gpu_neon: rework buffering to reduce flickering

... maybe

#324

frontend/plat_sdl.c

@@ -328,6 +328,13 @@ void plat_video_menu_end(void)
 
 void plat_video_menu_leave(void)
 {
+  void *fb = NULL;
+  if (plat_sdl_overlay != NULL || plat_sdl_gl_active)
+    fb = shadow_fb;
+  else if (plat_sdl_screen)
+    fb = plat_sdl_screen->pixels;
+  if (fb)
+    memset(fb, 0, g_menuscreen_w * g_menuscreen_h * 2);
   in_menu = 0;
 }
 

frontend/plugin_lib.c

@@ -134,7 +134,7 @@ static __attribute__((noinline)) void draw_active_chans(int vout_w, int vout_h)
 
 	static const unsigned short colors[2] = { 0x1fe3, 0x0700 };
 	unsigned short *dest = (unsigned short *)pl_vout_buf +
-		vout_w * (vout_h - HUD_HEIGHT) + vout_w / 2 - 192/2;
+		pl_vout_w * (vout_h - HUD_HEIGHT) + pl_vout_w / 2 - 192/2;
 	unsigned short *d, p;
 	int c, x, y;
 
@@ -149,7 +149,7 @@ static __attribute__((noinline)) void draw_active_chans(int vout_w, int vout_h)
 		     (fmod_chans & (1<<c)) ? 0xf000 :
 		     (noise_chans & (1<<c)) ? 0x001f :
 		     colors[c & 1];
-		for (y = 0; y < 8; y++, d += vout_w)
+		for (y = 0; y < 8; y++, d += pl_vout_w)
 			for (x = 0; x < 8; x++)
 				d[x] = p;
 	}
@@ -302,10 +302,16 @@ static void pl_vout_set_mode(int w, int h, int raw_w, int raw_h, int bpp)
 	menu_notify_mode_change(pl_vout_w, pl_vout_h, pl_vout_bpp);
 }
 
+static int flip_clear_counter;
+
+void pl_force_clear(void)
+{
+	flip_clear_counter = 2;
+}
+
 static void pl_vout_flip(const void *vram, int stride, int bgr24,
 	int x, int y, int w, int h, int dims_changed)
 {
-	static int clear_counter;
 	unsigned char *dest = pl_vout_buf;
 	const unsigned short *src = vram;
 	int dstride = pl_vout_w, h1 = h;
@@ -332,15 +338,15 @@ static void pl_vout_flip(const void *vram, int stride, int bgr24,
 	doffs = xoffs + y * dstride;
 
 	if (dims_changed)
-		clear_counter = 2;
+		flip_clear_counter = 2;
 
-	if (clear_counter > 0) {
+	if (flip_clear_counter > 0) {
 		if (pl_plat_clear)
 			pl_plat_clear();
 		else
 			memset(pl_vout_buf, 0,
 				dstride * h_full * pl_vout_bpp / 8);
-		clear_counter--;
+		flip_clear_counter--;
 	}
 
 	if (pl_plat_blit)

frontend/plugin_lib.h

@@ -45,6 +45,7 @@ void  pl_start_watchdog(void);
 void *pl_prepare_screenshot(int *w, int *h, int *bpp);
 void  pl_init(void);
 void  pl_switch_dispmode(void);
+void  pl_force_clear(void);
 
 void  pl_timing_prepare(int is_pal);
 void  pl_frame_limit(void);

plugins/dfxvideo/gpulib_if.c

@@ -299,7 +299,7 @@ void renderer_notify_res_change(void)
 {
 }
 
-void renderer_notify_scanout_x_change(int x, int w)
+void renderer_notify_scanout_change(int x, int y)
 {
 }
 

plugins/gpu-gles/gpulib_if.c

@@ -514,7 +514,7 @@ void renderer_notify_res_change(void)
 {
 }
 
-void renderer_notify_scanout_x_change(int x, int w)
+void renderer_notify_scanout_change(int x, int y)
 {
 }
 

plugins/gpu_neon/psx_gpu/psx_gpu.c

@@ -15,6 +15,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
+#include <stddef.h>
 #include <string.h>
 #include <assert.h>
 
@@ -23,6 +24,7 @@
 #include "vector_ops.h"
 #endif
 #include "psx_gpu_simd.h"
+#include "psx_gpu_offsets.h"
 
 #if 0
 void dump_r_d(const char *name, void *dump);
@@ -5012,6 +5014,9 @@ void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
   psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
 
   psx_gpu->saved_hres = 256;
+
+  // check some offset
+  psx_gpu->reserved_a[(offsetof(psx_gpu_struct, blocks) == psx_gpu_blocks_offset) - 1] = 0;
 }
 
 u64 get_us(void)

plugins/gpu_neon/psx_gpu/psx_gpu.h

@@ -141,15 +141,15 @@ typedef struct
   u32 triangle_color;
   u32 dither_table[4];
 
-  u32 uvrgb_phase;
-
   struct render_block_handler_struct *render_block_handler;
   void *texture_page_ptr;
   void *texture_page_base;
   u16 *clut_ptr;
   u16 *vram_ptr;
   u16 *vram_out_ptr;
 
+  u32 uvrgb_phase;
+
   u16 render_state_base;
   u16 render_state;
 
@@ -194,15 +194,15 @@ typedef struct
   s16 saved_viewport_start_y;
   s16 saved_viewport_end_x;
   s16 saved_viewport_end_y;
-  u8  enhancement_buf_by_x16[64];    // 0-3 specifying which buf
-  u16 enhancement_buf_start[4];      // x pos where buf[n] begins
-
-  u16 enhancement_scanout_x[4];
-  u16 enhancement_scanout_select;
+  struct psx_gpu_scanout {
+    u16 x, y, w, h;
+  } enhancement_scanouts[4];         // 0-3 specifying which buf to use
+  u16 enhancement_scanout_eselect;   // eviction selector
+  u16 enhancement_current_buf;
 
   // Align up to 64 byte boundary to keep the upcoming buffers cache line
   // aligned, also make reachable with single immediate addition
-  u8 reserved_a[142];
+  u8 reserved_a[188 + 9*4 - 9*sizeof(void *)];
 
   // 8KB
   block_struct blocks[MAX_BLOCKS_PER_ROW];

plugins/gpu_neon/psx_gpu/psx_gpu_4x.c

@@ -1,11 +1,3 @@
-#define select_enhancement_buf_index(psx_gpu, x) \
-  ((psx_gpu)->enhancement_buf_by_x16[(u32)(x) / \
-    (1024u / sizeof((psx_gpu)->enhancement_buf_by_x16))])
-
-#define select_enhancement_buf_ptr(psx_gpu, x) \
-  ((psx_gpu)->enhancement_buf_ptr + \
-    (select_enhancement_buf_index(psx_gpu, x) << 20))
-
 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
 
 #ifndef zip_4x32b
@@ -325,12 +317,12 @@ render_block_handler_struct render_sprite_block_handlers_4x[] =
   render_sprite_blocks_switch_block_4x()
 };
 
-
 void render_sprite_4x(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
  s32 width, s32 height, u32 flags, u32 color)
 {
   s32 x_right = x + width - 1;
   s32 y_bottom = y + height - 1;
+  s16 end_x;
 
 #ifdef PROFILE
   sprites++;
@@ -352,16 +344,22 @@ void render_sprite_4x(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
     height -= clip;
   }
 
-  if(x_right > psx_gpu->viewport_end_x)
-    width -= x_right - psx_gpu->viewport_end_x;
+  end_x = psx_gpu->viewport_end_x;
+  if (end_x - psx_gpu->viewport_start_x + 1 > 512)
+    end_x = psx_gpu->viewport_start_x + 511;
+
+  if(x_right > end_x)
+    width -= x_right - end_x;
 
   if(y_bottom > psx_gpu->viewport_end_y)
     height -= y_bottom - psx_gpu->viewport_end_y;
 
   if((width <= 0) || (height <= 0))
     return;
 
-  psx_gpu->vram_out_ptr = select_enhancement_buf_ptr(psx_gpu, x);
+  if (!psx_gpu->enhancement_current_buf_ptr)
+    return;
+  psx_gpu->vram_out_ptr = psx_gpu->enhancement_current_buf_ptr;
 
   x *= 2;
   y *= 2;

plugins/gpu_neon/psx_gpu/psx_gpu_offsets.h

@@ -1,7 +1,6 @@
 #ifndef __P_PSX_GPU_OFFSETS_H__
 #define __P_PSX_GPU_OFFSETS_H__
 
-#define psx_gpu_test_mask_offset                          0x0
 #define psx_gpu_uvrg_offset                               0x10
 #define psx_gpu_uvrg_dx_offset                            0x20
 #define psx_gpu_uvrg_dy_offset                            0x30
@@ -13,23 +12,18 @@
 #define psx_gpu_b_offset                                  0x90
 #define psx_gpu_b_dy_offset                               0x94
 #define psx_gpu_triangle_area_offset                      0x98
-#define psx_gpu_texture_window_settings_offset            0x9c
 #define psx_gpu_current_texture_mask_offset               0xa0
-#define psx_gpu_viewport_mask_offset                      0xa4
 #define psx_gpu_dirty_textures_4bpp_mask_offset           0xa8
 #define psx_gpu_dirty_textures_8bpp_mask_offset           0xac
 #define psx_gpu_dirty_textures_8bpp_alternate_mask_offset 0xb0
 #define psx_gpu_triangle_color_offset                     0xb4
 #define psx_gpu_dither_table_offset                       0xb8
-#define psx_gpu_uvrgb_phase_offset                        0xc8
-#define psx_gpu_render_block_handler_offset               0xcc
-#define psx_gpu_texture_page_ptr_offset                   0xd0
-#define psx_gpu_texture_page_base_offset                  0xd4
-#define psx_gpu_clut_ptr_offset                           0xd8
-#define psx_gpu_vram_ptr_offset                           0xdc
-#define psx_gpu_vram_out_ptr_offset                       0xe0
-#define psx_gpu_render_state_base_offset                  0xe4
-#define psx_gpu_render_state_offset                       0xe6
+#define psx_gpu_texture_page_ptr_offset                   0xcc
+#define psx_gpu_texture_page_base_offset                  0xd0
+#define psx_gpu_clut_ptr_offset                           0xd4
+#define psx_gpu_vram_ptr_offset                           0xd8
+#define psx_gpu_vram_out_ptr_offset                       0xdc
+#define psx_gpu_uvrgb_phase_offset                        0xe0
 #define psx_gpu_num_spans_offset                          0xe8
 #define psx_gpu_num_blocks_offset                         0xea
 #define psx_gpu_viewport_start_x_offset                   0xec
@@ -38,26 +32,13 @@
 #define psx_gpu_viewport_end_y_offset                     0xf2
 #define psx_gpu_mask_msb_offset                           0xf4
 #define psx_gpu_triangle_winding_offset                   0xf6
-#define psx_gpu_display_area_draw_enable_offset           0xf7
 #define psx_gpu_current_texture_page_offset               0xf8
-#define psx_gpu_last_8bpp_texture_page_offset             0xf9
 #define psx_gpu_texture_mask_width_offset                 0xfa
 #define psx_gpu_texture_mask_height_offset                0xfb
-#define psx_gpu_texture_window_x_offset                   0xfc
-#define psx_gpu_texture_window_y_offset                   0xfd
-#define psx_gpu_primitive_type_offset                     0xfe
-#define psx_gpu_render_mode_offset                        0xff
-#define psx_gpu_offset_x_offset                           0x100
-#define psx_gpu_offset_y_offset                           0x102
-#define psx_gpu_clut_settings_offset                      0x104
-#define psx_gpu_texture_settings_offset                   0x106
 #define psx_gpu_reciprocal_table_ptr_offset               0x108
 #define psx_gpu_blocks_offset                             0x200
 #define psx_gpu_span_uvrg_offset_offset                   0x2200
 #define psx_gpu_span_edge_data_offset                     0x4200
 #define psx_gpu_span_b_offset_offset                      0x5200
-#define psx_gpu_texture_4bpp_cache_offset                 0x5a00
-#define psx_gpu_texture_8bpp_even_cache_offset            0x205a00
-#define psx_gpu_texture_8bpp_odd_cache_offset             0x305a00
 
 #endif /* __P_PSX_GPU_OFFSETS_H__ */

plugins/gpu_neon/psx_gpu/psx_gpu_offsets_update.c

@@ -4,7 +4,7 @@
 #include "common.h"
 
 #define WRITE_OFFSET(f, member) \
-	fprintf(f, "#define %-50s0x%x\n", \
+	fprintf(f, "#define %-50s0x%zx\n", \
 		"psx_gpu_" #member "_offset", \
 		offsetof(psx_gpu_struct, member));
 
@@ -22,8 +22,10 @@ int main()
 		perror("fopen");
 		return 1;
 	}
+	fputs("#ifndef __P_PSX_GPU_OFFSETS_H__\n", f);
+	fputs("#define __P_PSX_GPU_OFFSETS_H__\n\n", f);
 
-	WRITE_OFFSET(f, test_mask);
+	//WRITE_OFFSET(f, test_mask);
 	WRITE_OFFSET(f, uvrg);
 	WRITE_OFFSET(f, uvrg_dx);
 	WRITE_OFFSET(f, uvrg_dy);
@@ -35,23 +37,23 @@ int main()
 	WRITE_OFFSET(f, b);
 	WRITE_OFFSET(f, b_dy);
 	WRITE_OFFSET(f, triangle_area);
-	WRITE_OFFSET(f, texture_window_settings);
+	//WRITE_OFFSET(f, texture_window_settings);
 	WRITE_OFFSET(f, current_texture_mask);
-	WRITE_OFFSET(f, viewport_mask);
+	//WRITE_OFFSET(f, viewport_mask);
 	WRITE_OFFSET(f, dirty_textures_4bpp_mask);
 	WRITE_OFFSET(f, dirty_textures_8bpp_mask);
 	WRITE_OFFSET(f, dirty_textures_8bpp_alternate_mask);
 	WRITE_OFFSET(f, triangle_color);
 	WRITE_OFFSET(f, dither_table);
-	WRITE_OFFSET(f, uvrgb_phase);
-	WRITE_OFFSET(f, render_block_handler);
+	//WRITE_OFFSET(f, render_block_handler);
 	WRITE_OFFSET(f, texture_page_ptr);
 	WRITE_OFFSET(f, texture_page_base);
 	WRITE_OFFSET(f, clut_ptr);
 	WRITE_OFFSET(f, vram_ptr);
 	WRITE_OFFSET(f, vram_out_ptr);
-	WRITE_OFFSET(f, render_state_base);
-	WRITE_OFFSET(f, render_state);
+	WRITE_OFFSET(f, uvrgb_phase);
+	//WRITE_OFFSET(f, render_state_base);
+	//WRITE_OFFSET(f, render_state);
 	WRITE_OFFSET(f, num_spans);
 	WRITE_OFFSET(f, num_blocks);
 	WRITE_OFFSET(f, viewport_start_x);
@@ -60,27 +62,29 @@ int main()
 	WRITE_OFFSET(f, viewport_end_y);
 	WRITE_OFFSET(f, mask_msb);
 	WRITE_OFFSET(f, triangle_winding);
-	WRITE_OFFSET(f, display_area_draw_enable);
+	//WRITE_OFFSET(f, display_area_draw_enable);
 	WRITE_OFFSET(f, current_texture_page);
-	WRITE_OFFSET(f, last_8bpp_texture_page);
+	//WRITE_OFFSET(f, last_8bpp_texture_page);
 	WRITE_OFFSET(f, texture_mask_width);
 	WRITE_OFFSET(f, texture_mask_height);
-	WRITE_OFFSET(f, texture_window_x);
-	WRITE_OFFSET(f, texture_window_y);
-	WRITE_OFFSET(f, primitive_type);
-	WRITE_OFFSET(f, render_mode);
-	WRITE_OFFSET(f, offset_x);
-	WRITE_OFFSET(f, offset_y);
-	WRITE_OFFSET(f, clut_settings);
-	WRITE_OFFSET(f, texture_settings);
+	//WRITE_OFFSET(f, texture_window_x);
+	//WRITE_OFFSET(f, texture_window_y);
+	//WRITE_OFFSET(f, primitive_type);
+	//WRITE_OFFSET(f, render_mode);
+	//WRITE_OFFSET(f, offset_x);
+	//WRITE_OFFSET(f, offset_y);
+	//WRITE_OFFSET(f, clut_settings);
+	//WRITE_OFFSET(f, texture_settings);
 	WRITE_OFFSET(f, reciprocal_table_ptr);
 	WRITE_OFFSET(f, blocks);
 	WRITE_OFFSET(f, span_uvrg_offset);
 	WRITE_OFFSET(f, span_edge_data);
 	WRITE_OFFSET(f, span_b_offset);
-	WRITE_OFFSET(f, texture_4bpp_cache);
-	WRITE_OFFSET(f, texture_8bpp_even_cache);
-	WRITE_OFFSET(f, texture_8bpp_odd_cache);
+	//WRITE_OFFSET(f, texture_4bpp_cache);
+	//WRITE_OFFSET(f, texture_8bpp_even_cache);
+	//WRITE_OFFSET(f, texture_8bpp_odd_cache);
+
+	fputs("\n#endif /* __P_PSX_GPU_OFFSETS_H__ */\n", f);
 	fclose(f);
 
 	return 0;