diff options
| author | Dan Gohman <gohman@apple.com> | 2010-04-23 01:55:05 +0000 |
|---|---|---|
| committer | Dan Gohman <gohman@apple.com> | 2010-04-23 01:55:05 +0000 |
| commit | 997bbc54d6b3bebee605cf229bb6fd8b56190a89 (patch) | |
| tree | 5584f3f8d3425457a65d329fca3d61ccdae0ab47 /llvm/lib/Analysis | |
| parent | ff3174e97f40670db05daf8077347a372b9581c3 (diff) | |
| download | bcm5719-llvm-997bbc54d6b3bebee605cf229bb6fd8b56190a89.tar.gz bcm5719-llvm-997bbc54d6b3bebee605cf229bb6fd8b56190a89.zip | |
Fix LSR to tolerate cases where ScalarEvolution initially
misses an opportunity to fold add operands, but folds them
after LSR has separated them out. This fixes rdar://7886751.
llvm-svn: 102157
Diffstat (limited to 'llvm/lib/Analysis')
| -rw-r--r-- | llvm/lib/Analysis/README.txt | 12 |
1 files changed, 12 insertions, 0 deletions
diff --git a/llvm/lib/Analysis/README.txt b/llvm/lib/Analysis/README.txt index c4010902729..88ea9f11ad7 100644 --- a/llvm/lib/Analysis/README.txt +++ b/llvm/lib/Analysis/README.txt @@ -16,3 +16,15 @@ In addition to being much more complicated, it involves i65 arithmetic, which is very inefficient when expanded into code. //===---------------------------------------------------------------------===// + +In test/CodeGen/X86/lsr-delayed-fold.ll, + +ScalarEvolution is forming this expression: + +((trunc i64 (-1 * %arg5) to i32) + (trunc i64 %arg5 to i32) + (-1 * (trunc i64 undef to i32))) + +This could be folded to + +(-1 * (trunc i64 undef to i32)) + +//===---------------------------------------------------------------------===// |
