OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v3d: Fix copy-propagation of input unpacks.

Browse Source 
I had a single function for "does this do float input unpacking" with two
major flaws: It was missing the most common thing to try to copy propagate
a f32 input nunpack to (the VFPACK to an FP16 render target) along with
several other ALU ops, and also would try to propagate an f32 unpack into
a VFMUL which only does f16 unpacks.

instructions in affected programs: 659232 -> 655895 (-0.51%)
uniforms in affected programs: 132613 -> 135336 (2.05%)

and a couple of programs increase their thread counts.

The uniforms hit appears to be a pattern in generated code of doing (-a >=
a) comparisons, which when a is abs(b) can result in the abs instruction
being copy propagated once but not fully DCEed.
This commit is contained in:
Eric Anholt
2019-02-04 10:35:40 -08:00
parent e5c6938590
commit 940501a446
5 changed files with 94 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/broadcom/compiler/v3d_compiler.h
View File

@@ -787,7 +787,6 @@ bool vir_is_raw_mov(struct qinst *inst);
bool vir_is_tex(struct qinst *inst);
bool vir_is_add(struct qinst *inst);
bool vir_is_mul(struct qinst *inst);
bool vir_is_float_input(struct qinst *inst);
bool vir_writes_r3(const struct v3d_device_info *devinfo, struct qinst *inst);
bool vir_writes_r4(const struct v3d_device_info *devinfo, struct qinst *inst);
struct qreg vir_follow_movs(struct v3d_compile *c, struct qreg reg);