I've landed most of the patches in the original series. What remains
is consolidating the two passes in a single one.
Since opt_copy_propagation.cpp would be gone in the end of the series,
I've dropped the patch that changes it to handle both paths in an
if-statement.
Patches 1 and 2 are already reviewed by Eric Anholt, I'm sending them
as a context for the patch 3. That last patch doesn't change the
results (instructions / cycles) but makes up for the performance time
we lost to process both paths of if-statement.
I've tried other optimization tricks without much significant results.
Caio Marcelo de Oliveira Filho (3):
glsl: teach copy_propagation_elements to deal with whole variables
glsl: use only copy_propagation_elements
glsl: propagate full variables eagerly
src/compiler/Makefile.sources | 1 -
src/compiler/glsl/glsl_parser_extras.cpp | 1 -
src/compiler/glsl/ir_optimization.h | 1 -
src/compiler/glsl/meson.build | 1 -
src/compiler/glsl/opt_copy_propagation.cpp| 369 --
.../glsl/opt_copy_propagation_elements.cpp| 168 ++--
src/compiler/glsl/test_optpass.cpp| 2 -
7 files changed, 139 insertions(+), 404 deletions(-)
delete mode 100644 src/compiler/glsl/opt_copy_propagation.cpp
--
2.18.0
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev