summaryrefslogtreecommitdiff
authorKenneth Graunke <kenneth@whitecape.org>2013-04-04 06:56:57 (GMT)
committer Kenneth Graunke <kenneth@whitecape.org>2013-04-04 22:38:19 (GMT)
commitedc52a8f280db8d9dc9a04314741319ea8b83e67 (patch) (side-by-side diff)
treeae9866c0c13ccfec896c5dd1ed433ad22956cced
parent967514ce680f7bf785ab544c6174786dd20425de (diff)
downloadmesa-edc52a8f280db8d9dc9a04314741319ea8b83e67.zip
mesa-edc52a8f280db8d9dc9a04314741319ea8b83e67.tar.gz
glsl: Add an optimization pass to flatten simple nested if blocks.
GLBenchmark 2.7's shaders contain conditional blocks like: if (x) { if (y) { ... } } where the outer conditional's then clause contains exactly one statement (the nested if) and there are no else clauses. This can easily be optimized into: if (x && y) { ... } This saves a few instructions in GLBenchmark 2.7: total instructions in shared programs: 11833 -> 11649 (-1.55%) instructions in affected programs: 8234 -> 8050 (-2.23%) It also helps CS:GO slightly (-0.05%/-0.22%). More importantly, however, it simplifies the control flow graph, which could enable other optimizations. Signed-off-by: Kenneth Graunke <kenneth@whitecape.org> Reviewed-by: Matt Turner <mattst88@gmail.com> Reviewed-by: Eric Anholt <eric@anholt.net>
Diffstat (more/less context) (ignore whitespace changes)
-rw-r--r--src/glsl/Makefile.sources1
-rw-r--r--src/glsl/glsl_parser_extras.cpp1
-rw-r--r--src/glsl/ir_optimization.h1
-rw-r--r--src/glsl/opt_flatten_nested_if_blocks.cpp103
4 files changed, 106 insertions, 0 deletions
diff --git a/src/glsl/Makefile.sources b/src/glsl/Makefile.sources
index c294aa4..b5282a6 100644
--- a/src/glsl/Makefile.sources
+++ b/src/glsl/Makefile.sources
@@ -80,6 +80,7 @@ LIBGLSL_FILES = \
$(GLSL_SRCDIR)/opt_dead_code.cpp \
$(GLSL_SRCDIR)/opt_dead_code_local.cpp \
$(GLSL_SRCDIR)/opt_dead_functions.cpp \
+ $(GLSL_SRCDIR)/opt_flatten_nested_if_blocks.cpp \
$(GLSL_SRCDIR)/opt_function_inlining.cpp \
$(GLSL_SRCDIR)/opt_if_simplification.cpp \
$(GLSL_SRCDIR)/opt_noop_swizzle.cpp \
diff --git a/src/glsl/glsl_parser_extras.cpp b/src/glsl/glsl_parser_extras.cpp
index 9740903..0992294 100644
--- a/src/glsl/glsl_parser_extras.cpp
+++ b/src/glsl/glsl_parser_extras.cpp
@@ -1218,6 +1218,7 @@ do_common_optimization(exec_list *ir, bool linked,
progress = do_structure_splitting(ir) || progress;
}
progress = do_if_simplification(ir) || progress;
+ progress = opt_flatten_nested_if_blocks(ir) || progress;
progress = do_copy_propagation(ir) || progress;
progress = do_copy_propagation_elements(ir) || progress;
if (linked)
diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index 2454bbe..a8885d7 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -82,6 +82,7 @@ bool do_function_inlining(exec_list *instructions);
bool do_lower_jumps(exec_list *instructions, bool pull_out_jumps = true, bool lower_sub_return = true, bool lower_main_return = false, bool lower_continue = false, bool lower_break = false);
bool do_lower_texture_projection(exec_list *instructions);
bool do_if_simplification(exec_list *instructions);
+bool opt_flatten_nested_if_blocks(exec_list *instructions);
bool do_discard_simplification(exec_list *instructions);
bool lower_if_to_cond_assign(exec_list *instructions, unsigned max_depth = 0);
bool do_mat_op_to_vec(exec_list *instructions);
diff --git a/src/glsl/opt_flatten_nested_if_blocks.cpp b/src/glsl/opt_flatten_nested_if_blocks.cpp
new file mode 100644
index 0000000..c702102
--- a/dev/null
+++ b/src/glsl/opt_flatten_nested_if_blocks.cpp
@@ -0,0 +1,103 @@
+/*
+ * Copyright © 2013 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 opt_flatten_nested_if_blocks.cpp
+ *
+ * Flattens nested if blocks such as:
+ *
+ * if (x) {
+ * if (y) {
+ * ...
+ * }
+ * }
+ *
+ * into a single if block with a combined condition:
+ *
+ * if (x && y) {
+ * ...
+ * }
+ */
+
+#include "ir.h"
+#include "ir_builder.h"
+
+using namespace ir_builder;
+
+namespace {
+
+class nested_if_flattener : public ir_hierarchical_visitor {
+public:
+ nested_if_flattener()
+ {
+ progress = false;
+ }
+
+ ir_visitor_status visit_leave(ir_if *);
+ ir_visitor_status visit_enter(ir_assignment *);
+
+ bool progress;
+};
+
+} /* unnamed namespace */
+
+/* We only care about the top level "if" instructions, so don't
+ * descend into expressions.
+ */
+ir_visitor_status
+nested_if_flattener::visit_enter(ir_assignment *ir)
+{
+ (void) ir;
+ return visit_continue_with_parent;
+}
+
+bool
+opt_flatten_nested_if_blocks(exec_list *instructions)
+{
+ nested_if_flattener v;
+
+ v.run(instructions);
+ return v.progress;
+}
+
+
+ir_visitor_status
+nested_if_flattener::visit_leave(ir_if *ir)
+{
+ /* Only handle a single ir_if within the then clause of an ir_if. No extra
+ * instructions, no else clauses, nothing.
+ */
+ if (ir->then_instructions.is_empty() || !ir->else_instructions.is_empty())
+ return visit_continue;
+
+ ir_if *inner = ((ir_instruction *) ir->then_instructions.head)->as_if();
+ if (!inner || !inner->next->is_tail_sentinel() ||
+ !inner->else_instructions.is_empty())
+ return visit_continue;
+
+ ir->condition = logic_and(ir->condition, inner->condition);
+ inner->then_instructions.move_nodes_to(&ir->then_instructions);
+
+ progress = true;
+ return visit_continue;
+}