[X86][SSE] Force execution domain of 32-bit extractps/pextrd in the stack folding...

llvm-svn: 288910

diff --git a/llvm/test/CodeGen/X86/stack-folding-fp-avx1.ll b/llvm/test/CodeGen/X86/stack-folding-fp-avx1.ll
index 98e4f99..5e939cc 100644 (file)
--- a/llvm/test/CodeGen/X86/stack-folding-fp-avx1.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-fp-avx1.ll
@@ -893,10 +893,12 @@ define i32 @stack_fold_extractps(<4 x float> %a0) {
   ;CHECK-LABEL: stack_fold_extractps
   ;CHECK:       vextractps $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
   ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
-  %1 = extractelement <4 x float> %a0, i32 1
-  %2 = bitcast float %1 to i32
-  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i32 %2
+  ; fadd forces execution domain
+  %1 = fadd <4 x float> %a0, <float 1.0, float 2.0, float 3.0, float 4.0>
+  %2 = extractelement <4 x float> %1, i32 1
+  %3 = bitcast float %2 to i32
+  %4 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i32 %3
 }
 
 define <2 x double> @stack_fold_haddpd(<2 x double> %a0, <2 x double> %a1) {

diff --git a/llvm/test/CodeGen/X86/stack-folding-fp-sse42.ll b/llvm/test/CodeGen/X86/stack-folding-fp-sse42.ll
index bc346a6..c577827 100644 (file)
--- a/llvm/test/CodeGen/X86/stack-folding-fp-sse42.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-fp-sse42.ll
@@ -613,10 +613,12 @@ define i32 @stack_fold_extractps(<4 x float> %a0) {
   ;CHECK-LABEL: stack_fold_extractps
   ;CHECK:       extractps $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
   ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
-  %1 = extractelement <4 x float> %a0, i32 1
-  %2 = bitcast float %1 to i32
-  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i32 %2
+  ; fadd forces execution domain
+  %1 = fadd <4 x float> %a0, <float 1.0, float 2.0, float 3.0, float 4.0>
+  %2 = extractelement <4 x float> %1, i32 1
+  %3 = bitcast float %2 to i32
+  %4 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i32 %3
 }
 
 define <2 x double> @stack_fold_haddpd(<2 x double> %a0, <2 x double> %a1) {

diff --git a/llvm/test/CodeGen/X86/stack-folding-int-avx1.ll b/llvm/test/CodeGen/X86/stack-folding-int-avx1.ll
index 3e3bbb9..6e8874c 100644 (file)
--- a/llvm/test/CodeGen/X86/stack-folding-int-avx1.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-int-avx1.ll
@@ -430,9 +430,11 @@ define i32 @stack_fold_pextrd(<4 x i32> %a0) {
   ;CHECK-LABEL: stack_fold_pextrd
   ;CHECK:       pextrd $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
   ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
-  %1 = extractelement <4 x i32> %a0, i32 1
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i32 %1
+  ; add forces execution domain
+  %1 = add <4 x i32> %a0, <i32 1, i32 2, i32 3, i32 4>
+  %2 = extractelement <4 x i32> %1, i32 1
+  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i32 %2
 }
 
 define i64 @stack_fold_pextrq(<2 x i64> %a0) {

diff --git a/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll b/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
index f732607..a839a31 100644 (file)
--- a/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
@@ -457,9 +457,11 @@ define i32 @stack_fold_pextrd(<4 x i32> %a0) {
   ;CHECK-LABEL: stack_fold_pextrd
   ;CHECK:       pextrd $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
   ;CHECK:       movl    {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
-  %1 = extractelement <4 x i32> %a0, i32 1
-  %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
-  ret i32 %1
+  ; add forces execution domain
+  %1 = add <4 x i32> %a0, <i32 1, i32 2, i32 3, i32 4>
+  %2 = extractelement <4 x i32> %1, i32 1
+  %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i32 %2
 }
 
 define i64 @stack_fold_pextrq(<2 x i64> %a0) {