projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e42021d) | patch
Enable support for floating-point division reductions
authorYichao Yu <yyc1992@gmail.com>
Tue, 24 Nov 2020 00:59:35 +0000 (19:59 -0500)
committerValentin Churavy <v.churavy@gmail.com>
Tue, 24 Nov 2020 01:00:58 +0000 (20:00 -0500)
commit4bc88a0e9a2ee29959a9053e867ae6f051348554
tree72a7088737ed8cda68c150c75960cb41afe6fc57tree | snapshot
parente42021d5cc25a8dc7e3efac1e7007cc0c1a7b2bdcommit | diff
Enable support for floating-point division reductions

Similar to fsub, fdiv can also be vectorized using fmul.

Also http://llvm.org/viewvc/llvm-project?view=revision&revision=215200

Differential Revision: https://reviews.llvm.org/D34078

Co-authored-by: Jameson Nash <jameson@juliacomputing.com>
llvm/lib/Analysis/IVDescriptors.cpp diff | blob | history
llvm/test/Transforms/LoopVectorize/float-reduction.ll diff | blob | history
Domain: System / Toolchain;