projects / gcc.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 5a36cae) | patch
amdgcn: Allow V64DFmode min/max reductions
authorAndrew Stubbs <ams@codesourcery.com>
Mon, 9 Nov 2020 17:42:34 +0000 (17:42 +0000)
committerAndrew Stubbs <ams@codesourcery.com>
Tue, 26 Jan 2021 10:23:43 +0000 (10:23 +0000)
commitd9f50366102a8ca3521e4854f7716bd013c8ea0a
treee461f789926aca8c445373b0fcb87eda66b7d726tree
parent5a36cae275ad84cc7e623f2f5829bdad767e3f6acommit | diff
amdgcn: Allow V64DFmode min/max reductions

I don't know why these were disabled. There're no direct min/max DPP
instructions for this mode, but the "use moves" strategy works fine.

gcc/ChangeLog:

* config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
for V64DFmode min/max reductions.
gcc/config/gcn/gcn.c diff | blob | history