Skip to content

Commit

Permalink
reuse cache when redrawing same map
Browse files Browse the repository at this point in the history
  • Loading branch information
@skyfloogle
skyfloogle committed Dec 8, 2023
1 parent 4db6459 commit d16521f
Showing 1 changed file with 78 additions and 70 deletions.
148 changes: 78 additions & 70 deletions source/3ds/video.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -286,7 +286,9 @@ void sceneRender() {
u16 *windows = (u16 *)(V810_DISPLAY_RAM.pmemory + 0x3d800);

uint8_t object_group_id = 3;
int cache_id = 0;

int cached_backgrounds[AFFINE_CACHE_SIZE];
for (int i = 0; i < AFFINE_CACHE_SIZE; i++) cached_backgrounds[i] = -1;

for (int8_t wnd = 31; wnd >= 0; wnd--) {
if (windows[wnd * 16] & 0x40)
Expand Down Expand Up @@ -393,75 +395,84 @@ void sceneRender() {
} else {
// hbias or affine world
for (uint8_t sub_bg = 0; sub_bg < scx * scy; sub_bg++) {
int cache_y1, cache_y2;
if ((windows[wnd * 16] & 0x3000) == 0x1000) {
cache_y1 = (sub_bg & (scy - 1)) == 0 ? my & ~7 : 0;
cache_y2 = (sub_bg & (scy - 1)) == scy - 1 ? my + h : 64 * 8;
} else {
cache_y1 = 0;
cache_y2 = 64 * 8;
}
// first, render a cache
// set up cache vertices

u16 *tilemap = (u16 *)(V810_DISPLAY_RAM.pmemory + 0x20000 + 8192 * (mapid + sub_bg)) + 64 * (cache_y1 >> 3);
for (int y = cache_y1; y < cache_y2; y += 8) {
for (int x = 0; x < 64 * 8; x += 8) {
uint16_t tile = *tilemap++;
uint16_t tileid = tile & 0x07ff;
if (!tileVisible[tileid]) continue;
bool hflip = (tile & 0x2000) != 0;
bool vflip = (tile & 0x1000) != 0;
short u = (tileid % 32) * 8;
short v = (tileid / 32) * 8;

vcur->x1 = x + 8 * hflip;
vcur->y1 = y + 8 * vflip;
vcur->x2 = x + 8 * !hflip;
vcur->y2 = y + 8 * !vflip;
vcur->u = u;
vcur->v = v;
vcur++->palette = tile >> 14;

vcount++;
int cache_id = (mapid + sub_bg) % AFFINE_CACHE_SIZE;
if (cached_backgrounds[cache_id] != mapid + sub_bg) {
cached_backgrounds[cache_id] = mapid + sub_bg;
int cache_y1, cache_y2;
if ((windows[wnd * 16] & 0x3000) == 0x1000) {
// with this caching thing, scanning only part of it isn't really viable
/*
cache_y1 = (sub_bg & (scy - 1)) == 0 ? my & ~7 : 0;
cache_y2 = (sub_bg & (scy - 1)) == scy - 1 ? my + h : 64 * 8;
*/
cache_y1 = 0;
cache_y2 = 64 * 8;
} else {
cache_y1 = 0;
cache_y2 = 64 * 8;
}
}
if (vcount == 0) {
// bail
continue;
}
if (vcur - vbuf > VBUF_SIZE) printf("VBUF OVERRUN - %i/%i\n", vcur - vbuf, VBUF_SIZE);

// set up cache texture
C3D_FrameDrawOn(tileMapCacheTarget[cache_id]);
C3D_FVUnifSet(GPU_VERTEX_SHADER, uLoc_posscale, 1.0 / (512 / 2), 1.0 / (512 / 2), -1.0, 1.0);
C3D_SetScissor(GPU_SCISSOR_DISABLE, 0, 0, 0, 0);
// first, render a cache
// set up cache vertices

// clear
C3D_BindProgram(&sFinal);
C3D_AlphaBlend(GPU_BLEND_ADD, GPU_BLEND_ADD, GPU_ONE, GPU_ZERO, GPU_ONE, GPU_ZERO);
C3D_AlphaTest(false, GPU_GREATER, 0);

C3D_TexEnv *env = C3D_GetTexEnv(0);
C3D_TexEnvInit(env);
C3D_TexEnvColor(env, 0);
C3D_TexEnvSrc(env, C3D_Both, GPU_CONSTANT, 0, 0);
C3D_TexEnvFunc(env, C3D_Both, GPU_REPLACE);
u16 *tilemap = (u16 *)(V810_DISPLAY_RAM.pmemory + 0x20000 + 8192 * (mapid + sub_bg)) + 64 * (cache_y1 >> 3);
for (int y = cache_y1; y < cache_y2; y += 8) {
for (int x = 0; x < 64 * 8; x += 8) {
uint16_t tile = *tilemap++;
uint16_t tileid = tile & 0x07ff;
if (!tileVisible[tileid]) continue;
bool hflip = (tile & 0x2000) != 0;
bool vflip = (tile & 0x1000) != 0;
short u = (tileid % 32) * 8;
short v = (tileid / 32) * 8;

C3D_ImmDrawBegin(GPU_GEOMETRY_PRIM);
C3D_ImmSendAttrib(1, 1, -1, 1);
C3D_ImmSendAttrib(0, 0, 0, 0);
C3D_ImmSendAttrib(-1, -1, -1, 1);
C3D_ImmSendAttrib(1, 1, 0, 0);
C3D_ImmDrawEnd();
vcur->x1 = x + 8 * hflip;
vcur->y1 = y + 8 * vflip;
vcur->x2 = x + 8 * !hflip;
vcur->y2 = y + 8 * !vflip;
vcur->u = u;
vcur->v = v;
vcur++->palette = tile >> 14;

// reset and draw cache
C3D_BindProgram(&sChar);
C3D_AlphaBlend(GPU_BLEND_ADD, GPU_BLEND_ADD, GPU_SRC_ALPHA, GPU_ONE_MINUS_SRC_ALPHA, GPU_SRC_ALPHA, GPU_ONE_MINUS_SRC_ALPHA);
C3D_AlphaTest(true, GPU_GREATER, 0);
setRegularTexEnv();

C3D_DrawArrays(GPU_GEOMETRY_PRIM, vcur - vbuf - vcount, vcount);
vcount++;
}
}
if (vcount == 0) {
// bail
continue;
}
if (vcur - vbuf > VBUF_SIZE) printf("VBUF OVERRUN - %i/%i\n", vcur - vbuf, VBUF_SIZE);

// set up cache texture
C3D_FrameDrawOn(tileMapCacheTarget[cache_id]);
C3D_FVUnifSet(GPU_VERTEX_SHADER, uLoc_posscale, 1.0 / (512 / 2), 1.0 / (512 / 2), -1.0, 1.0);
C3D_SetScissor(GPU_SCISSOR_DISABLE, 0, 0, 0, 0);

// clear
C3D_BindProgram(&sFinal);
C3D_AlphaBlend(GPU_BLEND_ADD, GPU_BLEND_ADD, GPU_ONE, GPU_ZERO, GPU_ONE, GPU_ZERO);
C3D_AlphaTest(false, GPU_GREATER, 0);

C3D_TexEnv *env = C3D_GetTexEnv(0);
C3D_TexEnvInit(env);
C3D_TexEnvColor(env, 0);
C3D_TexEnvSrc(env, C3D_Both, GPU_CONSTANT, 0, 0);
C3D_TexEnvFunc(env, C3D_Both, GPU_REPLACE);

C3D_ImmDrawBegin(GPU_GEOMETRY_PRIM);
C3D_ImmSendAttrib(1, 1, -1, 1);
C3D_ImmSendAttrib(0, 0, 0, 0);
C3D_ImmSendAttrib(-1, -1, -1, 1);
C3D_ImmSendAttrib(1, 1, 0, 0);
C3D_ImmDrawEnd();

// reset and draw cache
C3D_BindProgram(&sChar);
C3D_AlphaBlend(GPU_BLEND_ADD, GPU_BLEND_ADD, GPU_SRC_ALPHA, GPU_ONE_MINUS_SRC_ALPHA, GPU_SRC_ALPHA, GPU_ONE_MINUS_SRC_ALPHA);
C3D_AlphaTest(true, GPU_GREATER, 0);
setRegularTexEnv();

C3D_DrawArrays(GPU_GEOMETRY_PRIM, vcur - vbuf - vcount, vcount);
}

// set up wrapping for affine map
C3D_TexSetWrap(&tileMapCache[cache_id],
Expand Down Expand Up @@ -489,7 +500,7 @@ void sceneRender() {
BufInfo_Init(bufInfo);
BufInfo_Add(bufInfo, avbuf, sizeof(avertex), 3, 0x210);

env = C3D_GetTexEnv(0);
C3D_TexEnv *env = C3D_GetTexEnv(0);
C3D_TexEnvInit(env);
C3D_TexEnvSrc(env, C3D_Both, GPU_TEXTURE0, 0, 0);
C3D_TexEnvFunc(env, C3D_Both, GPU_REPLACE);
Expand Down Expand Up @@ -580,9 +591,6 @@ void sceneRender() {
C3D_TexBind(0, &tileTexture);
setRegularDrawing();
vcount = 0;
if (++cache_id == AFFINE_CACHE_SIZE) {
cache_id = 0;
}
}
}
} else {
Expand Down

0 comments on commit d16521f

Please sign in to comment.