alex/vello
1
0
Fork 0
mirror of https://github.com/italicsjenga/vello.git synced 2025-01-10 12:41:30 +11:00
Code Issues Packages Projects Releases Wiki Activity

kernel4: separate area from alpha in clip stack

Browse source 
This change prepares for kernel4 to output alpha. No functional changes.

Signed-off-by: Elias Naur <mail@eliasnaur.com>
This commit is contained in:
Elias Naur 2021-03-22 16:13:39 +01:00
parent 22507dea0e
commit ee4429a26f
5 changed files with 18 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
piet-gpu/shader/coarse.comp
View file

@ -418,8 +418,8 @@ void main() {
if (bin_tile_x + tile_x < conf.width_in_tiles && bin_tile_y + tile_y < conf.height_in_tiles) {
Cmd_End_write(cmd_alloc, cmd_ref);
if (num_begin_slots > 0) {
// Write scratch allocation: one word per BeginClip per rasterizer chunk.
uint scratch_size = num_begin_slots * TILE_WIDTH_PX * TILE_HEIGHT_PX * 4;
// Write scratch allocation: one state per BeginClip per rasterizer chunk.
uint scratch_size = num_begin_slots * TILE_WIDTH_PX * TILE_HEIGHT_PX * CLIP_STATE_SIZE * 4;
MallocResult scratch = malloc(scratch_size);
// Ignore scratch.failed; we don't use the allocation and kernel4
// checks for memory overflow before using it.

BIN
piet-gpu/shader/coarse.spv
View file

Binary file not shown.

21
piet-gpu/shader/kernel4.comp
View file

@ -201,25 +201,28 @@ void main() {
cmd_ref.offset += 4 + CmdImage_size;
break;
case Cmd_BeginClip:
uint base_ix = (scratch_alloc.offset >> 2) + clip_depth * TILE_WIDTH_PX * TILE_HEIGHT_PX +
gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y;
uint base_ix = (scratch_alloc.offset >> 2) + CLIP_STATE_SIZE * (clip_depth * TILE_WIDTH_PX * TILE_HEIGHT_PX +
gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y);
for (uint k = 0; k < CHUNK; k++) {
uvec2 offset = chunk_offset(k);
uint state = packsRGB(vec4(rgb[k], clamp(abs(area[k]), 0.0, 1.0)));
write_mem(scratch_alloc, base_ix + offset.x + offset.y * TILE_WIDTH_PX, state);
uint srgb = packsRGB(vec4(rgb[k], 1.0));
float alpha = clamp(abs(area[k]), 0.0, 1.0);
write_mem(scratch_alloc, base_ix + 0 + CLIP_STATE_SIZE * (offset.x + offset.y * TILE_WIDTH_PX), srgb);
write_mem(scratch_alloc, base_ix + 1 + CLIP_STATE_SIZE * (offset.x + offset.y * TILE_WIDTH_PX), floatBitsToUint(alpha));
}
clip_depth++;
cmd_ref.offset += 4;
break;
case Cmd_EndClip:
clip_depth--;
base_ix = (scratch_alloc.offset >> 2) + clip_depth * TILE_WIDTH_PX * TILE_HEIGHT_PX +
gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y;
base_ix = (scratch_alloc.offset >> 2) + CLIP_STATE_SIZE * (clip_depth * TILE_WIDTH_PX * TILE_HEIGHT_PX +
gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y);
for (uint k = 0; k < CHUNK; k++) {
uvec2 offset = chunk_offset(k);
uint state = read_mem(scratch_alloc, base_ix + offset.x + offset.y * TILE_WIDTH_PX);
vec4 rgba = unpacksRGB(state);
rgb[k] = mix(rgba.rgb, rgb[k], area[k] * rgba.a);
uint srgb = read_mem(scratch_alloc, base_ix + 0 + CLIP_STATE_SIZE * (offset.x + offset.y * TILE_WIDTH_PX));
uint alpha = read_mem(scratch_alloc, base_ix + 1 + CLIP_STATE_SIZE * (offset.x + offset.y * TILE_WIDTH_PX));
vec4 rgba = unpacksRGB(srgb);
rgb[k] = mix(rgba.rgb, rgb[k], area[k] * uintBitsToFloat(alpha));
}
cmd_ref.offset += 4;
break;

BIN
piet-gpu/shader/kernel4.spv
View file

Binary file not shown.

3
piet-gpu/shader/setup.h
View file

@ -44,6 +44,9 @@ struct Config {
#define MODE_NONZERO 0
#define MODE_STROKE 1
// Size of kernel4 clip state, in words.
#define CLIP_STATE_SIZE 2
// fill_mode_from_flags extracts the fill mode from tag flags.
uint fill_mode_from_flags(uint flags) {
return flags & 0x1;