 __U, __m128i __A, __m128i
> __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm_or_epi64 (__m128i __A, __m128i __B)
> +{
> + return (__m128i) ((__v2du)__A | (__v2du)__B);
> +}
> +
> extern __inline __m256i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm256_mask_xor_epi64 (__m256i __W, __mmask8 __U, __m256i __A,
> @@ -7383,6 +7419,12 @@ _mm256_maskz_xor_epi64 (__mmask8 __U, __m256i __A,
> __m256i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m256i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm256_xor_epi64 (__m256i __A, __m256i __B)
> +{
> + return (__m256i) ((__v4du)__A ^ (__v4du)__B);
> +}
> +
> extern __inline __m128i
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm_mask_xor_epi64 (__m128i __W, __mmask8 __U, __m128i __A,
> @@ -7405,6 +7447,12 @@ _mm_maskz_xor_epi64 (__mmask8 __U, __m128i __A,
> __m128i __B)
> (__mmask8) __U);
> }
>
> +extern __inline __m128i __attribute__((__gnu_inline__, __always_inline__,
> __artificial__))
> +_mm_xor_epi64 (__m128i __A, __m128i __B)
> +{
> + return (__m128i) ((__v2du)__A ^ (__v2du)__B);
> +}
> +
> extern __inline __m256d
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> _mm256_mask_max_pd (__m256d __W, __mmask8 __U, __m256d __A,
> diff --git a/gcc/testsuite/gcc.target/i386/pr87662.c
> b/gcc/testsuite/gcc.target/i386/pr87662.c
> new file mode 100644
> index 00000000000..c9110a04440
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr87662.c
> @@ -0,0 +1,76 @@
> +/* { dg-do compile { target { ! ia32 } } } */
> +/* { dg-options "-O2 -mavx512vl" } */
> +/* { dg-final { scan-assembler-times "vpord\[^\n\r\]*ymm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vpord\[^\n\r\]*xmm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vporq\[^\n\r\]*ymm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vporq\[^\n\r\]*xmm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vpxord\[^\n\r\]*ymm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vpxord\[^\n\r\]*xmm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vpxorq\[^\n\r\]*ymm16" 1 } } */
> +/* { dg-final { scan-assembler-times "vpxorq\[^\n\r\]*xmm16" 1 } } */
> +
> +#include <immintrin.h>
> +
> +__m256i
> +foo1 (__m256i x, __m256i y)
> +{
> + register __m256i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm256_or_epi32 (x, z);
> +}
> +
> +__m256i
> +foo2 (__m256i x, __m256i y)
> +{
> + register __m256i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm256_xor_epi32 (x, z);
> +}
> +
> +__m128i
> +foo3 (__m128i x, __m128i y)
> +{
> + register __m128i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm_or_epi32 (x, z);
> +}
> +
> +__m128i
> +foo4 (__m128i x, __m128i y)
> +{
> + register __m128i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm_xor_epi32 (x, z);
> +}
> +
> +__m256i
> +foo5 (__m256i x, __m256i y)
> +{
> + register __m256i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm256_or_epi64 (x, z);
> +}
> +
> +__m256i
> +foo6 (__m256i x, __m256i y)
> +{
> + register __m256i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm256_xor_epi64 (x, z);
> +}
> +
> +__m128i
> +foo7 (__m128i x, __m128i y)
> +{
> + register __m128i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm_or_epi64 (x, z);
> +}
> +
> +__m128i
> +foo8 (__m128i x, __m128i y)
> +{
> + register __m128i z __asm ("xmm16") = y;
> + asm volatile ("" : "+v" (z));
> + return _mm_xor_epi64 (x, z);
> +}
> --
> 2.17.2
>
