diff options
| author | David Green <david.green@arm.com> | 2023-09-04 12:49:13 +0100 |
|---|---|---|
| committer | David Green <david.green@arm.com> | 2023-09-04 12:49:13 +0100 |
| commit | 2955cc15ff2242b53cce5e0318ed6867f1b06714 (patch) | |
| tree | d284c949a9df146d387e5b059f0f320737006d7f /llvm/lib/Transforms/Utils/LoopUtils.cpp | |
| parent | 3dcf3cbc421cb5ee985b3b4a697c51247df8b5dc (diff) | |
| download | llvm-2955cc15ff2242b53cce5e0318ed6867f1b06714.zip llvm-2955cc15ff2242b53cce5e0318ed6867f1b06714.tar.gz llvm-2955cc15ff2242b53cce5e0318ed6867f1b06714.tar.bz2 | |
[ARM] Improve costs for FMin/Max reductions
Similar to the other reductions, this changes the cost of fmin/fmax reductions
under MVE/NEON to perform vector operations until the types need to be
scalarized. The fp16 vectors can perform a VREV+FMIN/FMAX to skip a step of the
reduction, and otherwise need lanewise extract fro the top lanes.
Diffstat (limited to 'llvm/lib/Transforms/Utils/LoopUtils.cpp')
0 files changed, 0 insertions, 0 deletions