Skip to content

Commit

Permalink
optimize normal layers (still broken)
Browse files Browse the repository at this point in the history
  • Loading branch information
@skyfloogle
skyfloogle committed Apr 4, 2024
1 parent 0b9cfef commit 7ec674c
Show file tree
Hide file tree
Showing 2 changed files with 88 additions and 67 deletions.
2 changes: 2 additions & 0 deletions source/3ds/char.v.pica
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
.fvec posscale ; third and fourth elements must be -1 and 1
.fvec palettes[8]
.fvec offset

.constf const2(0.03125, 0.015625, 0.0, 0.0)

Expand All @@ -26,6 +27,7 @@
mov r0, posscale
mov r1, posscale
mad r0.xy, r0.xy, v0.xy, posscale.zz
add r0.xy, offset.xy, r0.xy
add r1.xy, orients[a0.x].xy, r0.xy
mov outpos1, r1
add r1.xy, orients[a0.x].zw, r0.xy
Expand Down
153 changes: 86 additions & 67 deletions source/3ds/video_hard.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,8 @@ static shaderProgram_s sChar;
static s8 uLoc_posscale;
static s8 uLoc_palettes;
static C3D_FVec palettes[8];
static s8 uLoc_offset;
static C3D_FVec char_offset;
static s8 uLoc_bgmap_offsets;
static C3D_FVec bgmap_offsets[2];

Expand Down Expand Up @@ -87,6 +89,7 @@ void video_hard_init() {

uLoc_posscale = shaderInstanceGetUniformLocation(sChar.vertexShader, "posscale");
uLoc_palettes = shaderInstanceGetUniformLocation(sChar.vertexShader, "palettes");
uLoc_offset = shaderInstanceGetUniformLocation(sChar.vertexShader, "offset");

sAffine_dvlb = DVLB_ParseFile((u32 *)affine_shbin, affine_shbin_size);
shaderProgramInit(&sAffine);
Expand Down Expand Up @@ -284,6 +287,7 @@ static int render_affine_cache(int mapid, vertex *vbuf, vertex *vcur, int umin,
// set up cache texture
C3D_FrameDrawOn(cache->target);
C3D_FVUnifSet(GPU_VERTEX_SHADER, uLoc_posscale, 1.0 / (512 / 2), 1.0 / (512 / 2), -1.0, 1.0);
C3D_FVUnifSet(GPU_VERTEX_SHADER, uLoc_offset, 0, 0, 0, 0);
C3D_SetScissor(GPU_SCISSOR_DISABLE, 0, 0, 0, 0);

C3D_AlphaTest(false, GPU_GREATER, 0);
Expand Down Expand Up @@ -472,78 +476,93 @@ void video_hard_render() {

if ((windows[wnd * 16] & 0x3000) == 0) {
// normal world
for (int eye = start_eye; eye < end_eye; eye++) {
if (!(windows[wnd * 16] & (0x8000 >> eye)))
continue;
vertex *vstart = vcur;
vcount = 0;

int left_mx = base_mx - abs(mp);
int right_mx = base_mx + abs(mp);
int left_gx = base_gx - abs(gp);
int right_gx = base_gx + abs(gp);

u16 *tilemap = (u16 *)(V810_DISPLAY_RAM.pmemory + 0x20000);
int tsx = left_mx >> 3;
int ty = my >> 3;
int mapsx = tsx >> 6;
int mapy = ty >> 6;
tsx &= 63;
ty &= 63;
if (!over) {
mapsx &= scx - 1;
mapy &= scy - 1;
}
bool over_visible = !over || tileVisible[over_tile];

for (int y = gy - (my & 7); y < gy + h; y += 8) {
if (y >= 224) break;
if (y > -8 && (over_visible || (mapy & (scy - 1)) == mapy)) {
int tx = tsx;
int mapx = mapsx;
int current_map = mapid + scx * mapy + mapx;
for (int x = 0; x < 512; x += 8) {
if (x + left_gx >= 384) break;
bool use_over = over && ((mapx & (scx - 1)) != mapx || (mapy & (scy - 1)) != mapy);
uint16_t tile = tilemap[use_over ? over_tile : (64 * 64) * current_map + 64 * ty + tx];
if (++tx >= 64) {
tx = 0;
if ((++mapx & (scx - 1)) == 0 && !over) mapx = 0;
current_map = mapid + scx * mapy + mapx;
}
uint16_t tileid = tile & 0x07ff;
if (!tileVisible[tileid]) continue;
short u = (tileid % 32);
short v = (tileid / 32);

if (vcur >= vbuf + VBUF_SIZE) {
dprintf(0, "VBUF OVERRUN!\n");
break;
}

int gx = base_gx;
int mx = base_mx;
if (eye == 0) {
gx -= gp;
mx -= mp;
} else {
gx += gp;
mx += mp;
vcur->x = x;
vcur->y = y;
vcur->u = u;
vcur->v = v;
vcur->palette = tile >> 14;
vcur++->orient = (tile >> 12) & 3;

vcount++;
}
}
vcount = 0;

C3D_SetScissor(GPU_SCISSOR_NORMAL, gx >= 0 ? gx : 0, (gy >= 0 ? gy : 0) + 256 * eye, gx + w, (gy + h < 256 ? gy + h : 256) + 256 * eye);
u16 *tilemap = (u16 *)(V810_DISPLAY_RAM.pmemory + 0x20000);
int tsx = mx >> 3;
int ty = my >> 3;
int mapsx = tsx >> 6;
int mapy = ty >> 6;
tsx &= 63;
ty &= 63;
if (!over) {
mapsx &= scx - 1;
mapy &= scy - 1;
if (++ty >= 64) {
ty = 0;
if (++mapy >= scy && !over) mapy = 0;
}
bool over_visible = !over || tileVisible[over_tile];

for (int y = gy - (my & 7); y < gy + h; y += 8) {
if (y >= 224) break;
if (y > -8 && (over_visible || (mapy & (scy - 1)) == mapy)) {
int tx = tsx;
int mapx = mapsx;
int current_map = mapid + scx * mapy + mapx;
for (int x = gx - (mx & 7); x < gx + w; x += 8) {
if (x >= 384) continue;
bool use_over = over && ((mapx & (scx - 1)) != mapx || (mapy & (scy - 1)) != mapy);
uint16_t tile = tilemap[use_over ? over_tile : (64 * 64) * current_map + 64 * ty + tx];
if (++tx >= 64) {
tx = 0;
if ((++mapx & (scx - 1)) == 0 && !over) mapx = 0;
current_map = mapid + scx * mapy + mapx;
}
// doing it down here so as to not mess up the above
if (x < -8) continue;
uint16_t tileid = tile & 0x07ff;
if (!tileVisible[tileid]) continue;
short u = (tileid % 32);
short v = (tileid / 32);

if (vcur >= vbuf + VBUF_SIZE) {
dprintf(0, "VBUF OVERRUN!\n");
break;
}

vcur->x = x;
vcur->y = y + 256 * eye;
vcur->u = u;
vcur->v = v;
vcur->palette = tile >> 14;
vcur++->orient = (tile >> 12) & 3;

vcount++;
}
}
if (++ty >= 64) {
ty = 0;
if (++mapy >= scy && !over) mapy = 0;
}

if (vcount != 0) {
for (int eye = start_eye; eye < end_eye; eye++) {
if (!(windows[wnd * 16] & (0x8000 >> eye)))
continue;
int gx = base_gx;
int mx = base_mx;
if (eye == 0) {
gx -= gp;
mx -= mp;
} else {
gx += gp;
mx += mp;
}

int offset_x = gx - (mx & 7);
offset_x -= (mx & ~7) - (left_mx & ~7);

C3D_FVUnifSet(GPU_VERTEX_SHADER, uLoc_offset,
offset_x / 256.0,
eye, 0, 0);

C3D_SetScissor(GPU_SCISSOR_NORMAL, gx >= 0 ? gx : 0, (gy >= 0 ? gy : 0) + 256 * eye, gx + w, (gy + h < 256 ? gy + h : 256) + 256 * eye);

C3D_DrawArrays(GPU_GEOMETRY_PRIM, vstart - vbuf, vcount);
}
DRAW_VBUF;
}
} else {
// hbias or affine world
Expand Down

0 comments on commit 7ec674c

Please sign in to comment.