projects / platform / upstream / mesa.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d54f8ec) | patch
intel/fs: Use the original destination region for int MUL lowering
authorJason Ekstrand <jason.ekstrand@intel.com>
Wed, 18 Oct 2017 01:59:26 +0000 (18:59 -0700)
committerJason Ekstrand <jason.ekstrand@intel.com>
Tue, 7 Nov 2017 18:37:52 +0000 (10:37 -0800)
commit18fde36ced4279f2577097a1a7d31b55f2f5f141
tree5b53c9e5ad1e2b25b640e08b5785bb82274910d2tree | snapshot
parentd54f8ec744545673fd78f15ffce3cb4e47d4b5f1commit | diff
intel/fs: Use the original destination region for int MUL lowering

Some hardware (CHV, BXT) have special restrictions on register regions
when doing integer multiplication.  We want to respect those when we
lower to DxW multiplication.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Cc: mesa-stable@lists.freedesktop.org
src/intel/compiler/brw_fs.cpp diff | blob | history
Domain: Graphics System / GL;