Optimized load + SIGN_EXTEND patterns in the X86 backend.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@170506 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 98 insertions, 3 deletions

diff --git a/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll b/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
index a720753..da734d4 100644
--- a/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
+++ b/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
@@ -16,8 +16,8 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK: main
 define i32 @main() nounwind uwtable {
 entry:
-; CHECK: movsbq  j(%rip), %
-; CHECK: movsbq  i(%rip), %
+; CHECK: pmovsxbq  j(%rip), %
+; CHECK: pmovsxbq  i(%rip), %
   %0 = load <2 x i8>* @i, align 8
   %1 = load <2 x i8>* @j, align 8
   %div = sdiv <2 x i8> %1, %0
diff --git a/test/CodeGen/X86/avx-sext.ll b/test/CodeGen/X86/avx-sext.ll
index 3713a8c..425d09c 100755
--- a/test/CodeGen/X86/avx-sext.ll
+++ b/test/CodeGen/X86/avx-sext.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx | FileCheck %s
 
 define <8 x i32> @sext_8i16_to_8i32(<8 x i16> %A) nounwind uwtable readnone ssp {
 ;CHECK: sext_8i16_to_8i32
@@ -15,3 +15,57 @@ define <4 x i64> @sext_4i32_to_4i64(<4 x i32> %A) nounwind uwtable readnone ssp
   %B = sext <4 x i32> %A to <4 x i64>
   ret <4 x i64>%B
 }
+
+; CHECK: load_sext_test1
+; CHECK: vpmovsxwd (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <4 x i32> @load_sext_test1(<4 x i16> *%ptr) {
+ %X = load <4 x i16>* %ptr
+ %Y = sext <4 x i16> %X to <4 x i32>
+ ret <4 x i32>%Y
+}
+
+; CHECK: load_sext_test2
+; CHECK: vpmovsxbd (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <4 x i32> @load_sext_test2(<4 x i8> *%ptr) {
+ %X = load <4 x i8>* %ptr
+ %Y = sext <4 x i8> %X to <4 x i32>
+ ret <4 x i32>%Y
+}
+
+; CHECK: load_sext_test3
+; CHECK: vpmovsxbq (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <2 x i64> @load_sext_test3(<2 x i8> *%ptr) {
+ %X = load <2 x i8>* %ptr
+ %Y = sext <2 x i8> %X to <2 x i64>
+ ret <2 x i64>%Y
+}
+
+; CHECK: load_sext_test4
+; CHECK: vpmovsxwq (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <2 x i64> @load_sext_test4(<2 x i16> *%ptr) {
+ %X = load <2 x i16>* %ptr
+ %Y = sext <2 x i16> %X to <2 x i64>
+ ret <2 x i64>%Y
+}
+
+; CHECK: load_sext_test5
+; CHECK: vpmovsxdq (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <2 x i64> @load_sext_test5(<2 x i32> *%ptr) {
+ %X = load <2 x i32>* %ptr
+ %Y = sext <2 x i32> %X to <2 x i64>
+ ret <2 x i64>%Y
+}
+
+; CHECK: load_sext_test6
+; CHECK: vpmovsxbw (%r{{[^,]*}}), %xmm{{.*}}
+; CHECK: ret
+define <8 x i16> @load_sext_test6(<8 x i8> *%ptr) {
+ %X = load <8 x i8>* %ptr
+ %Y = sext <8 x i8> %X to <8 x i16>
+ ret <8 x i16>%Y
+}
diff --git a/test/CodeGen/X86/avx2-conversions.ll b/test/CodeGen/X86/avx2-conversions.ll
index b474913..3ce08dc 100755
--- a/test/CodeGen/X86/avx2-conversions.ll
+++ b/test/CodeGen/X86/avx2-conversions.ll
@@ -63,6 +63,47 @@ define <8 x i32> @zext_8i8_8i32(<8 x i8> %A) nounwind {
   ret <8 x i32>%B
 }
 
+; CHECK: load_sext_test1
+; CHECK: vpmovsxdq (%r{{[^,]*}}), %ymm{{.*}}
+; CHECK: ret 
+define <4 x i64> @load_sext_test1(<4 x i32> *%ptr) {
+ %X = load <4 x i32>* %ptr
+ %Y = sext <4 x i32> %X to <4 x i64>
+ ret <4 x i64>%Y
+}
+
+; CHECK: load_sext_test2
+; CHECK: vpmovsxbq (%r{{[^,]*}}), %ymm{{.*}}
+; CHECK: ret 
+define <4 x i64> @load_sext_test2(<4 x i8> *%ptr) {
+ %X = load <4 x i8>* %ptr
+ %Y = sext <4 x i8> %X to <4 x i64>
+ ret <4 x i64>%Y
+}
 
+; CHECK: load_sext_test3
+; CHECK: vpmovsxwq (%r{{[^,]*}}), %ymm{{.*}}
+; CHECK: ret 
+define <4 x i64> @load_sext_test3(<4 x i16> *%ptr) {
+ %X = load <4 x i16>* %ptr
+ %Y = sext <4 x i16> %X to <4 x i64>
+ ret <4 x i64>%Y
+}
 
+; CHECK: load_sext_test4
+; CHECK: vpmovsxwd (%r{{[^,]*}}), %ymm{{.*}}
+; CHECK: ret 
+define <8 x i32> @load_sext_test4(<8 x i16> *%ptr) {
+ %X = load <8 x i16>* %ptr
+ %Y = sext <8 x i16> %X to <8 x i32>
+ ret <8 x i32>%Y
+}
 
+; CHECK: load_sext_test5
+; CHECK: vpmovsxbd (%r{{[^,]*}}), %ymm{{.*}}
+; CHECK: ret 
+define <8 x i32> @load_sext_test5(<8 x i8> *%ptr) {
+ %X = load <8 x i8>* %ptr
+ %Y = sext <8 x i8> %X to <8 x i32>
+ ret <8 x i32>%Y
+}