diff options
author | Timothy Arceri <[email protected]> | 2016-07-31 22:39:26 +1000 |
---|---|---|
committer | Timothy Arceri <[email protected]> | 2017-04-24 12:08:14 +1000 |
commit | 21173194db12f5eff7361040c038dbfc9cde46e4 (patch) | |
tree | 427c6484432094b9167fd6650b3c15f565a92f34 /src/compiler/glsl | |
parent | eb8aa93c03ee89ffd3041d41b6293e4b282b6ce6 (diff) |
glsl: use ARB_enhahnced_layouts for packing where possible
If packing doesn't cross locations we can easily make use of
ARB_enhanced_layouts to do packing rather than using the GLSL IR
lowering pass lower_packed_varyings().
Shader-db Broadwell results:
total instructions in shared programs: 12977822 -> 12977819 (-0.00%)
instructions in affected programs: 1871 -> 1868 (-0.16%)
helped: 4
HURT: 3
total cycles in shared programs: 246567288 -> 246567668 (0.00%)
cycles in affected programs: 1370386 -> 1370766 (0.03%)
helped: 592
HURT: 733
Acked-by: Elie Tournier <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
Diffstat (limited to 'src/compiler/glsl')
-rw-r--r-- | src/compiler/glsl/link_varyings.cpp | 70 |
1 files changed, 70 insertions, 0 deletions
diff --git a/src/compiler/glsl/link_varyings.cpp b/src/compiler/glsl/link_varyings.cpp index 6bcec72f5fc..d04f546c243 100644 --- a/src/compiler/glsl/link_varyings.cpp +++ b/src/compiler/glsl/link_varyings.cpp @@ -1217,6 +1217,7 @@ class varying_matches { public: varying_matches(bool disable_varying_packing, bool xfb_enabled, + bool enhanced_layouts_enabled, gl_shader_stage producer_stage, gl_shader_stage consumer_stage); ~varying_matches(); @@ -1250,6 +1251,8 @@ private: */ const bool xfb_enabled; + const bool enhanced_layouts_enabled; + /** * Enum representing the order in which varyings are packed within a * packing class. @@ -1326,10 +1329,12 @@ private: varying_matches::varying_matches(bool disable_varying_packing, bool xfb_enabled, + bool enhanced_layouts_enabled, gl_shader_stage producer_stage, gl_shader_stage consumer_stage) : disable_varying_packing(disable_varying_packing), xfb_enabled(xfb_enabled), + enhanced_layouts_enabled(enhanced_layouts_enabled), producer_stage(producer_stage), consumer_stage(consumer_stage) { @@ -1622,6 +1627,12 @@ varying_matches::assign_locations(struct gl_shader_program *prog, void varying_matches::store_locations() const { + /* Check is location needs to be packed with lower_packed_varyings() or if + * we can just use ARB_enhanced_layouts packing. + */ + bool pack_loc[MAX_VARYINGS_INCL_PATCH] = { 0 }; + const glsl_type *loc_type[MAX_VARYINGS_INCL_PATCH][4] = { {NULL, NULL} }; + for (unsigned i = 0; i < this->num_matches; i++) { ir_variable *producer_var = this->matches[i].producer_var; ir_variable *consumer_var = this->matches[i].consumer_var; @@ -1639,6 +1650,64 @@ varying_matches::store_locations() const consumer_var->data.location = VARYING_SLOT_VAR0 + slot; consumer_var->data.location_frac = offset; } + + /* Find locations suitable for native packing via + * ARB_enhanced_layouts. + */ + if (producer_var && consumer_var) { + if (enhanced_layouts_enabled) { + const glsl_type *type = + get_varying_type(producer_var, producer_stage); + if (type->is_array() || type->is_matrix() || type->is_record() || + type->is_double()) { + unsigned comp_slots = type->component_slots() + offset; + unsigned slots = comp_slots / 4; + if (comp_slots % 4) + slots += 1; + + for (unsigned j = 0; j < slots; j++) { + pack_loc[slot + j] = true; + } + } else if (offset + type->vector_elements > 4) { + pack_loc[slot] = true; + pack_loc[slot + 1] = true; + } else { + loc_type[slot][offset] = type; + } + } + } + } + + /* Attempt to use ARB_enhanced_layouts for more efficient packing if + * suitable. + */ + if (enhanced_layouts_enabled) { + for (unsigned i = 0; i < this->num_matches; i++) { + ir_variable *producer_var = this->matches[i].producer_var; + ir_variable *consumer_var = this->matches[i].consumer_var; + unsigned generic_location = this->matches[i].generic_location; + unsigned slot = generic_location / 4; + + if (pack_loc[slot] || !producer_var || !consumer_var) + continue; + + const glsl_type *type = + get_varying_type(producer_var, producer_stage); + bool type_match = true; + for (unsigned j = 0; j < 4; j++) { + if (loc_type[slot][j]) { + if (type->base_type != loc_type[slot][j]->base_type) + type_match = false; + } + } + + if (type_match) { + producer_var->data.explicit_location = 1; + consumer_var->data.explicit_location = 1; + producer_var->data.explicit_component = 1; + consumer_var->data.explicit_component = 1; + } + } } } @@ -2100,6 +2169,7 @@ assign_varying_locations(struct gl_context *ctx, disable_varying_packing = true; varying_matches matches(disable_varying_packing, xfb_enabled, + ctx->Extensions.ARB_enhanced_layouts, producer ? producer->Stage : (gl_shader_stage)-1, consumer ? consumer->Stage : (gl_shader_stage)-1); hash_table *tfeedback_candidates = |