[PATCH] D18930: Adding avx512 (shuf, sqrt{ss|sd}, rsqrt ) builtin to clang

[Asaf Badouh via llvm-commits]

AsafBadouh added inline comments.

================
Comment at: lib/Headers/avx512fintrin.h:4287
@@ +4286,3 @@
+                 (__mmask8) __U,\
+                 _MM_FROUND_CUR_DIRECTION);\
+})
----------------
last parameter should be __R

================
Comment at: test/CodeGen/avx512f-builtins.c:2581
@@ +2580,3 @@
+  // CHECK-LABEL: @test_mm512_shuffle_f32x4
+  // CHECK: @llvm.x86.avx512.mask.shuf
+  return _mm512_shuffle_f32x4(__A, __B, 4); 
----------------
please add the suffix of the intrinsic

================
Comment at: test/CodeGen/avx512f-builtins.c:2587
@@ +2586,3 @@
+  // CHECK-LABEL: @test_mm512_mask_shuffle_f32x4
+  // CHECK: @llvm.x86.avx512.mask.shuf
+  return _mm512_mask_shuffle_f32x4(__W, __U, __A, __B, 4); 
----------------
please add the suffix of the intrinsic

================
Comment at: test/CodeGen/avx512vl-builtins.c:4302
@@ +4301,3 @@
+  // CHECK: @llvm.x86.avx512.mask.shuf
+  return _mm256_shuffle_f32x4(__A, __B, _MM_FROUND_CUR_DIRECTION); 
+}
----------------
please add the suffix the intrinsic

================
Comment at: test/CodeGen/avx512vl-builtins.c:4302
@@ +4301,3 @@
+  // CHECK: @llvm.x86.avx512.mask.shuf
+  return _mm256_shuffle_f32x4(__A, __B, _MM_FROUND_CUR_DIRECTION); 
+}
----------------
AsafBadouh wrote:
> please add the suffix the intrinsic
there is no rounding mode in shuffle.
I think you want to use the enum you declared in avx512fintrin.h (_MM_PERM_*)


http://reviews.llvm.org/D18930