From e31266ed3e3667c043bc5ad1abd65cfdb0fa7fdb Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 25 Jan 2011 10:28:13 +1000 Subject: 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. --- src/glsl/ir_optimization.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src/glsl/ir_optimization.h') 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); -- cgit v1.2.3