diff options
| author | Eric Anholt <[email protected]> | 2016-07-06 15:21:53 -0700 |
|---|---|---|
| committer | Eric Anholt <[email protected]> | 2016-07-12 15:47:25 -0700 |
| commit | 6858f059246b2cf6b969b050154dcb7307a03833 (patch) | |
| tree | ce67e9ee67849a3ae7039756078784a452185112 /src/gallium | |
| parent | e8959ba7afbda7a23805072efc15c6f11449103e (diff) | |
vc4: Also enable phi elimination.
This avoids a bunch of code gen regressions when enabling loops in vc4.
Prior to that, the GLSL that would have generated these optimizable phi
nodes was being lowered to csels between either (undef, a) or (a, a), and
those were being dealt with by nir_opt_undef and nir_opt_algebraic.
Diffstat (limited to 'src/gallium')
| -rw-r--r-- | src/gallium/drivers/vc4/vc4_program.c | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/src/gallium/drivers/vc4/vc4_program.c b/src/gallium/drivers/vc4/vc4_program.c index 8305459454b..20e2850e625 100644 --- a/src/gallium/drivers/vc4/vc4_program.c +++ b/src/gallium/drivers/vc4/vc4_program.c @@ -1336,6 +1336,7 @@ vc4_optimize_nir(struct nir_shader *s) NIR_PASS_V(s, nir_lower_phis_to_scalar); NIR_PASS(progress, s, nir_copy_prop); + NIR_PASS(progress, s, nir_opt_remove_phis); NIR_PASS(progress, s, nir_opt_dce); NIR_PASS(progress, s, nir_opt_dead_cf); NIR_PASS(progress, s, nir_opt_cse); |