lima/ppir: enable vectorize optimization

pp has vector units and some operations can be optimized when bundled
together.
Benchmarking this with piglit shaders shows that the instruction count
can be greatly reduced on many examples with vectorize.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Qiang Yu <yuq825@gmail.com>

diff --git a/src/gallium/drivers/lima/lima_program.c b/src/gallium/drivers/lima/lima_program.c
index cb05c6f..b9c4cbc 100644 (file)
--- a/src/gallium/drivers/lima/lima_program.c
+++ b/src/gallium/drivers/lima/lima_program.c
@@ -171,6 +171,11 @@ lima_program_optimize_fs_nir(struct nir_shader *s)
 
    do {
       progress = false;
+      NIR_PASS(progress, s, nir_opt_vectorize);
+   } while (progress);
+
+   do {
+      progress = false;
 
       NIR_PASS_V(s, nir_lower_vars_to_ssa);
       NIR_PASS(progress, s, nir_lower_alu_to_scalar, alu_lower);