gpu: start doing some basic gpu timing

minimum only for now, mostly based on Mednafen
libretro/pcsx_rearmed#573
libretro/pcsx_rearmed#783

diff --git a/plugins/dfxvideo/gpulib_if.c b/plugins/dfxvideo/gpulib_if.c
index d08ca67..20383ab 100644 (file)
--- a/plugins/dfxvideo/gpulib_if.c
+++ b/plugins/dfxvideo/gpulib_if.c
@@ -303,16 +303,19 @@ void renderer_notify_scanout_change(int x, int y)
 {
 }
 
+#include "../gpulib/gpu_timing.h"
 extern const unsigned char cmd_lengths[256];
 
-int do_cmd_list(uint32_t *list, int list_len, int *last_cmd)
+int do_cmd_list(uint32_t *list, int list_len, int *cpu_cycles_out, int *last_cmd)
 {
   unsigned int cmd = 0, len;
   uint32_t *list_start = list;
   uint32_t *list_end = list + list_len;
+  u32 cpu_cycles = 0;
 
   for (; list < list_end; list += 1 + len)
   {
+    short *slist = (void *)list;
     cmd = GETLE32(list) >> 24;
     len = cmd_lengths[cmd];
     if (list + 1 + len > list_end) {
@@ -338,6 +341,8 @@ int do_cmd_list(uint32_t *list, int list_len, int *last_cmd)
 
         while(1)
         {
+          cpu_cycles += gput_line(0);
+
           if(list_position >= list_end) {
             cmd = -1;
             goto breakloop;
@@ -361,6 +366,8 @@ int do_cmd_list(uint32_t *list, int list_len, int *last_cmd)
 
         while(1)
         {
+          cpu_cycles += gput_line(0);
+
           if(list_position >= list_end) {
             cmd = -1;
             goto breakloop;
@@ -380,7 +387,6 @@ int do_cmd_list(uint32_t *list, int list_len, int *last_cmd)
 #ifdef TEST
       case 0xA0:          //  sys -> vid
       {
-        short *slist = (void *)list;
         u32 load_width = LE2HOST32(slist[4]);
         u32 load_height = LE2HOST32(slist[5]);
         u32 load_size = load_width * load_height;
@@ -389,6 +395,35 @@ int do_cmd_list(uint32_t *list, int list_len, int *last_cmd)
         break;
       }
 #endif
+
+      // timing
+      case 0x02:
+        cpu_cycles += gput_fill(LE2HOST32(slist[4]) & 0x3ff,
+            LE2HOST32(slist[5]) & 0x1ff);
+        break;
+      case 0x20 ... 0x23: cpu_cycles += gput_poly_base();    break;
+      case 0x24 ... 0x27: cpu_cycles += gput_poly_base_t();  break;
+      case 0x28 ... 0x2B: cpu_cycles += gput_quad_base();    break;
+      case 0x2C ... 0x2F: cpu_cycles += gput_quad_base_t();  break;
+      case 0x30 ... 0x33: cpu_cycles += gput_poly_base_g();  break;
+      case 0x34 ... 0x37: cpu_cycles += gput_poly_base_gt(); break;
+      case 0x38 ... 0x3B: cpu_cycles += gput_quad_base_g();  break;
+      case 0x3C ... 0x3F: cpu_cycles += gput_quad_base_gt(); break;
+      case 0x40 ... 0x47: cpu_cycles += gput_line(0);        break;
+      case 0x50 ... 0x57: cpu_cycles += gput_line(0);        break;
+      case 0x60 ... 0x63:
+        cpu_cycles += gput_sprite(LE2HOST32(slist[4]) & 0x3ff,
+            LE2HOST32(slist[5]) & 0x1ff);
+        break;
+      case 0x64 ... 0x67:
+        cpu_cycles += gput_sprite(LE2HOST32(slist[6]) & 0x3ff,
+            LE2HOST32(slist[7]) & 0x1ff);
+        break;
+      case 0x68 ... 0x6B: cpu_cycles += gput_sprite(1, 1);   break;
+      case 0x70 ... 0x73:
+      case 0x74 ... 0x77: cpu_cycles += gput_sprite(8, 8);   break;
+      case 0x78 ... 0x7B:
+      case 0x7C ... 0x7F: cpu_cycles += gput_sprite(16, 16); break;
     }
   }
 
@@ -396,6 +431,7 @@ breakloop:
   gpu.ex_regs[1] &= ~0x1ff;
   gpu.ex_regs[1] |= lGPUstatusRet & 0x1ff;
 
+  *cpu_cycles_out += cpu_cycles;
   *last_cmd = cmd;
   return list - list_start;
 }
@@ -440,3 +476,5 @@ void renderer_set_config(const struct rearmed_cbs *cbs)
   cbs->pl_set_gpu_caps(0);
  set_vram(gpu.vram);
 }
+
+// vim:ts=2:shiftwidth=2:expandtab

diff --git a/plugins/gpu-gles/gpulib_if.c b/plugins/gpu-gles/gpulib_if.c
index a3a0c43..d440fdb 100644 (file)
--- a/plugins/gpu-gles/gpulib_if.c
+++ b/plugins/gpu-gles/gpulib_if.c
@@ -521,7 +521,7 @@ void renderer_notify_scanout_change(int x, int y)
 extern const unsigned char cmd_lengths[256];
 
 // XXX: mostly dupe code from soft peops
-int do_cmd_list(unsigned int *list, int list_len, int *last_cmd)
+int do_cmd_list(unsigned int *list, int list_len, int *cycles, int *last_cmd)
 {
   unsigned int cmd, len;
   unsigned int *list_start = list;

diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu.h b/plugins/gpu_neon/psx_gpu/psx_gpu.h
index da9e342..06514b9 100644 (file)
--- a/plugins/gpu_neon/psx_gpu/psx_gpu.h
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu.h
@@ -254,7 +254,8 @@ void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu);
 void flush_render_block_buffer(psx_gpu_struct *psx_gpu);
 
 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram);
-u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command);
+u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
+ s32 *cpu_cycles, u32 *last_command);
 
 void triangle_benchmark(psx_gpu_struct *psx_gpu);
 

diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_main.c b/plugins/gpu_neon/psx_gpu/psx_gpu_main.c
index c7ce0ee..435c51a 100644 (file)
--- a/plugins/gpu_neon/psx_gpu/psx_gpu_main.c
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_main.c
@@ -135,6 +135,8 @@ int main(int argc, char *argv[])
   FILE *state_file;
   FILE *list_file;
   u32 no_display = 0;
+  s32 dummy0 = 0;
+  u32 dummy1 = 0;
 
   if((argc != 3) && (argc != 4))
   {
@@ -213,7 +215,7 @@ int main(int argc, char *argv[])
   init_counter();
 #endif
 
-  gpu_parse(psx_gpu, list, size, NULL);
+  gpu_parse(psx_gpu, list, size, &dummy0, &dummy1);
   flush_render_block_buffer(psx_gpu);
 
   clear_stats();
@@ -222,7 +224,7 @@ int main(int argc, char *argv[])
   u32 cycles = get_counter();
 #endif
 
-  gpu_parse(psx_gpu, list, size, NULL);
+  gpu_parse(psx_gpu, list, size, &dummy0, &dummy1);
   flush_render_block_buffer(psx_gpu);
 
 #ifdef NEON_BUILD

diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c b/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c
index 5f69919..b0254af 100644 (file)
--- a/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c
@@ -15,6 +15,7 @@
 #include <stdio.h>
 
 #include "common.h"
+#include "../../gpulib/gpu_timing.h"
 
 #ifndef command_lengths
 const u8 command_lengths[256] =
@@ -250,30 +251,31 @@ static void do_fill(psx_gpu_struct *psx_gpu, u32 x, u32 y,
 #define SET_Ex(r, v)
 #endif
 
-u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
+u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
+ s32 *cpu_cycles_out, u32 *last_command)
 {
   vertex_struct vertexes[4] __attribute__((aligned(16))) = {};
-  u32 current_command = 0, command_length;
+  u32 current_command = 0, command_length, cpu_cycles = 0;
 
   u32 *list_start = list;
   u32 *list_end = list + (size / 4);
 
   for(; list < list_end; list += 1 + command_length)
   {
-       s16 *list_s16 = (void *)list;
-       current_command = *list >> 24;
-       command_length = command_lengths[current_command];
-       if (list + 1 + command_length > list_end) {
-         current_command = (u32)-1;
-         break;
-       }
-
-       switch(current_command)
-       {
-               case 0x00:
-                       break;
-  
-               case 0x02:
+    s16 *list_s16 = (void *)list;
+    current_command = *list >> 24;
+    command_length = command_lengths[current_command];
+    if (list + 1 + command_length > list_end) {
+      current_command = (u32)-1;
+      break;
+    }
+
+    switch(current_command)
+    {
+      case 0x00:
+        break;
+
+      case 0x02:
       {
         u32 x = list_s16[2] & 0x3FF;
         u32 y = list_s16[3] & 0x1FF;
@@ -282,10 +284,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         u32 color = list[0] & 0xFFFFFF;
 
         do_fill(psx_gpu, x, y, width, height, color);
-                       break;
+        cpu_cycles += gput_fill(width, height);
+        break;
       }
-  
-               case 0x20 ... 0x23:
+
+      case 0x20 ... 0x23:
       {
         set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
   
@@ -294,10 +297,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         get_vertex_data_xy(2, 6);
           
         render_triangle(psx_gpu, vertexes, current_command);
-                       break;
+        cpu_cycles += gput_poly_base();
+        break;
       }
   
-               case 0x24 ... 0x27:
+      case 0x24 ... 0x27:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[9]);
@@ -308,10 +312,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         get_vertex_data_xy_uv(2, 10);
   
         render_triangle(psx_gpu, vertexes, current_command);
-                       break;
+        cpu_cycles += gput_poly_base_t();
+        break;
       }
   
-               case 0x28 ... 0x2B:
+      case 0x28 ... 0x2B:
       {
         set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
   
@@ -322,10 +327,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
   
         render_triangle(psx_gpu, vertexes, current_command);
         render_triangle(psx_gpu, &(vertexes[1]), current_command);
-                       break;
+        cpu_cycles += gput_quad_base();
+        break;
       }
   
-               case 0x2C ... 0x2F:
+      case 0x2C ... 0x2F:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[9]);
@@ -338,23 +344,22 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
   
         render_triangle(psx_gpu, vertexes, current_command);
         render_triangle(psx_gpu, &(vertexes[1]), current_command);
-                       break;
+        cpu_cycles += gput_quad_base_t();
+        break;
       }
   
-               case 0x30 ... 0x33:
+      case 0x30 ... 0x33:
       {
         get_vertex_data_xy_rgb(0, 0);
         get_vertex_data_xy_rgb(1, 4);
         get_vertex_data_xy_rgb(2, 8);
   
         render_triangle(psx_gpu, vertexes, current_command);
-                       break;
+        cpu_cycles += gput_poly_base_g();
+        break;
       }
   
-               case 0x34:
-               case 0x35:
-               case 0x36:
-               case 0x37:
+      case 0x34 ... 0x37:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[11]);
@@ -364,13 +369,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         get_vertex_data_xy_uv_rgb(2, 12);
 
         render_triangle(psx_gpu, vertexes, current_command);
-                       break;
+        cpu_cycles += gput_poly_base_gt();
+        break;
       }
   
-               case 0x38:
-               case 0x39:
-               case 0x3A:
-               case 0x3B:
+      case 0x38 ... 0x3B:
       {
         get_vertex_data_xy_rgb(0, 0);
         get_vertex_data_xy_rgb(1, 4);
@@ -379,13 +382,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
   
         render_triangle(psx_gpu, vertexes, current_command);
         render_triangle(psx_gpu, &(vertexes[1]), current_command);
-                       break;
+        cpu_cycles += gput_quad_base_g();
+        break;
       }
   
-               case 0x3C:
-               case 0x3D:
-               case 0x3E:
-               case 0x3F:
+      case 0x3C ... 0x3F:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[11]);
@@ -397,10 +398,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
   
         render_triangle(psx_gpu, vertexes, current_command);
         render_triangle(psx_gpu, &(vertexes[1]), current_command);
-                       break;
+        cpu_cycles += gput_quad_base_gt();
+        break;
       }
   
-               case 0x40 ... 0x47:
+      case 0x40 ... 0x47:
       {
         vertexes[0].x = list_s16[2] + psx_gpu->offset_x;
         vertexes[0].y = list_s16[3] + psx_gpu->offset_y;
@@ -408,10 +410,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         vertexes[1].y = list_s16[5] + psx_gpu->offset_y;
 
         render_line(psx_gpu, vertexes, current_command, list[0], 0);
-                       break;
+        cpu_cycles += gput_line(0);
+        break;
       }
   
-               case 0x48 ... 0x4F:
+      case 0x48 ... 0x4F:
       {
         u32 num_vertexes = 1;
         u32 *list_position = &(list[2]);
@@ -429,6 +432,7 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
           vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
 
           render_line(psx_gpu, vertexes, current_command, list[0], 0);
+          cpu_cycles += gput_line(0);
 
           list_position++;
           num_vertexes++;
@@ -448,7 +452,7 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         break;
       }
   
-               case 0x50 ... 0x57:
+      case 0x50 ... 0x57:
       {
         vertexes[0].r = list[0] & 0xFF;
         vertexes[0].g = (list[0] >> 8) & 0xFF;
@@ -463,7 +467,8 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         vertexes[1].y = list_s16[7] + psx_gpu->offset_y;
 
         render_line(psx_gpu, vertexes, current_command, 0, 0);
-                       break;
+        cpu_cycles += gput_line(0);
+        break;
       }
  
       case 0x58 ... 0x5F:
@@ -493,6 +498,7 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
           vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
 
           render_line(psx_gpu, vertexes, current_command, 0, 0);
+          cpu_cycles += gput_line(0);
 
           list_position += 2;
           num_vertexes++;
@@ -512,7 +518,7 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         break;
       }
   
-               case 0x60 ... 0x63:
+      case 0x60 ... 0x63:
       {        
         u32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         u32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -520,10 +526,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         u32 height = list_s16[5] & 0x1FF;
 
         render_sprite(psx_gpu, x, y, 0, 0, width, height, current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(width, height);
+        break;
       }
   
-               case 0x64 ... 0x67:
+      case 0x64 ... 0x67:
       {        
         u32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         u32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -535,37 +542,31 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
 
         render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, width, height,
          current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(width, height);
+        break;
       }
   
-               case 0x68:
-               case 0x69:
-               case 0x6A:
-               case 0x6B:
+      case 0x68 ... 0x6B:
       {
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
 
         render_sprite(psx_gpu, x, y, 0, 0, 1, 1, current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(1, 1);
+        break;
       }
   
-               case 0x70:
-               case 0x71:
-               case 0x72:
-               case 0x73:
+      case 0x70 ... 0x73:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
 
         render_sprite(psx_gpu, x, y, 0, 0, 8, 8, current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(8, 8);
+        break;
       }
   
-               case 0x74:
-               case 0x75:
-               case 0x76:
-               case 0x77:
+      case 0x74 ... 0x77:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -575,25 +576,21 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
 
         render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, 8, 8,
          current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(8, 8);
+        break;
       }
   
-               case 0x78:
-               case 0x79:
-               case 0x7A:
-               case 0x7B:
+      case 0x78 ... 0x7B:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
 
         render_sprite(psx_gpu, x, y, 0, 0, 16, 16, current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(16, 16);
+        break;
       }
   
-               case 0x7C:
-               case 0x7D:
-               case 0x7E:
-               case 0x7F:
+      case 0x7C ... 0x7F:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -603,7 +600,8 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
 
         render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, 16, 16,
          current_command, list[0]);
-                       break;
+        cpu_cycles += gput_sprite(16, 16);
+        break;
       }
   
 #ifdef PCSX
@@ -643,14 +641,14 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
 
         render_block_copy(psx_gpu, (u16 *)&(list_s16[6]), load_x, load_y,
          load_width, load_height, load_width);
-                       break;
+        break;
       }
 
       case 0xC0 ... 0xDF:          //  vid -> sys
         break;
 #endif
 
-               case 0xE1:
+      case 0xE1:
         set_texture(psx_gpu, list[0]);
 
         if(list[0] & (1 << 9))
@@ -659,10 +657,10 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
           psx_gpu->render_state_base &= ~RENDER_STATE_DITHER;
 
         psx_gpu->display_area_draw_enable = (list[0] >> 10) & 0x1;
-                       SET_Ex(1, list[0]);
-                       break;
+        SET_Ex(1, list[0]);
+        break;
   
-               case 0xE2:
+      case 0xE2:
       {
         // TODO: Clean
         u32 texture_window_settings = list[0];
@@ -751,11 +749,11 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
         psx_gpu->offset_x = offset_x >> 21;
         psx_gpu->offset_y = offset_y >> 21; 
   
-                       SET_Ex(5, list[0]);
-                       break;
-               }
+        SET_Ex(5, list[0]);
+        break;
+      }
 
-               case 0xE6:
+      case 0xE6:
       {
         u32 mask_settings = list[0];
         u16 mask_msb = mask_settings << 15;
@@ -771,18 +769,18 @@ u32 gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size, u32 *last_command)
           psx_gpu->mask_msb = mask_msb;
         }
 
-                       SET_Ex(6, list[0]);
-                       break;
+        SET_Ex(6, list[0]);
+        break;
       }
   
-               default:
-                       break;
-       }
+      default:
+        break;
+    }
   }
 
 breakloop:
-  if (last_command != NULL)
-    *last_command = current_command;
+  *cpu_cycles_out += cpu_cycles;
+  *last_command = current_command;
   return list - list_start;
 }
 
@@ -1194,10 +1192,10 @@ static void do_sprite_enhanced(psx_gpu_struct *psx_gpu, int x, int y,
 #endif
 
 u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
- u32 *last_command)
+ s32 *cpu_cycles_out, u32 *last_command)
 {
   vertex_struct vertexes[4] __attribute__((aligned(16))) = {};
-  u32 current_command = 0, command_length;
+  u32 current_command = 0, command_length, cpu_cycles = 0;
 
   u32 *list_start = list;
   u32 *list_end = list + (size / 4);
@@ -1236,6 +1234,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         x &= ~0xF;
         width = ((width + 0xF) & ~0xF);
+        cpu_cycles += gput_fill(width, height);
         if (width == 0 || height == 0)
           break;
 
@@ -1266,6 +1265,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy(2, 6);
 
         do_triangle_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_poly_base();
         break;
       }
   
@@ -1280,6 +1280,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy_uv(2, 10);
   
         do_triangle_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_poly_base_t();
         break;
       }
   
@@ -1293,6 +1294,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy(3, 8);
 
         do_quad_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_quad_base();
         break;
       }
   
@@ -1309,6 +1311,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
   
         uv_hack(vertexes, 4);
         do_quad_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_quad_base_t();
         break;
       }
   
@@ -1319,13 +1322,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy_rgb(2, 8);
   
         do_triangle_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_poly_base_g();
         break;
       }
   
-      case 0x34:
-      case 0x35:
-      case 0x36:
-      case 0x37:
+      case 0x34 ... 0x37:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[11]);
@@ -1335,13 +1336,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy_uv_rgb(2, 12);
 
         do_triangle_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_poly_base_gt();
         break;
       }
   
-      case 0x38:
-      case 0x39:
-      case 0x3A:
-      case 0x3B:
+      case 0x38 ... 0x3B:
       {
         get_vertex_data_xy_rgb(0, 0);
         get_vertex_data_xy_rgb(1, 4);
@@ -1349,13 +1348,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         get_vertex_data_xy_rgb(3, 12);
   
         do_quad_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_quad_base_g();
         break;
       }
   
-      case 0x3C:
-      case 0x3D:
-      case 0x3E:
-      case 0x3F:
+      case 0x3C ... 0x3F:
       {
         set_clut(psx_gpu, list_s16[5]);
         set_texture(psx_gpu, list_s16[11]);
@@ -1367,6 +1364,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         uv_hack(vertexes, 4);
         do_quad_enhanced(psx_gpu, vertexes, current_command);
+        cpu_cycles += gput_quad_base_gt();
         break;
       }
   
@@ -1380,6 +1378,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         render_line(psx_gpu, vertexes, current_command, list[0], 0);
         if (enhancement_enable(psx_gpu))
           render_line(psx_gpu, vertexes, current_command, list[0], 1);
+        cpu_cycles += gput_line(0);
         break;
       }
   
@@ -1404,6 +1403,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
           render_line(psx_gpu, vertexes, current_command, list[0], 0);
           if (enhancement_enable(psx_gpu))
             render_line(psx_gpu, vertexes, current_command, list[0], 1);
+          cpu_cycles += gput_line(0);
 
           list_position++;
           num_vertexes++;
@@ -1440,6 +1440,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
         render_line(psx_gpu, vertexes, current_command, 0, 0);
         if (enhancement_enable(psx_gpu))
           render_line(psx_gpu, vertexes, current_command, 0, 1);
+        cpu_cycles += gput_line(0);
         break;
       }
  
@@ -1473,6 +1474,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
           render_line(psx_gpu, vertexes, current_command, 0, 0);
           if (enhancement_enable(psx_gpu))
             render_line(psx_gpu, vertexes, current_command, 0, 1);
+          cpu_cycles += gput_line(0);
 
           list_position += 2;
           num_vertexes++;
@@ -1503,6 +1505,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + width))
           do_sprite_enhanced(psx_gpu, x, y, 0, 0, width, height, list[0]);
+        cpu_cycles += gput_sprite(width, height);
         break;
       }
   
@@ -1522,13 +1525,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + width))
           do_sprite_enhanced(psx_gpu, x, y, u, v, width, height, list[0]);
+        cpu_cycles += gput_sprite(width, height);
         break;
       }
   
-      case 0x68:
-      case 0x69:
-      case 0x6A:
-      case 0x6B:
+      case 0x68 ... 0x6B:
       {
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -1537,13 +1538,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + 1))
           do_sprite_enhanced(psx_gpu, x, y, 0, 0, 1, 1, list[0]);
+        cpu_cycles += gput_sprite(1, 1);
         break;
       }
   
-      case 0x70:
-      case 0x71:
-      case 0x72:
-      case 0x73:
+      case 0x70 ... 0x73:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -1552,13 +1551,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + 8))
           do_sprite_enhanced(psx_gpu, x, y, 0, 0, 8, 8, list[0]);
+        cpu_cycles += gput_sprite(8, 8);
         break;
       }
   
-      case 0x74:
-      case 0x75:
-      case 0x76:
-      case 0x77:
+      case 0x74 ... 0x77:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -1572,13 +1569,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + 8))
           do_sprite_enhanced(psx_gpu, x, y, u, v, 8, 8, list[0]);
+        cpu_cycles += gput_sprite(8, 8);
         break;
       }
   
-      case 0x78:
-      case 0x79:
-      case 0x7A:
-      case 0x7B:
+      case 0x78 ... 0x7B:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -1587,13 +1582,11 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + 16))
           do_sprite_enhanced(psx_gpu, x, y, 0, 0, 16, 16, list[0]);
+        cpu_cycles += gput_sprite(16, 16);
         break;
       }
   
-      case 0x7C:
-      case 0x7D:
-      case 0x7E:
-      case 0x7F:
+      case 0x7C ... 0x7F:
       {        
         s32 x = sign_extend_11bit(list_s16[2] + psx_gpu->offset_x);
         s32 y = sign_extend_11bit(list_s16[3] + psx_gpu->offset_y);
@@ -1606,6 +1599,7 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
 
         if (check_enhanced_range(psx_gpu, x, x + 16))
           do_sprite_enhanced(psx_gpu, x, y, u, v, 16, 16, list[0]);
+        cpu_cycles += gput_sprite(16, 16);
         break;
       }
 
@@ -1759,8 +1753,8 @@ u32 gpu_parse_enhanced(psx_gpu_struct *psx_gpu, u32 *list, u32 size,
   enhancement_disable();
 
 breakloop:
-  if (last_command != NULL)
-    *last_command = current_command;
+  *cpu_cycles_out += cpu_cycles;
+  *last_command = current_command;
   return list - list_start;
 }
 

diff --git a/plugins/gpu_neon/psx_gpu_if.c b/plugins/gpu_neon/psx_gpu_if.c
index a1476f4..570cc5d 100644 (file)
--- a/plugins/gpu_neon/psx_gpu_if.c
+++ b/plugins/gpu_neon/psx_gpu_if.c
@@ -39,7 +39,7 @@ sync_enhancement_buffers(int x, int y, int w, int h);
 
 static psx_gpu_struct egpu __attribute__((aligned(256)));
 
-int do_cmd_list(uint32_t *list, int count, int *last_cmd)
+int do_cmd_list(uint32_t *list, int count, int *cycles, int *last_cmd)
 {
   int ret;
 
@@ -49,9 +49,9 @@ int do_cmd_list(uint32_t *list, int count, int *last_cmd)
 #endif
 
   if (gpu.state.enhancement_active)
-    ret = gpu_parse_enhanced(&egpu, list, count * 4, (u32 *)last_cmd);
+    ret = gpu_parse_enhanced(&egpu, list, count * 4, cycles, (u32 *)last_cmd);
   else
-    ret = gpu_parse(&egpu, list, count * 4, (u32 *)last_cmd);
+    ret = gpu_parse(&egpu, list, count * 4, cycles, (u32 *)last_cmd);
 
 #if defined(__arm__) && defined(NEON_BUILD) && !defined(SIMD_BUILD)
   __asm__ __volatile__("":::"q4","q5","q6","q7");
@@ -153,7 +153,9 @@ sync_enhancement_buffers(int x, int y, int w, int h)
 
 void renderer_sync_ecmds(uint32_t *ecmds)
 {
-  gpu_parse(&egpu, ecmds + 1, 6 * 4, NULL);
+  s32 dummy0 = 0;
+  u32 dummy1 = 0;
+  gpu_parse(&egpu, ecmds + 1, 6 * 4, &dummy0, &dummy1);
 }
 
 void renderer_update_caches(int x, int y, int w, int h, int state_changed)

diff --git a/plugins/gpu_unai/gpulib_if.cpp b/plugins/gpu_unai/gpulib_if.cpp
index 2079431..191108b 100644 (file)
--- a/plugins/gpu_unai/gpulib_if.cpp
+++ b/plugins/gpu_unai/gpulib_if.cpp
@@ -390,14 +390,16 @@ static void gpuGP0Cmd_0xEx(gpu_unai_t &gpu_unai, u32 cmd_word)
 }
 #endif
 
+#include "../gpulib/gpu_timing.h"
 extern const unsigned char cmd_lengths[256];
 
-int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
+int do_cmd_list(u32 *_list, int list_len, int *cpu_cycles_out, int *last_cmd)
 {
   u32 cmd = 0, len, i;
   le32_t *list = (le32_t *)_list;
   le32_t *list_start = list;
   le32_t *list_end = list + list_len;
+  u32 cpu_cycles = 0;
 
   //TODO: set ilace_mask when resolution changes instead of every time,
   // eliminate #ifdef below.
@@ -430,6 +432,8 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
     {
       case 0x02:
         gpuClearImage(packet);
+        cpu_cycles += gput_fill(le16_to_s16(packet.U2[4]) & 0x3ff,
+           le16_to_s16(packet.U2[5]) & 0x1ff);
         break;
 
       case 0x20:
@@ -442,6 +446,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | gpu_unai.PixelMSB
         ];
         gpuDrawPolyF(packet, driver, false);
+        cpu_cycles += gput_poly_base();
       } break;
 
       case 0x24:
@@ -466,6 +471,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
 
         PP driver = gpuPolySpanDrivers[driver_idx];
         gpuDrawPolyFT(packet, driver, false);
+        cpu_cycles += gput_poly_base_t();
       } break;
 
       case 0x28:
@@ -478,6 +484,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | gpu_unai.PixelMSB
         ];
         gpuDrawPolyF(packet, driver, true); // is_quad = true
+        cpu_cycles += gput_quad_base();
       } break;
 
       case 0x2C:
@@ -502,6 +509,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
 
         PP driver = gpuPolySpanDrivers[driver_idx];
         gpuDrawPolyFT(packet, driver, true); // is_quad = true
+        cpu_cycles += gput_quad_base_t();
       } break;
 
       case 0x30:
@@ -519,6 +527,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | 129 | gpu_unai.PixelMSB
         ];
         gpuDrawPolyG(packet, driver, false);
+        cpu_cycles += gput_poly_base_g();
       } break;
 
       case 0x34:
@@ -534,6 +543,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | ((Lighting)?129:0) | gpu_unai.PixelMSB
         ];
         gpuDrawPolyGT(packet, driver, false);
+        cpu_cycles += gput_poly_base_gt();
       } break;
 
       case 0x38:
@@ -548,6 +558,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | 129 | gpu_unai.PixelMSB
         ];
         gpuDrawPolyG(packet, driver, true); // is_quad = true
+        cpu_cycles += gput_quad_base_g();
       } break;
 
       case 0x3C:
@@ -563,6 +574,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.Masking | Blending | ((Lighting)?129:0) | gpu_unai.PixelMSB
         ];
         gpuDrawPolyGT(packet, driver, true); // is_quad = true
+        cpu_cycles += gput_quad_base_gt();
       } break;
 
       case 0x40:
@@ -573,6 +585,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         u32 driver_idx = (Blending_Mode | gpu_unai.Masking | Blending | (gpu_unai.PixelMSB>>3)) >> 1;
         PSD driver = gpuPixelSpanDrivers[driver_idx];
         gpuDrawLineF(packet, driver);
+        cpu_cycles += gput_line(0);
       } break;
 
       case 0x48 ... 0x4F: { // Monochrome line strip
@@ -589,6 +602,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.PacketBuffer.U4[1] = gpu_unai.PacketBuffer.U4[2];
           gpu_unai.PacketBuffer.U4[2] = *list_position++;
           gpuDrawLineF(packet, driver);
+          cpu_cycles += gput_line(0);
 
           num_vertexes++;
           if(list_position >= list_end) {
@@ -612,6 +626,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         driver_idx |= (1 << 5);
         PSD driver = gpuPixelSpanDrivers[driver_idx];
         gpuDrawLineG(packet, driver);
+        cpu_cycles += gput_line(0);
       } break;
 
       case 0x58 ... 0x5F: { // Gouraud-shaded line strip
@@ -632,6 +647,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           gpu_unai.PacketBuffer.U4[2] = *list_position++;
           gpu_unai.PacketBuffer.U4[3] = *list_position++;
           gpuDrawLineG(packet, driver);
+          cpu_cycles += gput_line(0);
 
           num_vertexes++;
           if(list_position >= list_end) {
@@ -651,6 +667,8 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
       case 0x63: {          // Monochrome rectangle (variable size)
         PT driver = gpuTileSpanDrivers[(Blending_Mode | gpu_unai.Masking | Blending | (gpu_unai.PixelMSB>>3)) >> 1];
         gpuDrawT(packet, driver);
+        cpu_cycles += gput_sprite(le16_to_u16(packet.U2[4]) & 0x3ff,
+                         le16_to_u16(packet.U2[5]) & 0x1ff);
       } break;
 
       case 0x64:
@@ -678,6 +696,8 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           driver_idx |= Lighting;
         PS driver = gpuSpriteSpanDrivers[driver_idx];
         gpuDrawS(packet, driver);
+        cpu_cycles += gput_sprite(le16_to_u16(packet.U2[6]) & 0x3ff,
+                         le16_to_u16(packet.U2[7]) & 0x1ff);
       } break;
 
       case 0x68:
@@ -687,6 +707,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         gpu_unai.PacketBuffer.U4[2] = u32_to_le32(0x00010001);
         PT driver = gpuTileSpanDrivers[(Blending_Mode | gpu_unai.Masking | Blending | (gpu_unai.PixelMSB>>3)) >> 1];
         gpuDrawT(packet, driver);
+        cpu_cycles += gput_sprite(1, 1);
       } break;
 
       case 0x70:
@@ -696,6 +717,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         gpu_unai.PacketBuffer.U4[2] = u32_to_le32(0x00080008);
         PT driver = gpuTileSpanDrivers[(Blending_Mode | gpu_unai.Masking | Blending | (gpu_unai.PixelMSB>>3)) >> 1];
         gpuDrawT(packet, driver);
+        cpu_cycles += gput_sprite(8, 8);
       } break;
 
       case 0x74:
@@ -713,6 +735,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           driver_idx |= Lighting;
         PS driver = gpuSpriteSpanDrivers[driver_idx];
         gpuDrawS(packet, driver);
+        cpu_cycles += gput_sprite(8, 8);
       } break;
 
       case 0x78:
@@ -722,6 +745,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         gpu_unai.PacketBuffer.U4[2] = u32_to_le32(0x00100010);
         PT driver = gpuTileSpanDrivers[(Blending_Mode | gpu_unai.Masking | Blending | (gpu_unai.PixelMSB>>3)) >> 1];
         gpuDrawT(packet, driver);
+        cpu_cycles += gput_sprite(16, 16);
       } break;
 
       case 0x7C:
@@ -731,6 +755,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
         {
           gpuSetCLUT    (le32_to_u32(gpu_unai.PacketBuffer.U4[2]) >> 16);
           gpuDrawS16(packet);
+          cpu_cycles += gput_sprite(16, 16);
           break;
         }
         // fallthrough
@@ -747,6 +772,7 @@ int do_cmd_list(u32 *_list, int list_len, int *last_cmd)
           driver_idx |= Lighting;
         PS driver = gpuSpriteSpanDrivers[driver_idx];
         gpuDrawS(packet, driver);
+        cpu_cycles += gput_sprite(16, 16);
       } break;
 
 #ifdef TEST
@@ -782,6 +808,7 @@ breakloop:
   gpu.ex_regs[1] &= ~0x1ff;
   gpu.ex_regs[1] |= gpu_unai.GPU_GP1 & 0x1ff;
 
+  *cpu_cycles_out += cpu_cycles;
   *last_cmd = cmd;
   return list - list_start;
 }
@@ -789,7 +816,7 @@ breakloop:
 void renderer_sync_ecmds(u32 *ecmds)
 {
   int dummy;
-  do_cmd_list(&ecmds[1], 6, &dummy);
+  do_cmd_list(&ecmds[1], 6, &dummy, &dummy);
 }
 
 void renderer_update_caches(int x, int y, int w, int h, int state_changed)

diff --git a/plugins/gpu_unai_old/gpulib_if.cpp b/plugins/gpu_unai_old/gpulib_if.cpp
index cc32802..ee694d3 100644 (file)
--- a/plugins/gpu_unai_old/gpulib_if.cpp
+++ b/plugins/gpu_unai_old/gpulib_if.cpp
@@ -169,7 +169,7 @@ void renderer_notify_scanout_change(int x, int y)
 
 extern const unsigned char cmd_lengths[256];
 
-int do_cmd_list(unsigned int *list, int list_len, int *last_cmd)
+int do_cmd_list(unsigned int *list, int list_len, int *cycles, int *last_cmd)
 {
   unsigned int cmd = 0, len, i;
   unsigned int *list_start = list;
@@ -523,7 +523,7 @@ breakloop:
 void renderer_sync_ecmds(uint32_t *ecmds)
 {
   int dummy;
-  do_cmd_list(&ecmds[1], 6, &dummy);
+  do_cmd_list(&ecmds[1], 6, &dummy, &dummy);
 }
 
 void renderer_update_caches(int x, int y, int w, int h, int state_changed)

diff --git a/plugins/gpulib/gpu.c b/plugins/gpulib/gpu.c
index 2ac36c1..7d40938 100644 (file)
--- a/plugins/gpulib/gpu.c
+++ b/plugins/gpulib/gpu.c
@@ -12,6 +12,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include "gpu.h"
+#include "gpu_timing.h"
 #include "../../libpcsxcore/gpu.h" // meh
 #include "../../frontend/plugin_lib.h"
 
@@ -33,13 +34,14 @@
 
 struct psx_gpu gpu;
 
-static noinline int do_cmd_buffer(uint32_t *data, int count);
+static noinline int do_cmd_buffer(uint32_t *data, int count, int *cpu_cycles);
 static void finish_vram_transfer(int is_read);
 
 static noinline void do_cmd_reset(void)
 {
+  int dummy = 0;
   if (unlikely(gpu.cmd_len > 0))
-    do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len);
+    do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len, &dummy);
   gpu.cmd_len = 0;
 
   if (unlikely(gpu.dma.h > 0))
@@ -172,8 +174,8 @@ static noinline void decide_frameskip(void)
     gpu.frameskip.active = 0;
 
   if (!gpu.frameskip.active && gpu.frameskip.pending_fill[0] != 0) {
-    int dummy;
-    do_cmd_list(gpu.frameskip.pending_fill, 3, &dummy);
+    int dummy = 0;
+    do_cmd_list(gpu.frameskip.pending_fill, 3, &dummy, &dummy);
     gpu.frameskip.pending_fill[0] = 0;
   }
 }
@@ -472,7 +474,7 @@ static void finish_vram_transfer(int is_read)
     gpu.gpu_state_change(PGS_VRAM_TRANSFER_END);
 }
 
-static void do_vram_copy(const uint32_t *params)
+static void do_vram_copy(const uint32_t *params, int *cpu_cycles)
 {
   const uint32_t sx =  LE32TOH(params[0]) & 0x3FF;
   const uint32_t sy = (LE32TOH(params[0]) >> 16) & 0x1FF;
@@ -484,6 +486,7 @@ static void do_vram_copy(const uint32_t *params)
   uint16_t lbuf[128];
   uint32_t x, y;
 
+  *cpu_cycles += gput_copy(w, h);
   if (sx == dx && sy == dy && msb == 0)
     return;
 
@@ -519,7 +522,7 @@ static void do_vram_copy(const uint32_t *params)
 
 static noinline int do_cmd_list_skip(uint32_t *data, int count, int *last_cmd)
 {
-  int cmd = 0, pos = 0, len, dummy, v;
+  int cmd = 0, pos = 0, len, dummy = 0, v;
   int skip = 1;
 
   gpu.frameskip.pending_fill[0] = 0;
@@ -533,7 +536,7 @@ static noinline int do_cmd_list_skip(uint32_t *data, int count, int *last_cmd)
       case 0x02:
         if ((LE32TOH(list[2]) & 0x3ff) > gpu.screen.w || ((LE32TOH(list[2]) >> 16) & 0x1ff) > gpu.screen.h)
           // clearing something large, don't skip
-          do_cmd_list(list, 3, &dummy);
+          do_cmd_list(list, 3, &dummy, &dummy);
         else
           memcpy(gpu.frameskip.pending_fill, list, 3 * 4);
         break;
@@ -583,7 +586,7 @@ static noinline int do_cmd_list_skip(uint32_t *data, int count, int *last_cmd)
   return pos;
 }
 
-static noinline int do_cmd_buffer(uint32_t *data, int count)
+static noinline int do_cmd_buffer(uint32_t *data, int count, int *cpu_cycles)
 {
   int cmd, pos;
   uint32_t old_e3 = gpu.ex_regs[3];
@@ -617,7 +620,7 @@ static noinline int do_cmd_buffer(uint32_t *data, int count)
         cmd = -1; // incomplete cmd, can't consume yet
         break;
       }
-      do_vram_copy(data + pos + 1);
+      do_vram_copy(data + pos + 1, cpu_cycles);
       vram_dirty = 1;
       pos += 4;
       continue;
@@ -627,7 +630,7 @@ static noinline int do_cmd_buffer(uint32_t *data, int count)
     if (gpu.frameskip.active && (gpu.frameskip.allow || ((LE32TOH(data[pos]) >> 24) & 0xf0) == 0xe0))
       pos += do_cmd_list_skip(data + pos, count - pos, &cmd);
     else {
-      pos += do_cmd_list(data + pos, count - pos, &cmd);
+      pos += do_cmd_list(data + pos, count - pos, cpu_cycles, &cmd);
       vram_dirty = 1;
     }
 
@@ -650,7 +653,8 @@ static noinline int do_cmd_buffer(uint32_t *data, int count)
 
 static noinline void flush_cmd_buffer(void)
 {
-  int left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len);
+  int dummy = 0, left;
+  left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len, &dummy);
   if (left > 0)
     memmove(gpu.cmd_buffer, gpu.cmd_buffer + gpu.cmd_len - left, left * 4);
   if (left != gpu.cmd_len) {
@@ -662,14 +666,14 @@ static noinline void flush_cmd_buffer(void)
 
 void GPUwriteDataMem(uint32_t *mem, int count)
 {
-  int left;
+  int dummy = 0, left;
 
   log_io("gpu_dma_write %p %d\n", mem, count);
 
   if (unlikely(gpu.cmd_len > 0))
     flush_cmd_buffer();
 
-  left = do_cmd_buffer(mem, count);
+  left = do_cmd_buffer(mem, count, &dummy);
   if (left)
     log_anomaly("GPUwriteDataMem: discarded %d/%d words\n", left, count);
 }
@@ -686,7 +690,7 @@ long GPUdmaChain(uint32_t *rambase, uint32_t start_addr, uint32_t *progress_addr
 {
   uint32_t addr, *list, ld_addr = 0;
   int len, left, count;
-  long cpu_cycles = 0;
+  int cpu_cycles = 0;
 
   preload(rambase + (start_addr & 0x1fffff) / 4);
 
@@ -720,7 +724,7 @@ long GPUdmaChain(uint32_t *rambase, uint32_t start_addr, uint32_t *progress_addr
     }
 
     if (len) {
-      left = do_cmd_buffer(list + 1, len);
+      left = do_cmd_buffer(list + 1, len, &cpu_cycles);
       if (left) {
         memcpy(gpu.cmd_buffer, list + 1 + len - left, left * 4);
         gpu.cmd_len = left;

diff --git a/plugins/gpulib/gpu.h b/plugins/gpulib/gpu.h
index 886bb1f..13e73c5 100644 (file)
--- a/plugins/gpulib/gpu.h
+++ b/plugins/gpulib/gpu.h
@@ -119,7 +119,7 @@ extern struct psx_gpu gpu;
 
 extern const unsigned char cmd_lengths[256];
 
-int do_cmd_list(uint32_t *list, int count, int *last_cmd);
+int do_cmd_list(uint32_t *list, int count, int *cycles, int *last_cmd);
 
 struct rearmed_cbs;
 

diff --git a/plugins/gpulib/gpu_timing.h b/plugins/gpulib/gpu_timing.h
new file mode 100644 (file)
index 0000000..0dfe0d6
--- /dev/null
+++ b/plugins/gpulib/gpu_timing.h
@@ -0,0 +1,15 @@
+
+// very conservative and wrong
+#define gput_fill(w, h)     (23 + (4 + (w) / 16u) * (h))
+#define gput_copy(w, h)     ((w) * (h))
+#define gput_poly_base()    (23)
+#define gput_poly_base_t()  (gput_poly_base() + 90)
+#define gput_poly_base_g()  (gput_poly_base() + 144)
+#define gput_poly_base_gt() (gput_poly_base() + 225)
+#define gput_quad_base()    gput_poly_base()
+#define gput_quad_base_t()  gput_poly_base_t()
+#define gput_quad_base_g()  gput_poly_base_g()
+#define gput_quad_base_gt() gput_poly_base_gt()
+#define gput_line(k)        (8 + (k))
+#define gput_sprite(w, h)   (8 + ((w) / 2u) * (h))
+

diff --git a/plugins/gpulib/test.c b/plugins/gpulib/test.c
index 80d0e9e..3f24cc4 100644 (file)
--- a/plugins/gpulib/test.c
+++ b/plugins/gpulib/test.c
@@ -88,13 +88,13 @@ int main(int argc, char *argv[])
  
   pcnt_init();
   renderer_init();
-  memcpy(gpu.vram, state.vram, sizeof(gpu.vram));
+  memcpy(gpu.vram, state.vram, 1024*512*2);
   if ((state.gpu_register[8] & 0x24) == 0x24)
     renderer_set_interlace(1, !(state.status >> 31));
 
   start_cycles = pcnt_get();
 
-  do_cmd_list(list, size / 4, &dummy);
+  do_cmd_list(list, size / 4, &dummy, &dummy);
   renderer_flush_queues();
 
   printf("%u\n", pcnt_get() - start_cycles);