2nd attempt, fixing SSE4.1 issues and implementing feedback from duncan.
PR2957
ISD::VECTOR_SHUFFLE now stores an array of integers representing the shuffle
mask internal to the node, rather than taking a BUILD_VECTOR of ConstantSDNodes
as the shuffle mask. A value of -1 represents UNDEF.
In addition to eliminating the creation of illegal BUILD_VECTORS just to
represent shuffle masks, we are better about canonicalizing the shuffle mask,
resulting in substantially better code for some classes of shuffles.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70225 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/Generic/vector-casts.ll b/test/CodeGen/Generic/vector-casts.ll
index 12104a3..e661e84 100644
--- a/test/CodeGen/Generic/vector-casts.ll
+++ b/test/CodeGen/Generic/vector-casts.ll
@@ -1,45 +1,9 @@
; RUN: llvm-as < %s | llc
; PR2671
-define void @a(<2 x double>* %p, <2 x i8>* %q) {
- %t = load <2 x double>* %p
- %r = fptosi <2 x double> %t to <2 x i8>
- store <2 x i8> %r, <2 x i8>* %q
- ret void
-}
-define void @b(<2 x double>* %p, <2 x i8>* %q) {
- %t = load <2 x double>* %p
- %r = fptoui <2 x double> %t to <2 x i8>
- store <2 x i8> %r, <2 x i8>* %q
- ret void
-}
-define void @c(<2 x i8>* %p, <2 x double>* %q) {
- %t = load <2 x i8>* %p
- %r = sitofp <2 x i8> %t to <2 x double>
- store <2 x double> %r, <2 x double>* %q
- ret void
-}
-define void @d(<2 x i8>* %p, <2 x double>* %q) {
- %t = load <2 x i8>* %p
- %r = uitofp <2 x i8> %t to <2 x double>
- store <2 x double> %r, <2 x double>* %q
- ret void
-}
-define void @e(<2 x i8>* %p, <2 x i16>* %q) {
- %t = load <2 x i8>* %p
- %r = sext <2 x i8> %t to <2 x i16>
- store <2 x i16> %r, <2 x i16>* %q
- ret void
-}
-define void @f(<2 x i8>* %p, <2 x i16>* %q) {
- %t = load <2 x i8>* %p
- %r = zext <2 x i8> %t to <2 x i16>
- store <2 x i16> %r, <2 x i16>* %q
- ret void
-}
define void @g(<2 x i16>* %p, <2 x i8>* %q) {
- %t = load <2 x i16>* %p
- %r = trunc <2 x i16> %t to <2 x i8>
- store <2 x i8> %r, <2 x i8>* %q
- ret void
+ %t = load <2 x i16>* %p
+ %r = trunc <2 x i16> %t to <2 x i8>
+ store <2 x i8> %r, <2 x i8>* %q
+ ret void
}