glsl/lower_ubo_reference: split array copies into element copies

Improves register pressure, since otherwise we end up emitting
loads for all the elements in the RHS and them emitting
stores for all elements in the LHS.

v2:
  - Mark progress properly. This also fixes some instances where the added
    nodes with individual element copies where not being lowered, which is
    expected behavior as explained in the documentation for
    visit_list_elements.
  - Only need to do this if the RHS is a buffer-backed variable.
  - We can also have arrays inside structs. A later patch will make it so
    we also split struct copies and end up with multiple
    ir_dereference_record assignments, so make sure that if any of these
    is an array copy, we also split it.

Fixes the following piglit tests:
tests/spec/arb_shader_storage_buffer_object/execution/large-field-copy.shader_test
tests/spec/arb_shader_storage_buffer_object/linker/copy-large-array.shader_test

Reviewed-by: Jordan Justen <[email protected]>

1 files changed, 67 insertions, 0 deletions

diff --git a/src/glsl/lower_ubo_reference.cpp b/src/glsl/lower_ubo_reference.cpp
index b74aa3d0630..b82d800eefb 100644
--- a/src/glsl/lower_ubo_reference.cpp
+++ b/src/glsl/lower_ubo_reference.cpp
@@ -154,6 +154,7 @@ public:
    ir_call *ssbo_load(const struct glsl_type *type,
                       ir_rvalue *offset);
 
+   bool check_for_buffer_array_copy(ir_assignment *ir);
    void check_for_ssbo_store(ir_assignment *ir);
    void write_to_memory(ir_dereference *deref,
                         ir_variable *var,
@@ -1132,10 +1133,76 @@ lower_ubo_reference_visitor::check_for_ssbo_store(ir_assignment *ir)
    progress = true;
 }
 
+static bool
+is_buffer_backed_variable(ir_variable *var)
+{
+   return var->is_in_buffer_block() ||
+          var->data.mode == ir_var_shader_shared;
+}
+
+bool
+lower_ubo_reference_visitor::check_for_buffer_array_copy(ir_assignment *ir)
+{
+   if (!ir || !ir->lhs || !ir->rhs)
+      return false;
+
+   /* LHS and RHS must be arrays
+    * FIXME: arrays of arrays?
+    */
+   if (!ir->lhs->type->is_array() || !ir->rhs->type->is_array())
+      return false;
+
+   /* RHS must be a buffer-backed variable. This is what can cause the problem
+    * since it would lead to a series of loads that need to live until we
+    * see the writes to the LHS.
+    */
+   ir_variable *rhs_var = ir->rhs->variable_referenced();
+   if (!rhs_var || !is_buffer_backed_variable(rhs_var))
+      return false;
+
+   /* Split the array copy into individual element copies to reduce
+    * register pressure
+    */
+   ir_dereference *rhs_deref = ir->rhs->as_dereference();
+   if (!rhs_deref)
+      return false;
+
+   ir_dereference *lhs_deref = ir->lhs->as_dereference();
+   if (!lhs_deref)
+      return false;
+
+   assert(lhs_deref->type->length == rhs_deref->type->length);
+   mem_ctx = ralloc_parent(shader->ir);
+
+   for (unsigned i = 0; i < lhs_deref->type->length; i++) {
+      ir_dereference *lhs_i =
+         new(mem_ctx) ir_dereference_array(lhs_deref->clone(mem_ctx, NULL),
+                                           new(mem_ctx) ir_constant(i));
+
+      ir_dereference *rhs_i =
+         new(mem_ctx) ir_dereference_array(rhs_deref->clone(mem_ctx, NULL),
+                                           new(mem_ctx) ir_constant(i));
+      ir->insert_after(assign(lhs_i, rhs_i));
+   }
+
+   ir->remove();
+   progress = true;
+   return true;
+}
 
 ir_visitor_status
 lower_ubo_reference_visitor::visit_enter(ir_assignment *ir)
 {
+   /* Array copies could involve large amounts of load/store
+    * operations. To improve register pressure we want to special-case
+    * these and split array copies into individual element copies.
+    * This way we avoid emitting all the loads for the RHS first and
+    * all the writes for the LHS second and register usage is more
+    * efficient.
+    */
+   if (check_for_buffer_array_copy(ir))
+      return visit_continue_with_parent;
+
    check_ssbo_unsized_array_length_assignment(ir);
    check_for_ssbo_store(ir);
    return rvalue_visit(ir);