merge from libretro fork

[pcsx_rearmed.git] / plugins / gpu_neon / psx_gpu / psx_gpu.c

diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu.c b/plugins/gpu_neon/psx_gpu/psx_gpu.c
index a79254d..85cf89f 100644 (file)
--- a/plugins/gpu_neon/psx_gpu/psx_gpu.c
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu.c
@@ -14,6 +14,7 @@
 
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdint.h>
 #include <string.h>
 
 #include "common.h"
@@ -22,6 +23,13 @@
 #endif
 #include "psx_gpu_simd.h"
 
+#if 0
+void dump_r_d(const char *name, void *dump);
+void dump_r_q(const char *name, void *dump);
+#define dumprd(n) dump_r_d(#n, n.e)
+#define dumprq(n) dump_r_q(#n, n.e)
+#endif
+
 u32 span_pixels = 0;
 u32 span_pixel_blocks = 0;
 u32 spans = 0;
@@ -51,6 +59,8 @@ u32 zero_block_spans = 0;
 u32 texture_cache_loads = 0;
 u32 false_modulated_blocks = 0;
 
+#define stats_add(stat, count) // stat += count
+
 /* double size for enhancement */
 u32 reciprocal_table[512 * 2];
 
@@ -767,13 +777,13 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
 {                                                                              \
   u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data;          \
   if (_num_spans > MAX_SPANS)                                                  \
-    *(int *)0 = 1;                                                             \
+    *(volatile int *)0 = 1;                                                    \
   if (_num_spans < psx_gpu->num_spans)                                         \
   {                                                                            \
     if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW)                 \
-      *(int *)0 = 1;                                                           \
-    if(span_edge_data_element.y > 2048)                                        \
-      *(int *)0 = 1;                                                           \
+      *(volatile int *)0 = 2;                                                  \
+    if(span_edge_data_element.y >= 2048)                                       \
+      *(volatile int *)0 = 3;                                                  \
   }                                                                            \
 }                                                                              \
 
@@ -786,7 +796,7 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
   vec_2x64s alternate_x;                                                       \
   vec_2x64s alternate_dx_dy;                                                   \
   vec_4x32s alternate_x_32;                                                    \
-  vec_2x32s alternate_x_16;                                                    \
+  vec_4x16u alternate_x_16;                                                    \
                                                                                \
   vec_4x16u alternate_select;                                                  \
   vec_4x16s y_mid_point;                                                       \
@@ -1842,7 +1852,7 @@ void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
   }                                                                            \
 
 #define setup_blocks_add_blocks_direct()                                       \
-  texel_blocks_untextured += span_num_blocks;                                  \
+  stats_add(texel_blocks_untextured, span_num_blocks);                         \
   span_pixel_blocks += span_num_blocks                                         \
 
 
@@ -1938,14 +1948,14 @@ setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
 {
   if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
-    texel_blocks_untextured += psx_gpu->num_blocks;
+    stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
 }
 
 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
 {
   block_struct *block = psx_gpu->blocks;
   u32 num_blocks = psx_gpu->num_blocks;
-  texel_blocks_4bpp += num_blocks;
+  stats_add(texel_blocks_4bpp, num_blocks);
 
   vec_8x8u texels_low;
   vec_8x8u texels_high;
@@ -1997,7 +2007,7 @@ void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
   block_struct *block = psx_gpu->blocks;
   u32 num_blocks = psx_gpu->num_blocks;
 
-  texel_blocks_8bpp += num_blocks;
+  stats_add(texel_blocks_8bpp, num_blocks);
 
   if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
     update_texture_8bpp_cache(psx_gpu);
@@ -2031,7 +2041,7 @@ void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
   block_struct *block = psx_gpu->blocks;
   u32 num_blocks = psx_gpu->num_blocks;
 
-  texel_blocks_16bpp += num_blocks;
+  stats_add(texel_blocks_16bpp, num_blocks);
 
   vec_8x16u texels;
 
@@ -3067,7 +3077,7 @@ void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
     render_triangle_p(psx_gpu, vertex_ptrs, flags);
 }
 
-#ifndef NEON_BUILD
+#if !defined(NEON_BUILD) || defined(SIMD_BUILD)
 
 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
 {
@@ -3120,7 +3130,7 @@ void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
 
 
 #define setup_sprite_tile_fetch_texel_block_8bpp(offset)                       \
-  texture_block_ptr = psx_gpu->texture_page_ptr +                              \
+  texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr +                        \
    ((texture_offset + offset) & texture_mask);                                 \
                                                                                \
   load_64b(texels, texture_block_ptr)                                          \
@@ -3228,7 +3238,7 @@ void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
 
 #define setup_sprite_tile_half_8bpp(edge)                                      \
 {                                                                              \
-  setup_sprite_tile_add_blocks(sub_tile_height * 2);                           \
+  setup_sprite_tile_add_blocks(sub_tile_height);                               \
                                                                                \
   while(sub_tile_height)                                                       \
   {                                                                            \
@@ -3682,7 +3692,7 @@ void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
   u32 num_blocks = psx_gpu->num_blocks;                                        \
   block_struct *block = psx_gpu->blocks + num_blocks;                          \
                                                                                \
-  u16 *texture_block_ptr;                                                      \
+  u8 *texture_block_ptr;                                                       \
   vec_8x8u texels;                                                             \
                                                                                \
   setup_sprite_tiled_initialize_##texture_mode##x4mode();                      \
@@ -3775,6 +3785,9 @@ setup_sprite_tiled_builder(8bpp,);
 
 setup_sprite_tiled_builder(4bpp,_4x);
 setup_sprite_tiled_builder(8bpp,_4x);
+#endif
+
+#if !defined(NEON_BUILD) || defined(SIMD_BUILD)
 
 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
  s32 v, s32 width, s32 height, u32 color)
@@ -3803,7 +3816,7 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
 
   texture_offset_base &= ~0x7;
 
-  sprites_16bpp++;
+  stats_add(sprites_16bpp, 1);
 
   if(block_width == 1)
   {
@@ -3824,7 +3837,7 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
       texture_block_ptr =
        texture_page_ptr + (texture_offset_base & texture_mask);
 
-      load_128b(block->texels, texture_block_ptr);
+      block->texels = *(vec_8x16u *)texture_block_ptr;
       block->draw_mask_bits = mask_bits;
       block->fb_ptr = fb_ptr;
 
@@ -3858,7 +3871,7 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
       texture_offset_base += 1024;
 
       texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
-      load_128b(block->texels, texture_block_ptr);
+      block->texels = *(vec_8x16u *)texture_block_ptr;
 
       block->draw_mask_bits = left_mask_bits;
       block->fb_ptr = fb_ptr;
@@ -3870,7 +3883,7 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
       while(blocks_remaining)
       {
         texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
-        load_128b(block->texels, texture_block_ptr);
+        block->texels = *(vec_8x16u *)texture_block_ptr;
 
         block->draw_mask_bits = 0;
         block->fb_ptr = fb_ptr;
@@ -3883,7 +3896,7 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
       }
 
       texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
-      load_128b(block->texels, texture_block_ptr);
+      block->texels = *(vec_8x16u *)texture_block_ptr;
 
       block->draw_mask_bits = right_mask_bits;
       block->fb_ptr = fb_ptr;
@@ -3897,6 +3910,10 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
   }
 }
 
+#endif
+
+#ifndef NEON_BUILD
+
 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
  s32 v, s32 width, s32 height, u32 color)
 {
@@ -4000,7 +4017,7 @@ void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y,
     num_width = width;
 
     vram_ptr = (void *)vram_ptr16;
-    if((long)vram_ptr16 & 2)
+    if((uintptr_t)vram_ptr16 & 2)
     {
       *vram_ptr16 = color_32bpp;
       vram_ptr = (void *)(vram_ptr16 + 1);