X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=pcsx_rearmed.git;a=blobdiff_plain;f=plugins%2Fgpu_neon%2Fpsx_gpu_if.c;h=ad017614b527758764a22177704643556063f358;hp=ea98ade85e9de4c39f4dc170b2a5a0695c8ae389;hb=HEAD;hpb=1587b08ecb80dd257ac255f76e716a6053c2785b

diff --git a/plugins/gpu_neon/psx_gpu_if.c b/plugins/gpu_neon/psx_gpu_if.c
index ea98ade8..f85155e3 100644
--- a/plugins/gpu_neon/psx_gpu_if.c
+++ b/plugins/gpu_neon/psx_gpu_if.c
@@ -10,7 +10,6 @@
 
 #include <stdio.h>
 #include <assert.h>
-#include <sys/mman.h>
 
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
 #ifndef min
@@ -39,23 +38,17 @@ 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_sum, int *cycles_last, int *last_cmd)
 {
   int ret;
 
-#if defined(__arm__) && defined(NEON_BUILD) && !defined(SIMD_BUILD)
-  // the asm doesn't bother to save callee-save vector regs, so do it here
-  __asm__ __volatile__("":::"q4","q5","q6","q7");
-#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_sum, cycles_last, (u32 *)last_cmd);
   else
-    ret = gpu_parse(&egpu, list, count * 4, (u32 *)last_cmd);
-
-#if defined(__arm__) && defined(NEON_BUILD) && !defined(SIMD_BUILD)
-  __asm__ __volatile__("":::"q4","q5","q6","q7");
-#endif
+    ret = gpu_parse(&egpu, list, count * 4,
+            cycles_sum, cycles_last, (u32 *)last_cmd);
 
   ex_regs[1] &= ~0x1ff;
   ex_regs[1] |= egpu.texture_settings & 0x1ff;
@@ -85,8 +78,9 @@ static void map_enhancement_buffer(void)
   // to be able to reuse 1024-width code better (triangle setup,
   // dithering phase, lines).
   egpu.enhancement_buf_ptr = gpu.mmap(ENHANCEMENT_BUF_SIZE);
-  if (egpu.enhancement_buf_ptr == NULL) {
+  if (egpu.enhancement_buf_ptr == NULL || egpu.enhancement_buf_ptr == (void *)(intptr_t)-1) {
     fprintf(stderr, "failed to map enhancement buffer\n");
+    egpu.enhancement_buf_ptr = NULL;
     gpu.get_enhancement_bufer = NULL;
   }
   else {
@@ -129,6 +123,8 @@ sync_enhancement_buffers(int x, int y, int w, int h)
   // due to intersection stuff, see the update_enhancement_buf_scanouts() mess
   int s_w = max(gpu.screen.hres, gpu.screen.w);
   int s_h = gpu.screen.vres;
+  if (gpu.screen.y < 0)
+    s_h -= gpu.screen.y;
   s_w = min(s_w, 512);
   for (i = 0; i < ARRAY_SIZE(egpu.enhancement_scanouts); i++) {
     const struct psx_gpu_scanout *s = &egpu.enhancement_scanouts[i];
@@ -143,6 +139,9 @@ sync_enhancement_buffers(int x, int y, int w, int h)
     x2 = min(right, s->x + s_w);
     y1 = max(y, s->y);
     y2 = min(bottom, s->y + s_h);
+    // 16-byte align for the asm version
+    x2 += x1 & 7;
+    x1 &= ~7;
     scale2x_tiles8(dst + y1 * 1024*2 + x1 * 2,
         src + y1 * 1024 + x1, (x2 - x1 + 7) / 8u, y2 - y1);
   }
@@ -150,7 +149,9 @@ 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, &dummy0, &dummy1);
 }
 
 void renderer_update_caches(int x, int y, int w, int h, int state_changed)
@@ -159,10 +160,13 @@ void renderer_update_caches(int x, int y, int w, int h, int state_changed)
 
   if (gpu.state.enhancement_active) {
     if (state_changed) {
+      int vres = gpu.screen.vres;
+      if (gpu.screen.y < 0)
+        vres -= gpu.screen.y;
       memset(egpu.enhancement_scanouts, 0, sizeof(egpu.enhancement_scanouts));
       egpu.enhancement_scanout_eselect = 0;
       update_enhancement_buf_scanouts(&egpu,
-        gpu.screen.src_x, gpu.screen.src_y, gpu.screen.hres, gpu.screen.vres);
+        gpu.screen.src_x, gpu.screen.src_y, gpu.screen.hres, vres);
       return;
     }
     sync_enhancement_buffers(x, y, w, h);
@@ -190,10 +194,13 @@ void renderer_notify_res_change(void)
 
 void renderer_notify_scanout_change(int x, int y)
 {
+  int vres = gpu.screen.vres;
   if (!gpu.state.enhancement_active || !egpu.enhancement_buf_ptr)
     return;
 
-  update_enhancement_buf_scanouts(&egpu, x, y, gpu.screen.hres, gpu.screen.vres);
+  if (gpu.screen.y < 0)
+    vres -= gpu.screen.y;
+  update_enhancement_buf_scanouts(&egpu, x, y, gpu.screen.hres, vres);
 }
 
 #include "../../frontend/plugin_lib.h"
@@ -206,12 +213,37 @@ void renderer_set_config(const struct rearmed_cbs *cbs)
   }
   if (cbs->pl_set_gpu_caps)
     cbs->pl_set_gpu_caps(GPU_CAP_SUPPORTS_2X);
+  
+  egpu.allow_dithering = cbs->dithering;
+  egpu.force_dithering = cbs->dithering >> 1;
+  /*
+  if (!egpu.allow_dithering) {
+    egpu.dither_table[0] = dither_table_row(0, 0, 0, 0);
+    egpu.dither_table[1] = dither_table_row(0, 0, 0, 0);
+    egpu.dither_table[2] = dither_table_row(0, 0, 0, 0);
+    egpu.dither_table[3] = dither_table_row(0, 0, 0, 0);
+  } else {
+    egpu.dither_table[0] = dither_table_row(-4, 0, -3, 1);
+    egpu.dither_table[1] = dither_table_row(2, -2, 3, -1);
+    egpu.dither_table[2] = dither_table_row(-3, 1, -4, 0);
+    egpu.dither_table[3] = dither_table_row(3, -1, 2, -2); 
+  }
+  */
 
-  disable_main_render = cbs->gpu_neon.enhancement_no_main;
+  egpu.hack_disable_main = cbs->gpu_neon.enhancement_no_main;
+  egpu.hack_texture_adj = cbs->gpu_neon.enhancement_tex_adj;
   if (gpu.state.enhancement_enable) {
     if (gpu.mmap != NULL && egpu.enhancement_buf_ptr == NULL)
       map_enhancement_buffer();
   }
 }
 
+void renderer_sync(void)
+{
+}
+
+void renderer_notify_update_lace(int updated)
+{
+}
+
 // vim:ts=2:sw=2:expandtab