[Mesa-dev] [PATCH 2/4] freedreno: a2xx: add partial lower_scalar pass for ir2
Jonathan Marek
jonathan at marek.ca
Mon Nov 19 21:49:59 UTC 2018
some instructions can only be scalar on a2xx, lower these only
Signed-off-by: Jonathan Marek <jonathan at marek.ca>
---
.../drivers/freedreno/Makefile.sources | 1 +
src/gallium/drivers/freedreno/a2xx/ir2_nir.c | 3 +
.../freedreno/a2xx/ir2_nir_lower_scalar.c | 174 ++++++++++++++++++
.../drivers/freedreno/a2xx/ir2_private.h | 1 +
src/gallium/drivers/freedreno/meson.build | 1 +
5 files changed, 180 insertions(+)
create mode 100644 src/gallium/drivers/freedreno/a2xx/ir2_nir_lower_scalar.c
diff --git a/src/gallium/drivers/freedreno/Makefile.sources b/src/gallium/drivers/freedreno/Makefile.sources
index 9061b26ba7..7f2b8e7b7d 100644
--- a/src/gallium/drivers/freedreno/Makefile.sources
+++ b/src/gallium/drivers/freedreno/Makefile.sources
@@ -87,6 +87,7 @@ a2xx_SOURCES := \
a2xx/instr-a2xx.h \
a2xx/ir2.c \
a2xx/ir2_nir.c \
+ a2xx/ir2_nir_lower_scalar.c \
a2xx/ir2_substitutions.c \
a2xx/ir2_ra.c \
a2xx/ir2_assemble.c \
diff --git a/src/gallium/drivers/freedreno/a2xx/ir2_nir.c b/src/gallium/drivers/freedreno/a2xx/ir2_nir.c
index f24debf140..12d4ee6653 100644
--- a/src/gallium/drivers/freedreno/a2xx/ir2_nir.c
+++ b/src/gallium/drivers/freedreno/a2xx/ir2_nir.c
@@ -1112,6 +1112,9 @@ ir2_nir_compile(struct ir2_context *ctx, unsigned variant)
/* postprocess */
OPT_V(ctx->nir, nir_opt_algebraic_late);
+ /* lower to scalar instructions that can only be scalar on a2xx */
+ OPT_V(ctx->nir, ir2_nir_lower_scalar);
+
OPT_V(ctx->nir, nir_lower_to_source_mods);
OPT_V(ctx->nir, nir_copy_prop);
OPT_V(ctx->nir, nir_opt_dce);
diff --git a/src/gallium/drivers/freedreno/a2xx/ir2_nir_lower_scalar.c b/src/gallium/drivers/freedreno/a2xx/ir2_nir_lower_scalar.c
new file mode 100644
index 0000000000..2b72a86b3e
--- /dev/null
+++ b/src/gallium/drivers/freedreno/a2xx/ir2_nir_lower_scalar.c
@@ -0,0 +1,174 @@
+/*
+ * Copyright (C) 2018 Jonathan Marek <jonathan at marek.ca>
+ *
+ * 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:
+ * Jonathan Marek <jonathan at marek.ca>
+ */
+
+/* some operations can only be scalar on a2xx:
+ * rsq, rcp, log2, exp2, cos, sin, sqrt
+ * mostly copy-pasted from nir_lower_alu_to_scalar.c
+ */
+
+#include "ir2_private.h"
+#include "compiler/nir/nir_builder.h"
+
+static void
+nir_alu_ssa_dest_init(nir_alu_instr * instr, unsigned num_components,
+ unsigned bit_size)
+{
+ nir_ssa_dest_init(&instr->instr, &instr->dest.dest, num_components,
+ bit_size, NULL);
+ instr->dest.write_mask = (1 << num_components) - 1;
+}
+
+static void
+lower_reduction(nir_alu_instr * instr, nir_op chan_op, nir_op merge_op,
+ nir_builder * builder)
+{
+ unsigned num_components = nir_op_infos[instr->op].input_sizes[0];
+
+ nir_ssa_def *last = NULL;
+ for (unsigned i = 0; i < num_components; i++) {
+ nir_alu_instr *chan =
+ nir_alu_instr_create(builder->shader, chan_op);
+ nir_alu_ssa_dest_init(chan, 1, instr->dest.dest.ssa.bit_size);
+ nir_alu_src_copy(&chan->src[0], &instr->src[0], chan);
+ chan->src[0].swizzle[0] = chan->src[0].swizzle[i];
+ if (nir_op_infos[chan_op].num_inputs > 1) {
+ assert(nir_op_infos[chan_op].num_inputs == 2);
+ nir_alu_src_copy(&chan->src[1], &instr->src[1], chan);
+ chan->src[1].swizzle[0] = chan->src[1].swizzle[i];
+ }
+ chan->exact = instr->exact;
+
+ nir_builder_instr_insert(builder, &chan->instr);
+
+ if (i == 0) {
+ last = &chan->dest.dest.ssa;
+ } else {
+ last = nir_build_alu(builder, merge_op,
+ last, &chan->dest.dest.ssa, NULL, NULL);
+ }
+ }
+
+ assert(instr->dest.write_mask == 1);
+ nir_ssa_def_rewrite_uses(&instr->dest.dest.ssa, nir_src_for_ssa(last));
+ nir_instr_remove(&instr->instr);
+}
+
+static bool lower_scalar(nir_alu_instr * instr, nir_builder * b)
+{
+ assert(instr->dest.dest.is_ssa);
+ assert(instr->dest.write_mask != 0);
+
+ b->cursor = nir_before_instr(&instr->instr);
+ b->exact = instr->exact;
+
+#define LOWER_REDUCTION(name, chan, merge) \
+ case name##2: \
+ case name##3: \
+ case name##4: \
+ lower_reduction(instr, chan, merge, b); \
+ return true;
+
+ switch (instr->op) {
+ /* TODO: handle these instead of lowering */
+ LOWER_REDUCTION(nir_op_fall_equal, nir_op_seq, nir_op_fand);
+ LOWER_REDUCTION(nir_op_fany_nequal, nir_op_sne, nir_op_for);
+
+ default:
+ return false;
+ case nir_op_frsq:
+ case nir_op_frcp:
+ case nir_op_flog2:
+ case nir_op_fexp2:
+ case nir_op_fcos:
+ case nir_op_fsin:
+ case nir_op_fsqrt:
+ break;
+ }
+
+ assert(nir_op_infos[instr->op].num_inputs == 1);
+
+ unsigned num_components = instr->dest.dest.ssa.num_components;
+ nir_ssa_def *comps[NIR_MAX_VEC_COMPONENTS] = { NULL };
+ unsigned chan;
+
+ if (num_components == 1)
+ return false;
+
+ for (chan = 0; chan < num_components; chan++) {
+ assert(instr->dest.write_mask & (1 << chan));
+
+ nir_alu_instr *lower = nir_alu_instr_create(b->shader, instr->op);
+
+ nir_alu_src_copy(&lower->src[0], &instr->src[0], lower);
+ lower->src[0].swizzle[0] = instr->src[0].swizzle[chan];
+
+ nir_alu_ssa_dest_init(lower, 1, instr->dest.dest.ssa.bit_size);
+ lower->dest.saturate = instr->dest.saturate;
+ comps[chan] = &lower->dest.dest.ssa;
+ lower->exact = instr->exact;
+
+ nir_builder_instr_insert(b, &lower->instr);
+ }
+
+ nir_ssa_def *vec = nir_vec(b, comps, num_components);
+
+ nir_ssa_def_rewrite_uses(&instr->dest.dest.ssa, nir_src_for_ssa(vec));
+
+ nir_instr_remove(&instr->instr);
+ return true;
+}
+
+static bool lower_scalar_impl(nir_function_impl * impl)
+{
+ nir_builder builder;
+ nir_builder_init(&builder, impl);
+ bool progress = false;
+
+ nir_foreach_block(block, impl) {
+ nir_foreach_instr_safe(instr, block) {
+ if (instr->type == nir_instr_type_alu)
+ progress = lower_scalar(nir_instr_as_alu(instr), &builder)
+ || progress;
+ }
+ }
+
+ nir_metadata_preserve(impl, nir_metadata_block_index |
+ nir_metadata_dominance);
+
+ return progress;
+}
+
+bool ir2_nir_lower_scalar(nir_shader * shader)
+{
+ bool progress = false;
+
+ nir_foreach_function(function, shader) {
+ if (function->impl)
+ progress = lower_scalar_impl(function->impl) || progress;
+ }
+
+ return progress;
+}
diff --git a/src/gallium/drivers/freedreno/a2xx/ir2_private.h b/src/gallium/drivers/freedreno/a2xx/ir2_private.h
index f45876be04..44bc9f8c62 100644
--- a/src/gallium/drivers/freedreno/a2xx/ir2_private.h
+++ b/src/gallium/drivers/freedreno/a2xx/ir2_private.h
@@ -190,6 +190,7 @@ void late_substitutions(struct ir2_context *ctx);
void assemble(struct ir2_context *ctx);
void ir2_nir_compile(struct ir2_context *ctx, unsigned variant);
+bool ir2_nir_lower_scalar(nir_shader * shader);
void ra_count_refs(struct ir2_context *ctx);
void ra_reg(struct ir2_context *ctx, struct ir2_reg *reg, int force_idx,
diff --git a/src/gallium/drivers/freedreno/meson.build b/src/gallium/drivers/freedreno/meson.build
index 064bea214e..b61e503c91 100644
--- a/src/gallium/drivers/freedreno/meson.build
+++ b/src/gallium/drivers/freedreno/meson.build
@@ -115,6 +115,7 @@ files_libfreedreno = files(
'a2xx/instr-a2xx.h',
'a2xx/ir2.c',
'a2xx/ir2_nir.c',
+ 'a2xx/ir2_nir_lower_scalar.c',
'a2xx/ir2_substitutions.c',
'a2xx/ir2_ra.c',
'a2xx/ir2_assemble.c',
--
2.17.1
More information about the mesa-dev
mailing list