I like this, but it's going to conflict with the universe as far as other i965 not changes are concerned so I'm not sure when we want to land it. Other than that and the handful of comments, this series is Reviewed-by: Jason Ekstrand <<a href="mailto:jason.ekstrand@intel.com">jason.ekstrand@intel.com</a>> <div class="gmail_quote">On Jan 22, 2015 3:41 AM, "Kenneth Graunke" <<a href="mailto:kenneth@whitecape.org">kenneth@whitecape.org</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">We want to run CSE and algebraic optimizations again after lowering IO. Some of the passes in the optimization loop don't handle saturates and other modifiers, so run it before lowering to source modifiers. total instructions in shared programs: 6046190 -> 6045768 (-0.01%) instructions in affected programs: 22406 -> 21984 (-1.88%) helped: 47 HURT: 0 GAINED: 0 LOST: 0 Signed-off-by: Kenneth Graunke <<a href="mailto:kenneth@whitecape.org">kenneth@whitecape.org</a>> --- src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 34 ++++++++++++++++++++------------ 1 file changed, 21 insertions(+), 13 deletions(-) I'm not sure about this patch. It's obviously beneficial, but... I wonder if we want some kind of optimization loop in core NIR, along the lines of do_common_optimization. Right now, NIR seems like a collection of tools to build your own compile process. Which works, and the flexibility is nice. But having everything in one place isn't a crazy plan either... diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp index 510092e..40a1673 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp @@ -26,20 +26,9 @@ #include "glsl/nir/glsl_to_nir.h" #include "brw_fs.h" -void -fs_visitor::emit_nir_code() +static void +nir_optimize(nir_shader *nir) { - /* first, lower the GLSL IR shader to NIR */ - lower_output_reads(shader-><a href="http://base.ir" target="_blank">base.ir</a>); - nir_shader *nir = glsl_to_nir(shader-><a href="http://base.ir" target="_blank">base.ir</a>, NULL, true); - nir_validate_shader(nir); - - nir_lower_global_vars_to_local(nir); - nir_validate_shader(nir); - - nir_split_var_copies(nir); - nir_validate_shader(nir); - bool progress; do { progress = false; @@ -58,6 +47,23 @@ fs_visitor::emit_nir_code() progress |= nir_opt_constant_folding(nir); nir_validate_shader(nir); } while (progress); +} + +void +fs_visitor::emit_nir_code() +{ + /* first, lower the GLSL IR shader to NIR */ + lower_output_reads(shader-><a href="http://base.ir" target="_blank">base.ir</a>); + nir_shader *nir = glsl_to_nir(shader-><a href="http://base.ir" target="_blank">base.ir</a>, NULL, true); + nir_validate_shader(nir); + + nir_lower_global_vars_to_local(nir); + nir_validate_shader(nir); + + nir_split_var_copies(nir); + nir_validate_shader(nir); + + nir_optimize(nir); /* Lower a bunch of stuff */ nir_lower_var_copies(nir); @@ -81,6 +87,8 @@ fs_visitor::emit_nir_code() nir_lower_atomics(nir); nir_validate_shader(nir); + nir_optimize(nir); + nir_lower_to_source_mods(nir); nir_validate_shader(nir); nir_copy_prop(nir); -- 2.2.2 _______________________________________________ mesa-dev mailing list <a href="mailto:mesa-dev@lists.freedesktop.org">mesa-dev@lists.freedesktop.org</a> <a href="http://lists.freedesktop.org/mailman/listinfo/mesa-dev" target="_blank">http://lists.freedesktop.org/mailman/listinfo/mesa-dev</a> </blockquote></div>