[CLANG][AVX512][BUILTIN] Adding PSRLV builtin

Differential Revision: http://reviews.llvm.org/D17718

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@262326 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index ac7ccdfde179f8a50cbf084a732cbe51ab9e77f2..fb6270548c1d888b7c5a1db83dffb6d33bd2ef57 100644 (file)
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -1661,6 +1661,13 @@ TARGET_BUILTIN(__builtin_ia32_psllq128_mask, "V2LLiV2LLiV2LLiV2LLiUc","","avx512
 TARGET_BUILTIN(__builtin_ia32_psllq256_mask, "V4LLiV4LLiV2LLiV4LLiUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_psllqi128_mask, "V2LLiV2LLiiV2LLiUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_psllqi256_mask, "V4LLiV4LLiiV4LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv32hi_mask, "V32sV32sV32sV32sUi","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_psrlv16hi_mask, "V16sV16sV16sV16sUs","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv8hi_mask, "V8sV8sV8sV8sUc","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv2di_mask, "V2LLiV2LLiV2LLiV2LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv4di_mask, "V4LLiV4LLiV4LLiV4LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv4si_mask, "V4iV4iV4iV4iUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_psrlv8si_mask, "V8iV8iV8iV8iUc","","avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

diff --git a/lib/Headers/avx512bwintrin.h b/lib/Headers/avx512bwintrin.h
index 1eda638452842fdd379b2fa454ee9168b8e1ebb8..c1523cfa276d6132aad9a4917800119910d789f2 100644 (file)
--- a/lib/Headers/avx512bwintrin.h
+++ b/lib/Headers/avx512bwintrin.h
@@ -1718,6 +1718,37 @@ __builtin_ia32_psllwi512_mask ((__v32hi)( __A),( __B),\
               (__mmask32)( __U));\
 })
 
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_srlv_epi16 (__m512i __A, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_psrlv32hi_mask ((__v32hi) __A,
+              (__v32hi) __B,
+              (__v32hi)
+              _mm512_setzero_hi (),
+              (__mmask32) -1);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_srlv_epi16 (__m512i __W, __mmask32 __U, __m512i __A,
+      __m512i __B)
+{
+  return (__m512i) __builtin_ia32_psrlv32hi_mask ((__v32hi) __A,
+              (__v32hi) __B,
+              (__v32hi) __W,
+              (__mmask32) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_srlv_epi16 (__mmask32 __U, __m512i __A, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_psrlv32hi_mask ((__v32hi) __A,
+              (__v32hi) __B,
+              (__v32hi)
+              _mm512_setzero_hi (),
+              (__mmask32) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif

diff --git a/lib/Headers/avx512vlbwintrin.h b/lib/Headers/avx512vlbwintrin.h
index 831ca8a6d3adf99b545262728c469b494480c53d..8f0a60384c357f57ca7d233629a8935e73b5451e 100644 (file)
--- a/lib/Headers/avx512vlbwintrin.h
+++ b/lib/Headers/avx512vlbwintrin.h
@@ -2593,6 +2593,68 @@ __builtin_ia32_psllwi256_mask ((__v16hi)( __A),( __B),\
               (__mmask16)( __U));\
 })
 
+
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_srlv_epi16 (__m256i __A, __m256i __B)
+{
+  return (__m256i) __builtin_ia32_psrlv16hi_mask ((__v16hi) __A,
+              (__v16hi) __B,
+              (__v16hi)
+              _mm256_setzero_si256 (),
+              (__mmask16) -1);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_srlv_epi16 (__m256i __W, __mmask16 __U, __m256i __A,
+      __m256i __B)
+{
+  return (__m256i) __builtin_ia32_psrlv16hi_mask ((__v16hi) __A,
+              (__v16hi) __B,
+              (__v16hi) __W,
+              (__mmask16) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_srlv_epi16 (__mmask16 __U, __m256i __A, __m256i __B)
+{
+  return (__m256i) __builtin_ia32_psrlv16hi_mask ((__v16hi) __A,
+              (__v16hi) __B,
+              (__v16hi)
+              _mm256_setzero_si256 (),
+              (__mmask16) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_srlv_epi16 (__m128i __A, __m128i __B)
+{
+  return (__m128i) __builtin_ia32_psrlv8hi_mask ((__v8hi) __A,
+             (__v8hi) __B,
+             (__v8hi)
+             _mm_setzero_hi (),
+             (__mmask8) -1);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_srlv_epi16 (__m128i __W, __mmask8 __U, __m128i __A,
+         __m128i __B)
+{
+  return (__m128i) __builtin_ia32_psrlv8hi_mask ((__v8hi) __A,
+             (__v8hi) __B,
+             (__v8hi) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_srlv_epi16 (__mmask8 __U, __m128i __A, __m128i __B)
+{
+  return (__m128i) __builtin_ia32_psrlv8hi_mask ((__v8hi) __A,
+             (__v8hi) __B,
+             (__v8hi)
+             _mm_setzero_si128 (),
+             (__mmask8) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __AVX512VLBWINTRIN_H */

diff --git a/lib/Headers/avx512vlintrin.h b/lib/Headers/avx512vlintrin.h
index 0812cd768cfc9178a0bd8ebcbf6c1e602ea45a3d..731d7fe2619e7fa794a1136fe6902759e3b0374c 100644 (file)
--- a/lib/Headers/avx512vlintrin.h
+++ b/lib/Headers/avx512vlintrin.h
@@ -5556,6 +5556,88 @@ __builtin_ia32_psllv8si_mask ((__v8si)( __X),\
              (__mmask8)( __U));\
 })
 
+
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_srlv_epi64 (__m128i __W, __mmask8 __U, __m128i __X,
+         __m128i __Y)
+{
+  return (__m128i) __builtin_ia32_psrlv2di_mask ((__v2di) __X,
+             (__v2di) __Y,
+             (__v2di) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_srlv_epi64 (__mmask8 __U, __m128i __X, __m128i __Y)
+{
+  return (__m128i) __builtin_ia32_psrlv2di_mask ((__v2di) __X,
+             (__v2di) __Y,
+             (__v2di)
+             _mm_setzero_di (),
+             (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_srlv_epi64 (__m256i __W, __mmask8 __U, __m256i __X,
+      __m256i __Y)
+{
+  return (__m256i) __builtin_ia32_psrlv4di_mask ((__v4di) __X,
+             (__v4di) __Y,
+             (__v4di) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_srlv_epi64 (__mmask8 __U, __m256i __X, __m256i __Y)
+{
+  return (__m256i) __builtin_ia32_psrlv4di_mask ((__v4di) __X,
+             (__v4di) __Y,
+             (__v4di)
+             _mm256_setzero_si256 (),
+             (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_srlv_epi32 (__m128i __W, __mmask8 __U, __m128i __X,
+         __m128i __Y)
+{
+  return (__m128i) __builtin_ia32_psrlv4si_mask ((__v4si) __X,
+             (__v4si) __Y,
+             (__v4si) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_srlv_epi32 (__mmask8 __U, __m128i __X, __m128i __Y)
+{
+  return (__m128i) __builtin_ia32_psrlv4si_mask ((__v4si) __X,
+             (__v4si) __Y,
+             (__v4si)
+             _mm_setzero_si128 (),
+             (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_srlv_epi32 (__m256i __W, __mmask8 __U, __m256i __X,
+      __m256i __Y)
+{
+  return (__m256i) __builtin_ia32_psrlv8si_mask ((__v8si) __X,
+             (__v8si) __Y,
+             (__v8si) __W,
+             (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_srlv_epi32 (__mmask8 __U, __m256i __X, __m256i __Y)
+{
+  return (__m256i) __builtin_ia32_psrlv8si_mask ((__v8si) __X,
+             (__v8si) __Y,
+             (__v8si)
+             _mm256_setzero_si256 (),
+             (__mmask8) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_BOTH
 

diff --git a/test/CodeGen/avx512bw-builtins.c b/test/CodeGen/avx512bw-builtins.c
index 8520f969e050ac51c6dc97591838fb344d23d029..4fceab70584828aeb0c310d3acb59bb12f115c67 100644 (file)
--- a/test/CodeGen/avx512bw-builtins.c
+++ b/test/CodeGen/avx512bw-builtins.c
@@ -1167,4 +1167,22 @@ __m512i test_mm512_maskz_slli_epi16(__mmask32 __U, __m512i __A) {
   return _mm512_maskz_slli_epi16(__U, __A, 5); 
 }
 
+__m512i test_mm512_srlv_epi16(__m512i __A, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_srlv_epi16
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm512_srlv_epi16(__A, __B); 
+}
+
+__m512i test_mm512_mask_srlv_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_mask_srlv_epi16
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm512_mask_srlv_epi16(__W, __U, __A, __B); 
+}
+
+__m512i test_mm512_maskz_srlv_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_srlv_epi16
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm512_maskz_srlv_epi16(__U, __A, __B); 
+}
+
 

diff --git a/test/CodeGen/avx512vl-builtins.c b/test/CodeGen/avx512vl-builtins.c
index f417aff23eba5ff409e435bbe932cd5e4db920a2..2901f2612039760b2a505f7fa1bb4c874aeba49e 100644 (file)
--- a/test/CodeGen/avx512vl-builtins.c
+++ b/test/CodeGen/avx512vl-builtins.c
@@ -3732,3 +3732,51 @@ __m256i test_mm256_maskz_sllv_epi32(__mmask8 __U, __m256i __X, __m256i __Y) {
   return _mm256_maskz_sllv_epi32(__U, __X, __Y); 
 }
 
+__m128i test_mm_mask_srlv_epi64(__m128i __W, __mmask8 __U, __m128i __X, __m128i __Y) {
+  // CHECK-LABEL: @test_mm_mask_srlv_epi64
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm_mask_srlv_epi64(__W, __U, __X, __Y); 
+}
+
+__m128i test_mm_maskz_srlv_epi64(__mmask8 __U, __m128i __X, __m128i __Y) {
+  // CHECK-LABEL: @test_mm_maskz_srlv_epi64
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm_maskz_srlv_epi64(__U, __X, __Y); 
+}
+
+__m256i test_mm256_mask_srlv_epi64(__m256i __W, __mmask8 __U, __m256i __X, __m256i __Y) {
+  // CHECK-LABEL: @test_mm256_mask_srlv_epi64
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm256_mask_srlv_epi64(__W, __U, __X, __Y); 
+}
+
+__m256i test_mm256_maskz_srlv_epi64(__mmask8 __U, __m256i __X, __m256i __Y) {
+  // CHECK-LABEL: @test_mm256_maskz_srlv_epi64
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm256_maskz_srlv_epi64(__U, __X, __Y); 
+}
+
+__m128i test_mm_mask_srlv_epi32(__m128i __W, __mmask8 __U, __m128i __X, __m128i __Y) {
+  // CHECK-LABEL: @test_mm_mask_srlv_epi32
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm_mask_srlv_epi32(__W, __U, __X, __Y); 
+}
+
+__m128i test_mm_maskz_srlv_epi32(__mmask8 __U, __m128i __X, __m128i __Y) {
+  // CHECK-LABEL: @test_mm_maskz_srlv_epi32
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm_maskz_srlv_epi32(__U, __X, __Y); 
+}
+
+__m256i test_mm256_mask_srlv_epi32(__m256i __W, __mmask8 __U, __m256i __X, __m256i __Y) {
+  // CHECK-LABEL: @test_mm256_mask_srlv_epi32
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm256_mask_srlv_epi32(__W, __U, __X, __Y); 
+}
+
+__m256i test_mm256_maskz_srlv_epi32(__mmask8 __U, __m256i __X, __m256i __Y) {
+  // CHECK-LABEL: @test_mm256_maskz_srlv_epi32
+  // CHECK: @llvm.x86.avx512.mask.psrlv
+  return _mm256_maskz_srlv_epi32(__U, __X, __Y); 
+}
+