An initial take on bsnes CPU filters.

Makefile

@@ -84,11 +84,7 @@ ifeq ($(HAVE_XML), 1)
 endif
 
 ifeq ($(HAVE_FILTER), 1)
-   OBJ += hqflt/hq.o
-   OBJ += hqflt/grayscale.o
-   OBJ += hqflt/bleed.o
-   OBJ += hqflt/ntsc.o
-   OBJ += hqflt/snes_ntsc/snes_ntsc.o
+   LIBS += -ldl
 endif
 
 ifeq ($(HAVE_FREETYPE), 1)

audio/sdl.c

@@ -51,19 +51,6 @@ static void sdl_audio_cb(void *data, Uint8 *stream, int len)
    memset(stream + write_size, 0, len - write_size);
 }
 
-// Interesting hack from http://www-graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2
-static inline uint32_t next_pow2(uint32_t v)
-{
-   v--;
-   v |= v >> 1;
-   v |= v >> 2;
-   v |= v >> 4;
-   v |= v >> 8;
-   v |= v >> 16;
-   v++;
-   return v;
-}
-
 static inline int find_num_frames(int rate, int latency)
 {
    int frames = (rate * latency) / 1000;

audio/xaudio.c

@@ -67,19 +67,6 @@ static bool init_lib(void)
    return true;
 }
 
-// Interesting hack from http://www-graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2
-static inline uint32_t next_pow2(uint32_t v)
-{
-   v--;
-   v |= v >> 1;
-   v |= v >> 2;
-   v |= v >> 4;
-   v |= v >> 8;
-   v |= v >> 16;
-   v++;
-   return v;
-}
-
 static void* __xa_init(const char* device, int rate, int latency)
 {
    if (!init_lib())

driver.c

@@ -20,8 +20,8 @@
 #include "general.h"
 #include <stdio.h>
 #include <string.h>
-#include "hqflt/filters.h"
 #include <assert.h>
+#include <math.h>
 
 #ifdef HAVE_CONFIG_H
 #include "config.h"
@@ -212,28 +212,90 @@ void uninit_audio(void)
    free(g_extern.audio_data.conv_outsamples); g_extern.audio_data.conv_outsamples = NULL;
 }
 
+#ifdef HAVE_FILTER
+static void init_filter(void)
+{
+   if (g_extern.filter.active)
+      return;
+   if (*g_settings.video.filter_path == '\0')
+      return;
+
+   g_extern.filter.lib = dylib_load(g_settings.video.filter_path);
+   if (!g_extern.filter.lib)
+   {
+      SSNES_ERR("Failed to load filter \"%s\"\n", g_settings.video.filter_path);
+      return;
+   }
+
+   g_extern.filter.psize = dylib_proc(g_extern.filter.lib, "video_size");
+   g_extern.filter.prender = dylib_proc(g_extern.filter.lib, "video_render");
+   if (!g_extern.filter.psize || !g_extern.filter.prender)
+   {
+      SSNES_ERR("Failed to find functions in filter...\n");
+      dylib_close(g_extern.filter.lib);
+      g_extern.filter.lib = NULL;
+      return;
+   }
+
+   g_extern.filter.active = true;
+
+   unsigned width = 512;
+   unsigned height = 512;
+   g_extern.filter.psize(&width, &height);
+
+   unsigned pow2_x = next_pow2(ceil(width));
+   unsigned pow2_y = next_pow2(ceil(height));
+   unsigned maxsize = pow2_x > pow2_y ? pow2_x : pow2_y; 
+   g_extern.filter.scale = maxsize / 256;
+
+   g_extern.filter.buffer = malloc(256 * 256 * g_extern.filter.scale * g_extern.filter.scale * sizeof(uint32_t));
+   g_extern.filter.pitch = 256 * g_extern.filter.scale * sizeof(uint32_t);
+   assert(g_extern.filter.buffer);
+
+   g_extern.filter.colormap = malloc(32768 * sizeof(uint32_t));
+   assert(g_extern.filter.colormap);
+
+   // TODO: Taken from bSNES source. WTF does this do?
+   for (int i = 0; i < 32768; i++)
+   {
+      unsigned r = (i >> 10) & 31;
+      unsigned g = (i >>  5) & 31;
+      unsigned b = (i >>  0) & 31;
+
+      r = (r << 3) | (r >> 2);
+      g = (g << 3) | (g >> 2);
+      b = (b << 3) | (b >> 2);
+      g_extern.filter.colormap[i] = (r << 16) | (g << 8) | (b << 0);
+   }
+}
+#endif
+
+static void deinit_filter(void)
+{
+   if (g_extern.filter.active)
+   {
+      g_extern.filter.active = false;
+      dylib_close(g_extern.filter.lib);
+      g_extern.filter.lib = NULL;
+      free(g_extern.filter.buffer);
+      free(g_extern.filter.colormap);
+   }
+}
+
 void init_video_input(void)
 {
-   int scale = 2;
+#ifdef HAVE_FILTER
+   init_filter();
+#endif
+
+   // We use at least 512x512 textures to accomodate for hi-res games.
+   unsigned scale = 2;
 
    find_video_driver();
    find_input_driver();
 
-   // We multiply scales with 2 to allow for hi-res games.
-#if HAVE_FILTER
-   switch (g_settings.video.filter)
-   {
-      case FILTER_HQ2X:
-         scale = 4;
-         break;
-      case FILTER_HQ4X:
-      case FILTER_NTSC:
-         scale = 8;
-         break;
-      default:
-         break;
-   }
-#endif
+   if (g_extern.filter.active)
+      scale = g_extern.filter.scale;
 
    video_info_t video = {
       .width = (g_settings.video.fullscreen) ? g_settings.video.fullscreen_x : (296 * g_settings.video.xscale),

dynamic.c

@@ -25,31 +25,24 @@
 
 #include <libsnes.hpp>
 
-#ifdef HAVE_DYNAMIC
-
 #ifdef _WIN32
 #include <windows.h>
-#define SYM(x) do { \
-   p##x = ((void*)GetProcAddress(lib_handle, #x)); \
-   if (p##x == NULL) { SSNES_ERR("Failed to load symbol: \"%s\"\n", #x); exit(1); } \
-} while(0)
 #else
 #include <dlfcn.h>
-#define SYM(x) do { \
-   p##x = dlsym(lib_handle, #x); \
-   if (p##x == NULL) { SSNES_ERR("Failed to load symbol: \"%s\"\n", #x); exit(1); } \
-} while(0)
-#endif // _WIN32
-#endif // HAVE_DYNAMIC
+#endif
 
 #ifdef HAVE_DYNAMIC
-#ifdef _WIN32
-static HMODULE lib_handle;
-#else
-static void *lib_handle = NULL;
-#endif
+#define DLSYM(lib, x) dylib_proc(lib, #x)
+
+#define SYM(x) do { \
+   p##x = DLSYM(lib_handle, x); \
+   if (p##x == NULL) { SSNES_ERR("Failed to load symbol: \"%s\"\n", #x); exit(1); } \
+} while(0)
+
+static dylib_t lib_handle = NULL;
 #endif
 
+
 void (*psnes_init)(void);
 
 void (*psnes_set_video_refresh)(snes_video_refresh_t);
@@ -97,11 +90,7 @@ void (*psnes_term)(void);
 static void load_dynamic(void)
 {
    SSNES_LOG("Loading dynamic libsnes from: \"%s\"\n", g_settings.libsnes);
-#ifdef _WIN32
-   lib_handle = LoadLibrary(g_settings.libsnes);
-#else
-   lib_handle = dlopen(g_settings.libsnes, RTLD_LAZY);
-#endif
+   lib_handle = dylib_load(g_settings.libsnes);
    if (!lib_handle)
    {
       SSNES_ERR("Failed to open dynamic library: \"%s\"\n", g_settings.libsnes);
@@ -186,12 +175,35 @@ void uninit_dlsym(void)
 {
 #ifdef HAVE_DYNAMIC
    if (lib_handle)
-   {
-#ifdef _WIN32
-      FreeLibrary(lib_handle);
-#else
-      dlclose(lib_handle);
-#endif
-   }
+      dylib_close(lib_handle);
 #endif
 }
+
+// Platform independent dylib loading.
+dylib_t dylib_load(const char *path)
+{
+#ifdef _WIN32
+   return LoadLibrary(path);
+#else
+   return dlopen(path, RTLD_LAZY);
+#endif
+}
+
+void* dylib_proc(dylib_t lib, const char *proc)
+{
+#ifdef _WIN32
+   return GetProcAddress(lib, proc);
+#else
+   return dlsym(lib, proc);
+#endif
+}
+
+void dylib_close(dylib_t lib)
+{
+#ifdef _WIN32
+   FreeLibrary(lib);
+#else
+   dlclose(lib);
+#endif
+}
+

dynamic.h

@@ -24,6 +24,16 @@
 void init_dlsym(void);
 void uninit_dlsym(void);
 
+#ifdef _WIN32
+typedef HMODULE dylib_t;
+#else
+typedef void* dylib_t;
+#endif
+
+dylib_t dylib_load(const char *path);
+void dylib_close(dylib_t lib);
+void* dylib_proc(dylib_t lib, const char *proc);
+
 extern void (*psnes_init)(void);
 
 extern void (*psnes_set_video_refresh)(snes_video_refresh_t);

general.h

@@ -28,6 +28,7 @@
 #include "movie.h"
 #include "autosave.h"
 #include "netplay.h"
+#include "dynamic.h"
 
 #ifdef HAVE_CONFIG_H
 #include "config.h"
@@ -184,6 +185,20 @@ struct global
       int16_t *conv_outsamples;
    } audio_data;
 
+   struct
+   {
+      bool active;
+      uint32_t *buffer;
+      uint32_t *colormap;
+      unsigned pitch;
+      dylib_t lib;
+      unsigned scale;
+
+      void (*psize)(unsigned *width, unsigned *height);
+      void (*prender)(uint32_t *colormap, uint32_t *output, unsigned outpitch,
+            const uint16_t *input, unsigned pitch, unsigned width, unsigned height);
+   } filter;
+
    msg_queue_t *msg_queue;
 
    // Rewind support.
@@ -240,4 +255,18 @@ extern struct global g_extern;
       fflush(stderr); \
    } while(0)
 
+static inline uint32_t next_pow2(uint32_t v)
+{
+   v--;
+   v |= v >> 1;
+   v |= v >> 2;
+   v |= v >> 4;
+   v |= v >> 8;
+   v |= v >> 16;
+   v++;
+   return v;
+}
+
 #endif
+
+

gfx/gl.c

@@ -141,6 +141,10 @@ typedef struct gl
    GLfloat tex_coords[8];
    GLfloat fbo_tex_coords[8];
 
+   GLenum texture_type; // XBGR1555 or RGBA
+   GLenum texture_fmt;
+   unsigned base_size; // 2 or 4
+
 #ifdef HAVE_FREETYPE
    font_renderer_t *font;
    GLuint font_tex;
@@ -263,18 +267,6 @@ static inline void gl_init_font(gl_t *gl, const char *font_path, unsigned font_s
 #endif
 }
 
-static inline unsigned next_pow_2(unsigned v)
-{
-   v--;
-   v |= v >> 1;
-   v |= v >> 2;
-   v |= v >> 4;
-   v |= v >> 8;
-   v |= v >> 16;
-   v++;
-   return v;
-}
-
 static void gl_init_fbo(gl_t *gl, unsigned width, unsigned height)
 {
    if (!g_settings.video.render_to_texture)
@@ -288,8 +280,8 @@ static void gl_init_fbo(gl_t *gl, unsigned width, unsigned height)
 
    float scale_x = g_settings.video.fbo_scale_x;
    float scale_y = g_settings.video.fbo_scale_y;
-   unsigned xscale = next_pow_2(ceil(scale_x));
-   unsigned yscale = next_pow_2(ceil(scale_y));
+   unsigned xscale = next_pow2(ceil(scale_x));
+   unsigned yscale = next_pow2(ceil(scale_y));
    SSNES_LOG("Internal FBO scale: (%u, %u)\n", xscale, yscale);
 
    gl->fbo_width = width * xscale;
@@ -485,7 +477,7 @@ static void show_fps(void)
    frames++;
 }
 
-static bool gl_frame(void *data, const uint16_t* frame, unsigned width, unsigned height, unsigned pitch, const char *msg)
+static bool gl_frame(void *data, const void* frame, unsigned width, unsigned height, unsigned pitch, const char *msg)
 {
    gl_t *gl = data;
 
@@ -517,10 +509,11 @@ static bool gl_frame(void *data, const uint16_t* frame, unsigned width, unsigned
       gl->last_height = height;
       glPixelStorei(GL_UNPACK_ALIGNMENT, get_alignment(pitch));
       glPixelStorei(GL_UNPACK_ROW_LENGTH, gl->tex_w);
-      uint8_t *tmp = calloc(1, gl->tex_w * gl->tex_h * sizeof(uint16_t));
+
+      void *tmp = calloc(1, gl->tex_w * gl->tex_h * gl->base_size);
       glTexSubImage2D(GL_TEXTURE_2D,
-            0, 0, 0, gl->tex_w, gl->tex_h, GL_BGRA,
-            GL_UNSIGNED_SHORT_1_5_5_5_REV, tmp);
+            0, 0, 0, gl->tex_w, gl->tex_h, gl->texture_type,
+            gl->texture_fmt, tmp);
       free(tmp);
 
       gl->tex_coords[0] = 0;
@@ -534,10 +527,10 @@ static bool gl_frame(void *data, const uint16_t* frame, unsigned width, unsigned
    }
 
 
-   glPixelStorei(GL_UNPACK_ROW_LENGTH, pitch >> 1);
+   glPixelStorei(GL_UNPACK_ROW_LENGTH, pitch / gl->base_size);
    glTexSubImage2D(GL_TEXTURE_2D,
-         0, 0, 0, width, height, GL_BGRA,
-         GL_UNSIGNED_SHORT_1_5_5_5_REV, frame);
+         0, 0, 0, width, height, gl->texture_type,
+         gl->texture_fmt, frame);
 
    glFlush();
    glDrawArrays(GL_QUADS, 0, 4);
@@ -681,6 +674,10 @@ static void* gl_init(video_info_t *video, const input_driver_t **input, void **i
    else
       gl->tex_filter = GL_NEAREST;
 
+   gl->texture_type = video->rgb32 ? GL_RGBA : GL_BGRA;
+   gl->texture_fmt = video->rgb32 ? GL_UNSIGNED_INT_8_8_8_8 : GL_UNSIGNED_SHORT_1_5_5_5_REV;
+   gl->base_size = video->rgb32 ? sizeof(uint32_t) : sizeof(uint16_t);
+
    glEnable(GL_TEXTURE_2D);
    glDisable(GL_DITHER);
    glDisable(GL_DEPTH_TEST);
@@ -710,11 +707,13 @@ static void* gl_init(video_info_t *video, const input_driver_t **input, void **i
 
    gl->tex_w = 256 * video->input_scale;
    gl->tex_h = 256 * video->input_scale;
-   uint8_t *tmp = calloc(1, gl->tex_w * gl->tex_h * sizeof(uint32_t));
+
+   void *tmp = calloc(1, gl->tex_w * gl->tex_h * gl->base_size);
    glTexImage2D(GL_TEXTURE_2D,
-         0, GL_RGBA, gl->tex_w, gl->tex_h, 0, GL_BGRA,
-         GL_UNSIGNED_SHORT_1_5_5_5_REV, tmp);
+         0, GL_RGBA, gl->tex_w, gl->tex_h, 0, gl->texture_type,
+         gl->texture_fmt, tmp);
    free(tmp);
+
    gl->last_width = gl->tex_w;
    gl->last_height = gl->tex_h;
 

qb/config.params.sh

@@ -10,7 +10,7 @@ PACKAGE_VERSION=0.3.1
 add_command_line_enable DYNAMIC "Enable dynamic loading of libsnes library." no
 add_command_line_string LIBSNES "libsnes library used" "-lsnes"
 add_command_line_enable FFMPEG "Enable FFmpeg recording support" no
-add_command_line_enable FILTER "Enable CPU filter support" no
+add_command_line_enable FILTER "Disable CPU filter support" yes
 add_command_line_enable SRC "Enable libsamplerate support" no
 add_command_line_enable CONFIGFILE "Disable support for config file" yes
 add_command_line_enable CG "Enable Cg shader support" auto

settings.c

@@ -19,7 +19,6 @@
 #include "conf/config_file.h"
 #include <assert.h>
 #include <string.h>
-#include "hqflt/filters.h"
 #include "config.def.h"
 
 #ifdef HAVE_CONFIG_H

ssnes.c

@@ -25,7 +25,6 @@
 #include <time.h>
 #include "driver.h"
 #include "file.h"
-#include "filters.h"
 #include "general.h"
 #include "dynamic.h"
 #include "record/ffemu.h"
@@ -100,12 +99,14 @@ static void video_frame(const uint16_t *data, unsigned width, unsigned height)
    const char *msg = msg_queue_pull(g_extern.msg_queue);
 
 #ifdef HAVE_FILTER
-   unsigned owidth = width;
-   unsigned oheight = height;
-   video_filter_size(&owidth, &oheight);
    if (g_extern.filter.active)
    {
-      video_filter_render(g_extern.filter.buffer, g_extern.filter.pitch >> 2, data, (height == 448 || height == 478) ? 512 : 1024, width, height);
+      unsigned owidth = width;
+      unsigned oheight = height;
+      g_extern.filter.psize(&owidth, &oheight);
+
+      g_extern.filter.prender(g_extern.filter.colormap, g_extern.filter.buffer, 
+            g_extern.filter.pitch >> 2, data, (height == 448 || height == 478) ? 512 : 1024, width, height);
       if (!driver.video->frame(driver.video_data, g_extern.filter.buffer, owidth, oheight, g_extern.filter.pitch, msg))
          g_extern.video_active = false;
    }