diff options
author | Eli Friedman <eli.friedman@gmail.com> | 2009-06-06 06:05:10 +0000 |
---|---|---|
committer | Eli Friedman <eli.friedman@gmail.com> | 2009-06-06 06:05:10 +0000 |
commit | 1041553424cb0a9af03398ee0bff0d1a23fccc6f (patch) | |
tree | 1c4be2704e733e7a4d4c08a10e1e39daa9930c65 /test/CodeGen/X86/2009-06-05-VZextByteShort.ll | |
parent | 9d47b8d8ea5f7cc714fb56acc4711ba25324b454 (diff) | |
download | llvm-1041553424cb0a9af03398ee0bff0d1a23fccc6f.tar.gz llvm-1041553424cb0a9af03398ee0bff0d1a23fccc6f.tar.bz2 llvm-1041553424cb0a9af03398ee0bff0d1a23fccc6f.tar.xz |
Get rid of some bogus patterns for X86vzmovl. Don't create VZEXT_MOVL
nodes for vectors with an i16 element type. Add an optimization for
building a vector which is all zeros/undef except for the bottom
element, where the bottom element is an i8 or i16.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72988 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/2009-06-05-VZextByteShort.ll')
-rw-r--r-- | test/CodeGen/X86/2009-06-05-VZextByteShort.ll | 37 |
1 files changed, 37 insertions, 0 deletions
diff --git a/test/CodeGen/X86/2009-06-05-VZextByteShort.ll b/test/CodeGen/X86/2009-06-05-VZextByteShort.ll new file mode 100644 index 0000000000..220423aa98 --- /dev/null +++ b/test/CodeGen/X86/2009-06-05-VZextByteShort.ll @@ -0,0 +1,37 @@ +; RUN: llvm-as < %s | llc -march=x86 -mattr=+mmx,+sse2 > %t1 +; RUN: grep movzwl %t1 | count 2 +; RUN: grep movzbl %t1 | count 2 +; RUN: grep movd %t1 | count 4 + +define <4 x i16> @a(i32* %x1) nounwind { + %x2 = load i32* %x1 + %x3 = lshr i32 %x2, 1 + %x = trunc i32 %x3 to i16 + %r = insertelement <4 x i16> zeroinitializer, i16 %x, i32 0 + ret <4 x i16> %r +} + +define <8 x i16> @b(i32* %x1) nounwind { + %x2 = load i32* %x1 + %x3 = lshr i32 %x2, 1 + %x = trunc i32 %x3 to i16 + %r = insertelement <8 x i16> zeroinitializer, i16 %x, i32 0 + ret <8 x i16> %r +} + +define <8 x i8> @c(i32* %x1) nounwind { + %x2 = load i32* %x1 + %x3 = lshr i32 %x2, 1 + %x = trunc i32 %x3 to i8 + %r = insertelement <8 x i8> zeroinitializer, i8 %x, i32 0 + ret <8 x i8> %r +} + +define <16 x i8> @d(i32* %x1) nounwind { + %x2 = load i32* %x1 + %x3 = lshr i32 %x2, 1 + %x = trunc i32 %x3 to i8 + %r = insertelement <16 x i8> zeroinitializer, i8 %x, i32 0 + ret <16 x i8> %r +} + |