[Mesa-dev] [PATCH 4/4] radeonsi: add radeonsi_debug_disassembly option

[Nicolai Hähnle]

From: Nicolai Hähnle <nicolai.haehnle at amd.com>

This dumps disassembly to the pipe_debug_callback together with shader
stats.

Can be used together with shader-db to get full disassembly of all shaders
in the database.
---
 src/gallium/drivers/radeonsi/si_debug_options.h |  1 +
 src/gallium/drivers/radeonsi/si_shader.c        | 15 +++++++++------
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/src/gallium/drivers/radeonsi/si_debug_options.h b/src/gallium/drivers/radeonsi/si_debug_options.h
index db642366ca6..ef8435804fb 100644
--- a/src/gallium/drivers/radeonsi/si_debug_options.h
+++ b/src/gallium/drivers/radeonsi/si_debug_options.h
@@ -1,8 +1,9 @@
 OPT_BOOL(clear_db_cache_before_clear, false, "Clear DB cache before fast depth clear")
 OPT_BOOL(enable_nir, false, "Enable NIR")
 OPT_BOOL(aux_debug, false, "Generate ddebug_dumps for the auxiliary context")
 OPT_BOOL(sync_compile, false, "Always compile synchronously (will cause stalls)")
 OPT_BOOL(dump_shader_binary, false, "Dump shader binary as part of ddebug_dumps")
+OPT_BOOL(debug_disassembly, false, "Report shader disassembly as part of driver debug messages (for shader db)")
 OPT_BOOL(vs_fetch_always_opencode, false, "Always open code vertex fetches (less efficient, purely for testing)")
 
 #undef OPT_BOOL
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 2186938fec9..4c321dc60dc 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -5154,27 +5154,22 @@ static void si_shader_dump_disassembly(struct si_screen *screen,
 			.elf_ptrs = &binary->elf_buffer,
 			.elf_sizes = &binary->elf_size }))
 		return;
 
 	const char *disasm;
 	uint64_t nbytes;
 
 	if (!ac_rtld_get_section_by_name(&rtld_binary, ".AMDGPU.disasm", &disasm, &nbytes))
 		goto out;
 
-	fprintf(file, "Shader %s disassembly:\n", name);
-	if (nbytes > INT_MAX) {
-		fprintf(file, "too long\n");
+	if (nbytes > INT_MAX)
 		goto out;
-	}
-
-	fprintf(file, "%*s", (int)nbytes, disasm);
 
 	if (debug && debug->debug_message) {
 		/* Very long debug messages are cut off, so send the
 		 * disassembly one line at a time. This causes more
 		 * overhead, but on the plus side it simplifies
 		 * parsing of resulting logs.
 		 */
 		pipe_debug_message(debug, SHADER_INFO,
 				   "Shader Disassembly Begin");
 
@@ -5190,20 +5185,25 @@ static void si_shader_dump_disassembly(struct si_screen *screen,
 						   "%.*s", count, disasm + line);
 			}
 
 			line += count + 1;
 		}
 
 		pipe_debug_message(debug, SHADER_INFO,
 				   "Shader Disassembly End");
 	}
 
+	if (file) {
+		fprintf(file, "Shader %s disassembly:\n", name);
+		fprintf(file, "%*s", (int)nbytes, disasm);
+	}
+
 out:
 	ac_rtld_close(&rtld_binary);
 }
 
 static void si_calculate_max_simd_waves(struct si_shader *shader)
 {
 	struct si_screen *sscreen = shader->selector->screen;
 	struct ac_shader_config *conf = &shader->config;
 	unsigned num_inputs = shader->selector->info.num_inputs;
 	unsigned lds_increment = sscreen->info.chip_class >= CIK ? 512 : 256;
@@ -5255,20 +5255,23 @@ static void si_calculate_max_simd_waves(struct si_shader *shader)
 
 	shader->max_simd_waves = max_simd_waves;
 }
 
 void si_shader_dump_stats_for_shader_db(struct si_screen *screen,
 					struct si_shader *shader,
 					struct pipe_debug_callback *debug)
 {
 	const struct ac_shader_config *conf = &shader->config;
 
+	if (screen->options.debug_disassembly)
+		si_shader_dump_disassembly(screen, &shader->binary, debug, "main", NULL);
+
 	pipe_debug_message(debug, SHADER_INFO,
 			   "Shader Stats: SGPRS: %d VGPRS: %d Code Size: %d "
 			   "LDS: %d Scratch: %d Max Waves: %d Spilled SGPRs: %d "
 			   "Spilled VGPRs: %d PrivMem VGPRs: %d",
 			   conf->num_sgprs, conf->num_vgprs,
 			   si_get_shader_binary_size(screen, shader),
 			   conf->lds_size, conf->scratch_bytes_per_wave,
 			   shader->max_simd_waves, conf->spilled_sgprs,
 			   conf->spilled_vgprs, shader->private_mem_vgprs);
 }
-- 
2.20.1