vendor/llvm/llvm-trunk-r305145

2 files changed, 162 insertions, 24 deletions

diff --git a/test/Transforms/SLPVectorizer/X86/arith-fp.ll b/test/Transforms/SLPVectorizer/X86/arith-fp.ll
index 7eec13e535d4..e00ed849ee4b 100644
--- a/test/Transforms/SLPVectorizer/X86/arith-fp.ll
+++ b/test/Transforms/SLPVectorizer/X86/arith-fp.ll
@@ -10,7 +10,7 @@
 
 define <2 x double> @buildvector_add_2f64(<2 x double> %a, <2 x double> %b) {
 ; CHECK-LABEL: @buildvector_add_2f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
@@ -30,7 +30,7 @@ define <2 x double> @buildvector_add_2f64(<2 x double> %a, <2 x double> %b) {
 
 define <2 x double> @buildvector_sub_2f64(<2 x double> %a, <2 x double> %b) {
 ; CHECK-LABEL: @buildvector_sub_2f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <2 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <2 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
@@ -50,7 +50,7 @@ define <2 x double> @buildvector_sub_2f64(<2 x double> %a, <2 x double> %b) {
 
 define <2 x double> @buildvector_mul_2f64(<2 x double> %a, <2 x double> %b) {
 ; CHECK-LABEL: @buildvector_mul_2f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <2 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <2 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
@@ -70,7 +70,7 @@ define <2 x double> @buildvector_mul_2f64(<2 x double> %a, <2 x double> %b) {
 
 define <2 x double> @buildvector_div_2f64(<2 x double> %a, <2 x double> %b) {
 ; CHECK-LABEL: @buildvector_div_2f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <2 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <2 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
@@ -90,7 +90,7 @@ define <2 x double> @buildvector_div_2f64(<2 x double> %a, <2 x double> %b) {
 
 define <4 x float> @buildvector_add_4f32(<4 x float> %a, <4 x float> %b) {
 ; CHECK-LABEL: @buildvector_add_4f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
@@ -122,7 +122,7 @@ define <4 x float> @buildvector_add_4f32(<4 x float> %a, <4 x float> %b) {
 
 define <4 x float> @buildvector_sub_4f32(<4 x float> %a, <4 x float> %b) {
 ; CHECK-LABEL: @buildvector_sub_4f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <4 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <4 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
@@ -154,7 +154,7 @@ define <4 x float> @buildvector_sub_4f32(<4 x float> %a, <4 x float> %b) {
 
 define <4 x float> @buildvector_mul_4f32(<4 x float> %a, <4 x float> %b) {
 ; CHECK-LABEL: @buildvector_mul_4f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <4 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <4 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
@@ -186,7 +186,7 @@ define <4 x float> @buildvector_mul_4f32(<4 x float> %a, <4 x float> %b) {
 
 define <4 x float> @buildvector_div_4f32(<4 x float> %a, <4 x float> %b) {
 ; CHECK-LABEL: @buildvector_div_4f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <4 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <4 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP1]], i32 1
@@ -222,7 +222,7 @@ define <4 x float> @buildvector_div_4f32(<4 x float> %a, <4 x float> %b) {
 
 define <4 x double> @buildvector_add_4f64(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: @buildvector_add_4f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x double> [[TMP1]], i32 1
@@ -254,7 +254,7 @@ define <4 x double> @buildvector_add_4f64(<4 x double> %a, <4 x double> %b) {
 
 define <4 x double> @buildvector_sub_4f64(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: @buildvector_sub_4f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <4 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <4 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x double> [[TMP1]], i32 1
@@ -286,7 +286,7 @@ define <4 x double> @buildvector_sub_4f64(<4 x double> %a, <4 x double> %b) {
 
 define <4 x double> @buildvector_mul_4f64(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: @buildvector_mul_4f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <4 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <4 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x double> [[TMP1]], i32 1
@@ -318,7 +318,7 @@ define <4 x double> @buildvector_mul_4f64(<4 x double> %a, <4 x double> %b) {
 
 define <4 x double> @buildvector_div_4f64(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: @buildvector_div_4f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <4 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <4 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <4 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x double> [[TMP1]], i32 1
@@ -350,7 +350,7 @@ define <4 x double> @buildvector_div_4f64(<4 x double> %a, <4 x double> %b) {
 
 define <8 x float> @buildvector_add_8f32(<8 x float> %a, <8 x float> %b) {
 ; CHECK-LABEL: @buildvector_add_8f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <8 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <8 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x float> [[TMP1]], i32 1
@@ -406,7 +406,7 @@ define <8 x float> @buildvector_add_8f32(<8 x float> %a, <8 x float> %b) {
 
 define <8 x float> @buildvector_sub_8f32(<8 x float> %a, <8 x float> %b) {
 ; CHECK-LABEL: @buildvector_sub_8f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <8 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <8 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x float> [[TMP1]], i32 1
@@ -462,7 +462,7 @@ define <8 x float> @buildvector_sub_8f32(<8 x float> %a, <8 x float> %b) {
 
 define <8 x float> @buildvector_mul_8f32(<8 x float> %a, <8 x float> %b) {
 ; CHECK-LABEL: @buildvector_mul_8f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <8 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <8 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x float> [[TMP1]], i32 1
@@ -518,7 +518,7 @@ define <8 x float> @buildvector_mul_8f32(<8 x float> %a, <8 x float> %b) {
 
 define <8 x float> @buildvector_div_8f32(<8 x float> %a, <8 x float> %b) {
 ; CHECK-LABEL: @buildvector_div_8f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <8 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <8 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x float> [[TMP1]], i32 1
@@ -578,7 +578,7 @@ define <8 x float> @buildvector_div_8f32(<8 x float> %a, <8 x float> %b) {
 
 define <8 x double> @buildvector_add_8f64(<8 x double> %a, <8 x double> %b) {
 ; CHECK-LABEL: @buildvector_add_8f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <8 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <8 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x double> [[TMP1]], i32 1
@@ -634,7 +634,7 @@ define <8 x double> @buildvector_add_8f64(<8 x double> %a, <8 x double> %b) {
 
 define <8 x double> @buildvector_sub_8f64(<8 x double> %a, <8 x double> %b) {
 ; CHECK-LABEL: @buildvector_sub_8f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <8 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <8 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x double> [[TMP1]], i32 1
@@ -690,7 +690,7 @@ define <8 x double> @buildvector_sub_8f64(<8 x double> %a, <8 x double> %b) {
 
 define <8 x double> @buildvector_mul_8f64(<8 x double> %a, <8 x double> %b) {
 ; CHECK-LABEL: @buildvector_mul_8f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <8 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <8 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x double> [[TMP1]], i32 1
@@ -746,7 +746,7 @@ define <8 x double> @buildvector_mul_8f64(<8 x double> %a, <8 x double> %b) {
 
 define <8 x double> @buildvector_div_8f64(<8 x double> %a, <8 x double> %b) {
 ; CHECK-LABEL: @buildvector_div_8f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <8 x double> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <8 x double> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x double> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <8 x double> undef, double [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x double> [[TMP1]], i32 1
@@ -802,7 +802,7 @@ define <8 x double> @buildvector_div_8f64(<8 x double> %a, <8 x double> %b) {
 
 define <16 x float> @buildvector_add_16f32(<16 x float> %a, <16 x float> %b) {
 ; CHECK-LABEL: @buildvector_add_16f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <16 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <16 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <16 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <16 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <16 x float> [[TMP1]], i32 1
@@ -906,7 +906,7 @@ define <16 x float> @buildvector_add_16f32(<16 x float> %a, <16 x float> %b) {
 
 define <16 x float> @buildvector_sub_16f32(<16 x float> %a, <16 x float> %b) {
 ; CHECK-LABEL: @buildvector_sub_16f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <16 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <16 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <16 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <16 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <16 x float> [[TMP1]], i32 1
@@ -1010,7 +1010,7 @@ define <16 x float> @buildvector_sub_16f32(<16 x float> %a, <16 x float> %b) {
 
 define <16 x float> @buildvector_mul_16f32(<16 x float> %a, <16 x float> %b) {
 ; CHECK-LABEL: @buildvector_mul_16f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fmul <16 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <16 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <16 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <16 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <16 x float> [[TMP1]], i32 1
@@ -1114,7 +1114,7 @@ define <16 x float> @buildvector_mul_16f32(<16 x float> %a, <16 x float> %b) {
 
 define <16 x float> @buildvector_div_16f32(<16 x float> %a, <16 x float> %b) {
 ; CHECK-LABEL: @buildvector_div_16f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <16 x float> %a, %b
+; CHECK-NEXT:    [[TMP1:%.*]] = fdiv <16 x float> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <16 x float> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[R0:%.*]] = insertelement <16 x float> undef, float [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <16 x float> [[TMP1]], i32 1
diff --git a/test/Transforms/SLPVectorizer/X86/reverse_extract_elements.ll b/test/Transforms/SLPVectorizer/X86/reverse_extract_elements.ll
new file mode 100644
index 000000000000..4c8748e220fd
--- /dev/null
+++ b/test/Transforms/SLPVectorizer/X86/reverse_extract_elements.ll
@@ -0,0 +1,138 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 | FileCheck %s
+
+define float @dotf(<4 x float> %x, <4 x float> %y) {
+; CHECK-LABEL: @dotf(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = fmul fast <4 x float> [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP1]], [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x float> [[BIN_RDX]], <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x float> [[BIN_RDX]], [[RDX_SHUF1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[BIN_RDX2]], i32 0
+; CHECK-NEXT:    ret float [[TMP2]]
+;
+entry:
+  %vecext = extractelement <4 x float> %x, i32 0
+  %vecext1 = extractelement <4 x float> %y, i32 0
+  %mul = fmul fast float %vecext, %vecext1
+  %vecext.1 = extractelement <4 x float> %x, i32 1
+  %vecext1.1 = extractelement <4 x float> %y, i32 1
+  %mul.1 = fmul fast float %vecext.1, %vecext1.1
+  %add.1 = fadd fast float %mul.1, %mul
+  %vecext.2 = extractelement <4 x float> %x, i32 2
+  %vecext1.2 = extractelement <4 x float> %y, i32 2
+  %mul.2 = fmul fast float %vecext.2, %vecext1.2
+  %add.2 = fadd fast float %mul.2, %add.1
+  %vecext.3 = extractelement <4 x float> %x, i32 3
+  %vecext1.3 = extractelement <4 x float> %y, i32 3
+  %mul.3 = fmul fast float %vecext.3, %vecext1.3
+  %add.3 = fadd fast float %mul.3, %add.2
+  ret float %add.3
+}
+
+define double @dotd(<4 x double>* byval nocapture readonly align 32, <4 x double>* byval nocapture readonly align 32) {
+; CHECK-LABEL: @dotd(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[X:%.*]] = load <4 x double>, <4 x double>* [[TMP0:%.*]], align 32
+; CHECK-NEXT:    [[Y:%.*]] = load <4 x double>, <4 x double>* [[TMP1:%.*]], align 32
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x double> [[X]], [[Y]]
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[TMP2]], <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x double> [[TMP3]], <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x double> [[TMP3]], [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x double> [[BIN_RDX]], <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x double> [[BIN_RDX]], [[RDX_SHUF1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[BIN_RDX2]], i32 0
+; CHECK-NEXT:    ret double [[TMP4]]
+;
+entry:
+  %x = load <4 x double>, <4 x double>* %0, align 32
+  %y = load <4 x double>, <4 x double>* %1, align 32
+  %vecext = extractelement <4 x double> %x, i32 0
+  %vecext1 = extractelement <4 x double> %y, i32 0
+  %mul = fmul fast double %vecext, %vecext1
+  %vecext.1 = extractelement <4 x double> %x, i32 1
+  %vecext1.1 = extractelement <4 x double> %y, i32 1
+  %mul.1 = fmul fast double %vecext.1, %vecext1.1
+  %add.1 = fadd fast double %mul.1, %mul
+  %vecext.2 = extractelement <4 x double> %x, i32 2
+  %vecext1.2 = extractelement <4 x double> %y, i32 2
+  %mul.2 = fmul fast double %vecext.2, %vecext1.2
+  %add.2 = fadd fast double %mul.2, %add.1
+  %vecext.3 = extractelement <4 x double> %x, i32 3
+  %vecext1.3 = extractelement <4 x double> %y, i32 3
+  %mul.3 = fmul fast double %vecext.3, %vecext1.3
+  %add.3 = fadd fast double %mul.3, %add.2
+  ret double %add.3
+}
+
+define float @dotfq(<4 x float>* nocapture readonly %x, <4 x float>* nocapture readonly %y) {
+; CHECK-LABEL: @dotfq(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, <4 x float>* [[X:%.*]], align 16
+; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[Y:%.*]], align 16
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x float> [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP3]], [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x float> [[BIN_RDX]], <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x float> [[BIN_RDX]], [[RDX_SHUF1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x float> [[BIN_RDX2]], i32 0
+; CHECK-NEXT:    ret float [[TMP4]]
+;
+entry:
+  %0 = load <4 x float>, <4 x float>* %x, align 16
+  %1 = load <4 x float>, <4 x float>* %y, align 16
+  %vecext = extractelement <4 x float> %0, i32 0
+  %vecext1 = extractelement <4 x float> %1, i32 0
+  %mul = fmul fast float %vecext1, %vecext
+  %vecext.1 = extractelement <4 x float> %0, i32 1
+  %vecext1.1 = extractelement <4 x float> %1, i32 1
+  %mul.1 = fmul fast float %vecext1.1, %vecext.1
+  %add.1 = fadd fast float %mul.1, %mul
+  %vecext.2 = extractelement <4 x float> %0, i32 2
+  %vecext1.2 = extractelement <4 x float> %1, i32 2
+  %mul.2 = fmul fast float %vecext1.2, %vecext.2
+  %add.2 = fadd fast float %mul.2, %add.1
+  %vecext.3 = extractelement <4 x float> %0, i32 3
+  %vecext1.3 = extractelement <4 x float> %1, i32 3
+  %mul.3 = fmul fast float %vecext1.3, %vecext.3
+  %add.3 = fadd fast float %mul.3, %add.2
+  ret float %add.3
+}
+
+define double @dotdq(<4 x double>* nocapture readonly %x, <4 x double>* nocapture readonly %y) {
+; CHECK-LABEL: @dotdq(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x double>, <4 x double>* [[X:%.*]], align 32
+; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x double>, <4 x double>* [[Y:%.*]], align 32
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x double> [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[TMP2]], <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x double> [[TMP3]], <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x double> [[TMP3]], [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x double> [[BIN_RDX]], <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x double> [[BIN_RDX]], [[RDX_SHUF1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[BIN_RDX2]], i32 0
+; CHECK-NEXT:    ret double [[TMP4]]
+;
+entry:
+  %0 = load <4 x double>, <4 x double>* %x, align 32
+  %1 = load <4 x double>, <4 x double>* %y, align 32
+  %vecext = extractelement <4 x double> %0, i32 0
+  %vecext1 = extractelement <4 x double> %1, i32 0
+  %mul = fmul fast double %vecext1, %vecext
+  %vecext.1 = extractelement <4 x double> %0, i32 1
+  %vecext1.1 = extractelement <4 x double> %1, i32 1
+  %mul.1 = fmul fast double %vecext1.1, %vecext.1
+  %add.1 = fadd fast double %mul.1, %mul
+  %vecext.2 = extractelement <4 x double> %0, i32 2
+  %vecext1.2 = extractelement <4 x double> %1, i32 2
+  %mul.2 = fmul fast double %vecext1.2, %vecext.2
+  %add.2 = fadd fast double %mul.2, %add.1
+  %vecext.3 = extractelement <4 x double> %0, i32 3
+  %vecext1.3 = extractelement <4 x double> %1, i32 3
+  %mul.3 = fmul fast double %vecext1.3, %vecext.3
+  %add.3 = fadd fast double %mul.3, %add.2
+  ret double %add.3
+}