From d299c2eb0bdaf96f3c43d351688933c67cadd0cb Mon Sep 17 00:00:00 2001
From: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Date: Fri, 13 Nov 2020 17:34:25 -0500
Subject: [PATCH] pan/bi: Implement shader-db stats

v2: Drop register tracking since it was wrong, and meaningful accounting
is tricky for Bifrost (which wants round robin RA for at least some
registers)... we'll cross that bridge when we get there, possibly
preferring a "max liveness" estimate to the raw RA output.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7615>
---
 src/panfrost/bifrost/bifrost.h         |  1 +
 src/panfrost/bifrost/bifrost_compile.c | 57 ++++++++++++++++++++++++++
 2 files changed, 58 insertions(+)

diff --git a/src/panfrost/bifrost/bifrost.h b/src/panfrost/bifrost/bifrost.h
index 8ffd52b7fc9..76d370892d4 100644
--- a/src/panfrost/bifrost/bifrost.h
+++ b/src/panfrost/bifrost/bifrost.h
@@ -31,6 +31,7 @@
 
 #define BIFROST_DBG_MSGS        0x0001
 #define BIFROST_DBG_SHADERS     0x0002
+#define BIFROST_DBG_SHADERDB    0x0004
 
 extern int bifrost_debug;
 
diff --git a/src/panfrost/bifrost/bifrost_compile.c b/src/panfrost/bifrost/bifrost_compile.c
index ddcba5c9a7f..7b51543190f 100644
--- a/src/panfrost/bifrost/bifrost_compile.c
+++ b/src/panfrost/bifrost/bifrost_compile.c
@@ -40,11 +40,15 @@
 static const struct debug_named_value debug_options[] = {
         {"msgs",      BIFROST_DBG_MSGS,		"Print debug messages"},
         {"shaders",   BIFROST_DBG_SHADERS,	"Dump shaders in NIR and MIR"},
+        {"shaderdb",  BIFROST_DBG_SHADERDB,	"Print statistics"},
         DEBUG_NAMED_VALUE_END
 };
 
 DEBUG_GET_ONCE_FLAGS_OPTION(bifrost_debug, "BIFROST_MESA_DEBUG", debug_options, 0)
 
+/* TODO: This is not thread safe!! */
+static unsigned SHADER_DB_COUNT = 0;
+
 int bifrost_debug = 0;
 
 #define DBG(fmt, ...) \
@@ -2356,6 +2360,54 @@ emit_cf_list(bi_context *ctx, struct exec_list *list)
         return start_block;
 }
 
+/* shader-db stuff */
+
+static void
+bi_print_stats(bi_context *ctx, FILE *fp)
+{
+        unsigned nr_clauses = 0, nr_tuples = 0, nr_ins = 0;
+
+        /* Count instructions, clauses, and tuples */
+        bi_foreach_block(ctx, _block) {
+                bi_block *block = (bi_block *) _block;
+
+                bi_foreach_clause_in_block(block, clause) {
+                        nr_clauses++;
+                        nr_tuples += clause->bundle_count;
+
+                        for (unsigned i = 0; i < clause->bundle_count; ++i) {
+                                if (clause->bundles[i].fma)
+                                        nr_ins++;
+
+                                if (clause->bundles[i].add)
+                                        nr_ins++;
+                        }
+                }
+        }
+
+        /* tuples = ((# of instructions) + (# of nops)) / 2 */
+        unsigned nr_nops = (2 * nr_tuples) - nr_ins;
+
+        /* In the future, we'll calculate thread count for v7. For now we
+         * always use fewer threads than we should (v6 style) due to missing
+         * piping, TODO: fix that for a nice perf win */
+        unsigned nr_threads = 1;
+
+        /* Dump stats */
+
+        fprintf(stderr, "shader%d - %s shader: "
+                        "%u inst, %u nops, %u clauses, "
+                        "%u threads, %u loops, "
+                        "%u:%u spills:fills\n",
+                        SHADER_DB_COUNT++,
+                        ctx->is_blend ? "PAN_SHADER_BLEND" :
+                        gl_shader_stage_name(ctx->stage),
+                        nr_ins, nr_nops, nr_clauses,
+                        nr_threads,
+                        ctx->loop_count,
+                        ctx->spills, ctx->fills);
+}
+
 static int
 glsl_type_size(const struct glsl_type *type, bool bindless)
 {
@@ -2539,6 +2591,11 @@ bifrost_compile_shader_nir(void *mem_ctx, nir_shader *nir,
 
         program->tls_size = ctx->tls_size;
 
+        if ((bifrost_debug & BIFROST_DBG_SHADERDB || inputs->shaderdb) &&
+            !nir->info.internal) {
+                bi_print_stats(ctx, stderr);
+        }
+
         ralloc_free(ctx);
 
         return program;