[Mesa-dev] [RFC PATCH 14/17] eir: add nir alu to scalar lowering pass

[Christian Gmeiner]

Some operations are scalar and need special treatment. Those
are exp and log2.

  dst := exp2(src2.x)
  dst := log2(src2.x)

The result is broadcasted over all active destination components.

Signed-off-by: Christian Gmeiner <christian.gmeiner at gmail.com>
---
 src/etnaviv/compiler/eir_nir.h                |   3 +
 .../compiler/eir_nir_lower_alu_to_scalar.c    | 131 ++++++++++++++++++
 src/etnaviv/compiler/meson.build              |   1 +
 3 files changed, 135 insertions(+)
 create mode 100644 src/etnaviv/compiler/eir_nir_lower_alu_to_scalar.c

diff --git a/src/etnaviv/compiler/eir_nir.h b/src/etnaviv/compiler/eir_nir.h
index 0ba5fa8e6a3..cd21cae38e4 100644
--- a/src/etnaviv/compiler/eir_nir.h
+++ b/src/etnaviv/compiler/eir_nir.h
@@ -37,4 +37,7 @@ eir_get_compiler_options(void);
 struct nir_shader *
 eir_optimize_nir(struct nir_shader *s);
 
+void
+eir_nir_lower_alu_to_scalar(struct nir_shader *shader);
+
 #endif // H_EIR_NIR
diff --git a/src/etnaviv/compiler/eir_nir_lower_alu_to_scalar.c b/src/etnaviv/compiler/eir_nir_lower_alu_to_scalar.c
new file mode 100644
index 00000000000..8dc264d083f
--- /dev/null
+++ b/src/etnaviv/compiler/eir_nir_lower_alu_to_scalar.c
@@ -0,0 +1,131 @@
+/*
+ * Copyright (c) 2019 Etnaviv Project
+ * Copyright (C) 2019 Zodiac Inflight Innovations
+ *
+ * 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, sub license,
+ * 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 NON-INFRINGEMENT. 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:
+ *    Christian Gmeiner <christian.gmeiner at gmail.com>
+ */
+
+#include "eir_nir.h"
+#include "nir.h"
+#include "nir_builder.h"
+
+/** @file eir_nir_lower_to_scalar.c
+ *
+ * Replaces nir alu operations with individual per-channel operations.
+ */
+
+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 nir_ssa_def *
+replace(nir_builder *b, nir_alu_instr *instr)
+{
+   unsigned num_src = nir_op_infos[instr->op].num_inputs;
+   unsigned num_components = instr->dest.dest.ssa.num_components;
+   nir_ssa_def *comps[NIR_MAX_VEC_COMPONENTS] = { NULL };
+   unsigned i, chan;
+
+   for (chan = 0; chan < NIR_MAX_VEC_COMPONENTS; chan++) {
+      if (!(instr->dest.write_mask & (1 << chan)))
+         continue;
+
+      nir_alu_instr *lower = nir_alu_instr_create(b->shader, instr->op);
+      for (i = 0; i < num_src; i++) {
+         /* We only handle same-size-as-dest (input_sizes[] == 0) or scalar
+          * args (input_sizes[] == 1).
+          */
+         assert(nir_op_infos[instr->op].input_sizes[i] < 2);
+         unsigned src_chan = (nir_op_infos[instr->op].input_sizes[i] == 1 ?
+                              0 : chan);
+
+         nir_alu_src_copy(&lower->src[i], &instr->src[i], lower);
+         for (int j = 0; j < NIR_MAX_VEC_COMPONENTS; j++)
+            lower->src[i].swizzle[j] = instr->src[i].swizzle[src_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);
+   }
+
+   return nir_vec(b, comps, num_components);
+}
+
+static void
+lower_alu_to_scalar_impl(nir_function_impl *impl)
+{
+   bool progress = false;
+
+   nir_builder b;
+   nir_builder_init(&b, impl);
+
+   nir_foreach_block(block, impl) {
+      nir_foreach_instr_safe(instr, block) {
+         if (instr->type != nir_instr_type_alu)
+            continue;
+
+         nir_alu_instr *alu_instr = nir_instr_as_alu(instr);
+         nir_ssa_def *vec;
+
+         b.cursor = nir_before_instr(instr);
+
+         switch (alu_instr->op) {
+         case nir_op_fexp2:
+            /* fall-through */
+         case nir_op_flog2:
+            vec = replace(&b, alu_instr);
+            break;
+         default:
+            continue;
+         }
+
+         nir_ssa_def_rewrite_uses(&alu_instr->dest.dest.ssa,
+                                  nir_src_for_ssa(vec));
+         nir_instr_remove(instr);
+         progress = true;
+      }
+   }
+
+   if (progress) {
+      nir_metadata_preserve(impl, nir_metadata_block_index |
+                                  nir_metadata_dominance);
+   }
+}
+
+void
+eir_nir_lower_alu_to_scalar(struct nir_shader *shader)
+{
+   nir_foreach_function(function, shader) {
+      if (function->impl)
+         lower_alu_to_scalar_impl(function->impl);
+   }
+}
diff --git a/src/etnaviv/compiler/meson.build b/src/etnaviv/compiler/meson.build
index f8af6c8b35d..3431cb48c36 100644
--- a/src/etnaviv/compiler/meson.build
+++ b/src/etnaviv/compiler/meson.build
@@ -28,6 +28,7 @@ libetnaviv_compiler_files = files(
   'eir_compiler_nir.c',
   'eir_legalize.c',
   'eir_live_variables.c',
+  'eir_nir_lower_alu_to_scalar.c',
   'eir_nir.c',
   'eir_nir.h',
   'eir_print.c',
-- 
2.21.0