[Clang][Builtin][AVX512] Adding builtin to clang . [cvt{b|d|q}2mask{128|256|512} cvtmask2{b|d|q}{128|256|512}]

[AsafBadouh]

LGTM

[delena]

include/clang/Basic/BuiltinsX86.def

1958 ↗

(On Diff #53376)

vl and dq

1964 ↗

(On Diff #53376)

the same

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

 TARGET_BUILTIN(__builtin_ia32_shufps128_mask, "V4fV4fV4fIiV4fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_shufps256_mask, "V8fV8fV8fIiV8fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_sqrtsd_round_mask, "V2dV2dV2dV2dUcIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_sqrtss_round_mask, "V4fV4fV4fV4fUcIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_rsqrt14pd128_mask, "V2dV2dV2dUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_rsqrt14pd256_mask, "V4dV4dV4dUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_rsqrt14ps128_mask, "V4fV4fV4fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_rsqrt14ps256_mask, "V8fV8fV8fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtb2mask512, "ULLiV64c","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2b512, "V64cULLi","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2w512, "V32sUi","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_cvtd2mask512, "UsV16i","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2d512, "V16iUs","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2q512, "V8LLiUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_cvtq2mask512, "UcV8LLi","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_cvtb2mask128, "UsV16c","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtb2mask256, "UiV32c","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2b128, "V16cUs","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2b256, "V32cUi","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2w128, "V8sUc","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2w256, "V16sUs","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtd2mask128, "UcV4i","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtd2mask256, "UcV8i","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2d128, "V4iUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2d256, "V8iUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2q128, "V2LLiUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtmask2q256, "V4LLiUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtq2mask128, "UcV2LLi","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_cvtq2mask256, "UcV4LLi","","avx512dq,avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512bwintrin.h

 
 static __inline__ __mmask32 __DEFAULT_FN_ATTRS
 _mm512_mask_testn_epi16_mask (__mmask32 __U, __m512i __A, __m512i __B)
 {
   return (__mmask32) __builtin_ia32_ptestnmw512 ((__v32hi) __A,
              (__v32hi) __B, __U);
 }
 
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS
+_mm512_movepi8_mask (__m512i __A)
+{
+  return (__mmask64) __builtin_ia32_cvtb2mask512 ((__v64qi) __A);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_movm_epi8 (__mmask64 __A)
+{
+  return (__m512i) __builtin_ia32_cvtmask2b512 (__A);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_movm_epi16 (__mmask32 __A)
+{
+  return (__m512i) __builtin_ia32_cvtmask2w512 (__A);
+}
+
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/lib/Headers/avx512dqintrin.h

 #define _mm512_mask_reduce_round_ps(__W, __U, __A, __B, __R) __extension__ ({\
   (__m512) __builtin_ia32_reduceps512_mask ((__v16sf) __A, __B,      \
                (__v16sf) __W, (__mmask16) __U, __R);})
 
 #define _mm512_maskz_reduce_round_ps(__U, __A, __B, __R) __extension__ ({\
   (__m512) __builtin_ia32_reduceps512_mask ((__v16sf) __A, __B,      \
                (__v16sf) _mm512_setzero_ps(), (__mmask16) __U, __R);})
 
+static __inline__ __mmask16 __DEFAULT_FN_ATTRS
+_mm512_movepi32_mask (__m512i __A)
+{
+  return (__mmask16) __builtin_ia32_cvtd2mask512 ((__v16si) __A);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_movm_epi32 (__mmask16 __A)
+{
+  return (__m512i) __builtin_ia32_cvtmask2d512 (__A);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_movm_epi64 (__mmask8 __A)
+{
+  return (__m512i) __builtin_ia32_cvtmask2q512 (__A);
+}
+
+static __inline__ __mmask8 __DEFAULT_FN_ATTRS
+_mm512_movepi64_mask (__m512i __A)
+{
+  return (__mmask8) __builtin_ia32_cvtq2mask512 ((__v8di) __A);
+}
+
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/lib/Headers/avx512vlbwintrin.h

 
 static __inline__ __mmask16 __DEFAULT_FN_ATTRS
 _mm256_mask_testn_epi16_mask (__mmask16 __U, __m256i __A, __m256i __B)
 {
   return (__mmask16) __builtin_ia32_ptestnmw256 ((__v16hi) __A,
              (__v16hi) __B, __U);
 }
 
+static __inline__ __mmask16 __DEFAULT_FN_ATTRS
+_mm_movepi8_mask (__m128i __A)
+{
+  return (__mmask16) __builtin_ia32_cvtb2mask128 ((__v16qi) __A);
+}
+
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS
+_mm256_movepi8_mask (__m256i __A)
+{
+  return (__mmask32) __builtin_ia32_cvtb2mask256 ((__v32qi) __A);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_movm_epi8 (__mmask16 __A)
+{
+  return (__m128i) __builtin_ia32_cvtmask2b128 (__A);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_movm_epi8 (__mmask32 __A)
+{
+  return (__m256i) __builtin_ia32_cvtmask2b256 (__A);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_movm_epi16 (__mmask8 __A)
+{
+  return (__m128i) __builtin_ia32_cvtmask2w128 (__A);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_movm_epi16 (__mmask16 __A)
+{
+  return (__m256i) __builtin_ia32_cvtmask2w256 (__A);
+}
+
 
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __AVX512VLBWINTRIN_H */

cfe/trunk/lib/Headers/avx512vldqintrin.h

 #define _mm256_mask_reduce_ps(__W, __U, __A, __B) __extension__ ({ \
   (__m256) __builtin_ia32_reduceps256_mask ((__v8sf) __A, __B,     \
                 (__v8sf) __W, (__mmask8) __U); })
 
 #define _mm256_maskz_reduce_ps(__U, __A, __B) __extension__ ({     \
   (__m256) __builtin_ia32_reduceps256_mask ((__v8sf) __A, __B,     \
                 (__v8sf) _mm256_setzero_ps(), (__mmask8) __U); })
 
+static __inline__ __mmask8 __DEFAULT_FN_ATTRS
+_mm_movepi32_mask (__m128i __A)
+{
+  return (__mmask8) __builtin_ia32_cvtd2mask128 ((__v4si) __A);
+}
+
+static __inline__ __mmask8 __DEFAULT_FN_ATTRS
+_mm256_movepi32_mask (__m256i __A)
+{
+  return (__mmask8) __builtin_ia32_cvtd2mask256 ((__v8si) __A);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_movm_epi32 (__mmask8 __A)
+{
+  return (__m128i) __builtin_ia32_cvtmask2d128 (__A);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_movm_epi32 (__mmask8 __A)
+{
+  return (__m256i) __builtin_ia32_cvtmask2d256 (__A);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_movm_epi64 (__mmask8 __A)
+{
+  return (__m128i) __builtin_ia32_cvtmask2q128 (__A);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_movm_epi64 (__mmask8 __A)
+{
+  return (__m256i) __builtin_ia32_cvtmask2q256 (__A);
+}
+
+static __inline__ __mmask8 __DEFAULT_FN_ATTRS
+_mm_movepi64_mask (__m128i __A)
+{
+  return (__mmask8) __builtin_ia32_cvtq2mask128 ((__v2di) __A);
+}
+
+static __inline__ __mmask8 __DEFAULT_FN_ATTRS
+_mm256_movepi64_mask (__m256i __A)
+{
+  return (__mmask8) __builtin_ia32_cvtq2mask256 ((__v4di) __A);
+}
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/test/CodeGen/avx512bw-builtins.c

 }
 
 __mmask32 test_mm512_mask_testn_epi16_mask(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_testn_epi16_mask
   // CHECK: @llvm.x86.avx512.ptestnm.w.
   return _mm512_mask_testn_epi16_mask(__U, __A, __B); 
 }
 
+__mmask64 test_mm512_movepi8_mask(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_movepi8_mask
+  // CHECK: @llvm.x86.avx512.cvtb2mask.512
+  return _mm512_movepi8_mask(__A); 
+}
+
+__m512i test_mm512_movm_epi8(__mmask64 __A) {
+  // CHECK-LABEL: @test_mm512_movm_epi8
+  // CHECK: @llvm.x86.avx512.cvtmask2b.512
+  return _mm512_movm_epi8(__A); 
+}
+
+__m512i test_mm512_movm_epi16(__mmask32 __A) {
+  // CHECK-LABEL: @test_mm512_movm_epi16
+  // CHECK: @llvm.x86.avx512.cvtmask2w.512
+  return _mm512_movm_epi16(__A); 
+}
+
+

cfe/trunk/test/CodeGen/avx512dq-builtins.c

 }
 
 __m512 test_mm512_maskz_reduce_round_ps(__mmask16 __U, __m512 __A) {
   // CHECK-LABEL: @test_mm512_maskz_reduce_round_ps
   // CHECK: @llvm.x86.avx512.mask.reduce.ps.512
   return _mm512_maskz_reduce_round_ps(__U, __A, 4, 8); 
 }
 
+__mmask16 test_mm512_movepi32_mask(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_movepi32_mask
+  // CHECK: @llvm.x86.avx512.cvtd2mask.512
+  return _mm512_movepi32_mask(__A); 
+}
+
+__m512i test_mm512_movm_epi32(__mmask16 __A) {
+  // CHECK-LABEL: @test_mm512_movm_epi32
+  // CHECK: @llvm.x86.avx512.cvtmask2d.512
+  return _mm512_movm_epi32(__A); 
+}
+
+__m512i test_mm512_movm_epi64(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm512_movm_epi64
+  // CHECK: @llvm.x86.avx512.cvtmask2q.512
+  return _mm512_movm_epi64(__A); 
+}
+
+__mmask8 test_mm512_movepi64_mask(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_movepi64_mask
+  // CHECK: @llvm.x86.avx512.cvtq2mask.512
+  return _mm512_movepi64_mask(__A); 
+}
+

cfe/trunk/test/CodeGen/avx512vlbw-builtins.c

 }
 
 __mmask16 test_mm256_mask_testn_epi16_mask(__mmask16 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_testn_epi16_mask
   // CHECK: @llvm.x86.avx512.ptestnm.w.256
   return _mm256_mask_testn_epi16_mask(__U, __A, __B); 
 }
 
+__mmask16 test_mm_movepi8_mask(__m128i __A) {
+  // CHECK-LABEL: @test_mm_movepi8_mask
+  // CHECK: @llvm.x86.avx512.cvtb2mask.128
+  return _mm_movepi8_mask(__A); 
+}
+
+__mmask32 test_mm256_movepi8_mask(__m256i __A) {
+  // CHECK-LABEL: @test_mm256_movepi8_mask
+  // CHECK: @llvm.x86.avx512.cvtb2mask.256
+  return _mm256_movepi8_mask(__A); 
+}
+
+__m128i test_mm_movm_epi8(__mmask16 __A) {
+  // CHECK-LABEL: @test_mm_movm_epi8
+  // CHECK: @llvm.x86.avx512.cvtmask2b.128
+  return _mm_movm_epi8(__A); 
+}
+
+__m256i test_mm256_movm_epi8(__mmask32 __A) {
+  // CHECK-LABEL: @test_mm256_movm_epi8
+  // CHECK: @llvm.x86.avx512.cvtmask2b.256
+  return _mm256_movm_epi8(__A); 
+}
+
+__m128i test_mm_movm_epi16(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm_movm_epi16
+  // CHECK: @llvm.x86.avx512.cvtmask2w.128
+  return _mm_movm_epi16(__A); 
+}
+
+__m256i test_mm256_movm_epi16(__mmask16 __A) {
+  // CHECK-LABEL: @test_mm256_movm_epi16
+  // CHECK: @llvm.x86.avx512.cvtmask2w.256
+  return _mm256_movm_epi16(__A); 
+}
+
+
+

cfe/trunk/test/CodeGen/avx512vldq-builtins.c

__m256 test_mm256_mask_reduce_ps(__m256 __W, __mmask8 __U, __m256 __A) {
   return _mm256_mask_reduce_ps(__W, __U, __A, 4); 
 }
 
 __m256 test_mm256_maskz_reduce_ps(__mmask8 __U, __m256 __A) {
   // CHECK-LABEL: @test_mm256_maskz_reduce_ps
   // CHECK: @llvm.x86.avx512.mask.reduce.ps.256
   return _mm256_maskz_reduce_ps(__U, __A, 4); 
 }
+
+__mmask8 test_mm_movepi32_mask(__m128i __A) {
+  // CHECK-LABEL: @test_mm_movepi32_mask
+  // CHECK: @llvm.x86.avx512.cvtd2mask.128
+  return _mm_movepi32_mask(__A); 
+}
+
+__mmask8 test_mm256_movepi32_mask(__m256i __A) {
+  // CHECK-LABEL: @test_mm256_movepi32_mask
+  // CHECK: @llvm.x86.avx512.cvtd2mask.256
+  return _mm256_movepi32_mask(__A); 
+}
+
+__m128i test_mm_movm_epi32(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm_movm_epi32
+  // CHECK: @llvm.x86.avx512.cvtmask2d.128
+  return _mm_movm_epi32(__A); 
+}
+
+__m256i test_mm256_movm_epi32(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm256_movm_epi32
+  // CHECK: @llvm.x86.avx512.cvtmask2d.256
+  return _mm256_movm_epi32(__A); 
+}
+
+__m128i test_mm_movm_epi64(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm_movm_epi64
+  // CHECK: @llvm.x86.avx512.cvtmask2q.128
+  return _mm_movm_epi64(__A); 
+}
+
+__m256i test_mm256_movm_epi64(__mmask8 __A) {
+  // CHECK-LABEL: @test_mm256_movm_epi64
+  // CHECK: @llvm.x86.avx512.cvtmask2q.256
+  return _mm256_movm_epi64(__A); 
+}
+
+__mmask8 test_mm_movepi64_mask(__m128i __A) {
+  // CHECK-LABEL: @test_mm_movepi64_mask
+  // CHECK: @llvm.x86.avx512.cvtq2mask.128
+  return _mm_movepi64_mask(__A); 
+}
+
+__mmask8 test_mm256_movepi64_mask(__m256i __A) {
+  // CHECK-LABEL: @test_mm256_movepi64_mask
+  // CHECK: @llvm.x86.avx512.cvtq2mask.256
+  return _mm256_movepi64_mask(__A); 
+}
+