Fix several more entries in the x86 reload/remat folding tables.

[oota-llvm.git] / test / CodeGen / X86 / sse_reload_fold.ll

diff --git a/test/CodeGen/X86/sse_reload_fold.ll b/test/CodeGen/X86/sse_reload_fold.ll
new file mode 100644 (file)
index 0000000..1100eb7
--- /dev/null
+++ b/test/CodeGen/X86/sse_reload_fold.ll
@@ -0,0 +1,124 @@
+; RUN: llvm-as < %s | \
+; RUN: llc -march=x86-64 -mattr=+64bit,+sse3 -print-failed-fuse-candidates |& \
+; RUN:   grep fail | count 1
+
+declare void @bar()
+declare <4 x float> @qux()
+declare <2 x double> @pin()
+declare float @llvm.sqrt.f32(float)
+declare double @llvm.sqrt.f64(double)
+
+declare <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float>)
+declare <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float>)
+declare <4 x float> @llvm.x86.sse.rcp.ps(<4 x float>)
+declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>)
+declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>)
+declare <4 x float> @llvm.x86.sse.cmp.ps(<4 x float>, <4 x float>, i8)
+declare <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float>, <4 x float>)
+declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>)
+declare <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float>, <4 x float>)
+declare <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double>)
+declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>)
+declare <2 x double> @llvm.x86.sse2.max.pd(<2 x double>, <2 x double>)
+declare <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double>, <2 x double>, i8)
+declare <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double>, <2 x double>)
+declare <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double>, <2 x double>)
+declare <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double>, <2 x double>)
+
+define float @foo(float %f) {
+  call void @bar()
+  %t = call float @llvm.sqrt.f32(float %f)
+  ret float %t
+}
+define double @doo(double %f) {
+  call void @bar()
+  %t = call double @llvm.sqrt.f64(double %f)
+  ret double %t
+}
+define <4 x float> @a0(<4 x float> %f) {
+  call void @bar()
+  %t = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @a1(<4 x float> %f) {
+  call void @bar()
+  %t = call <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @a2(<4 x float> %f) {
+  call void @bar()
+  %t = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @b3(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %y, <4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @b4(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %y, <4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @b5(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse.cmp.ps(<4 x float> %y, <4 x float> %f, i8 7)
+  ret <4 x float> %t
+}
+define <4 x float> @b6(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> %y, <4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @b7(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %y, <4 x float> %f)
+  ret <4 x float> %t
+}
+define <4 x float> @b8(<4 x float> %f) {
+  %y = call <4 x float> @qux()
+  %t = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %y, <4 x float> %f)
+  ret <4 x float> %t
+}
+define <2 x double> @c1(<2 x double> %f) {
+  call void @bar()
+  %t = call <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double> %f)
+  ret <2 x double> %t
+}
+define <2 x double> @d3(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %y, <2 x double> %f)
+  ret <2 x double> %t
+}
+define <2 x double> @d4(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse2.max.pd(<2 x double> %y, <2 x double> %f)
+  ret <2 x double> %t
+}
+define <2 x double> @d5(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double> %y, <2 x double> %f, i8 7)
+  ret <2 x double> %t
+}
+define <2 x double> @d6(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> %y, <2 x double> %f)
+  ret <2 x double> %t
+}
+define <2 x double> @d7(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %y, <2 x double> %f)
+  ret <2 x double> %t
+}
+define <2 x double> @d8(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %y, <2 x double> %f)
+  ret <2 x double> %t
+}
+
+; This one should fail to fuse.
+define <2 x double> @z0(<2 x double> %f) {
+  %y = call <2 x double> @pin()
+  %t = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %f, <2 x double> %y)
+  ret <2 x double> %t
+}