Refactor benchmarking to self-contained C implementation

- Add mknes_bench.c with direct PMC access via perf_event_open()
- Remove dependency on external perf/awk for statistics
- Add RT priority and CPU affinity control in C code
- Use static BSS allocation (page-aligned) instead of malloc
- Add stalled-cycles-backend counter (gracefully handles AMD unavailability)
- Add throughput metrics (MIPS, Mcycles/sec)
- Optimize Bench.sh to only regenerate profile data when needed
- Add -n and -f flags for configurable runs and frames
- Suppress mapper messages during benchmark
- ~6x faster benchmark workflow (20s first run, 16s subsequent)

1 files changed, 26 insertions, 29 deletions

diff --git a/mknes.c b/mknes.c
index f8850b0..567d000 100644
--- a/mknes.c
+++ b/mknes.c
@@ -1,4 +1,5 @@
 #define GL_SILENCE_DEPRECATION
+#define _GNU_SOURCE
 
 #ifdef _WIN32
 #define NOMINMAX
@@ -57,8 +58,8 @@ static void audio_callback(int16_t *data, size_t frames) { }
 #ifdef BENCHMARK
 // Embed the ROM for benchmarking to eliminate file I/O overhead
 // Uncomment the ROM you want to benchmark:
-INCBIN_BYTES(benchmark_rom, "data/Life Force (USA).nes");
-// INCBIN_BYTES(benchmark_rom, "data/0000/Super Mario Bros. (World) (HVC-SM).nes");
+// INCBIN_BYTES(benchmark_rom, "data/Life Force (USA).nes");
+INCBIN_BYTES(benchmark_rom, "data/0000/Super Mario Bros. (World) (HVC-SM).nes");
 // INCBIN_BYTES(benchmark_rom, "data/0003/Gradius (USA).zip");
 #endif
 
@@ -177,6 +178,10 @@ static uint32_t frames;		// debug information
 #include "mknes_ines2.c"
 #include "mknes_mapper.c"
 
+#ifdef BENCHMARK
+#include "mknes_bench.c"
+#endif
+
 // struct nes_state nstate;
 
 static void framebuffer_callback(struct mkfw_state *mkfw_window, int32_t width, int32_t height, float aspect_ratio) {
@@ -240,14 +245,28 @@ int main(int argc, char **argv) {
 	// protect_opcode_lut();
 
 	struct nes_state *nstate = aligned_alloc(4096, (sizeof(struct nes_state) + 4095) & ~4095);
-	memset(nstate, 0, sizeof(struct nes_state));
 
-	ppu_reset(nstate);
 #ifdef BENCHMARK
-	// Use embedded ROM for consistent benchmarking without file I/O overhead
-	ines2_load_from_memory(nstate, benchmark_rom, INCBIN_SIZE(benchmark_rom));
+	// Run benchmark with configurable parameters
+	uint32_t num_runs = 10;
+	uint32_t frames_per_run = 0x1000;
+
+	// Parse command line arguments
+	for(int i = 1; i < argc; i++) {
+		if(strcmp(argv[i], "-n") == 0 && i + 1 < argc) {
+			num_runs = atoi(argv[i + 1]);
+			i++;
+		} else if(strcmp(argv[i], "-f") == 0 && i + 1 < argc) {
+			frames_per_run = atoi(argv[i + 1]);
+			i++;
+		}
+	}
 
+	run_benchmark(nstate, num_runs, frames_per_run);
+	return 0;
 #else
+	memset(nstate, 0, sizeof(struct nes_state));
+	ppu_reset(nstate);
 	// ines2_load(nstate, "data/0000/10-Yard Fight (USA, Europe).nes");
 	// ines2_load(nstate, "data/0000/Balloon Fight (USA).nes");
 	// ines2_load(nstate, "data/0000/Excitebike (Japan, USA).nes");
@@ -299,32 +318,10 @@ int main(int argc, char **argv) {
 
 	// ines2_load(nstate, "data/Blaster Master (USA).zip");	// mapper 1
 	// ines2_load(nstate, "AccuracyCoin.nes");	// mapper 1
-#endif
 
 	mapper_setup(nstate);
 	cpu_reset(nstate);
 
-#ifdef BENCHMARK
-	for(uint32_t i = 0; i < 0x1000; ++i) {
-		while(!nstate->ppu.frame_ready) {
-			// PROFILE_NAMED("nes emulator");
-			cpu_tick(nstate);
-		}
-		nstate->ppu.frame_ready = 0;
-		frames++;
-	}
-
-	// for(size_t i = 0; i < 9; ++i) {
-	// 	printf("count %d: %lld\n", i, sprite_counts[i]);
-	// }
-
-	// for(size_t i = 0; i < 256; ++i) {
-	// 	printf("instr %2.2x: %lld\n", i, instr_count[i]);
-	// }
-
-	return 0;
-#else
-
 //   WINDOW SETUP
 	struct mkfw_state *window = mkfw_init(WINDOW_WIDTH, WINDOW_HEIGHT);
 	mkfw_set_window_title(window, "mknes");
@@ -408,7 +405,7 @@ int main(int argc, char **argv) {
 	// free_nes_state(&nstate);
 	timer_shutdown();
 	mkfw_cleanup(window);
-#endif
 
 	return 0;
+#endif
 }