gpu_neon: rework buffering to reduce flickering

[pcsx_rearmed.git] / plugins / gpulib / gpu.c

diff --git a/plugins/gpulib/gpu.c b/plugins/gpulib/gpu.c
index fec4688..fdb109c 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 "../../libpcsxcore/gpu.h" // meh
 
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
 #ifdef __GNUC__
@@ -259,6 +260,7 @@ long GPUshutdown(void)
 void GPUwriteStatus(uint32_t data)
 {
   uint32_t cmd = data >> 24;
+  int src_x, src_y;
 
   if (cmd < ARRAY_SIZE(gpu.regs)) {
     if (cmd > 1 && cmd != 5 && gpu.regs[cmd] == data)
@@ -288,14 +290,17 @@ void GPUwriteStatus(uint32_t data)
       gpu.status |= PSX_GPU_STATUS_DMA(data & 3);
       break;
     case 0x05:
-      gpu.screen.src_x = data & 0x3ff;
-      gpu.screen.src_y = (data >> 10) & 0x1ff;
-      renderer_notify_scanout_x_change(gpu.screen.src_x, gpu.screen.hres);
-      if (gpu.frameskip.set) {
-        decide_frameskip_allow(gpu.ex_regs[3]);
-        if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) {
-          decide_frameskip();
-          gpu.frameskip.last_flip_frame = *gpu.state.frame_count;
+      src_x = data & 0x3ff; src_y = (data >> 10) & 0x1ff;
+      if (src_x != gpu.screen.src_x || src_y != gpu.screen.src_y) {
+        gpu.screen.src_x = src_x;
+        gpu.screen.src_y = src_y;
+        renderer_notify_scanout_change(src_x, src_y);
+        if (gpu.frameskip.set) {
+          decide_frameskip_allow(gpu.ex_regs[3]);
+          if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) {
+            decide_frameskip();
+            gpu.frameskip.last_flip_frame = *gpu.state.frame_count;
+          }
         }
       }
       break;
@@ -441,6 +446,8 @@ static void start_vram_transfer(uint32_t pos_word, uint32_t size_word, int is_re
 
   log_io("start_vram_transfer %c (%d, %d) %dx%d\n", is_read ? 'r' : 'w',
     gpu.dma.x, gpu.dma.y, gpu.dma.w, gpu.dma.h);
+  if (gpu.gpu_state_change)
+    gpu.gpu_state_change(PGS_VRAM_TRANSFER_START);
 }
 
 static void finish_vram_transfer(int is_read)
@@ -452,6 +459,8 @@ static void finish_vram_transfer(int is_read)
     renderer_update_caches(gpu.dma_start.x, gpu.dma_start.y,
                            gpu.dma_start.w, gpu.dma_start.h, 0);
   }
+  if (gpu.gpu_state_change)
+    gpu.gpu_state_change(PGS_VRAM_TRANSFER_END);
 }
 
 static void do_vram_copy(const uint32_t *params)
@@ -630,12 +639,16 @@ static noinline int do_cmd_buffer(uint32_t *data, int count)
   return count - pos;
 }
 
-static void flush_cmd_buffer(void)
+static noinline void flush_cmd_buffer(void)
 {
   int left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len);
   if (left > 0)
     memmove(gpu.cmd_buffer, gpu.cmd_buffer + gpu.cmd_len - left, left * 4);
-  gpu.cmd_len = left;
+  if (left != gpu.cmd_len) {
+    if (!gpu.dma.h && gpu.gpu_state_change)
+      gpu.gpu_state_change(PGS_PRIMITIVE_START);
+    gpu.cmd_len = left;
+  }
 }
 
 void GPUwriteDataMem(uint32_t *mem, int count)
@@ -816,7 +829,7 @@ long GPUfreeze(uint32_t type, struct GPUFreeze *freeze)
         GPUwriteStatus((i << 24) | (gpu.regs[i] ^ 1));
       }
       renderer_sync_ecmds(gpu.ex_regs);
-      renderer_update_caches(0, 0, 1024, 512, 1);
+      renderer_update_caches(0, 0, 1024, 512, 0);
       break;
   }
 
@@ -914,6 +927,7 @@ void GPUrearmedCallbacks(const struct rearmed_cbs *cbs)
 
   gpu.mmap = cbs->mmap;
   gpu.munmap = cbs->munmap;
+  gpu.gpu_state_change = cbs->gpu_state_change;
 
   // delayed vram mmap
   if (gpu.vram == NULL)