[Mesa-dev] [PATCH v2] glsl: Reduce memory consumption of copy propagation passes.

[Kenneth Graunke]

opt_copy_propagation and opt_copy_propagation_elements create new ACP
and Kill sets each time they enter a new control flow block.  For if
blocks, they also copy the entire existing ACP set contents into the
new set.

When we exit the control flow block, we discard the new sets.  However,
we weren't freeing them - so they lived on until the pass finished.
This can waste a lot of memory (57MB on one pessimal shader).

This patch makes the pass allocate ACP entries using this->acp as the
memory context, and Kill entries out of this->kill.  It also steals
kill entries when moving them from the inner kill list to the parent.

It then frees the lists, including their contents.

Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
---
 src/glsl/opt_copy_propagation.cpp          | 17 ++++++++++++++---
 src/glsl/opt_copy_propagation_elements.cpp | 18 +++++++++++++++---
 2 files changed, 29 insertions(+), 6 deletions(-)

After sending the first patch, I realized that opt_copy_propagation has
the exact same problem as opt_copy_propagation_elements.  This fixes both,
effectively halving the peak memory consumption for this shader.

diff --git a/src/glsl/opt_copy_propagation.cpp b/src/glsl/opt_copy_propagation.cpp
index 5c65af6..a9f3c3e 100644
--- a/src/glsl/opt_copy_propagation.cpp
+++ b/src/glsl/opt_copy_propagation.cpp
@@ -128,6 +128,9 @@ ir_copy_propagation_visitor::visit_enter(ir_function_signature *ir)
 
    visit_list_elements(this, &ir->body);
 
+   ralloc_free(this->acp);
+   ralloc_free(this->kills);
+
    this->kills = orig_kills;
    this->acp = orig_acp;
    this->killed_all = orig_killed_all;
@@ -215,11 +218,13 @@ ir_copy_propagation_visitor::handle_if_block(exec_list *instructions)
 
    /* Populate the initial acp with a copy of the original */
    foreach_in_list(acp_entry, a, orig_acp) {
-      this->acp->push_tail(new(this->mem_ctx) acp_entry(a->lhs, a->rhs));
+      this->acp->push_tail(new(this->acp) acp_entry(a->lhs, a->rhs));
    }
 
    visit_list_elements(this, instructions);
 
+   ralloc_free(this->acp);
+
    if (this->killed_all) {
       orig_acp->make_empty();
    }
@@ -232,6 +237,8 @@ ir_copy_propagation_visitor::handle_if_block(exec_list *instructions)
    foreach_in_list(kill_entry, k, new_kills) {
       kill(k->var);
    }
+
+   ralloc_free(new_kills);
 }
 
 ir_visitor_status
@@ -263,6 +270,8 @@ ir_copy_propagation_visitor::visit_enter(ir_loop *ir)
 
    visit_list_elements(this, &ir->body_instructions);
 
+   ralloc_free(this->acp);
+
    if (this->killed_all) {
       orig_acp->make_empty();
    }
@@ -276,6 +285,8 @@ ir_copy_propagation_visitor::visit_enter(ir_loop *ir)
       kill(k->var);
    }
 
+   ralloc_free(new_kills);
+
    /* already descended into the children. */
    return visit_continue_with_parent;
 }
@@ -294,7 +305,7 @@ ir_copy_propagation_visitor::kill(ir_variable *var)
 
    /* Add the LHS variable to the list of killed variables in this block.
     */
-   this->kills->push_tail(new(this->mem_ctx) kill_entry(var));
+   this->kills->push_tail(new(this->kills) kill_entry(var));
 }
 
 /**
@@ -322,7 +333,7 @@ ir_copy_propagation_visitor::add_copy(ir_assignment *ir)
 	 ir->condition = new(ralloc_parent(ir)) ir_constant(false);
 	 this->progress = true;
       } else {
-	 entry = new(this->mem_ctx) acp_entry(lhs_var, rhs_var);
+	 entry = new(this->acp) acp_entry(lhs_var, rhs_var);
 	 this->acp->push_tail(entry);
       }
    }
diff --git a/src/glsl/opt_copy_propagation_elements.cpp b/src/glsl/opt_copy_propagation_elements.cpp
index c3e55bc..dcebd4c 100644
--- a/src/glsl/opt_copy_propagation_elements.cpp
+++ b/src/glsl/opt_copy_propagation_elements.cpp
@@ -156,6 +156,9 @@ ir_copy_propagation_elements_visitor::visit_enter(ir_function_signature *ir)
 
    visit_list_elements(this, &ir->body);
 
+   ralloc_free(this->acp);
+   ralloc_free(this->kills);
+
    this->kills = orig_kills;
    this->acp = orig_acp;
    this->killed_all = orig_killed_all;
@@ -173,9 +176,9 @@ ir_copy_propagation_elements_visitor::visit_leave(ir_assignment *ir)
       kill_entry *k;
 
       if (lhs)
-	 k = new(mem_ctx) kill_entry(var, ir->write_mask);
+	 k = new(this->kills) kill_entry(var, ir->write_mask);
       else
-	 k = new(mem_ctx) kill_entry(var, ~0);
+	 k = new(this->kills) kill_entry(var, ~0);
 
       kill(k);
    }
@@ -334,11 +337,13 @@ ir_copy_propagation_elements_visitor::handle_if_block(exec_list *instructions)
 
    /* Populate the initial acp with a copy of the original */
    foreach_in_list(acp_entry, a, orig_acp) {
-      this->acp->push_tail(new(this->mem_ctx) acp_entry(a));
+      this->acp->push_tail(new(this->acp) acp_entry(a));
    }
 
    visit_list_elements(this, instructions);
 
+   ralloc_free(this->acp);
+
    if (this->killed_all) {
       orig_acp->make_empty();
    }
@@ -354,6 +359,8 @@ ir_copy_propagation_elements_visitor::handle_if_block(exec_list *instructions)
    foreach_in_list_safe(kill_entry, k, new_kills) {
       kill(k);
    }
+
+   ralloc_free(new_kills);
 }
 
 ir_visitor_status
@@ -385,6 +392,8 @@ ir_copy_propagation_elements_visitor::visit_enter(ir_loop *ir)
 
    visit_list_elements(this, &ir->body_instructions);
 
+   ralloc_free(this->acp);
+
    if (this->killed_all) {
       orig_acp->make_empty();
    }
@@ -398,6 +407,8 @@ ir_copy_propagation_elements_visitor::visit_enter(ir_loop *ir)
       kill(k);
    }
 
+   ralloc_free(new_kills);
+
    /* already descended into the children. */
    return visit_continue_with_parent;
 }
@@ -423,6 +434,7 @@ ir_copy_propagation_elements_visitor::kill(kill_entry *k)
    if (k->next)
       k->remove();
 
+   ralloc_steal(this->kills, k);
    this->kills->push_tail(k);
 }
 
-- 
2.1.3