diff options
| author | Eric Anholt <eric@anholt.net> | 2011-01-25 10:28:13 +1000 |
|---|---|---|
| committer | Eric Anholt <eric@anholt.net> | 2011-02-04 12:18:38 -0600 |
| commit | e31266ed3e3667c043bc5ad1abd65cfdb0fa7fdb (patch) | |
| tree | a9ddee7289bb07b33c2749a555f8df57055294cb /src/glsl/ir_optimization.h | |
| parent | cde443e0b94ec6d8d0a99284e72ca7f9e2591174 (diff) | |
glsl: Add a new opt_copy_propagation variant that does it channel-wise.
This patch cleans up many of the extra copies in GLSL IR introduced by
i965's scalarizing passes. It doesn't result in a statistically
significant performance difference on nexuiz high settings (n=3) or my
demo (n=10), due to brw_fs.cpp's register coalescing covering most of
those extra moves anyway. However, it does make the debug of wine's
GLSL shaders much more tractable, and reduces instruction count of
glsl-fs-convolution-2 from 376 to 288.
Diffstat (limited to 'src/glsl/ir_optimization.h')
| -rw-r--r-- | src/glsl/ir_optimization.h | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h index dbc9f4ab97..dd265673c5 100644 --- a/src/glsl/ir_optimization.h +++ b/src/glsl/ir_optimization.h @@ -43,6 +43,7 @@ bool do_constant_folding(exec_list *instructions); bool do_constant_variable(exec_list *instructions); bool do_constant_variable_unlinked(exec_list *instructions); bool do_copy_propagation(exec_list *instructions); +bool do_copy_propagation_elements(exec_list *instructions); bool do_constant_propagation(exec_list *instructions); bool do_dead_code(exec_list *instructions); bool do_dead_code_local(exec_list *instructions); |