pan/{mdg, bi}: Lower load_sample_pos

Conceptually same lowering as the DDK, although we're missing a number
of relevant compiler optimizations so the generated code is awful.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8774>

src/panfrost/Makefile.sources

@@ -103,6 +103,7 @@ util_FILES := \
         util/pan_liveness.c \
         util/pan_lower_framebuffer.c \
         util/pan_lower_helper_invocation.c \
+        util/pan_lower_sample_position.c \
         util/pan_lower_writeout.c \
         util/pan_lower_64bit_intrin.c \
         util/pan_sysval.c \

src/panfrost/bifrost/bifrost_compile.c

@@ -2547,6 +2547,7 @@ bifrost_compile_shader_nir(void *mem_ctx, nir_shader *nir,
                         glsl_type_size, 0);
         NIR_PASS_V(nir, nir_lower_ssbo);
         NIR_PASS_V(nir, pan_nir_lower_zs_store);
+        NIR_PASS_V(nir, pan_lower_sample_pos);
         NIR_PASS_V(nir, bifrost_nir_lower_i8_fragout);
         // TODO: re-enable when fp16 is flipped on
         // NIR_PASS_V(nir, nir_lower_mediump_outputs);

src/panfrost/midgard/midgard_compile.c

@@ -319,6 +319,7 @@ optimise_nir(nir_shader *nir, unsigned quirks, bool is_blend)
         NIR_PASS(progress, nir, midgard_nir_lower_image_bitsize);
         NIR_PASS(progress, nir, midgard_nir_lower_helper_writes);
         NIR_PASS(progress, nir, pan_lower_helper_invocation);
+        NIR_PASS(progress, nir, pan_lower_sample_pos);
 
         NIR_PASS(progress, nir, midgard_nir_lower_algebraic_early);
 
@@ -1964,6 +1965,10 @@ emit_intrinsic(compiler_context *ctx, nir_intrinsic_instr *instr)
                 emit_sysval_read(ctx, &instr->instr, 1, 0);
                 break;
 
+        case nir_intrinsic_load_sample_positions_pan:
+                emit_sysval_read(ctx, &instr->instr, 2, 0);
+                break;
+
         case nir_intrinsic_get_ssbo_size:
                 emit_sysval_read(ctx, &instr->instr, 1, 8);
                 break;

src/panfrost/util/meson.build

@@ -28,6 +28,7 @@ libpanfrost_util_files = files(
   'pan_liveness.c',
   'pan_lower_framebuffer.c',
   'pan_lower_helper_invocation.c',
+  'pan_lower_sample_position.c',
   'pan_lower_writeout.c',
   'pan_lower_64bit_intrin.c',
   'pan_sysval.c',

src/panfrost/util/pan_ir.h

@@ -284,5 +284,6 @@ bool pan_nir_lower_zs_store(nir_shader *nir);
 bool pan_nir_lower_64bit_intrin(nir_shader *shader);
 
 bool pan_lower_helper_invocation(nir_shader *shader);
+bool pan_lower_sample_pos(nir_shader *shader);
 
 #endif

src/panfrost/util/pan_lower_sample_position.c

@@ -0,0 +1,75 @@
+/*
+ * Copyright (C) 2021 Collabora, Ltd.
+ *
+ * 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.
+ */
+
+#include "pan_ir.h"
+#include "compiler/nir/nir_builder.h"
+
+/* Sample positions are supplied in a packed 8:8 fixed-point vec2 format in GPU
+ * memory indexed by the sample. We lower in NIR to take advantage of possible
+ * ALU optimizations at the end. This is convenient for Bifrost, since the
+ * sample positions are passed in this format and it saves the driver from any
+ * system value handling. For Midgard, it's a bit suboptimal (fp16 positions
+ * could be supplied directly), but this lets us unify the implementation, and
+ * it's a pretty trivial difference */
+
+static bool
+pan_lower_sample_pos_impl(struct nir_builder *b,
+                nir_instr *instr, UNUSED void *data)
+{
+        if (instr->type != nir_instr_type_intrinsic)
+                return false;
+
+        nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
+        if (intr->intrinsic != nir_intrinsic_load_sample_pos)
+                return false;
+
+        b->cursor = nir_before_instr(instr);
+
+        /* Elements are 4 bytes */
+        nir_ssa_def *addr = nir_iadd(b,
+                        nir_load_sample_positions_pan(b),
+                        nir_u2u64(b, nir_imul_imm(b, nir_load_sample_id(b), 4)));
+
+        /* Decode 8:8 fixed-point */
+        nir_ssa_def *raw = nir_load_global(b, addr, 2, 2, 16);
+        nir_ssa_def *decoded = nir_fmul_imm(b, nir_i2f16(b, raw), 1.0 / 256.0);
+
+        /* Make NIR validator happy */
+        if (decoded->bit_size != nir_dest_bit_size(intr->dest))
+                decoded = nir_f2fN(b, decoded, nir_dest_bit_size(intr->dest));
+
+        nir_ssa_def_rewrite_uses(&intr->dest.ssa, nir_src_for_ssa(decoded));
+        return true;
+}
+
+bool
+pan_lower_sample_pos(nir_shader *shader)
+{
+        if (shader->info.stage != MESA_SHADER_FRAGMENT)
+                return false;
+
+        return nir_shader_instructions_pass(shader,
+                        pan_lower_sample_pos_impl,
+                        nir_metadata_block_index | nir_metadata_dominance,
+                        NULL);
+}