KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vc4: Enable the nir_opt_algebraic_late() pass. 

Upcoming changes to sub optimization will make this pass required.  Over
the course of that series, we see uniforms +.46%, instructions -.24%
(seems like a fine tradeoff -- uniforms are 1/2 the size of instructions
as far as cache occupancy)

Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
This commit is contained in:
Eric Anholt 2019-09-26 11:03:46 -07:00 committed by Daniel Schürmann
parent f2b8051d69
commit d54ae70ee7
1 changed files with 15 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/gallium/drivers/vc4/vc4_program.c
View File

@ -2335,6 +2335,21 @@ vc4_shader_ntq(struct vc4_context *vc4, enum qstage stage,
vc4_optimize_nir(c->s);
/* Do late algebraic optimization to turn add(a, neg(b)) back into
* subs, then the mandatory cleanup after algebraic. Note that it may
* produce fnegs, and if so then we need to keep running to squash
* fneg(fneg(a)).
*/
bool more_late_algebraic = true;
while (more_late_algebraic) {
more_late_algebraic = false;
NIR_PASS(more_late_algebraic, c->s, nir_opt_algebraic_late);
NIR_PASS_V(c->s, nir_opt_constant_folding);
NIR_PASS_V(c->s, nir_copy_prop);
NIR_PASS_V(c->s, nir_opt_dce);
NIR_PASS_V(c->s, nir_opt_cse);
}
NIR_PASS_V(c->s, nir_lower_bool_to_int32);
NIR_PASS_V(c->s, nir_convert_from_ssa, true);