diff options
| author | Bill Wendling <isanbard@gmail.com> | 2011-05-12 19:02:15 +0000 |
|---|---|---|
| committer | Bill Wendling <isanbard@gmail.com> | 2011-05-12 19:02:15 +0000 |
| commit | e106c3481784ecfda9cb014cb2cfea5b3d13bbc9 (patch) | |
| tree | f258b9fd8a6a63d651fe1a1fc52cfcafd0acd20d /clang/test | |
| parent | 3f125fe2eaab19eac991fbef562c85896b725e14 (diff) | |
| download | bcm5719-llvm-e106c3481784ecfda9cb014cb2cfea5b3d13bbc9.tar.gz bcm5719-llvm-e106c3481784ecfda9cb014cb2cfea5b3d13bbc9.zip | |
LLVM doesn't always optimize away the four loads from this:
(__m128){ p[0], p[1], p[2], p[3] }
which produces really bad code. This could be done in instcombine, but it's
probably better to do it in the front-end instead.
<rdar://problem/9424836>
llvm-svn: 131237
Diffstat (limited to 'clang/test')
| -rw-r--r-- | clang/test/CodeGen/builtins-x86.c | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/clang/test/CodeGen/builtins-x86.c b/clang/test/CodeGen/builtins-x86.c index bb63048b616..190fa55c575 100644 --- a/clang/test/CodeGen/builtins-x86.c +++ b/clang/test/CodeGen/builtins-x86.c @@ -273,6 +273,7 @@ void f0() { #endif tmp_V2i = __builtin_ia32_cvttps2pi(tmp_V4f); (void) __builtin_ia32_maskmovq(tmp_V8c, tmp_V8c, tmp_cp); + tmp_V4f = __builtin_ia32_loadups(tmp_fCp); (void) __builtin_ia32_storeups(tmp_fp, tmp_V4f); (void) __builtin_ia32_storehps(tmp_V2ip, tmp_V4f); (void) __builtin_ia32_storelps(tmp_V2ip, tmp_V4f); @@ -290,6 +291,7 @@ void f0() { tmp_V4f = __builtin_ia32_sqrtps(tmp_V4f); tmp_V4f = __builtin_ia32_sqrtss(tmp_V4f); (void) __builtin_ia32_maskmovdqu(tmp_V16c, tmp_V16c, tmp_cp); + tmp_V2d = __builtin_ia32_loadupd(tmp_dCp); (void) __builtin_ia32_storeupd(tmp_dp, tmp_V2d); tmp_i = __builtin_ia32_movmskpd(tmp_V2d); tmp_i = __builtin_ia32_pmovmskb128(tmp_V16c); |
