Re-apply 40504, but with a fix for the segfault it caused in oggenc:

Make the alignedload and alignedstore patterns always require 16-byte
alignment. This way when they are used in the "Fs" instructions, in which
a vector instruction is used for a scalar purpose, they can still require
the full vector alignment. And add a regression test for this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@40555 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 14 insertions, 0 deletions

diff --git a/test/CodeGen/X86/fsxor-alignment.ll b/test/CodeGen/X86/fsxor-alignment.ll
new file mode 100644
index 0000000..bb55e44
--- /dev/null
+++ b/test/CodeGen/X86/fsxor-alignment.ll
@@ -0,0 +1,14 @@
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse -enable-unsafe-fp-math | \
+; RUN:  grep -v sp | grep xorps | wc -l | grep 2
+
+; Don't fold the incoming stack arguments into the xorps instructions used
+; to do floating-point negations, because the arguments aren't vectors
+; and aren't vector-aligned.
+
+define void @foo(float* %p, float* %q, float %s, float %y) {
+  %ss = sub float -0.0, %s
+  %yy = sub float -0.0, %y
+  store float %ss, float* %p
+  store float %yy, float* %q
+  ret void
+}