From 95580d6ed2760914110bee903e8a9dc985649b4b Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 23 Aug 2016 16:11:21 +0000
Subject: [PATCH] [X86][SSE] Demonstrate inability to recognise that
 (v)cvtpd2dq & (v)cvttpd2dq intrinsics implicitly zeroes the upper half of the
 xmm

llvm-svn: 279527
---
 llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll | 38 ++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
index e66c504..c5ed21c 100644
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-x86.ll
@@ -210,6 +210,25 @@ define <4 x i32> @test_x86_sse2_cvtpd2dq(<2 x double> %a0) {
 declare <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double>) nounwind readnone
 
 
+define <2 x i64> @test_mm_cvtpd_epi32_zext(<2 x double> %a0) nounwind {
+; SSE-LABEL: test_mm_cvtpd_epi32_zext:
+; SSE:       ## BB#0:
+; SSE-NEXT:    cvtpd2dq %xmm0, %xmm0
+; SSE-NEXT:    movq {{.*#+}} xmm0 = xmm0[0],zero
+; SSE-NEXT:    retl
+;
+; KNL-LABEL: test_mm_cvtpd_epi32_zext:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcvtpd2dq %xmm0, %xmm0
+; KNL-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero
+; KNL-NEXT:    retl
+  %cvt = call <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double> %a0)
+  %res = shufflevector <4 x i32> %cvt, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %bc = bitcast <4 x i32> %res to <2 x i64>
+  ret <2 x i64> %bc
+}
+
+
 define <4 x float> @test_x86_sse2_cvtpd2ps(<2 x double> %a0) {
 ; SSE-LABEL: test_x86_sse2_cvtpd2ps:
 ; SSE:       ## BB#0:
@@ -412,6 +431,25 @@ define <4 x i32> @test_x86_sse2_cvttpd2dq(<2 x double> %a0) {
 declare <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double>) nounwind readnone
 
 
+define <2 x i64> @test_mm_cvttpd_epi32_zext(<2 x double> %a0) nounwind {
+; SSE-LABEL: test_mm_cvttpd_epi32_zext:
+; SSE:       ## BB#0:
+; SSE-NEXT:    cvttpd2dq %xmm0, %xmm0
+; SSE-NEXT:    movq {{.*#+}} xmm0 = xmm0[0],zero
+; SSE-NEXT:    retl
+;
+; KNL-LABEL: test_mm_cvttpd_epi32_zext:
+; KNL:       ## BB#0:
+; KNL-NEXT:    vcvttpd2dq %xmm0, %xmm0
+; KNL-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero
+; KNL-NEXT:    retl
+  %cvt = call <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double> %a0)
+  %res = shufflevector <4 x i32> %cvt, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+  %bc = bitcast <4 x i32> %res to <2 x i64>
+  ret <2 x i64> %bc
+}
+
+
 define <4 x i32> @test_x86_sse2_cvttps2dq(<4 x float> %a0) {
 ; SSE-LABEL: test_x86_sse2_cvttps2dq:
 ; SSE:       ## BB#0:
-- 
2.7.4