[Mesa-dev] [PATCH 072/133] nir: Add a basic CSE pass
Jason Ekstrand
jason at jlekstrand.net
Mon Dec 15 22:05:22 PST 2014
This pass is still fairly basic. It only handles ALU operations, constant
loads, and phi nodes. No texture ops or intrinsics yet.
---
src/glsl/Makefile.sources | 1 +
src/glsl/nir/nir.h | 2 +
src/glsl/nir/nir_opt_cse.c | 269 +++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 272 insertions(+)
create mode 100644 src/glsl/nir/nir_opt_cse.c
diff --git a/src/glsl/Makefile.sources b/src/glsl/Makefile.sources
index 31608dc..20496ff 100644
--- a/src/glsl/Makefile.sources
+++ b/src/glsl/Makefile.sources
@@ -30,6 +30,7 @@ NIR_FILES = \
$(GLSL_SRCDIR)/nir/nir_opcodes.c \
$(GLSL_SRCDIR)/nir/nir_opcodes.h \
$(GLSL_SRCDIR)/nir/nir_opt_copy_propagate.c \
+ $(GLSL_SRCDIR)/nir/nir_opt_cse.c \
$(GLSL_SRCDIR)/nir/nir_opt_dce.c \
$(GLSL_SRCDIR)/nir/nir_opt_global_to_local.c \
$(GLSL_SRCDIR)/nir/nir_opt_peephole_ffma.c \
diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
index e20eb7c..f09bdd4 100644
--- a/src/glsl/nir/nir.h
+++ b/src/glsl/nir/nir.h
@@ -1370,6 +1370,8 @@ bool nir_opt_global_to_local(nir_shader *shader);
bool nir_copy_prop_impl(nir_function_impl *impl);
bool nir_copy_prop(nir_shader *shader);
+bool nir_opt_cse(nir_shader *shader);
+
bool nir_opt_dce_impl(nir_function_impl *impl);
bool nir_opt_dce(nir_shader *shader);
diff --git a/src/glsl/nir/nir_opt_cse.c b/src/glsl/nir/nir_opt_cse.c
new file mode 100644
index 0000000..a27d412
--- /dev/null
+++ b/src/glsl/nir/nir_opt_cse.c
@@ -0,0 +1,269 @@
+/*
+ * Copyright © 2014 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.
+ *
+ * Authors:
+ * Jason Ekstrand (jason at jlekstrand.net)
+ *
+ */
+
+#include "nir.h"
+
+/*
+ * Implements common subexpression elimination
+ */
+
+struct cse_state {
+ void *mem_ctx;
+ bool progress;
+};
+
+static bool
+nir_alu_srcs_equal(nir_alu_src src1, nir_alu_src src2, uint8_t read_mask)
+{
+ if (src1.abs != src2.abs || src1.negate != src2.negate)
+ return false;
+
+ for (int i = 0; i < 4; ++i) {
+ if (!(read_mask & (1 << i)))
+ continue;
+
+ if (src1.swizzle[i] != src2.swizzle[i])
+ return false;
+ }
+
+ return nir_srcs_equal(src1.src, src2.src);
+}
+
+static bool
+nir_instrs_equal(nir_instr *instr1, nir_instr *instr2)
+{
+ if (instr1->type != instr2->type)
+ return false;
+
+ switch (instr1->type) {
+ case nir_instr_type_alu: {
+ nir_alu_instr *alu1 = nir_instr_as_alu(instr1);
+ nir_alu_instr *alu2 = nir_instr_as_alu(instr2);
+
+ if (alu1->op != alu2->op)
+ return false;
+
+ if (alu1->has_predicate != alu2->has_predicate)
+ return false;
+
+ if (alu1->has_predicate &&
+ !nir_srcs_equal(alu1->predicate, alu2->predicate))
+ return false;
+
+ /* TODO: We can probably acutally do something more inteligent such
+ * as allowing different numbers and taking a maximum or something
+ * here */
+ if (alu1->dest.dest.ssa.num_components != alu2->dest.dest.ssa.num_components)
+ return false;
+
+ for (unsigned i = 0; i < nir_op_infos[alu1->op].num_inputs; i++) {
+ if (!nir_alu_srcs_equal(alu1->src[i], alu2->src[i],
+ (1 << alu1->dest.dest.ssa.num_components) - 1))
+ return false;
+ }
+ return true;
+ }
+ case nir_instr_type_texture:
+ return false;
+ case nir_instr_type_load_const: {
+ nir_load_const_instr *load1 = nir_instr_as_load_const(instr1);
+ nir_load_const_instr *load2 = nir_instr_as_load_const(instr2);
+
+ if (load1->num_components != load2->num_components)
+ return false;
+
+ return memcmp(load1->value.f, load2->value.f,
+ load1->num_components * sizeof load2->value.f) == 0;
+ }
+ case nir_instr_type_phi: {
+ nir_phi_instr *phi1 = nir_instr_as_phi(instr1);
+ nir_phi_instr *phi2 = nir_instr_as_phi(instr2);
+
+ if (phi1->instr.block != phi2->instr.block)
+ return false;
+
+ foreach_list_typed(nir_phi_src, src1, node, &phi1->srcs) {
+ foreach_list_typed(nir_phi_src, src2, node, &phi2->srcs) {
+ if (src1->pred == src2->pred) {
+ if (!nir_srcs_equal(src1->src, src2->src))
+ return false;
+
+ break;
+ }
+ }
+ }
+
+ return true;
+ }
+ case nir_instr_type_intrinsic:
+ case nir_instr_type_call:
+ case nir_instr_type_jump:
+ case nir_instr_type_ssa_undef:
+ case nir_instr_type_parallel_copy:
+ default:
+ unreachable("Invalid instruction type");
+ }
+
+ return false;
+}
+
+static bool
+src_is_ssa(nir_src *src, void *data)
+{
+ return src->is_ssa;
+}
+
+static bool
+dest_is_ssa(nir_dest *dest, void *data)
+{
+ return dest->is_ssa;
+}
+
+static bool
+nir_instr_can_cse(nir_instr *instr)
+{
+ switch (instr->type) {
+ case nir_instr_type_alu:
+ case nir_instr_type_load_const:
+ case nir_instr_type_phi:
+ return nir_foreach_dest(instr, dest_is_ssa, NULL) &&
+ nir_foreach_src(instr, src_is_ssa, NULL);
+ case nir_instr_type_texture:
+ return false; /* TODO */
+ case nir_instr_type_intrinsic:
+ case nir_instr_type_call:
+ case nir_instr_type_jump:
+ case nir_instr_type_ssa_undef:
+ return false;
+ case nir_instr_type_parallel_copy:
+ default:
+ unreachable("Invalid instruction type");
+ }
+
+ return false;
+}
+
+static nir_ssa_def *
+nir_instr_get_dest_ssa_def(nir_instr *instr)
+{
+ switch (instr->type) {
+ case nir_instr_type_alu:
+ assert(nir_instr_as_alu(instr)->dest.dest.is_ssa);
+ return &nir_instr_as_alu(instr)->dest.dest.ssa;
+ case nir_instr_type_load_const:
+ assert(nir_instr_as_load_const(instr)->dest.is_ssa);
+ return &nir_instr_as_load_const(instr)->dest.ssa;
+ case nir_instr_type_phi:
+ assert(nir_instr_as_phi(instr)->dest.is_ssa);
+ return &nir_instr_as_phi(instr)->dest.ssa;
+ default:
+ unreachable("We never ask for any of these");
+ }
+}
+
+static void
+nir_opt_cse_instr(nir_instr *instr, struct cse_state *state)
+{
+ if (!nir_instr_can_cse(instr))
+ return;
+
+ for (struct exec_node *node = instr->node.prev;
+ !exec_node_is_head_sentinel(node); node = node->prev) {
+ nir_instr *other = exec_node_data(nir_instr, node, node);
+ if (nir_instrs_equal(instr, other)) {
+ nir_src other_dest_src = {
+ .is_ssa = true,
+ .ssa = nir_instr_get_dest_ssa_def(other),
+ };
+ nir_ssa_def_rewrite_uses(nir_instr_get_dest_ssa_def(instr),
+ other_dest_src, state->mem_ctx);
+ nir_instr_remove(instr);
+ state->progress = true;
+ return;
+ }
+ }
+
+ for (nir_block *block = instr->block->imm_dom;
+ block != NULL; block = block->imm_dom) {
+ nir_foreach_instr_reverse(block, other) {
+ if (nir_instrs_equal(instr, other)) {
+ nir_src other_dest_src = {
+ .is_ssa = true,
+ .ssa = nir_instr_get_dest_ssa_def(other),
+ };
+ nir_ssa_def_rewrite_uses(nir_instr_get_dest_ssa_def(instr),
+ other_dest_src, state->mem_ctx);
+ nir_instr_remove(instr);
+ state->progress = true;
+ return;
+ }
+ }
+ }
+}
+
+static bool
+nir_opt_cse_block(nir_block *block, void *void_state)
+{
+ struct cse_state *state = void_state;
+
+ nir_foreach_instr_safe(block, instr)
+ nir_opt_cse_instr(instr, state);
+
+ return true;
+}
+
+static bool
+nir_opt_cse_impl(nir_function_impl *impl)
+{
+ struct cse_state state;
+
+ state.mem_ctx = ralloc_parent(impl);
+ state.progress = false;
+
+ nir_metadata_require(impl, nir_metadata_dominance);
+
+ nir_foreach_block(impl, nir_opt_cse_block, &state);
+
+ if (state.progress)
+ nir_metadata_dirty(impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+
+ return state.progress;
+}
+
+bool
+nir_opt_cse(nir_shader *shader)
+{
+ bool progress = false;
+
+ nir_foreach_overload(shader, overload) {
+ if (overload->impl)
+ progress |= nir_opt_cse_impl(overload->impl);
+ }
+
+ return progress;
+}
--
2.2.0
More information about the mesa-dev
mailing list