[X86] Add test case that shows a scalar sqrtsd intrinsic of a 128-bit vector load using the load form of the sqrtsd instruction which violates the intrinsic semantics.

The sqrtsd instruction only loads 64-bits and writes bits 63:0 with the sqrt result. Bits 127:64 are preserved in the destination register. The semantics of the intrinsic indicate bits 127:64 should come from the intrinsic argument which in this case is a 128-bit load. So the generated code should have a 128-bit load and use a register form of sqrtsd.

llvm-svn: 288780

1 files changed, 26 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
index 1a78729a7f6..22f25c19b0b 100644
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
@@ -1500,6 +1500,32 @@ define <2 x double> @test_x86_sse2_sqrt_sd(<2 x double> %a0) {
 declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone
 
 
+define <2 x double> @test_x86_sse2_sqrt_sd_vec_load(<2 x double>* %a0) {
+; SSE-LABEL: test_x86_sse2_sqrt_sd_vec_load:
+; SSE:       ## BB#0:
+; SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
+; SSE-NEXT:    sqrtsd (%eax), %xmm0 ## encoding: [0xf2,0x0f,0x51,0x00]
+; SSE-NEXT:    retl ## encoding: [0xc3]
+;
+; AVX2-LABEL: test_x86_sse2_sqrt_sd_vec_load:
+; AVX2:       ## BB#0:
+; AVX2-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
+; AVX2-NEXT:    vmovaps (%eax), %xmm0 ## encoding: [0xc5,0xf8,0x28,0x00]
+; AVX2-NEXT:    vsqrtsd %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xfb,0x51,0xc0]
+; AVX2-NEXT:    retl ## encoding: [0xc3]
+;
+; SKX-LABEL: test_x86_sse2_sqrt_sd_vec_load:
+; SKX:       ## BB#0:
+; SKX-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
+; SKX-NEXT:    vmovaps (%eax), %xmm0 ## encoding: [0x62,0xf1,0x7c,0x08,0x28,0x00]
+; SKX-NEXT:    vsqrtsd %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xfb,0x51,0xc0]
+; SKX-NEXT:    retl ## encoding: [0xc3]
+  %a1 = load <2 x double>, <2 x double>* %a0, align 16
+  %res = call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+
+
 define i32 @test_x86_sse2_ucomieq_sd(<2 x double> %a0, <2 x double> %a1) {
 ; SSE-LABEL: test_x86_sse2_ucomieq_sd:
 ; SSE:       ## BB#0: