glsl2: Add a pass to strip out noop swizzles.

With the glsl2-965 branch, the optimization of glsl-algebraic-rcp-rcp
regressed due to noop swizzles hiding information from ir_algebraic.
This cleans up those noop swizzles for us.

diff --git a/src/glsl/Makefile b/src/glsl/Makefile
index 48b7c8f66b52d3f816696b783bc915688bb0eae5..110228e72a15124919bcff4646a9b27a23b6d0f1 100644 (file)
--- a/src/glsl/Makefile
+++ b/src/glsl/Makefile
@@ -55,6 +55,7 @@ CXX_SOURCES = \
        ir_import_prototypes.cpp \
        ir_mat_op_to_vec.cpp \
        ir_mod_to_fract.cpp \
+       ir_noop_swizzle.cpp \
        ir_print_visitor.cpp \
        ir_reader.cpp \
        ir_rvalue_visitor.cpp \

diff --git a/src/glsl/glsl_parser_extras.cpp b/src/glsl/glsl_parser_extras.cpp
index 2ed3905abc98e491cc725e64251213b1b45a5ad3..d1bb1ae5ecce29763bbb0b00966252557a62e240 100644 (file)
--- a/src/glsl/glsl_parser_extras.cpp
+++ b/src/glsl/glsl_parser_extras.cpp
@@ -737,6 +737,7 @@ do_common_optimization(exec_list *ir, bool linked)
    progress = do_if_return(ir) || progress;
    progress = do_vec_index_to_swizzle(ir) || progress;
    progress = do_swizzle_swizzle(ir) || progress;
+   progress = do_noop_swizzle(ir) || progress;
 
    return progress;
 }

diff --git a/src/glsl/ir_noop_swizzle.cpp b/src/glsl/ir_noop_swizzle.cpp
new file mode 100644 (file)
index 0000000..b78c87b
--- /dev/null
+++ b/src/glsl/ir_noop_swizzle.cpp
@@ -0,0 +1,80 @@
+/*
+ * Copyright © 2010 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 ir_noop_swizzle.cpp
+ *
+ * If a swizzle doesn't change the order or count of components, then
+ * remove the swizzle so that other optimization passes see the value
+ * behind it.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_rvalue_visitor.h"
+#include "ir_print_visitor.h"
+#include "glsl_types.h"
+
+class ir_noop_swizzle_visitor : public ir_rvalue_visitor {
+public:
+   ir_noop_swizzle_visitor()
+   {
+      this->progress = false;
+   }
+
+   void handle_rvalue(ir_rvalue **rvalue);
+   bool progress;
+};
+
+void
+ir_noop_swizzle_visitor::handle_rvalue(ir_rvalue **rvalue)
+{
+   if (!*rvalue)
+      return;
+
+   ir_swizzle *swiz = (*rvalue)->as_swizzle();
+   if (!swiz || swiz->type != swiz->val->type)
+      return;
+
+   int elems = swiz->val->type->vector_elements;
+   if (swiz->mask.x != 0)
+      return;
+   if (elems >= 2 && swiz->mask.y != 1)
+      return;
+   if (elems >= 3 && swiz->mask.z != 1)
+      return;
+   if (elems >= 4 && swiz->mask.w != 1)
+      return;
+
+   this->progress = true;
+   *rvalue = swiz->val;
+}
+
+bool
+do_noop_swizzle(exec_list *instructions)
+{
+   ir_noop_swizzle_visitor v;
+   visit_list_elements(&v, instructions);
+
+   return v.progress;
+}

diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index 0c4e548e44c47b6582a7b65dbe747fe2d1a472f2..33f4bc78f79d045977d9d3e8c9940b11476a2103 100644 (file)
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -48,6 +48,7 @@ bool do_if_simplification(exec_list *instructions);
 bool do_if_to_cond_assign(exec_list *instructions);
 bool do_mat_op_to_vec(exec_list *instructions);
 bool do_mod_to_fract(exec_list *instructions);
+bool do_noop_swizzle(exec_list *instructions);
 bool do_structure_splitting(exec_list *instructions);
 bool do_sub_to_add_neg(exec_list *instructions);
 bool do_swizzle_swizzle(exec_list *instructions);