[Mesa-dev] [PATCH] i965/gen6+: Use non-normalized coordinates for GL_TEXTURE_RECTANGLE.

[Eric Anholt]

Improves performance of a GL_TEXTURE_RECTANGLE microbenchmark by 1.84%
+/- .15% (n=3)
---
 src/mesa/drivers/dri/i965/brw_fs_visitor.cpp     |    3 ++-
 src/mesa/drivers/dri/i965/brw_wm_fp.c            |    4 +++-
 src/mesa/drivers/dri/i965/brw_wm_sampler_state.c |    7 +++++++
 src/mesa/drivers/dri/i965/gen7_sampler_state.c   |    7 +++++++
 4 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
index 2dc9132..828cb63 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp
@@ -1025,7 +1025,8 @@ fs_visitor::visit(ir_texture *ir)
     * texture coordinates.  We use the program parameter state
     * tracking to get the scaling factor.
     */
-   if (ir->sampler->type->sampler_dimensionality == GLSL_SAMPLER_DIM_RECT) {
+   if (intel->gen < 6 &&
+       ir->sampler->type->sampler_dimensionality == GLSL_SAMPLER_DIM_RECT) {
       struct gl_program_parameter_list *params = c->fp->program.Base.Parameters;
       int tokens[STATE_LENGTH] = {
 	 STATE_INTERNAL,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index d52a958..bd46bd8 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -664,6 +664,8 @@ static void precalc_lit( struct brw_wm_compile *c,
 static void precalc_tex( struct brw_wm_compile *c,
 			 const struct prog_instruction *inst )
 {
+   struct brw_compile *p = &c->func;
+   struct intel_context *intel = &p->brw->intel;
    struct prog_src_register coord;
    struct prog_dst_register tmpcoord = { 0 };
    const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
@@ -727,7 +729,7 @@ static void precalc_tex( struct brw_wm_compile *c,
        release_temp(c, tmp0);
        release_temp(c, tmp1);
    }
-   else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
+   else if (intel->gen < 6 && inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
       struct prog_src_register scale = 
 	 search_or_add_param5( c, 
 			       STATE_INTERNAL, 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index 9814613..6834eba 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -289,6 +289,13 @@ static void brw_update_sampler_state(struct brw_context *brw,
    sampler->ss1.max_lod = U_FIXED(CLAMP(gl_sampler->MaxLod, 0, 13), 6);
    sampler->ss1.min_lod = U_FIXED(CLAMP(gl_sampler->MinLod, 0, 13), 6);
 
+   /* On Gen6+, the sampler can handle non-normalized texture
+    * rectangle coordinates natively
+    */
+   if (intel->gen >= 6 && texObj->Target == GL_TEXTURE_RECTANGLE) {
+      sampler->ss3.non_normalized_coord = 1;
+   }
+
    upload_default_color(brw, gl_sampler, unit);
 
    if (intel->gen >= 6) {
diff --git a/src/mesa/drivers/dri/i965/gen7_sampler_state.c b/src/mesa/drivers/dri/i965/gen7_sampler_state.c
index e787c21..dca705b 100644
--- a/src/mesa/drivers/dri/i965/gen7_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_sampler_state.c
@@ -157,6 +157,13 @@ gen7_update_sampler_state(struct brw_context *brw, int unit,
    sampler->ss1.max_lod = U_FIXED(CLAMP(gl_sampler->MaxLod, 0, 13), 8);
    sampler->ss1.min_lod = U_FIXED(CLAMP(gl_sampler->MinLod, 0, 13), 8);
 
+   /* On Gen6+, the sampler can handle non-normalized texture
+    * rectangle coordinates natively
+    */
+   if (intel->gen >= 6 && texObj->Target == GL_TEXTURE_RECTANGLE) {
+      sampler->ss3.non_normalized_coord = 1;
+   }
+
    upload_default_color(brw, gl_sampler, unit);
 
    sampler->ss2.default_color_pointer = brw->wm.sdc_offset[unit] >> 5;
-- 
1.7.5.4