[CLANG][AVX512][BUILTIN]movap{d|s}{128|256|512}

[delena]

include/clang/Basic/BuiltinsX86.def

1696 ↗

(On Diff #49640)

it should be under VL

lib/Headers/avx512fintrin.h

3605 ↗

(On Diff #49640)

please remove 2 empty lines

lib/Headers/avx512vlintrin.h

5875 ↗

(On Diff #49640)

and here

cfe/trunk/include/clang/Basic/BuiltinsX86.def

 TARGET_BUILTIN(__builtin_ia32_expanddi512_mask, "V8LLiV8LLiV8LLiUc","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandloaddf512_mask, "V8dvC*V8dUc","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandloaddi512_mask, "V8LLivC*V8LLiUc","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandloadsf512_mask, "V16fvC*V16fUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandloadsi512_mask, "V16ivC*V16iUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandsf512_mask, "V16fV16fV16fUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_expandsi512_mask, "V16iV16iV16iUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_cvtps2pd512_mask, "V8dV8fV8dUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_movapd512_mask, "V8dV8dV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_movaps512_mask, "V16fV16fV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_movapd128_mask, "V2dV2dV2dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_movapd256_mask, "V4dV4dV4dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_movaps128_mask, "V4fV4fV4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_movaps256_mask, "V8fV8fV8fUc","","avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512fintrin.h

 {
   return (__m512d) __builtin_ia32_cvtps2pd512_mask ((__v8sf) __A,
                 (__v8df)
                 _mm512_setzero_pd (),
                 (__mmask8) __U,
                 _MM_FROUND_CUR_DIRECTION);
 }
 
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_mask_mov_pd (__m512d __W, __mmask8 __U, __m512d __A)
+{
+  return (__m512d) __builtin_ia32_movapd512_mask ((__v8df) __A,
+              (__v8df) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_mov_pd (__mmask8 __U, __m512d __A)
+{
+  return (__m512d) __builtin_ia32_movapd512_mask ((__v8df) __A,
+              (__v8df)
+              _mm512_setzero_pd (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_mov_ps (__m512 __W, __mmask16 __U, __m512 __A)
+{
+  return (__m512) __builtin_ia32_movaps512_mask ((__v16sf) __A,
+             (__v16sf) __W,
+             (__mmask16) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_mov_ps (__mmask16 __U, __m512 __A)
+{
+  return (__m512) __builtin_ia32_movaps512_mask ((__v16sf) __A,
+             (__v16sf)
+             _mm512_setzero_ps (),
+             (__mmask16) __U);
+}
+
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif // __AVX512FINTRIN_H

cfe/trunk/lib/Headers/avx512vlintrin.h

 
 #define _mm_maskz_shuffle_epi32( __U,  __A, __I) __extension__({\
   __builtin_ia32_pshufd128_mask ((__v4si) (__A), (__I),\
               (__v4si)\
               _mm_setzero_si128 (),\
               (__mmask8) (__U));\
 })
 
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_mask_mov_pd (__m128d __W, __mmask8 __U, __m128d __A)
+{
+  return (__m128d) __builtin_ia32_movapd128_mask ((__v2df) __A,
+              (__v2df) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_maskz_mov_pd (__mmask8 __U, __m128d __A)
+{
+  return (__m128d) __builtin_ia32_movapd128_mask ((__v2df) __A,
+              (__v2df)
+              _mm_setzero_pd (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_mask_mov_pd (__m256d __W, __mmask8 __U, __m256d __A)
+{
+  return (__m256d) __builtin_ia32_movapd256_mask ((__v4df) __A,
+              (__v4df) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_maskz_mov_pd (__mmask8 __U, __m256d __A)
+{
+  return (__m256d) __builtin_ia32_movapd256_mask ((__v4df) __A,
+              (__v4df)
+              _mm256_setzero_pd (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_mask_mov_ps (__m128 __W, __mmask8 __U, __m128 __A)
+{
+  return (__m128) __builtin_ia32_movaps128_mask ((__v4sf) __A,
+             (__v4sf) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_maskz_mov_ps (__mmask8 __U, __m128 __A)
+{
+  return (__m128) __builtin_ia32_movaps128_mask ((__v4sf) __A,
+             (__v4sf)
+             _mm_setzero_ps (),
+             (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_mask_mov_ps (__m256 __W, __mmask8 __U, __m256 __A)
+{
+  return (__m256) __builtin_ia32_movaps256_mask ((__v8sf) __A,
+             (__v8sf) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_maskz_mov_ps (__mmask8 __U, __m256 __A)
+{
+  return (__m256) __builtin_ia32_movaps256_mask ((__v8sf) __A,
+             (__v8sf)
+             _mm256_setzero_ps (),
+             (__mmask8) __U);
+}
+
+
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_BOTH
 
 #endif /* __AVX512VLINTRIN_H */

cfe/trunk/test/CodeGen/avx512f-builtins.c

__m512d test_mm512_mask_cvtps_pd(__m512d __W, __mmask8 __U, __m256 __A) {
   return _mm512_mask_cvtps_pd(__W, __U, __A); 
 }
 
 __m512d test_mm512_maskz_cvtps_pd(__mmask8 __U, __m256 __A) {
   // CHECK-LABEL: @test_mm512_maskz_cvtps_pd
   // CHECK: @llvm.x86.avx512.mask.cvtps2pd.512
   return _mm512_maskz_cvtps_pd(__U, __A); 
 }
+__m512d test_mm512_mask_mov_pd(__m512d __W, __mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_mask_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.512
+  return _mm512_mask_mov_pd(__W, __U, __A); 
+}
+
+__m512d test_mm512_maskz_mov_pd(__mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_maskz_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.512
+  return _mm512_maskz_mov_pd(__U, __A); 
+}
+
+__m512 test_mm512_mask_mov_ps(__m512 __W, __mmask16 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_mask_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.512
+  return _mm512_mask_mov_ps(__W, __U, __A); 
+}
+
+__m512 test_mm512_maskz_mov_ps(__mmask16 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_maskz_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.512
+  return _mm512_maskz_mov_ps(__U, __A); 
+}

cfe/trunk/test/CodeGen/avx512vl-builtins.c

 }
 
 __m256i test_mm256_maskz_shuffle_epi32(__mmask8 __U, __m256i __A) {
   // CHECK-LABEL: @test_mm256_maskz_shuffle_epi32
   // CHECK: @llvm.x86.avx512.mask.pshuf.d.256
   return _mm256_maskz_shuffle_epi32(__U, __A, 2); 
 }
 
+__m128d test_mm_mask_mov_pd(__m128d __W, __mmask8 __U, __m128d __A) {
+  // CHECK-LABEL: @test_mm_mask_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.128
+  return _mm_mask_mov_pd(__W, __U, __A); 
+}
+
+__m128d test_mm_maskz_mov_pd(__mmask8 __U, __m128d __A) {
+  // CHECK-LABEL: @test_mm_maskz_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.128
+  return _mm_maskz_mov_pd(__U, __A); 
+}
+
+__m256d test_mm256_mask_mov_pd(__m256d __W, __mmask8 __U, __m256d __A) {
+  // CHECK-LABEL: @test_mm256_mask_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.256
+  return _mm256_mask_mov_pd(__W, __U, __A); 
+}
+
+__m256d test_mm256_maskz_mov_pd(__mmask8 __U, __m256d __A) {
+  // CHECK-LABEL: @test_mm256_maskz_mov_pd
+  // CHECK: @llvm.x86.avx512.mask.mova.pd.256
+  return _mm256_maskz_mov_pd(__U, __A); 
+}
+
+__m128 test_mm_mask_mov_ps(__m128 __W, __mmask8 __U, __m128 __A) {
+  // CHECK-LABEL: @test_mm_mask_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.128
+  return _mm_mask_mov_ps(__W, __U, __A); 
+}
+
+__m128 test_mm_maskz_mov_ps(__mmask8 __U, __m128 __A) {
+  // CHECK-LABEL: @test_mm_maskz_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.128
+  return _mm_maskz_mov_ps(__U, __A); 
+}
+
+__m256 test_mm256_mask_mov_ps(__m256 __W, __mmask8 __U, __m256 __A) {
+  // CHECK-LABEL: @test_mm256_mask_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.256
+  return _mm256_mask_mov_ps(__W, __U, __A); 
+}
+
+__m256 test_mm256_maskz_mov_ps(__mmask8 __U, __m256 __A) {
+  // CHECK-LABEL: @test_mm256_maskz_mov_ps
+  // CHECK: @llvm.x86.avx512.mask.mova.ps.256
+  return _mm256_maskz_mov_ps(__U, __A); 
+}
+