diff options
| author | Craig Topper <craig.topper@intel.com> | 2018-08-16 21:54:05 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2018-08-16 21:54:05 +0000 |
| commit | 883ff69c9365a7eb2fb146f029bb33afcb98524b (patch) | |
| tree | bd9c35208105e67da8361bbd1055d86dfc21389c /llvm/lib/IR/Dominators.cpp | |
| parent | bde2b43cb3386534dc2f66d6de29567cf1347b9d (diff) | |
| download | bcm5719-llvm-883ff69c9365a7eb2fb146f029bb33afcb98524b.tar.gz bcm5719-llvm-883ff69c9365a7eb2fb146f029bb33afcb98524b.zip | |
[DAGCombiner] Don't reassociate operations that have the vector reduction flag set.
When nodes are reassociated the vector-reduction flag gets lost.
The test case is here is what would happen if you had a sum of absolute differences loop that started with a non-zero but contant sum and that loop was unrolled. The vectorizer will generate a constant vector for the initial value. And DAGCombiner reassociate tries to move it down the addition tree erasing the vector-reduction flag. Interestingly this moves constants the opposite direction of the reassociate IR pass.
I've chosen to just punt on the reassociate, but I suppose we could maybe preserve the flag if both nodes have it set.
Differential Revision: https://reviews.llvm.org/D50827
llvm-svn: 339946
Diffstat (limited to 'llvm/lib/IR/Dominators.cpp')
0 files changed, 0 insertions, 0 deletions
