Add a -printsimstats switch.

Start working in a "extra" mode for debugging and other features
that we don't want on the main path. As a demonstration of what we
can do with this extra mode, print out a bunch of simulation info
that can help us optimize offline.

cen64.c

@@ -17,7 +17,7 @@ static int load_roms(const char *pifrom_path, const char *cart_path,
   struct rom_file *pifrom, struct rom_file *cart);
 
 static int setup_and_run_device(struct cen64_device *device,
-  const char *pifrom_path, const char *cart_path);
+  const char *pifrom_path, const char *cart_path, bool extra_mode);
 
 // Called when a simulation instance is terminating.
 void cen64_cleanup(struct cen64_device *device) {
@@ -28,7 +28,8 @@ void cen64_cleanup(struct cen64_device *device) {
 // Called when another simulation instance is desired.
 int cen64_main(struct cen64_device *device, int argc, const char *argv[]) {
   struct gl_window_hints hints;
-  int status;
+  bool extra_mode = false;
+  int status, i;
 
   // Prevent debugging tools from raising warnings
   // about uninitialized memory being read, etc.
@@ -36,7 +37,11 @@ int cen64_main(struct cen64_device *device, int argc, const char *argv[]) {
   get_default_gl_window_hints(&hints);
 
   if (argc < 3) {
-    printf("%s <pifrom.bin> <rom>\n", argv[0]);
+    printf("%s <pifrom.bin> <rom>\n\n"
+      "Options:\n"
+      "  -printsimstats             : Print simulation statistics at exit.\n",
+      argv[0]);
+
     return 255;
   }
 
@@ -45,8 +50,13 @@ int cen64_main(struct cen64_device *device, int argc, const char *argv[]) {
     return 1;
   }
 
+  // Parse arguments.
+  for (i = 1; i < argc - 2; i++)
+    if (!strcmp(argv[i], "-printsimstats"))
+      extra_mode = true;
+
   // Start simulation.
-  if ((status = setup_and_run_device(device, argv[1], argv[2])))
+  if ((status = setup_and_run_device(device, argv[i], argv[i + 1], extra_mode)))
     destroy_gl_window(&device->vi.gl_window);
 
   return status;
@@ -79,7 +89,7 @@ int load_roms(const char *pifrom_path, const char *cart_path,
 
 // Create a device, Load ROM images, etc. and run.
 int setup_and_run_device(struct cen64_device *device,
-  const char *pifrom_path, const char *cart_path) {
+  const char *pifrom_path, const char *cart_path, bool extra_mode) {
   struct rom_file pifrom, cart;
   int status;
 
@@ -90,8 +100,11 @@ int setup_and_run_device(struct cen64_device *device,
     device->cart_size = cart.size;
 
     // Create a device and roceed to the main application loop.
-    if (!(status = device_create(device) == NULL))
-      status = device_run(device);
+    if (!(status = device_create(device) == NULL)) {
+      status = unlikely(extra_mode)
+        ? device_run_extra(device)
+        : device_run(device);
+    }
 
     close_rom_file(&cart);
     close_rom_file(&pifrom);

device.c

@@ -118,3 +118,28 @@ int device_run(struct cen64_device *device) {
   return 0;
 }
 
+// Kicks off threads and starts the device.
+int device_run_extra(struct cen64_device *device) {
+  unsigned i;
+
+  if (setjmp(device->bus.unwind_data)) {
+    vr4300_print_summary(&device->vr4300_stats);
+    return 0;
+  }
+
+  while (1) {
+    for (i = 0; i < 2; i++) {
+      vi_cycle(&device->vi);
+
+      rsp_cycle(&device->rsp);
+      vr4300_cycle(&device->vr4300);
+      vr4300_cycle_extra(&device->vr4300, &device->vr4300_stats);
+    }
+
+    vr4300_cycle(&device->vr4300);
+    vr4300_cycle_extra(&device->vr4300, &device->vr4300_stats);
+  }
+
+  return 0;
+}
+

device.h

@@ -45,11 +45,16 @@ struct cen64_device {
 
   const uint8_t *pifrom;
   const uint8_t *cart;
+
+  // Debugging/statistical data.
+  struct vr4300_stats vr4300_stats;
 };
 
 struct cen64_device *device_create(struct cen64_device *device);
 void device_destroy(struct cen64_device *device);
+
 int device_run(struct cen64_device *device);
+int device_run_extra(struct cen64_device *device);
 
 #endif
 

vr4300/cpu.c

@@ -48,3 +48,51 @@ int vr4300_init(struct vr4300 *vr4300, struct bus_controller *bus) {
   return 0;
 }
 
+// Prints out simulation information to stdout.
+void vr4300_print_summary(struct vr4300_stats *stats) {
+  unsigned i, j;
+  float secs;
+  float cpi;
+
+  // Print banner.
+  printf("###############################\n"
+         " NEC VR4300 Simulation Summary\n"
+         "###############################\n"
+         "\n"
+  );
+
+  // Print configuration, summary, whatever.
+  secs = stats->total_cycles / 93750000.0f;
+
+  printf("   %16s: %.1f sec.\n"
+         "\n\n",
+
+    "Actual runtime", secs
+  );
+
+  // Print performance statistics.
+  cpi = (float) stats->executed_instructions / stats->total_cycles;
+
+  printf(" * Performance statistics:\n\n"
+         "   %16s: %llu\n"
+         "   %16s: %llu\n"
+         "   %16s: %1.2f\n"
+         "\n\n",
+
+    "Elapsed pcycles", stats->total_cycles,
+    "Insns executed", stats->executed_instructions,
+    "Average CPI", cpi
+  );
+
+  // Print executed opcode counts.
+  printf(" * Executed instruction counts:\n\n");
+
+  for (i = 1; i < NUM_VR4300_OPCODES; i += 2) {
+    for (j = 0; i + j < NUM_VR4300_OPCODES && j < 2; j++)
+      printf("   %16s: %16llu\t", vr4300_opcode_mnemonics[i + j],
+        stats->opcode_counts[i + j]);
+
+    printf("\n");
+  }
+}
+

vr4300/cpu.h

@@ -15,6 +15,7 @@
 #include "vr4300/cp1.h"
 #include "vr4300/dcache.h"
 #include "vr4300/icache.h"
+#include "vr4300/opcodes.h"
 #include "vr4300/pipeline.h"
 
 struct bus_controller;
@@ -110,8 +111,18 @@ struct vr4300 {
 
 };
 
+struct vr4300_stats {
+  unsigned long long executed_instructions;
+  unsigned long long total_cycles;
+
+  unsigned long long opcode_counts[NUM_VR4300_OPCODES];
+};
+
 void vr4300_cycle(struct vr4300 *vr4300);
+void vr4300_cycle_extra(struct vr4300 *vr4300, struct vr4300_stats *stats);
 int vr4300_init(struct vr4300 *vr4300, struct bus_controller *bus);
 
+void vr4300_print_summary(struct vr4300_stats *stats);
+
 #endif
 

vr4300/pipeline.c

@@ -460,6 +460,22 @@ void vr4300_cycle(struct vr4300 *vr4300) {
     return;
 }
 
+// Collects additional information about the pipeline each cycle.
+void vr4300_cycle_extra(struct vr4300 *vr4300, struct vr4300_stats *stats) {
+  struct vr4300_dcwb_latch *dcwb_latch = &vr4300->pipeline.dcwb_latch;
+  struct vr4300_rfex_latch *rfex_latch = &vr4300->pipeline.rfex_latch;
+
+  // Collect information for CPI.
+  stats->executed_instructions +=
+    !dcwb_latch->common.fault &&
+    !vr4300->pipeline.cycles_to_stall;
+
+  stats->total_cycles++;
+
+  // Collect information about executed instructions.
+  stats->opcode_counts[rfex_latch->opcode.id]++;
+}
+
 // Initializes the pipeline with default values.
 void vr4300_pipeline_init(struct vr4300_pipeline *pipeline) {
   pipeline->icrf_latch.segment = get_default_segment();