projects / gcc.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 4e6e941) | patch
Avoid uniform lane BB vectorization
authorRichard Biener <rguenther@suse.de>
Tue, 27 Oct 2020 10:03:27 +0000 (11:03 +0100)
committerRichard Biener <rguenther@suse.de>
Tue, 27 Oct 2020 12:16:35 +0000 (13:16 +0100)
commit5af1e827bbb624eb28f80d2c5e0da46185af3708
tree14498c1e161ef861cfe7bc4b6f7e72e1eb3c39e0tree
parent4e6e941db7ced1fb967beed5f15b69b4d56bce61commit | diff
Avoid uniform lane BB vectorization

This makes sure to use splats early when facing uniform internal
operands in BB SLP discovery rather than relying on the late
heuristincs re-building nodes from scratch.

2020-10-27  Richard Biener  <rguenther@suse.de>

* tree-vect-slp.c (vect_build_slp_tree_2): When vectorizing
BBs splat uniform operands and stop SLP discovery.

* gcc.target/i386/pr95866-1.c: Adjust.
gcc/testsuite/gcc.target/i386/pr95866-1.c diff | blob | history
gcc/tree-vect-slp.c diff | blob | history