i965: Create NIR during LinkShader() and ProgramStringNotify().

Previously, we translated into NIR and did all the optimizations and
lowering as part of running fs_visitor.  This meant that we did all of
that work twice for fragment shaders - once for SIMD8, and again for
SIMD16.  We also had to redo it every time we hit a state based
recompile.

We now generate NIR once at link time.  ARB programs don't have linking,
so we instead generate it at ProgramStringNotify time.

Mesa's fixed function vertex program handling doesn't bother to inform
the driver about new programs at all (which is rather mean), so we
generate NIR at the last minute, if it hasn't happened already.

shader-db runs ~9.4% faster on my i7-5600U, with a release build.

v2: Check NirOptions != NULL in ProgramStringNotify().  Don't bother
    using _mesa_program_enum_to_shader_stage as we already know it.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

1 files changed, 6 insertions, 0 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_nir.h b/src/mesa/drivers/dri/i965/brw_nir.h
index 27782a3..3131109 100644
--- a/src/mesa/drivers/dri/i965/brw_nir.h
+++ b/src/mesa/drivers/dri/i965/brw_nir.h
@@ -23,6 +23,7 @@
 
 #pragma once
 
+#include "brw_context.h"
 #include "glsl/nir/nir.h"
 
 #ifdef __cplusplus
@@ -73,6 +74,11 @@ enum {
 
 void brw_nir_analyze_boolean_resolves(nir_shader *nir);
 
+nir_shader *brw_create_nir(struct brw_context *brw,
+                           const struct gl_shader_program *shader_prog,
+                           const struct gl_program *prog,
+                           gl_shader_stage stage);
+
 #ifdef __cplusplus
 }
 #endif