fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-04 22:10:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
164359 commits 64 branches 915 tags 1,023 MiB
Commit graph

3 commits

Author SHA1 Message Date
Emma Anholt
22f7f167cd nir/opt_phi_precision: Fix missing swizzles when narrowing phi srcs. 
This NIR:

	vec4 32 ssa_169 = phi block_1: ssa_168, block_2: ssa_138
	vec1 16 ssa_209 = f2fmp ssa_169.x
	vec1 16 ssa_210 = f2fmp ssa_169.y
	vec1 16 ssa_211 = f2fmp ssa_169.z
	vec1 16 ssa_212 = f2fmp ssa_169.w
	vec4 16 ssa_213 = vec4 ssa_209, ssa_210, ssa_211, ssa_212
	intrinsic store_output (ssa_213, ssa_171) (base=0, wrmask=xyzw /*15*/, component=0, src_type=float16 /*144*/, io location=4 slots=1 mediump /*8388740*/, xfb() /*0*/, xfb2() /*0*/)

would turn into:

	vec4 32 ssa_169 = phi block_1: ssa_168, block_2: ssa_138
	vec4 16 ssa_216 = phi block_1: ssa_214, block_2: ssa_215
	vec1 16 ssa_209 = f2fmp ssa_169.x
	vec1 16 ssa_210 = f2fmp ssa_169.y
	vec1 16 ssa_211 = f2fmp ssa_169.z
	vec1 16 ssa_212 = f2fmp ssa_169.w
	vec4 16 ssa_213 = vec4 ssa_216.x, ssa_216.x, ssa_216.x, ssa_216.x
	intrinsic store_output (ssa_213, ssa_171) (base=0, wrmask=xyzw /*15*/, component=0, src_type=float16 /*144*/, io location=4 slots=1 mediump /*8388740*/, xfb() /*0*/, xfb2() /*0*/)

ignoring the swizzles from the f2fmp srcs.  Fixes failures in
dEQP-GLES2.functional.shaders.random.all_features.fragment.20 on
turnip+ANGLE.

Fixes: c7b935962b ("nir: Add pass to lower phi precision")
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19179>
 2022-10-22 03:06:31 +00:00
Emma Anholt
673cc9323a nir: Move phi src setup to a helper. 
Cleans up the ralloc/list push code all over the tree.

Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11772>
 2021-08-13 16:11:57 +00:00
Rob Clark
c7b935962b nir: Add pass to lower phi precision 
In addition to register pressure benefits from getting more fp16/int16,
this avoids i2imp's from standing in the way of loop unrolling.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11545>
 2021-06-29 23:27:28 +00:00