diff options
| author | Richard Sandiford <richard.sandiford@linaro.org> | 2018-05-24 12:34:18 +0000 |
|---|---|---|
| committer | Richard Sandiford <rsandifo@gcc.gnu.org> | 2018-05-24 12:34:18 +0000 |
| commit | c453ccc2335bf4267a154c9385eb50a8c45235a1 (patch) | |
| tree | a370c4868f5adf7c1f2a37c67c0b608933c93782 /gcc/tree-streamer-in.c | |
| parent | 0c08e1f85d96dd50de4289f05beb87b83b6d1d83 (diff) | |
| download | gcc-c453ccc2335bf4267a154c9385eb50a8c45235a1.zip gcc-c453ccc2335bf4267a154c9385eb50a8c45235a1.tar.gz gcc-c453ccc2335bf4267a154c9385eb50a8c45235a1.tar.bz2 | |
Use canonicalize_math_after_vectorization_p for FMA folds
The folds in r260348 kicked in before vectorisation, which hurts
for two reasons:
(1) the current suboptimal handling of nothrow meant that we could
drop the flag early and so prevent if-conversion
(2) some architectures provide more scalar forms than vector forms
(true for Advanced SIMD)
(1) is a bug in itself that needs to be fixed eventually, but delaying
the folds is still needed for (2).
2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
* match.pd: Delay FMA folds until after vectorization.
gcc/testsuite/
* gcc.dg/vect/vect-fma-1.c: New test.
From-SVN: r260639
Diffstat (limited to 'gcc/tree-streamer-in.c')
0 files changed, 0 insertions, 0 deletions