From: Craig Topper <craig.topper@intel.com>
Date: Wed, 30 May 2018 18:27:07 +0000 (+0000)
Subject: [X86] Simplify the implementation of _mm_sqrt_ss, _mm_rcp_ss, and _mm_rsqrt_ss.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=9107c536df587b118caf56544b8fd3ad986dcd38;p=clang

[X86] Simplify the implementation of _mm_sqrt_ss, _mm_rcp_ss, and _mm_rsqrt_ss.

We don't need the insertion back into the original vector at the end. The builtin already understands that.

This is different than _mm_sqrt_sd which takes two arguments and we do need to insert.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@333572 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Headers/xmmintrin.h b/lib/Headers/xmmintrin.h
index 75d1c9adb3..500e1026fa 100644
--- a/lib/Headers/xmmintrin.h
+++ b/lib/Headers/xmmintrin.h
@@ -224,8 +224,7 @@ _mm_div_ps(__m128 __a, __m128 __b)
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_sqrt_ss(__m128 __a)
 {
-  __m128 __c = __builtin_ia32_sqrtss((__v4sf)__a);
-  return (__m128) { __c[0], __a[1], __a[2], __a[3] };
+  return (__m128)__builtin_ia32_sqrtss((__v4sf)__a);
 }
 
 /// Calculates the square roots of the values stored in a 128-bit vector
@@ -260,8 +259,7 @@ _mm_sqrt_ps(__m128 __a)
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_rcp_ss(__m128 __a)
 {
-  __m128 __c = __builtin_ia32_rcpss((__v4sf)__a);
-  return (__m128) { __c[0], __a[1], __a[2], __a[3] };
+  return (__m128)__builtin_ia32_rcpss((__v4sf)__a);
 }
 
 /// Calculates the approximate reciprocals of the values stored in a
@@ -278,7 +276,7 @@ _mm_rcp_ss(__m128 __a)
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_rcp_ps(__m128 __a)
 {
-  return __builtin_ia32_rcpps((__v4sf)__a);
+  return (__m128)__builtin_ia32_rcpps((__v4sf)__a);
 }
 
 /// Calculates the approximate reciprocal of the square root of the value
@@ -297,8 +295,7 @@ _mm_rcp_ps(__m128 __a)
 static __inline__ __m128 __DEFAULT_FN_ATTRS
 _mm_rsqrt_ss(__m128 __a)
 {
-  __m128 __c = __builtin_ia32_rsqrtss((__v4sf)__a);
-  return (__m128) { __c[0], __a[1], __a[2], __a[3] };
+  return __builtin_ia32_rsqrtss((__v4sf)__a);
 }
 
 /// Calculates the approximate reciprocals of the square roots of the
diff --git a/test/CodeGen/sse-builtins.c b/test/CodeGen/sse-builtins.c
index 7e446ba2b0..18e51349fb 100644
--- a/test/CodeGen/sse-builtins.c
+++ b/test/CodeGen/sse-builtins.c
@@ -508,14 +508,6 @@ __m128 test_mm_rcp_ps(__m128 x) {
 __m128 test_mm_rcp_ss(__m128 x) {
   // CHECK-LABEL: test_mm_rcp_ss
   // CHECK: call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> {{.*}})
-  // CHECK: extractelement <4 x float> {{.*}}, i32 0
-  // CHECK: insertelement <4 x float> undef, float {{.*}}, i32 0
-  // CHECK: extractelement <4 x float> {{.*}}, i32 1
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 1
-  // CHECK: extractelement <4 x float> {{.*}}, i32 2
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 2
-  // CHECK: extractelement <4 x float> {{.*}}, i32 3
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 3
   return _mm_rcp_ss(x);
 }
 
@@ -528,14 +520,6 @@ __m128 test_mm_rsqrt_ps(__m128 x) {
 __m128 test_mm_rsqrt_ss(__m128 x) {
   // CHECK-LABEL: test_mm_rsqrt_ss
   // CHECK: call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> {{.*}})
-  // CHECK: extractelement <4 x float> {{.*}}, i32 0
-  // CHECK: insertelement <4 x float> undef, float {{.*}}, i32 0
-  // CHECK: extractelement <4 x float> {{.*}}, i32 1
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 1
-  // CHECK: extractelement <4 x float> {{.*}}, i32 2
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 2
-  // CHECK: extractelement <4 x float> {{.*}}, i32 3
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 3
   return _mm_rsqrt_ss(x);
 }
 
@@ -662,14 +646,6 @@ __m128 test_mm_sqrt_ps(__m128 x) {
 __m128 test_sqrt_ss(__m128 x) {
   // CHECK: define {{.*}} @test_sqrt_ss
   // CHECK: call <4 x float> @llvm.x86.sse.sqrt.ss
-  // CHECK: extractelement <4 x float> {{.*}}, i32 0
-  // CHECK: insertelement <4 x float> undef, float {{.*}}, i32 0
-  // CHECK: extractelement <4 x float> {{.*}}, i32 1
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 1
-  // CHECK: extractelement <4 x float> {{.*}}, i32 2
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 2
-  // CHECK: extractelement <4 x float> {{.*}}, i32 3
-  // CHECK: insertelement <4 x float> {{.*}}, float {{.*}}, i32 3
   return _mm_sqrt_ss(x);
 }