summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJordan Justen <jordan.l.justen@intel.com>2014-10-10 08:28:24 -0700
committerJordan Justen <jordan.l.justen@intel.com>2015-07-09 14:37:22 -0700
commit3b2a58c4890738efacf86ba25ed17441a319761d (patch)
tree90adac9ccb477f2cc4acb4ada61bfec3d1bb9e2a
parentdd39d992050c51f6320b7fb1800675bdacec4186 (diff)
glsl/cs: Lower gl_GlobalInvocationID variable
We lower this based on the extension spec formula: gl_GlobalInvocationID = gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID Suggested-by: Kenneth Graunke <kenneth@whitecape.org> Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
-rw-r--r--src/glsl/Makefile.sources1
-rw-r--r--src/glsl/ir_optimization.h1
-rw-r--r--src/glsl/linker.cpp2
-rw-r--r--src/glsl/lower_cs_global_id.cpp178
4 files changed, 182 insertions, 0 deletions
diff --git a/src/glsl/Makefile.sources b/src/glsl/Makefile.sources
index d784a81072..368c4845cf 100644
--- a/src/glsl/Makefile.sources
+++ b/src/glsl/Makefile.sources
@@ -143,6 +143,7 @@ LIBGLSL_FILES = \
loop_unroll.cpp \
lower_clip_distance.cpp \
lower_const_arrays_to_uniforms.cpp \
+ lower_cs_global_id.cpp \
lower_discard.cpp \
lower_discard_flow.cpp \
lower_if_to_cond_assign.cpp \
diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index e6939f3fe1..f377835a72 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -134,6 +134,7 @@ void optimize_dead_builtin_variables(exec_list *instructions,
enum ir_variable_mode other);
bool lower_vertex_id(gl_shader *shader);
+bool lower_cs_global_id(gl_shader *shader);
ir_rvalue *
compare_index_block(exec_list *instructions, ir_variable *index,
diff --git a/src/glsl/linker.cpp b/src/glsl/linker.cpp
index b7a783c098..0a2bf9d90e 100644
--- a/src/glsl/linker.cpp
+++ b/src/glsl/linker.cpp
@@ -1825,6 +1825,8 @@ link_intrastage_shaders(void *mem_ctx,
if (ctx->Const.VertexID_is_zero_based)
lower_vertex_id(linked);
+ lower_cs_global_id(linked);
+
/* Make a pass over all variable declarations to ensure that arrays with
* unspecified sizes have a size specified. The size is inferred from the
* max_array_access field.
diff --git a/src/glsl/lower_cs_global_id.cpp b/src/glsl/lower_cs_global_id.cpp
new file mode 100644
index 0000000000..8a00880d2f
--- /dev/null
+++ b/src/glsl/lower_cs_global_id.cpp
@@ -0,0 +1,178 @@
+/*
+ * Copyright (c) 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file lower_cs_global_id.cpp
+ *
+ * We lower this based on the extension spec formula:
+ * gl_GlobalInvocationID =
+ * gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID
+ */
+
+#include "glsl_symbol_table.h"
+#include "ir_hierarchical_visitor.h"
+#include "ir.h"
+#include "ir_builder.h"
+#include "linker.h"
+#include "program/prog_statevars.h"
+
+namespace {
+
+class lower_cs_global_id_visitor : public ir_hierarchical_visitor {
+public:
+ explicit lower_cs_global_id_visitor(ir_function_signature *main_sig,
+ exec_list *ir_list)
+ : progress(false), GlobalInvocationID(NULL), gl_WorkGroupID(NULL),
+ gl_WorkGroupSize(NULL), gl_LocalInvocationID(NULL),
+ main_sig(main_sig), ir_list(ir_list)
+ {
+ foreach_in_list(ir_instruction, ir, ir_list) {
+ ir_variable *const var = ir->as_variable();
+
+ if (var == NULL)
+ continue;
+
+ if (var->data.mode == ir_var_system_value) {
+ if (var->data.location == SYSTEM_VALUE_WORK_GROUP_ID)
+ gl_WorkGroupID = var;
+ if (var->data.location == SYSTEM_VALUE_LOCAL_INVOCATION_ID)
+ gl_LocalInvocationID = var;
+ } else if (gl_WorkGroupSize == NULL &&
+ /* FIXME: Clean this up */
+ /*var->data.mode == ir_var_auto &&
+ var->data.read_only &&
+ var->data.how_declared == ir_var_declared_implicitly &&*/
+ var->name != NULL &&
+ strcmp(var->name, "gl_WorkGroupSize") == 0) {
+ gl_WorkGroupSize = var;
+ } else
+ continue;
+ if (gl_WorkGroupID != NULL && gl_LocalInvocationID != NULL &&
+ gl_WorkGroupSize != NULL)
+ break;
+ }
+ }
+
+ virtual ir_visitor_status visit(ir_dereference_variable *);
+
+ bool progress;
+
+private:
+ ir_variable *GlobalInvocationID;
+ ir_variable *gl_WorkGroupID;
+ ir_variable *gl_WorkGroupSize;
+ ir_variable *gl_LocalInvocationID;
+
+ ir_function_signature *main_sig;
+ exec_list *ir_list;
+};
+
+} /* anonymous namespace */
+
+ir_visitor_status
+lower_cs_global_id_visitor::visit(ir_dereference_variable *ir)
+{
+ if (ir->var->data.mode != ir_var_system_value ||
+ ir->var->data.location != SYSTEM_VALUE_GLOBAL_INVOCATION_ID)
+ return visit_continue;
+
+ if (GlobalInvocationID == NULL) {
+ const glsl_type *const uvec3_t = glsl_type::uvec3_type;
+ void *const mem_ctx = ralloc_parent(ir);
+
+ GlobalInvocationID =
+ new(mem_ctx) ir_variable(uvec3_t, "__GlobalInvocationID",
+ ir_var_temporary);
+ ir_list->push_head(GlobalInvocationID);
+
+ if (gl_WorkGroupID == NULL) {
+ gl_WorkGroupID = new(mem_ctx) ir_variable(uvec3_t, "gl_WorkGroupID",
+ ir_var_system_value);
+ gl_WorkGroupID->data.how_declared = ir_var_declared_implicitly;
+ gl_WorkGroupID->data.read_only = true;
+ gl_WorkGroupID->data.location = SYSTEM_VALUE_GLOBAL_INVOCATION_ID;
+ gl_WorkGroupID->data.explicit_location = true;
+ gl_WorkGroupID->data.explicit_index = 0;
+ ir_list->push_head(gl_WorkGroupID);
+ }
+
+ if (gl_WorkGroupSize == NULL) {
+ gl_WorkGroupSize = new(mem_ctx) ir_variable(uvec3_t, "gl_WorkGroupSize",
+ ir_var_auto);
+ gl_WorkGroupSize->data.how_declared = ir_var_declared_implicitly;
+ gl_WorkGroupSize->data.read_only = true;
+ ir_list->push_head(gl_WorkGroupSize);
+ }
+
+ if (gl_LocalInvocationID == NULL) {
+ gl_LocalInvocationID =
+ new(mem_ctx) ir_variable(uvec3_t, "gl_LocalInvocationID",
+ ir_var_system_value);
+ gl_LocalInvocationID->data.how_declared = ir_var_declared_implicitly;
+ gl_LocalInvocationID->data.read_only = true;
+ gl_LocalInvocationID->data.location = SYSTEM_VALUE_LOCAL_INVOCATION_ID;
+ gl_LocalInvocationID->data.explicit_location = true;
+ gl_LocalInvocationID->data.explicit_index = 0;
+ ir_list->push_head(gl_LocalInvocationID);
+ }
+
+ /**
+ * gl_GlobalInvocationID =
+ * gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID
+ */
+ ir_instruction *const inst =
+ ir_builder::assign(GlobalInvocationID,
+ ir_builder::add(ir_builder::mul(gl_WorkGroupID,
+ gl_WorkGroupSize),
+ gl_LocalInvocationID));
+
+ main_sig->body.push_head(inst);
+ }
+
+ ir->var = GlobalInvocationID;
+ progress = true;
+
+ return visit_continue;
+}
+
+bool
+lower_cs_global_id(gl_shader *shader)
+{
+ /* gl_GlobalInvocationID only exists in the compute shader.
+ */
+ if (shader->Stage != MESA_SHADER_COMPUTE)
+ return false;
+
+ ir_function_signature *const main_sig =
+ link_get_main_function_signature(shader);
+ if (main_sig == NULL) {
+ assert(main_sig != NULL);
+ return false;
+ }
+
+ lower_cs_global_id_visitor v(main_sig, shader->ir);
+
+ v.run(shader->ir);
+
+ return v.progress;
+}