[Mesa-dev] [PATCH 35/46] glsl: pass shader stage to lower_output_reads and handle tess control

[Marek Olšák]

From: Ilia Mirkin <imirkin at alum.mit.edu>

Tessellation control outputs can be read in directly without first
having been written. Accessing these will require some special logic
anyways, so just let them through.

V2: Never lower tess control output reads, whether patch or not -- both
can be read back by other threads.

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
Signed-off-by: Chris Forbes <chrisf at ijw.co.nz>
---
 src/glsl/ir_optimization.h                 |  2 +-
 src/glsl/lower_output_reads.cpp            | 13 +++++++++----
 src/mesa/drivers/dri/i965/brw_shader.cpp   |  2 +-
 src/mesa/state_tracker/st_glsl_to_tgsi.cpp |  2 +-
 4 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index a174c96..766b723 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -120,7 +120,7 @@ bool lower_variable_index_to_cond_assign(gl_shader_stage stage,
 bool lower_quadop_vector(exec_list *instructions, bool dont_lower_swz);
 bool lower_const_arrays_to_uniforms(exec_list *instructions);
 bool lower_clip_distance(gl_shader *shader);
-void lower_output_reads(exec_list *instructions);
+void lower_output_reads(unsigned stage, exec_list *instructions);
 bool lower_packing_builtins(exec_list *instructions, int op_mask);
 void lower_ubo_reference(struct gl_shader *shader, exec_list *instructions);
 void lower_packed_varyings(void *mem_ctx,
diff --git a/src/glsl/lower_output_reads.cpp b/src/glsl/lower_output_reads.cpp
index 1ee815d..79488df 100644
--- a/src/glsl/lower_output_reads.cpp
+++ b/src/glsl/lower_output_reads.cpp
@@ -48,8 +48,10 @@ protected:
    hash_table *replacements;
 
    void *mem_ctx;
+
+   unsigned stage;
 public:
-   output_read_remover();
+   output_read_remover(unsigned stage);
    ~output_read_remover();
    virtual ir_visitor_status visit(class ir_dereference_variable *);
    virtual ir_visitor_status visit_leave(class ir_emit_vertex *);
@@ -75,8 +77,9 @@ hash_table_var_hash(const void *key)
    return hash_table_string_hash(var->name);
 }
 
-output_read_remover::output_read_remover()
+output_read_remover::output_read_remover(unsigned stage)
 {
+   this->stage = stage;
    mem_ctx = ralloc_context(NULL);
    replacements =
       hash_table_ctor(0, hash_table_var_hash, hash_table_pointer_compare);
@@ -93,6 +96,8 @@ output_read_remover::visit(ir_dereference_variable *ir)
 {
    if (ir->var->data.mode != ir_var_shader_out)
       return visit_continue;
+   if (stage == MESA_SHADER_TESS_CTRL)
+      return visit_continue;
 
    ir_variable *temp = (ir_variable *) hash_table_find(replacements, ir->var);
 
@@ -166,8 +171,8 @@ output_read_remover::visit_leave(ir_function_signature *sig)
 }
 
 void
-lower_output_reads(exec_list *instructions)
+lower_output_reads(unsigned stage, exec_list *instructions)
 {
-   output_read_remover v;
+   output_read_remover v(stage);
    visit_list_elements(&v, instructions);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp
index 8b5bb72..37bba17 100644
--- a/src/mesa/drivers/dri/i965/brw_shader.cpp
+++ b/src/mesa/drivers/dri/i965/brw_shader.cpp
@@ -220,7 +220,7 @@ process_glsl_ir(gl_shader_stage stage,
    } while (progress);
 
    if (options->NirOptions != NULL)
-      lower_output_reads(shader->ir);
+      lower_output_reads(stage, shader->ir);
 
    validate_ir_tree(shader->ir);
 
diff --git a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
index e440aef..9d10eef 100644
--- a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
+++ b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
@@ -5639,7 +5639,7 @@ get_mesa_program(struct gl_context *ctx,
                                                prog->Parameters);
 
    /* Remove reads from output registers. */
-   lower_output_reads(shader->ir);
+   lower_output_reads(shader->Stage, shader->ir);
 
    /* Emit intermediate IR for main(). */
    visit_exec_list(shader->ir, v);
-- 
2.1.0