[AVX512] Add sub-vector FP extracts

Analogous to AVX2, these need to be implemented as macros to properly
propagate the immediate index operand.

Part of <rdar://problem/17688758>

llvm-svn: 226496

diff --git a/clang/include/clang/Basic/BuiltinsX86.def b/clang/include/clang/Basic/BuiltinsX86.def
index 5a28aaa..9a6b685 100644 (file)
--- a/clang/include/clang/Basic/BuiltinsX86.def
+++ b/clang/include/clang/Basic/BuiltinsX86.def
@@ -912,6 +912,8 @@ BUILTIN(__builtin_ia32_vpermt2varps512_mask, "V16fV16iV16fV16fUs", "")
 BUILTIN(__builtin_ia32_vpermt2varpd512_mask, "V8dV8LLiV8dV8dUc", "")
 BUILTIN(__builtin_ia32_alignq512_mask, "V8LLiV8LLiV8LLiUcV8LLiUc", "")
 BUILTIN(__builtin_ia32_alignd512_mask, "V16iV16iV16iUcV16iUc", "")
+BUILTIN(__builtin_ia32_extractf64x4_mask, "V4dV8dIcV4dUc", "")
+BUILTIN(__builtin_ia32_extractf32x4_mask, "V4fV16fIcV4fUc", "")
 BUILTIN(__builtin_ia32_gathersiv8df, "V8dV8dvC*V8iUciC", "")
 BUILTIN(__builtin_ia32_gathersiv16sf, "V16fV16fvC*UsiC", "")
 BUILTIN(__builtin_ia32_gatherdiv8df, "V8dV8dvC*V8LLiUciC", "")

diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index e6f128d..eda1b5c 100644 (file)
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -633,6 +633,24 @@ _mm512_valign_epi32(__m512i __A, __m512i __B, const int __I)
                                                 (__mmask16) -1);
 }
 
+/* Vector Extract */
+
+#define _mm512_extractf64x4_pd(A, I) __extension__ ({                    \
+      __m512d __A = (A);                                                 \
+      (__m256d)                                                          \
+        __builtin_ia32_extractf64x4_mask((__v8df)__A,                    \
+                                         (I),                            \
+                                         (__v4df)_mm256_setzero_si256(), \
+                                         (__mmask8) -1); })
+
+#define _mm512_extractf32x4_ps(A, I) __extension__ ({                    \
+      __m512 __A = (A);                                                  \
+      (__m128)                                                           \
+        __builtin_ia32_extractf32x4_mask((__v16sf)__A,                   \
+                                         (I),                            \
+                                         (__v4sf)_mm_setzero_ps(),       \
+                                         (__mmask8) -1); })
+
 /* Vector Blend */
 
 static __inline __m512d __attribute__ ((__always_inline__, __nodebug__))

diff --git a/clang/test/CodeGen/avx512f-builtins.c b/clang/test/CodeGen/avx512f-builtins.c
index 5caf314..04cf3cc 100644 (file)
--- a/clang/test/CodeGen/avx512f-builtins.c
+++ b/clang/test/CodeGen/avx512f-builtins.c
@@ -264,3 +264,17 @@ __mmask8 test_mm512_cmp_pd_mask(__m512 __a, __m512 __b) {
   // CHECK: @llvm.x86.avx512.mask.cmp.pd.512
   return _mm512_cmp_pd_mask(__a, __b, 0);
 }
+
+__m256d test_mm512_extractf64x4_pd(__m512d a)
+{
+  // CHECK-LABEL: @test_mm512_extractf64x4_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x4.512
+  return _mm512_extractf64x4_pd(a, 1);
+}
+
+__m128 test_mm512_extractf32x4_ps(__m512 a)
+{
+  // CHECK-LABEL: @test_mm512_extractf32x4_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x4.512
+  return _mm512_extractf32x4_ps(a, 1);
+}