summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorEric Anholt <[email protected]>2010-08-09 21:22:17 -0700
committerEric Anholt <[email protected]>2010-08-09 21:41:14 -0700
commit5854d4583c6e8885185e12a0636f77489a62e24c (patch)
tree38a132c8c1d5a1f09fc4733da1584fe5dbdd47a3
parent8bebbeb7c5b26ec9166a4644a2c051238d18509b (diff)
glsl2: Add a pass to transform ir_binop_sub to add(op0, neg(op1))
All the current HW backends transform subtract to adding the negation, so I haven't bothered peepholing it back out in Mesa IR. This allows some subtract of subtract to get removed in ir_algebraic.
-rw-r--r--src/glsl/Makefile1
-rw-r--r--src/glsl/ir_optimization.h1
-rw-r--r--src/glsl/ir_sub_to_add_neg.cpp76
-rw-r--r--src/glsl/linker.cpp1
-rw-r--r--src/mesa/program/ir_to_mesa.cpp1
5 files changed, 80 insertions, 0 deletions
diff --git a/src/glsl/Makefile b/src/glsl/Makefile
index 841e2b9ce9c..85298d06a01 100644
--- a/src/glsl/Makefile
+++ b/src/glsl/Makefile
@@ -59,6 +59,7 @@ CXX_SOURCES = \
ir_reader.cpp \
ir_set_program_inouts.cpp \
ir_structure_splitting.cpp \
+ ir_sub_to_add_neg.cpp \
ir_swizzle_swizzle.cpp \
ir_tree_grafting.cpp \
ir_validate.cpp \
diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
index 97a0c25216e..5997a30eab3 100644
--- a/src/glsl/ir_optimization.h
+++ b/src/glsl/ir_optimization.h
@@ -47,6 +47,7 @@ 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_structure_splitting(exec_list *instructions);
+bool do_sub_to_add_neg(exec_list *instructions);
bool do_swizzle_swizzle(exec_list *instructions);
bool do_tree_grafting(exec_list *instructions);
bool do_vec_index_to_cond_assign(exec_list *instructions);
diff --git a/src/glsl/ir_sub_to_add_neg.cpp b/src/glsl/ir_sub_to_add_neg.cpp
new file mode 100644
index 00000000000..7ed8c1495e3
--- /dev/null
+++ b/src/glsl/ir_sub_to_add_neg.cpp
@@ -0,0 +1,76 @@
+/*
+ * 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_sub_to_add_neg.cpp
+ *
+ * Breaks an ir_binop_sub expression down to add(op0, neg(op1))
+ *
+ * This simplifies expression reassociation, and for many backends
+ * there is no subtract operation separate from adding the negation.
+ * For backends with native subtract operations, they will probably
+ * want to recognize add(op0, neg(op1)) or the other way around to
+ * produce a subtract anyway.
+ */
+
+#include "ir.h"
+
+class ir_sub_to_add_neg_visitor : public ir_hierarchical_visitor {
+public:
+ ir_sub_to_add_neg_visitor()
+ {
+ this->progress = false;
+ }
+
+ ir_visitor_status visit_leave(ir_expression *);
+
+ bool progress;
+};
+
+bool
+do_sub_to_add_neg(exec_list *instructions)
+{
+ ir_sub_to_add_neg_visitor v;
+
+ visit_list_elements(&v, instructions);
+ return v.progress;
+}
+
+ir_visitor_status
+ir_sub_to_add_neg_visitor::visit_leave(ir_expression *ir)
+{
+ if (ir->operation != ir_binop_sub)
+ return visit_continue;
+
+ void *mem_ctx = talloc_parent(ir);
+
+ ir->operation = ir_binop_add;
+ ir->operands[1] = new(mem_ctx) ir_expression(ir_unop_neg,
+ ir->operands[1]->type,
+ ir->operands[1],
+ NULL);
+
+ this->progress = true;
+
+ return visit_continue;
+}
diff --git a/src/glsl/linker.cpp b/src/glsl/linker.cpp
index 52c93227888..c462d31ef3d 100644
--- a/src/glsl/linker.cpp
+++ b/src/glsl/linker.cpp
@@ -1284,6 +1284,7 @@ link_shaders(struct gl_shader_program *prog)
do_mod_to_fract(ir);
do_div_to_mul_rcp(ir);
do_explog_to_explog2(ir);
+ do_sub_to_add_neg(ir);
do {
progress = false;
diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp
index 5a272ab88a1..a9a6f977c01 100644
--- a/src/mesa/program/ir_to_mesa.cpp
+++ b/src/mesa/program/ir_to_mesa.cpp
@@ -2570,6 +2570,7 @@ _mesa_glsl_compile_shader(GLcontext *ctx, struct gl_shader *shader)
do_mat_op_to_vec(shader->ir);
do_mod_to_fract(shader->ir);
do_div_to_mul_rcp(shader->ir);
+ do_sub_to_add_neg(shader->ir);
/* Optimization passes */
bool progress;