cso: make memcmp and hash computation use a literal key size for blend state
authorMarek Olšák <marek.olsak@amd.com>
Sun, 7 Aug 2022 23:41:16 +0000 (19:41 -0400)
committerMarge Bot <emma+marge@anholt.net>
Wed, 19 Oct 2022 04:56:55 +0000 (04:56 +0000)
commit211c2aebd2e12fef65ef4a397cb106ef98783fff
treedb329974591b0fc70f464ee966565a8a93153ec8
parentf7b513c9b1efa56514bf3d0aa778a141a96f98ff
cso: make memcmp and hash computation use a literal key size for blend state

This results in faster lookups because memcmp and the hash computation
can be unrolled.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19129>
src/gallium/auxiliary/cso_cache/cso_context.c