agx: Use 16-bit immediates
authorAlyssa Rosenzweig <alyssa@rosenzweig.io>
Sat, 10 Sep 2022 19:57:43 +0000 (15:57 -0400)
committerMarge Bot <emma+marge@anholt.net>
Thu, 22 Sep 2022 03:23:36 +0000 (03:23 +0000)
commit14f2be1f33827c73e98d5605daf20d5569b29fcf
tree8b2b581a42c66a83ca0f9e117d254f064fcc75c4
parente302e5d527fb2bee1d5863983de8f631259f0470
agx: Use 16-bit immediates

This is slightly more accurate in the IR, and means we instruction
select the current 16-bit size floating point instructions when all
non-immediate operands are 16-bit.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18687>
src/asahi/compiler/agx_compile.c
src/asahi/compiler/agx_compiler.h
src/asahi/compiler/agx_optimizer.c