+2017-11-29 Julia Koval <julia.koval@intel.com>
+
+ * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
+ _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
+ _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
+ _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
+ * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
+ _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
+ _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
+ _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
+ _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
+ _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
+ _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
+ * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
+ __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
+ __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
+ __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
+ __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
+ __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
+ __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
+ __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
+ __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
+ __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
+ __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
+ __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
+ __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
+ __builtin_ia32_vpshldv_v2di_mask,
+ __builtin_ia32_vpshldv_v2di_maskz): New builtins.
+ * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
+ vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
+
2017-11-29 Julia Koval <julia.koval@intel.com>
* config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
return (__m512i)__builtin_ia32_vpshrdv_v8di_maskz ((__v8di)__B, (__v8di) __C,
(__v8di) __D, (__mmask8)__A);
}
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_shldv_epi16 (__m512i __A, __m512i __B, __m512i __C)
+{
+ return (__m512i) __builtin_ia32_vpshldv_v32hi ((__v32hi)__A, (__v32hi) __B,
+ (__v32hi) __C);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask_shldv_epi16 (__m512i __A, __mmask32 __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v32hi_mask ((__v32hi)__A,
+ (__v32hi) __C, (__v32hi) __D, (__mmask32)__B);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_maskz_shldv_epi16 (__mmask32 __A, __m512i __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v32hi_maskz ((__v32hi)__B,
+ (__v32hi) __C, (__v32hi) __D, (__mmask32)__A);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_shldv_epi32 (__m512i __A, __m512i __B, __m512i __C)
+{
+ return (__m512i) __builtin_ia32_vpshldv_v16si ((__v16si)__A, (__v16si) __B,
+ (__v16si) __C);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask_shldv_epi32 (__m512i __A, __mmask16 __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v16si_mask ((__v16si)__A,
+ (__v16si) __C, (__v16si) __D, (__mmask16)__B);
+}
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_maskz_shldv_epi32 (__mmask16 __A, __m512i __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v16si_maskz ((__v16si)__B,
+ (__v16si) __C, (__v16si) __D, (__mmask16)__A);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_shldv_epi64 (__m512i __A, __m512i __B, __m512i __C)
+{
+ return (__m512i) __builtin_ia32_vpshldv_v8di ((__v8di)__A, (__v8di) __B,
+ (__v8di) __C);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask_shldv_epi64 (__m512i __A, __mmask8 __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v8di_mask ((__v8di)__A, (__v8di) __C,
+ (__v8di) __D, (__mmask8)__B);
+}
+
+extern __inline __m512i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_maskz_shldv_epi64 (__mmask8 __A, __m512i __B, __m512i __C, __m512i __D)
+{
+ return (__m512i)__builtin_ia32_vpshldv_v8di_maskz ((__v8di)__B, (__v8di) __C,
+ (__v8di) __D, (__mmask8)__A);
+}
#ifdef __DISABLE_AVX512VBMI2BW__
#undef __DISABLE_AVX512VBMI2BW__
(__v2di) __D, (__mmask8)__A);
}
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_shldv_epi16 (__m256i __A, __m256i __B, __m256i __C)
+{
+ return (__m256i) __builtin_ia32_vpshldv_v16hi ((__v16hi)__A, (__v16hi) __B,
+ (__v16hi) __C);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_shldv_epi16 (__m256i __A, __mmask16 __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v16hi_mask ((__v16hi)__A,
+ (__v16hi) __C, (__v16hi) __D, (__mmask16)__B);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_shldv_epi16 (__mmask16 __A, __m256i __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v16hi_maskz ((__v16hi)__B,
+ (__v16hi) __C, (__v16hi) __D, (__mmask16)__A);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_shldv_epi32 (__m256i __A, __m256i __B, __m256i __C)
+{
+ return (__m256i) __builtin_ia32_vpshldv_v8si ((__v8si)__A, (__v8si) __B,
+ (__v8si) __C);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_shldv_epi32 (__m256i __A, __mmask8 __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v8si_mask ((__v8si)__A, (__v8si) __C,
+ (__v8si) __D, (__mmask8)__B) ;
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_shldv_epi32 (__mmask8 __A, __m256i __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v8si_maskz ((__v8si)__B, (__v8si) __C,
+ (__v8si) __D, (__mmask8)__A);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_shldv_epi64 (__m256i __A, __m256i __B, __m256i __C)
+{
+ return (__m256i) __builtin_ia32_vpshldv_v4di ((__v4di)__A, (__v4di) __B,
+ (__v4di) __C);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_shldv_epi64 (__m256i __A, __mmask8 __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v4di_mask ((__v4di)__A, (__v4di) __C,
+ (__v4di) __D, (__mmask8)__B);
+}
+
+extern __inline __m256i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_shldv_epi64 (__mmask8 __A, __m256i __B, __m256i __C, __m256i __D)
+{
+ return (__m256i)__builtin_ia32_vpshldv_v4di_maskz ((__v4di)__B, (__v4di) __C,
+ (__v4di) __D, (__mmask8)__A);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shldv_epi16 (__m128i __A, __m128i __B, __m128i __C)
+{
+ return (__m128i) __builtin_ia32_vpshldv_v8hi ((__v8hi)__A, (__v8hi) __B,
+ (__v8hi) __C);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_mask_shldv_epi16 (__m128i __A, __mmask8 __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v8hi_mask ((__v8hi)__A, (__v8hi) __C,
+ (__v8hi) __D, (__mmask8)__B);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_maskz_shldv_epi16 (__mmask8 __A, __m128i __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v8hi_maskz ((__v8hi)__B, (__v8hi) __C,
+ (__v8hi) __D, (__mmask8)__A);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shldv_epi32 (__m128i __A, __m128i __B, __m128i __C)
+{
+ return (__m128i) __builtin_ia32_vpshldv_v4si ((__v4si)__A, (__v4si) __B,
+ (__v4si) __C);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_mask_shldv_epi32 (__m128i __A, __mmask8 __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v4si_mask ((__v4si)__A, (__v4si) __C,
+ (__v4si) __D, (__mmask8)__B);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_maskz_shldv_epi32 (__mmask8 __A, __m128i __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v4si_maskz ((__v4si)__B, (__v4si) __C,
+ (__v4si) __D, (__mmask8)__A);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shldv_epi64 (__m128i __A, __m128i __B, __m128i __C)
+{
+ return (__m128i) __builtin_ia32_vpshldv_v2di ((__v2di)__A, (__v2di) __B,
+ (__v2di) __C);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_mask_shldv_epi64 (__m128i __A, __mmask8 __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v2di_mask ((__v2di)__A, (__v2di) __C,
+ (__v2di) __D, (__mmask8)__B);
+}
+
+extern __inline __m128i
+__attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_maskz_shldv_epi64 (__mmask8 __A, __m128i __B, __m128i __C, __m128i __D)
+{
+ return (__m128i)__builtin_ia32_vpshldv_v2di_maskz ((__v2di)__B, (__v2di) __C,
+ (__v2di) __D, (__mmask8)__A);
+}
+
+
#ifdef __DISABLE_AVX512VBMI2VL__
BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshrdv_v2di_mask, "__builtin_ia32_vpshrdv_v2di_mask", IX86_BUILTIN_VPSHRDVV2DI_MASK, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_V2DI_INT)
BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshrdv_v2di_maskz, "__builtin_ia32_vpshrdv_v2di_maskz", IX86_BUILTIN_VPSHRDVV2DI_MASKZ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_V2DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v32hi, "__builtin_ia32_vpshldv_v32hi", IX86_BUILTIN_VPSHLDVV32HI, UNKNOWN, (int) V32HI_FTYPE_V32HI_V32HI_V32HI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v32hi_mask, "__builtin_ia32_vpshldv_v32hi_mask", IX86_BUILTIN_VPSHLDVV32HI_MASK, UNKNOWN, (int) V32HI_FTYPE_V32HI_V32HI_V32HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v32hi_maskz, "__builtin_ia32_vpshldv_v32hi_maskz", IX86_BUILTIN_VPSHLDVV32HI_MASKZ, UNKNOWN, (int) V32HI_FTYPE_V32HI_V32HI_V32HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16hi, "__builtin_ia32_vpshldv_v16hi", IX86_BUILTIN_VPSHLDVV16HI, UNKNOWN, (int) V16HI_FTYPE_V16HI_V16HI_V16HI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16hi_mask, "__builtin_ia32_vpshldv_v16hi_mask", IX86_BUILTIN_VPSHLDVV16HI_MASK, UNKNOWN, (int) V16HI_FTYPE_V16HI_V16HI_V16HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16hi_maskz, "__builtin_ia32_vpshldv_v16hi_maskz", IX86_BUILTIN_VPSHLDVV16HI_MASKZ, UNKNOWN, (int) V16HI_FTYPE_V16HI_V16HI_V16HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8hi, "__builtin_ia32_vpshldv_v8hi", IX86_BUILTIN_VPSHLDVV8HI, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_V8HI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8hi_mask, "__builtin_ia32_vpshldv_v8hi_mask", IX86_BUILTIN_VPSHLDVV8HI_MASK, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_V8HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8hi_maskz, "__builtin_ia32_vpshldv_v8hi_maskz", IX86_BUILTIN_VPSHLDVV8HI_MASKZ, UNKNOWN, (int) V8HI_FTYPE_V8HI_V8HI_V8HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16si, "__builtin_ia32_vpshldv_v16si", IX86_BUILTIN_VPSHLDVV16SI, UNKNOWN, (int) V16SI_FTYPE_V16SI_V16SI_V16SI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16si_mask, "__builtin_ia32_vpshldv_v16si_mask", IX86_BUILTIN_VPSHLDVV16SI_MASK, UNKNOWN, (int) V16SI_FTYPE_V16SI_V16SI_V16SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v16si_maskz, "__builtin_ia32_vpshldv_v16si_maskz", IX86_BUILTIN_VPSHLDVV16SI_MASKZ, UNKNOWN, (int) V16SI_FTYPE_V16SI_V16SI_V16SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8si, "__builtin_ia32_vpshldv_v8si", IX86_BUILTIN_VPSHLDVV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8si_mask, "__builtin_ia32_vpshldv_v8si_mask", IX86_BUILTIN_VPSHLDVV8SI_MASK, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8si_maskz, "__builtin_ia32_vpshldv_v8si_maskz", IX86_BUILTIN_VPSHLDVV8SI_MASKZ, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4si, "__builtin_ia32_vpshldv_v4si", IX86_BUILTIN_VPSHLDVV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4si_mask, "__builtin_ia32_vpshldv_v4si_mask", IX86_BUILTIN_VPSHLDVV4SI_MASK, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4si_maskz, "__builtin_ia32_vpshldv_v4si_maskz", IX86_BUILTIN_VPSHLDVV4SI_MASKZ, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8di, "__builtin_ia32_vpshldv_v8di", IX86_BUILTIN_VPSHLDVV8DI, UNKNOWN, (int) V8DI_FTYPE_V8DI_V8DI_V8DI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8di_mask, "__builtin_ia32_vpshldv_v8di_mask", IX86_BUILTIN_VPSHLDVV8DI_MASK, UNKNOWN, (int) V8DI_FTYPE_V8DI_V8DI_V8DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v8di_maskz, "__builtin_ia32_vpshldv_v8di_maskz", IX86_BUILTIN_VPSHLDVV8DI_MASKZ, UNKNOWN, (int) V8DI_FTYPE_V8DI_V8DI_V8DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4di, "__builtin_ia32_vpshldv_v4di", IX86_BUILTIN_VPSHLDVV4DI, UNKNOWN, (int) V4DI_FTYPE_V4DI_V4DI_V4DI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4di_mask, "__builtin_ia32_vpshldv_v4di_mask", IX86_BUILTIN_VPSHLDVV4DI_MASK, UNKNOWN, (int) V4DI_FTYPE_V4DI_V4DI_V4DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v4di_maskz, "__builtin_ia32_vpshldv_v4di_maskz", IX86_BUILTIN_VPSHLDVV4DI_MASKZ, UNKNOWN, (int) V4DI_FTYPE_V4DI_V4DI_V4DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v2di, "__builtin_ia32_vpshldv_v2di", IX86_BUILTIN_VPSHLDVV2DI, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_V2DI)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v2di_mask, "__builtin_ia32_vpshldv_v2di_mask", IX86_BUILTIN_VPSHLDVV2DI_MASK, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_V2DI_INT)
+BDESC (OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_vpshldv_v2di_maskz, "__builtin_ia32_vpshldv_v2di_maskz", IX86_BUILTIN_VPSHLDVV2DI_MASKZ, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_V2DI_INT)
+
BDESC_END (ARGS2, SPECIAL_ARGS2)
BDESC_FIRST (special_args2, SPECIAL_ARGS2, OPTION_MASK_ISA_AVX512VBMI2, CODE_FOR_compressstorev64qi_mask, "__builtin_ia32_compressstoreuqi512_mask", IX86_BUILTIN_PCOMPRESSBSTORE512, UNKNOWN, (int) VOID_FTYPE_PV64QI_V64QI_UDI)
UNSPEC_VPSHLD
UNSPEC_VPSHRD
UNSPEC_VPSHRDV
+ UNSPEC_VPSHLDV
])
(define_c_enum "unspecv" [
"vpshrdv<ssemodesuffix>\t{%3, %2, %0%{%5%}%{z%}|%0%{%5%}%{z%}, %2, %3 }"
[(set_attr ("prefix") ("evex"))
(set_attr "mode" "<sseinsnmode>")])
+
+(define_insn "vpshldv_<mode>"
+ [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
+ (unspec:VI248_VLBW
+ [(match_operand:VI248_VLBW 1 "register_operand" "0")
+ (match_operand:VI248_VLBW 2 "register_operand" "v")
+ (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")
+] UNSPEC_VPSHLDV))]
+ "TARGET_AVX512VBMI2"
+ "vpshldv<ssemodesuffix>\t{%3, %2, %0|%0, %2, %3 }"
+ [(set_attr ("prefix") ("evex"))
+ (set_attr "mode" "<sseinsnmode>")])
+
+(define_insn "vpshldv_<mode>_mask"
+ [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
+ (vec_merge:VI248_VLBW (unspec:VI248_VLBW
+ [(match_operand:VI248_VLBW 1 "register_operand" "0")
+ (match_operand:VI248_VLBW 2 "register_operand" "v")
+ (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")
+ ] UNSPEC_VPSHLDV)
+ (match_dup 1)
+ (match_operand:<avx512fmaskmode> 4 "register_operand" "Yk"))
+)]
+ "TARGET_AVX512VBMI2"
+ "vpshldv<ssemodesuffix>\t{%3, %2, %0%{%4%}|%0%{%4%}, %2, %3 }"
+ [(set_attr ("prefix") ("evex"))
+ (set_attr "mode" "<sseinsnmode>")])
+
+(define_expand "vpshldv_<mode>_maskz"
+ [(match_operand:VI248_VLBW 0 "register_operand")
+ (match_operand:VI248_VLBW 1 "register_operand")
+ (match_operand:VI248_VLBW 2 "register_operand")
+ (match_operand:VI248_VLBW 3 "nonimmediate_operand")
+ (match_operand:<avx512fmaskmode> 4 "register_operand")]
+ "TARGET_AVX512VBMI2"
+{
+ emit_insn (gen_vpshldv_<mode>_maskz_1 (
+ operands[0], operands[1], operands[2], operands[3],
+ CONST0_RTX (<MODE>mode), operands[4]));
+ DONE;
+})
+
+(define_insn "vpshldv_<mode>_maskz_1"
+ [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
+ (vec_merge:VI248_VLBW (unspec:VI248_VLBW
+ [(match_operand:VI248_VLBW 1 "register_operand" "0")
+ (match_operand:VI248_VLBW 2 "register_operand" "v")
+ (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")
+ ] UNSPEC_VPSHLDV)
+ (match_operand:VI248_VLBW 4 "const0_operand" "C")
+ (match_operand:<avx512fmaskmode> 5 "register_operand" "Yk"))
+)]
+ "TARGET_AVX512VBMI2"
+ "vpshldv<ssemodesuffix>\t{%3, %2, %0%{%5%}%{z%}|%0%{%5%}%{z%}, %2, %3 }"
+ [(set_attr ("prefix") ("evex"))
+ (set_attr "mode" "<sseinsnmode>")])
+2017-11-29 Julia Koval <julia.koval@intel.com>
+
+ * gcc.target/i386/avx512f-vpshldv-1.c: New test.
+ * gcc.target/i386/avx512f-vpshldvd-2.c: Ditto.
+ * gcc.target/i386/avx512f-vpshldvq-2.c: Ditto.
+ * gcc.target/i386/avx512f-vpshldvw-2.c: Ditto.
+ * gcc.target/i386/avx512vl-vpshldv-1.c: Ditto.
+ * gcc.target/i386/avx512vl-vpshldvd-2.c: Ditto.
+ * gcc.target/i386/avx512vl-vpshldvq-2.c: Ditto.
+ * gcc.target/i386/avx512vl-vpshldvw-2.c: Ditto.
+
2017-11-29 Julia Koval <julia.koval@intel.com>
* gcc.target/i386/avx512f-vpshrdv-1.c: New test.
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512f -mavx512vbmi2 -mavx512bw -O2" } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+\[^\n\r]*%zmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\\n\\r]*%zmm\[0-9\]+\[^\\n\\r\]*%zmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+
+
+
+#include <immintrin.h>
+
+volatile __m512i x,y,z,z1;
+volatile __mmask32 m32;
+volatile __mmask16 m16;
+volatile __mmask8 m8;
+
+void extern
+avx512f_test (void)
+{
+ x = _mm512_shldv_epi16 (x, y, z);
+ x = _mm512_mask_shldv_epi16 (x, m32, y, z);
+ x = _mm512_maskz_shldv_epi16 (m32, x, y, z);
+
+ x = _mm512_shldv_epi32 (x, y, z);
+ x = _mm512_mask_shldv_epi32 (x, m16, y, z);
+ x = _mm512_maskz_shldv_epi32 (m16, x, y, z);
+
+ x = _mm512_shldv_epi64 (x, y, z);
+ x = _mm512_mask_shldv_epi64 (x, m8, y, z);
+ x = _mm512_maskz_shldv_epi64 (m8, x, y, z);
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512f -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512f } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512F
+
+#define AVX512VBMI2
+#include "avx512f-helper.h"
+
+#define SIZE (AVX512F_LEN / 32)
+#include "avx512f-mask-type.h"
+
+static void
+CALC (int *r, int *dst, int *s1, int *s2)
+{
+ int i;
+ for (i = 0; i < SIZE; i++)
+ {
+ r[i] = (dst[i] << (s2[i] & 31)) | (s1[i] >> (32 - (s2[i] & 31)));
+ }
+}
+
+void
+TEST (void)
+{
+ int i;
+ UNION_TYPE (AVX512F_LEN, i_d) res1, res2, res3, src1, src2;
+ MASK_TYPE mask = MASK_VALUE;
+ int res_ref[SIZE];
+
+ for (i = 0; i < SIZE; i++)
+ {
+ src1.a[i] = 2 + i;
+ src2.a[i] = 1 + 3*i;
+ }
+
+ for (i = 0; i < SIZE; i++)
+ {
+ res1.a[i] = DEFAULT_VALUE;
+ res2.a[i] = DEFAULT_VALUE;
+ res3.a[i] = DEFAULT_VALUE;
+ }
+
+ CALC (res_ref, res1.a, src1.a, src2.a);
+
+ res1.x = INTRINSIC (_shldv_epi32) (res1.x, src1.x, src2.x);
+ res2.x = INTRINSIC (_mask_shldv_epi32) (res2.x, mask, src1.x, src2.x);
+ res3.x = INTRINSIC (_maskz_shldv_epi32) (mask, res3.x, src1.x, src2.x);
+
+ if (UNION_CHECK (AVX512F_LEN, i_d) (res1, res_ref))
+ abort ();
+
+ MASK_MERGE (i_d) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_d) (res2, res_ref))
+ abort ();
+
+ MASK_ZERO (i_d) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_d) (res3, res_ref))
+ abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512f -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512f } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512F
+
+#define AVX512VBMI2
+#include "avx512f-helper.h"
+
+#define SIZE (AVX512F_LEN / 64)
+#include "avx512f-mask-type.h"
+
+static void
+CALC (unsigned long long *r, unsigned long long *dst, unsigned long long *s1, unsigned long long *s2)
+{
+ int i;
+ for (i = 0; i < SIZE; i++)
+ {
+ r[i] = (dst[i] << (s2[i] & 63)) | (s1[i] >> (64 - (s2[i] & 63)));
+ }
+}
+
+void
+TEST (void)
+{
+ int i;
+ UNION_TYPE (AVX512F_LEN, i_q) res1, res2, res3, src1, src2;
+ MASK_TYPE mask = MASK_VALUE;
+ unsigned long long res_ref[SIZE];
+
+ for (i = 0; i < SIZE; i++)
+ {
+ src1.a[i] = 2 + i;
+ src2.a[i] = 1 + 3*i;
+ }
+
+ for (i = 0; i < SIZE; i++)
+ {
+ res1.a[i] = DEFAULT_VALUE;
+ res2.a[i] = DEFAULT_VALUE;
+ res3.a[i] = DEFAULT_VALUE;
+ }
+
+ CALC (res_ref, res1.a, src1.a, src2.a);
+
+ res1.x = INTRINSIC (_shldv_epi64) (res1.x, src1.x, src2.x);
+ res2.x = INTRINSIC (_mask_shldv_epi64) (res2.x, mask, src1.x, src2.x);
+ res3.x = INTRINSIC (_maskz_shldv_epi64) (mask, res3.x, src1.x, src2.x);
+
+ if (UNION_CHECK (AVX512F_LEN, i_q) (res1, res_ref))
+ abort ();
+
+ MASK_MERGE (i_q) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_q) (res2, res_ref))
+ abort ();
+
+ MASK_ZERO (i_q) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_q) (res3, res_ref))
+ abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512f -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512f } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512F
+
+#define AVX512VBMI2
+#include "avx512f-helper.h"
+
+#define SIZE (AVX512F_LEN / 16)
+#include "avx512f-mask-type.h"
+
+static void
+CALC (short *r, short *dst, short *s1, short *s2)
+{
+ int i;
+ for (i = 0; i < SIZE; i++)
+ {
+ r[i] = (dst[i] << (s2[i] & 15)) | (s1[i] >> (16 - (s2[i] & 15)));
+ }
+}
+
+void
+TEST (void)
+{
+ int i;
+ UNION_TYPE (AVX512F_LEN, i_w) res1, res2, res3, src1, src2;
+ MASK_TYPE mask = MASK_VALUE;
+ short res_ref[SIZE];
+
+ for (i = 0; i < SIZE; i++)
+ {
+ src1.a[i] = 2 + i;
+ src2.a[i] = 1 + 3*i;
+ }
+
+ for (i = 0; i < SIZE; i++)
+ {
+ res1.a[i] = DEFAULT_VALUE;
+ res2.a[i] = DEFAULT_VALUE;
+ res3.a[i] = DEFAULT_VALUE;
+ }
+
+ CALC (res_ref, res1.a, src1.a, src2.a);
+
+ res1.x = INTRINSIC (_shldv_epi16) (res1.x, src1.x, src2.x);
+ res2.x = INTRINSIC (_mask_shldv_epi16) (res2.x, mask, src1.x, src2.x);
+ res3.x = INTRINSIC (_maskz_shldv_epi16) (mask, res3.x, src1.x, src2.x);
+
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res1, res_ref))
+ abort ();
+
+ MASK_MERGE (i_w) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res2, res_ref))
+ abort ();
+
+ MASK_ZERO (i_w) (res_ref, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res3, res_ref))
+ abort ();
+}
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-mavx512vl -mavx512vbmi2 -mavx512bw -O2" } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\\n\\r]*%ymm\[0-9\]+\[^\\n\\r\]*%ymm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvw\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vpshldvq\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\\n\\r]*%xmm\[0-9\]+\[^\\n\\r\]*%xmm\[0-9\]+\{%k\[1-7\]\}\{z\}(?:\n|\[ \\t\]+#)" 1 } } */
+
+#include <immintrin.h>
+
+volatile __m256i x,y,z;
+volatile __m128i x_,y_,z_;
+volatile __mmask32 m;
+
+void extern
+avx512f_test (void)
+{
+ x = _mm256_shldv_epi16 (x, y, z);
+ x = _mm256_mask_shldv_epi16 (x, m, y, z);
+ x = _mm256_maskz_shldv_epi16 (m, x, y, z);
+
+ x = _mm256_shldv_epi32 (x, y, z);
+ x = _mm256_mask_shldv_epi32 (x, m, y, z);
+ x = _mm256_maskz_shldv_epi32 (m, x, y, z);
+
+ x = _mm256_shldv_epi64 (x, y, z);
+ x = _mm256_mask_shldv_epi64 (x, m, y, z);
+ x = _mm256_maskz_shldv_epi64 (m, x, y, z);
+
+ x_ = _mm_shldv_epi16 (x_, y_, z_);
+ x_ = _mm_mask_shldv_epi16 (x_, m, y_, z_);
+ x_ = _mm_maskz_shldv_epi16 (m, x_, y_, z_);
+
+ x_ = _mm_shldv_epi32 (x_, y_, z_);
+ x_ = _mm_mask_shldv_epi32 (x_, m, y_, z_);
+ x_ = _mm_maskz_shldv_epi32 (m, x_, y_, z_);
+
+ x_ = _mm_shldv_epi64 (x_, y_, z_);
+ x_ = _mm_mask_shldv_epi64 (x_, m, y_, z_);
+ x_ = _mm_maskz_shldv_epi64 (m, x_, y_, z_);
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512vl -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512vl } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512VL
+#define AVX512F_LEN 256
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvd-2.c"
+
+#undef AVX512F_LEN
+#undef AVX512F_LEN_HALF
+
+#define AVX512F_LEN 128
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvd-2.c"
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512vl -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512vl } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512VL
+#define AVX512F_LEN 256
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvq-2.c"
+
+#undef AVX512F_LEN
+#undef AVX512F_LEN_HALF
+
+#define AVX512F_LEN 128
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvq-2.c"
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512vl -mavx512vbmi2 -mavx512bw" } */
+/* { dg-require-effective-target avx512vl } */
+/* { dg-require-effective-target avx512vbmi2 } */
+
+#define AVX512VL
+#define AVX512F_LEN 256
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvw-2.c"
+
+#undef AVX512F_LEN
+#undef AVX512F_LEN_HALF
+
+#define AVX512F_LEN 128
+#define AVX512F_LEN_HALF 128
+#include "avx512f-vpshldvw-2.c"
projects / gcc.git / commitdiff