From 7bc389b6b00e26e07fa5ac64e43d9b06b73828c1 Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Tue, 8 Nov 2011 00:31:58 +0000
Subject: Add x86 isel logic and patterns to match movlps from clang generated
 IR for _mm_loadl_pi(). rdar://10134392, rdar://10050222

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144052 91177308-0d34-0410-b5e6-96231b3b80d8
---
 test/CodeGen/X86/vec_shuffle-38.ll |  3 +--
 test/CodeGen/X86/vec_shuffle-39.ll | 51 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 52 insertions(+), 2 deletions(-)
 create mode 100644 test/CodeGen/X86/vec_shuffle-39.ll

(limited to 'test/CodeGen')

diff --git a/test/CodeGen/X86/vec_shuffle-38.ll b/test/CodeGen/X86/vec_shuffle-38.ll
index 69a2ede..3531e32 100644
--- a/test/CodeGen/X86/vec_shuffle-38.ll
+++ b/test/CodeGen/X86/vec_shuffle-38.ll
@@ -48,8 +48,7 @@ entry:
 ; CHECK: f
 define <4 x float> @f(<4 x float> %x, double* nocapture %y) nounwind uwtable readonly ssp {
 entry:
-  ; CHECK: movsd  (%
-  ; CHECK-NEXT: movsd  %xmm
+  ; CHECK: movlps  (%rdi), %xmm0
   %u110.i = load double* %y, align 1
   %tmp8.i = insertelement <2 x double> undef, double %u110.i, i32 0
   %tmp9.i = bitcast <2 x double> %tmp8.i to <4 x float>
diff --git a/test/CodeGen/X86/vec_shuffle-39.ll b/test/CodeGen/X86/vec_shuffle-39.ll
new file mode 100644
index 0000000..6c889b8
--- /dev/null
+++ b/test/CodeGen/X86/vec_shuffle-39.ll
@@ -0,0 +1,51 @@
+; RUN: llc < %s -march=x86-64 | FileCheck %s
+; rdar://10050222, rdar://10134392
+
+define <4 x float> @t1(<4 x float> %a, <1 x i64>* nocapture %p) nounwind {
+entry:
+; CHECK: t1:
+; CHECK: movlps (%rdi), %xmm0
+; CHECK: ret
+  %p.val = load <1 x i64>* %p, align 1
+  %0 = bitcast <1 x i64> %p.val to <2 x float>
+  %shuffle.i = shufflevector <2 x float> %0, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+  %shuffle1.i = shufflevector <4 x float> %a, <4 x float> %shuffle.i, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
+  ret <4 x float> %shuffle1.i
+}
+
+define <4 x float> @t1a(<4 x float> %a, <1 x i64>* nocapture %p) nounwind {
+entry:
+; CHECK: t1a:
+; CHECK: movlps (%rdi), %xmm0
+; CHECK: ret
+  %0 = bitcast <1 x i64>* %p to double*
+  %1 = load double* %0
+  %2 = insertelement <2 x double> undef, double %1, i32 0
+  %3 = bitcast <2 x double> %2 to <4 x float>
+  %4 = shufflevector <4 x float> %a, <4 x float> %3, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
+  ret <4 x float> %4
+}
+
+define void @t2(<1 x i64>* nocapture %p, <4 x float> %a) nounwind {
+entry:
+; CHECK: t2:
+; CHECK: movlps %xmm0, (%rdi)
+; CHECK: ret
+  %cast.i = bitcast <4 x float> %a to <2 x i64>
+  %extract.i = extractelement <2 x i64> %cast.i, i32 0
+  %0 = getelementptr inbounds <1 x i64>* %p, i64 0, i64 0
+  store i64 %extract.i, i64* %0, align 8
+  ret void
+}
+
+define void @t2a(<1 x i64>* nocapture %p, <4 x float> %a) nounwind {
+entry:
+; CHECK: t2a:
+; CHECK: movlps %xmm0, (%rdi)
+; CHECK: ret
+  %0 = bitcast <1 x i64>* %p to double*
+  %1 = bitcast <4 x float> %a to <2 x double>
+  %2 = extractelement <2 x double> %1, i32 0
+  store double %2, double* %0
+  ret void
+}
-- 
cgit v1.1