alex/vello
1
0
Fork 0
mirror of https://github.com/italicsjenga/vello.git synced 2025-01-09 20:31:29 +11:00
Code Issues Packages Projects Releases Wiki Activity

Fix blend memory issues

Browse source
This commit is contained in:
Raph Levien 2022-05-20 14:02:24 -07:00
parent 368954a643
commit 9df486ef35
14 changed files with 647 additions and 603 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
piet-gpu/shader/coarse.comp
View file

@ -151,6 +151,7 @@ void main() {
uint part_start_ix = 0;
uint ready_ix = 0;
Alloc scratch_alloc = slice_mem(cmd_alloc, 0, Alloc_size);
cmd_ref.offset += 4;
// Accounting for allocation of blend memory
uint render_blend_depth = 0;
@ -460,7 +461,9 @@ 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 (max_blend_depth > BLEND_STACK_SPLIT) {
// TODO: allocate blend memory and write result
uint scratch_size = max_blend_depth * TILE_WIDTH_PX * TILE_HEIGHT_PX * CLIP_STATE_SIZE * 4;
MallocResult scratch = malloc(scratch_size);
alloc_write(scratch_alloc, scratch_alloc.offset, scratch.alloc);
}
}
}

BIN
piet-gpu/shader/gen/coarse.dxil generated
View file

Binary file not shown.

596
piet-gpu/shader/gen/coarse.hlsl generated
View file

@ -175,9 +175,9 @@ struct Config
static const uint3 gl_WorkGroupSize = uint3(256u, 1u, 1u);
RWByteAddressBuffer _260 : register(u0, space0);
ByteAddressBuffer _1005 : register(t1, space0);
ByteAddressBuffer _1378 : register(t2, space0);
RWByteAddressBuffer _266 : register(u0, space0);
ByteAddressBuffer _1020 : register(t1, space0);
ByteAddressBuffer _1399 : register(t2, space0);
static uint3 gl_WorkGroupID;
static uint3 gl_LocalInvocationID;
@ -200,8 +200,8 @@ groupshared uint sh_tile_count[256];
Alloc slice_mem(Alloc a, uint offset, uint size)
{
Alloc _337 = { a.offset + offset };
return _337;
Alloc _343 = { a.offset + offset };
return _343;
}
bool touch_mem(Alloc alloc, uint offset)
@ -217,7 +217,7 @@ uint read_mem(Alloc alloc, uint offset)
{
return 0u;
}
uint v = _260.Load(offset * 4 + 8);
uint v = _266.Load(offset * 4 + 8);
return v;
}
@ -230,8 +230,8 @@ Alloc new_alloc(uint offset, uint size, bool mem_ok)
BinInstanceRef BinInstance_index(BinInstanceRef ref, uint index)
{
BinInstanceRef _346 = { ref.offset + (index * 4u) };
return _346;
BinInstanceRef _361 = { ref.offset + (index * 4u) };
return _361;
}
BinInstance BinInstance_read(Alloc a, BinInstanceRef ref)
@ -259,8 +259,8 @@ Path Path_read(Alloc a, PathRef ref)
uint raw2 = read_mem(param_4, param_5);
Path s;
s.bbox = uint4(raw0 & 65535u, raw0 >> uint(16), raw1 & 65535u, raw1 >> uint(16));
TileRef _409 = { raw2 };
s.tiles = _409;
TileRef _424 = { raw2 };
s.tiles = _424;
return s;
}
@ -270,11 +270,11 @@ void write_tile_alloc(uint el_ix, Alloc a)
Alloc read_tile_alloc(uint el_ix, bool mem_ok)
{
uint _892;
_260.GetDimensions(_892);
_892 = (_892 - 8) / 4;
uint _907;
_266.GetDimensions(_907);
_907 = (_907 - 8) / 4;
uint param = 0u;
uint param_1 = uint(int(_892) * 4);
uint param_1 = uint(int(_907) * 4);
bool param_2 = mem_ok;
return new_alloc(param, param_1, param_2);
}
@ -288,31 +288,31 @@ Tile Tile_read(Alloc a, TileRef ref)
Alloc param_2 = a;
uint param_3 = ix + 1u;
uint raw1 = read_mem(param_2, param_3);
TileSegRef _434 = { raw0 };
TileSegRef _449 = { raw0 };
Tile s;
s.tile = _434;
s.tile = _449;
s.backdrop = int(raw1);
return s;
}
MallocResult malloc(uint size)
{
uint _266;
_260.InterlockedAdd(0, size, _266);
uint offset = _266;
uint _273;
_260.GetDimensions(_273);
_273 = (_273 - 8) / 4;
uint _272;
_266.InterlockedAdd(0, size, _272);
uint offset = _272;
uint _279;
_266.GetDimensions(_279);
_279 = (_279 - 8) / 4;
MallocResult r;
r.failed = (offset + size) > uint(int(_273) * 4);
r.failed = (offset + size) > uint(int(_279) * 4);
uint param = offset;
uint param_1 = size;
bool param_2 = !r.failed;
r.alloc = new_alloc(param, param_1, param_2);
if (r.failed)
{
uint _295;
_260.InterlockedMax(4, 1u, _295);
uint _301;
_266.InterlockedMax(4, 1u, _301);
return r;
}
return r;
@ -326,7 +326,7 @@ void write_mem(Alloc alloc, uint offset, uint val)
{
return;
}
_260.Store(offset * 4 + 8, val);
_266.Store(offset * 4 + 8, val);
}
void CmdJump_write(Alloc a, CmdJumpRef ref, CmdJump s)
@ -344,9 +344,9 @@ void Cmd_Jump_write(Alloc a, CmdRef ref, CmdJump s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 11u;
write_mem(param, param_1, param_2);
CmdJumpRef _885 = { ref.offset + 4u };
CmdJumpRef _900 = { ref.offset + 4u };
Alloc param_3 = a;
CmdJumpRef param_4 = _885;
CmdJumpRef param_4 = _900;
CmdJump param_5 = s;
CmdJump_write(param_3, param_4, param_5);
}
@ -358,21 +358,21 @@ bool alloc_cmd(inout Alloc cmd_alloc, inout CmdRef cmd_ref, inout uint cmd_limit
return true;
}
uint param = 1024u;
MallocResult _913 = malloc(param);
MallocResult new_cmd = _913;
MallocResult _928 = malloc(param);
MallocResult new_cmd = _928;
if (new_cmd.failed)
{
return false;
}
CmdJump _923 = { new_cmd.alloc.offset };
CmdJump jump = _923;
CmdJump _938 = { new_cmd.alloc.offset };
CmdJump jump = _938;
Alloc param_1 = cmd_alloc;
CmdRef param_2 = cmd_ref;
CmdJump param_3 = jump;
Cmd_Jump_write(param_1, param_2, param_3);
cmd_alloc = new_cmd.alloc;
CmdRef _935 = { cmd_alloc.offset };
cmd_ref = _935;
CmdRef _950 = { cmd_alloc.offset };
cmd_ref = _950;
cmd_limit = (cmd_alloc.offset + 1024u) - 144u;
return true;
}
@ -396,9 +396,9 @@ void Cmd_Fill_write(Alloc a, CmdRef ref, CmdFill s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 1u;
write_mem(param, param_1, param_2);
CmdFillRef _742 = { ref.offset + 4u };
CmdFillRef _757 = { ref.offset + 4u };
Alloc param_3 = a;
CmdFillRef param_4 = _742;
CmdFillRef param_4 = _757;
CmdFill param_5 = s;
CmdFill_write(param_3, param_4, param_5);
}
@ -430,9 +430,9 @@ void Cmd_Stroke_write(Alloc a, CmdRef ref, CmdStroke s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 2u;
write_mem(param, param_1, param_2);
CmdStrokeRef _760 = { ref.offset + 4u };
CmdStrokeRef _775 = { ref.offset + 4u };
Alloc param_3 = a;
CmdStrokeRef param_4 = _760;
CmdStrokeRef param_4 = _775;
CmdStroke param_5 = s;
CmdStroke_write(param_3, param_4, param_5);
}
@ -443,8 +443,8 @@ void write_fill(Alloc alloc, inout CmdRef cmd_ref, Tile tile, float linewidth)
{
if (tile.tile.offset != 0u)
{
CmdFill _958 = { tile.tile.offset, tile.backdrop };
CmdFill cmd_fill = _958;
CmdFill _973 = { tile.tile.offset, tile.backdrop };
CmdFill cmd_fill = _973;
Alloc param = alloc;
CmdRef param_1 = cmd_ref;
CmdFill param_2 = cmd_fill;
@ -461,8 +461,8 @@ void write_fill(Alloc alloc, inout CmdRef cmd_ref, Tile tile, float linewidth)
}
else
{
CmdStroke _988 = { tile.tile.offset, 0.5f * linewidth };
CmdStroke cmd_stroke = _988;
CmdStroke _1003 = { tile.tile.offset, 0.5f * linewidth };
CmdStroke cmd_stroke = _1003;
Alloc param_5 = alloc;
CmdRef param_6 = cmd_ref;
CmdStroke param_7 = cmd_stroke;
@ -486,9 +486,9 @@ void Cmd_Color_write(Alloc a, CmdRef ref, CmdColor s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 5u;
write_mem(param, param_1, param_2);
CmdColorRef _786 = { ref.offset + 4u };
CmdColorRef _801 = { ref.offset + 4u };
Alloc param_3 = a;
CmdColorRef param_4 = _786;
CmdColorRef param_4 = _801;
CmdColor param_5 = s;
CmdColor_write(param_3, param_4, param_5);
}
@ -520,9 +520,9 @@ void Cmd_LinGrad_write(Alloc a, CmdRef ref, CmdLinGrad s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 6u;
write_mem(param, param_1, param_2);
CmdLinGradRef _804 = { ref.offset + 4u };
CmdLinGradRef _819 = { ref.offset + 4u };
Alloc param_3 = a;
CmdLinGradRef param_4 = _804;
CmdLinGradRef param_4 = _819;
CmdLinGrad param_5 = s;
CmdLinGrad_write(param_3, param_4, param_5);
}
@ -582,9 +582,9 @@ void Cmd_RadGrad_write(Alloc a, CmdRef ref, CmdRadGrad s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 7u;
write_mem(param, param_1, param_2);
CmdRadGradRef _822 = { ref.offset + 4u };
CmdRadGradRef _837 = { ref.offset + 4u };
Alloc param_3 = a;
CmdRadGradRef param_4 = _822;
CmdRadGradRef param_4 = _837;
CmdRadGrad param_5 = s;
CmdRadGrad_write(param_3, param_4, param_5);
}
@ -608,9 +608,9 @@ void Cmd_Image_write(Alloc a, CmdRef ref, CmdImage s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 8u;
write_mem(param, param_1, param_2);
CmdImageRef _840 = { ref.offset + 4u };
CmdImageRef _855 = { ref.offset + 4u };
Alloc param_3 = a;
CmdImageRef param_4 = _840;
CmdImageRef param_4 = _855;
CmdImage param_5 = s;
CmdImage_write(param_3, param_4, param_5);
}
@ -638,9 +638,9 @@ void Cmd_EndClip_write(Alloc a, CmdRef ref, CmdEndClip s)
uint param_1 = ref.offset >> uint(2);
uint param_2 = 10u;
write_mem(param, param_1, param_2);
CmdEndClipRef _866 = { ref.offset + 4u };
CmdEndClipRef _881 = { ref.offset + 4u };
Alloc param_3 = a;
CmdEndClipRef param_4 = _866;
CmdEndClipRef param_4 = _881;
CmdEndClip param_5 = s;
CmdEndClip_write(param_3, param_4, param_5);
}
@ -653,27 +653,35 @@ void Cmd_End_write(Alloc a, CmdRef ref)
write_mem(param, param_1, param_2);
}
void alloc_write(Alloc a, uint offset, Alloc alloc)
{
Alloc param = a;
uint param_1 = offset >> uint(2);
uint param_2 = alloc.offset;
write_mem(param, param_1, param_2);
}
void comp_main()
{
uint width_in_bins = ((_1005.Load(8) + 16u) - 1u) / 16u;
uint width_in_bins = ((_1020.Load(8) + 16u) - 1u) / 16u;
uint bin_ix = (width_in_bins * gl_WorkGroupID.y) + gl_WorkGroupID.x;
uint partition_ix = 0u;
uint n_partitions = ((_1005.Load(0) + 256u) - 1u) / 256u;
uint n_partitions = ((_1020.Load(0) + 256u) - 1u) / 256u;
uint th_ix = gl_LocalInvocationID.x;
uint bin_tile_x = 16u * gl_WorkGroupID.x;
uint bin_tile_y = 16u * gl_WorkGroupID.y;
uint tile_x = gl_LocalInvocationID.x % 16u;
uint tile_y = gl_LocalInvocationID.x / 16u;
uint this_tile_ix = (((bin_tile_y + tile_y) * _1005.Load(8)) + bin_tile_x) + tile_x;
Alloc _1070;
_1070.offset = _1005.Load(24);
uint this_tile_ix = (((bin_tile_y + tile_y) * _1020.Load(8)) + bin_tile_x) + tile_x;
Alloc _1085;
_1085.offset = _1020.Load(24);
Alloc param;
param.offset = _1070.offset;
param.offset = _1085.offset;
uint param_1 = this_tile_ix * 1024u;
uint param_2 = 1024u;
Alloc cmd_alloc = slice_mem(param, param_1, param_2);
CmdRef _1079 = { cmd_alloc.offset };
CmdRef cmd_ref = _1079;
CmdRef _1094 = { cmd_alloc.offset };
CmdRef cmd_ref = _1094;
uint cmd_limit = (cmd_ref.offset + 1024u) - 144u;
uint clip_depth = 0u;
uint clip_zero_depth = 0u;
@ -681,21 +689,25 @@ void comp_main()
uint wr_ix = 0u;
uint part_start_ix = 0u;
uint ready_ix = 0u;
Alloc param_3 = cmd_alloc;
uint param_4 = 0u;
uint param_5 = 8u;
Alloc scratch_alloc = slice_mem(param_3, param_4, param_5);
cmd_ref.offset += 4u;
uint render_blend_depth = 0u;
uint max_blend_depth = 0u;
uint drawmonoid_start = _1005.Load(44) >> uint(2);
uint drawtag_start = _1005.Load(100) >> uint(2);
uint drawdata_start = _1005.Load(104) >> uint(2);
uint drawinfo_start = _1005.Load(68) >> uint(2);
bool mem_ok = _260.Load(4) == 0u;
Alloc param_3;
Alloc param_5;
uint _1310;
uint drawmonoid_start = _1020.Load(44) >> uint(2);
uint drawtag_start = _1020.Load(100) >> uint(2);
uint drawdata_start = _1020.Load(104) >> uint(2);
uint drawinfo_start = _1020.Load(68) >> uint(2);
bool mem_ok = _266.Load(4) == 0u;
Alloc param_6;
Alloc param_8;
uint _1331;
uint element_ix;
Alloc param_14;
Alloc param_17;
uint tile_count;
uint _1611;
uint _1632;
float linewidth;
CmdLinGrad cmd_lin;
CmdRadGrad cmd_rad;
@ -705,40 +717,40 @@ void comp_main()
{
sh_bitmaps[i][th_ix] = 0u;
}
bool _1362;
bool _1383;
for (;;)
{
if ((ready_ix == wr_ix) && (partition_ix < n_partitions))
{
part_start_ix = ready_ix;
uint count = 0u;
bool _1160 = th_ix < 256u;
bool _1168;
if (_1160)
bool _1181 = th_ix < 256u;
bool _1189;
if (_1181)
{
_1168 = (partition_ix + th_ix) < n_partitions;
_1189 = (partition_ix + th_ix) < n_partitions;
}
else
{
_1168 = _1160;
_1189 = _1181;
}
if (_1168)
if (_1189)
{
uint in_ix = (_1005.Load(20) >> uint(2)) + ((((partition_ix + th_ix) * 256u) + bin_ix) * 2u);
Alloc _1185;
_1185.offset = _1005.Load(20);
param_3.offset = _1185.offset;
uint param_4 = in_ix;
count = read_mem(param_3, param_4);
Alloc _1196;
_1196.offset = _1005.Load(20);
param_5.offset = _1196.offset;
uint param_6 = in_ix + 1u;
uint offset = read_mem(param_5, param_6);
uint param_7 = offset;
uint param_8 = count * 4u;
bool param_9 = mem_ok;
sh_part_elements[th_ix] = new_alloc(param_7, param_8, param_9);
uint in_ix = (_1020.Load(20) >> uint(2)) + ((((partition_ix + th_ix) * 256u) + bin_ix) * 2u);
Alloc _1206;
_1206.offset = _1020.Load(20);
param_6.offset = _1206.offset;
uint param_7 = in_ix;
count = read_mem(param_6, param_7);
Alloc _1217;
_1217.offset = _1020.Load(20);
param_8.offset = _1217.offset;
uint param_9 = in_ix + 1u;
uint offset = read_mem(param_8, param_9);
uint param_10 = offset;
uint param_11 = count * 4u;
bool param_12 = mem_ok;
sh_part_elements[th_ix] = new_alloc(param_10, param_11, param_12);
}
for (uint i_1 = 0u; i_1 < 8u; i_1++)
{
@ -778,35 +790,35 @@ void comp_main()
}
if (part_ix > 0u)
{
_1310 = sh_part_count[part_ix - 1u];
_1331 = sh_part_count[part_ix - 1u];
}
else
{
_1310 = part_start_ix;
_1331 = part_start_ix;
}
ix -= _1310;
ix -= _1331;
Alloc bin_alloc = sh_part_elements[part_ix];
BinInstanceRef _1329 = { bin_alloc.offset };
BinInstanceRef inst_ref = _1329;
BinInstanceRef param_10 = inst_ref;
uint param_11 = ix;
Alloc param_12 = bin_alloc;
BinInstanceRef param_13 = BinInstance_index(param_10, param_11);
BinInstance inst = BinInstance_read(param_12, param_13);
BinInstanceRef _1350 = { bin_alloc.offset };
BinInstanceRef inst_ref = _1350;
BinInstanceRef param_13 = inst_ref;
uint param_14 = ix;
Alloc param_15 = bin_alloc;
BinInstanceRef param_16 = BinInstance_index(param_13, param_14);
BinInstance inst = BinInstance_read(param_15, param_16);
sh_elements[th_ix] = inst.element_ix;
}
GroupMemoryBarrierWithGroupSync();
wr_ix = min((rd_ix + 256u), ready_ix);
bool _1352 = (wr_ix - rd_ix) < 256u;
if (_1352)
bool _1373 = (wr_ix - rd_ix) < 256u;
if (_1373)
{
_1362 = (wr_ix < ready_ix) || (partition_ix < n_partitions);
_1383 = (wr_ix < ready_ix) || (partition_ix < n_partitions);
}
else
{
_1362 = _1352;
_1383 = _1373;
}
if (_1362)
if (_1383)
{
continue;
}
@ -819,7 +831,7 @@ void comp_main()
if ((th_ix + rd_ix) < wr_ix)
{
element_ix = sh_elements[th_ix];
tag = _1378.Load((drawtag_start + element_ix) * 4 + 0);
tag = _1399.Load((drawtag_start + element_ix) * 4 + 0);
}
switch (tag)
{
@ -831,13 +843,13 @@ void comp_main()
case 37u:
{
uint drawmonoid_base = drawmonoid_start + (4u * element_ix);
uint path_ix = _260.Load(drawmonoid_base * 4 + 8);
PathRef _1403 = { _1005.Load(16) + (path_ix * 12u) };
Alloc _1406;
_1406.offset = _1005.Load(16);
param_14.offset = _1406.offset;
PathRef param_15 = _1403;
Path path = Path_read(param_14, param_15);
uint path_ix = _266.Load(drawmonoid_base * 4 + 8);
PathRef _1424 = { _1020.Load(16) + (path_ix * 12u) };
Alloc _1427;
_1427.offset = _1020.Load(16);
param_17.offset = _1427.offset;
PathRef param_18 = _1424;
Path path = Path_read(param_17, param_18);
uint stride = path.bbox.z - path.bbox.x;
sh_tile_stride[th_ix] = stride;
int dx = int(path.bbox.x) - int(bin_tile_x);
@ -852,13 +864,13 @@ void comp_main()
tile_count = uint(x1 - x0) * uint(y1 - y0);
uint base = path.tiles.offset - (((uint(dy) * stride) + uint(dx)) * 8u);
sh_tile_base[th_ix] = base;
uint param_16 = path.tiles.offset;
uint param_17 = ((path.bbox.z - path.bbox.x) * (path.bbox.w - path.bbox.y)) * 8u;
bool param_18 = mem_ok;
Alloc path_alloc = new_alloc(param_16, param_17, param_18);
uint param_19 = th_ix;
Alloc param_20 = path_alloc;
write_tile_alloc(param_19, param_20);
uint param_19 = path.tiles.offset;
uint param_20 = ((path.bbox.z - path.bbox.x) * (path.bbox.w - path.bbox.y)) * 8u;
bool param_21 = mem_ok;
Alloc path_alloc = new_alloc(param_19, param_20, param_21);
uint param_22 = th_ix;
Alloc param_23 = path_alloc;
write_tile_alloc(param_22, param_23);
break;
}
default:
@ -892,56 +904,56 @@ void comp_main()
}
}
uint element_ix_1 = sh_elements[el_ix];
uint tag_1 = _1378.Load((drawtag_start + element_ix_1) * 4 + 0);
uint tag_1 = _1399.Load((drawtag_start + element_ix_1) * 4 + 0);
if (el_ix > 0u)
{
_1611 = sh_tile_count[el_ix - 1u];
_1632 = sh_tile_count[el_ix - 1u];
}
else
{
_1611 = 0u;
_1632 = 0u;
}
uint seq_ix = ix_1 - _1611;
uint seq_ix = ix_1 - _1632;
uint width = sh_tile_width[el_ix];
uint x = sh_tile_x0[el_ix] + (seq_ix % width);
uint y = sh_tile_y0[el_ix] + (seq_ix / width);
bool include_tile = false;
if (mem_ok)
{
uint param_21 = el_ix;
bool param_22 = mem_ok;
TileRef _1663 = { sh_tile_base[el_ix] + (((sh_tile_stride[el_ix] * y) + x) * 8u) };
Alloc param_23 = read_tile_alloc(param_21, param_22);
TileRef param_24 = _1663;
Tile tile = Tile_read(param_23, param_24);
uint param_24 = el_ix;
bool param_25 = mem_ok;
TileRef _1684 = { sh_tile_base[el_ix] + (((sh_tile_stride[el_ix] * y) + x) * 8u) };
Alloc param_26 = read_tile_alloc(param_24, param_25);
TileRef param_27 = _1684;
Tile tile = Tile_read(param_26, param_27);
bool is_clip = (tag_1 & 1u) != 0u;
bool is_blend = false;
if (is_clip)
{
uint drawmonoid_base_1 = drawmonoid_start + (4u * element_ix_1);
uint scene_offset = _260.Load((drawmonoid_base_1 + 2u) * 4 + 8);
uint scene_offset = _266.Load((drawmonoid_base_1 + 2u) * 4 + 8);
uint dd = drawdata_start + (scene_offset >> uint(2));
uint blend = _1378.Load(dd * 4 + 0);
uint blend = _1399.Load(dd * 4 + 0);
is_blend = blend != 32771u;
}
bool _1699 = tile.tile.offset != 0u;
bool _1708;
if (!_1699)
bool _1720 = tile.tile.offset != 0u;
bool _1729;
if (!_1720)
{
_1708 = (tile.backdrop == 0) == is_clip;
_1729 = (tile.backdrop == 0) == is_clip;
}
else
{
_1708 = _1699;
_1729 = _1720;
}
include_tile = _1708 || is_blend;
include_tile = _1729 || is_blend;
}
if (include_tile)
{
uint el_slice = el_ix / 32u;
uint el_mask = 1u << (el_ix & 31u);
uint _1730;
InterlockedOr(sh_bitmaps[el_slice][(y * 16u) + x], el_mask, _1730);
uint _1751;
InterlockedOr(sh_bitmaps[el_slice][(y * 16u) + x], el_mask, _1751);
}
}
GroupMemoryBarrierWithGroupSync();
@ -965,178 +977,178 @@ void comp_main()
uint element_ref_ix = (slice_ix * 32u) + uint(int(firstbitlow(bitmap)));
uint element_ix_2 = sh_elements[element_ref_ix];
bitmap &= (bitmap - 1u);
uint drawtag = _1378.Load((drawtag_start + element_ix_2) * 4 + 0);
uint drawtag = _1399.Load((drawtag_start + element_ix_2) * 4 + 0);
if (clip_zero_depth == 0u)
{
uint param_25 = element_ref_ix;
bool param_26 = mem_ok;
TileRef _1807 = { sh_tile_base[element_ref_ix] + (((sh_tile_stride[element_ref_ix] * tile_y) + tile_x) * 8u) };
Alloc param_27 = read_tile_alloc(param_25, param_26);
TileRef param_28 = _1807;
Tile tile_1 = Tile_read(param_27, param_28);
uint param_28 = element_ref_ix;
bool param_29 = mem_ok;
TileRef _1828 = { sh_tile_base[element_ref_ix] + (((sh_tile_stride[element_ref_ix] * tile_y) + tile_x) * 8u) };
Alloc param_30 = read_tile_alloc(param_28, param_29);
TileRef param_31 = _1828;
Tile tile_1 = Tile_read(param_30, param_31);
uint drawmonoid_base_2 = drawmonoid_start + (4u * element_ix_2);
uint scene_offset_1 = _260.Load((drawmonoid_base_2 + 2u) * 4 + 8);
uint info_offset = _260.Load((drawmonoid_base_2 + 3u) * 4 + 8);
uint scene_offset_1 = _266.Load((drawmonoid_base_2 + 2u) * 4 + 8);
uint info_offset = _266.Load((drawmonoid_base_2 + 3u) * 4 + 8);
uint dd_1 = drawdata_start + (scene_offset_1 >> uint(2));
uint di = drawinfo_start + (info_offset >> uint(2));
switch (drawtag)
{
case 68u:
{
linewidth = asfloat(_260.Load(di * 4 + 8));
Alloc param_29 = cmd_alloc;
CmdRef param_30 = cmd_ref;
uint param_31 = cmd_limit;
bool _1855 = alloc_cmd(param_29, param_30, param_31);
cmd_alloc = param_29;
cmd_ref = param_30;
cmd_limit = param_31;
if (!_1855)
linewidth = asfloat(_266.Load(di * 4 + 8));
Alloc param_32 = cmd_alloc;
CmdRef param_33 = cmd_ref;
uint param_34 = cmd_limit;
bool _1876 = alloc_cmd(param_32, param_33, param_34);
cmd_alloc = param_32;
cmd_ref = param_33;
cmd_limit = param_34;
if (!_1876)
{
break;
}
Alloc param_32 = cmd_alloc;
CmdRef param_33 = cmd_ref;
Tile param_34 = tile_1;
float param_35 = linewidth;
write_fill(param_32, param_33, param_34, param_35);
cmd_ref = param_33;
uint rgba = _1378.Load(dd_1 * 4 + 0);
CmdColor _1878 = { rgba };
Alloc param_36 = cmd_alloc;
CmdRef param_37 = cmd_ref;
CmdColor param_38 = _1878;
Cmd_Color_write(param_36, param_37, param_38);
Alloc param_35 = cmd_alloc;
CmdRef param_36 = cmd_ref;
Tile param_37 = tile_1;
float param_38 = linewidth;
write_fill(param_35, param_36, param_37, param_38);
cmd_ref = param_36;
uint rgba = _1399.Load(dd_1 * 4 + 0);
CmdColor _1899 = { rgba };
Alloc param_39 = cmd_alloc;
CmdRef param_40 = cmd_ref;
CmdColor param_41 = _1899;
Cmd_Color_write(param_39, param_40, param_41);
cmd_ref.offset += 8u;
break;
}
case 276u:
{
Alloc param_39 = cmd_alloc;
CmdRef param_40 = cmd_ref;
uint param_41 = cmd_limit;
bool _1896 = alloc_cmd(param_39, param_40, param_41);
cmd_alloc = param_39;
cmd_ref = param_40;
cmd_limit = param_41;
if (!_1896)
Alloc param_42 = cmd_alloc;
CmdRef param_43 = cmd_ref;
uint param_44 = cmd_limit;
bool _1917 = alloc_cmd(param_42, param_43, param_44);
cmd_alloc = param_42;
cmd_ref = param_43;
cmd_limit = param_44;
if (!_1917)
{
break;
}
linewidth = asfloat(_260.Load(di * 4 + 8));
Alloc param_42 = cmd_alloc;
CmdRef param_43 = cmd_ref;
Tile param_44 = tile_1;
float param_45 = linewidth;
write_fill(param_42, param_43, param_44, param_45);
cmd_ref = param_43;
cmd_lin.index = _1378.Load(dd_1 * 4 + 0);
cmd_lin.line_x = asfloat(_260.Load((di + 1u) * 4 + 8));
cmd_lin.line_y = asfloat(_260.Load((di + 2u) * 4 + 8));
cmd_lin.line_c = asfloat(_260.Load((di + 3u) * 4 + 8));
Alloc param_46 = cmd_alloc;
CmdRef param_47 = cmd_ref;
CmdLinGrad param_48 = cmd_lin;
Cmd_LinGrad_write(param_46, param_47, param_48);
linewidth = asfloat(_266.Load(di * 4 + 8));
Alloc param_45 = cmd_alloc;
CmdRef param_46 = cmd_ref;
Tile param_47 = tile_1;
float param_48 = linewidth;
write_fill(param_45, param_46, param_47, param_48);
cmd_ref = param_46;
cmd_lin.index = _1399.Load(dd_1 * 4 + 0);
cmd_lin.line_x = asfloat(_266.Load((di + 1u) * 4 + 8));
cmd_lin.line_y = asfloat(_266.Load((di + 2u) * 4 + 8));
cmd_lin.line_c = asfloat(_266.Load((di + 3u) * 4 + 8));
Alloc param_49 = cmd_alloc;
CmdRef param_50 = cmd_ref;
CmdLinGrad param_51 = cmd_lin;
Cmd_LinGrad_write(param_49, param_50, param_51);
cmd_ref.offset += 20u;
break;
}
case 732u:
{
Alloc param_49 = cmd_alloc;
CmdRef param_50 = cmd_ref;
uint param_51 = cmd_limit;
bool _1960 = alloc_cmd(param_49, param_50, param_51);
cmd_alloc = param_49;
cmd_ref = param_50;
cmd_limit = param_51;
if (!_1960)
Alloc param_52 = cmd_alloc;
CmdRef param_53 = cmd_ref;
uint param_54 = cmd_limit;
bool _1981 = alloc_cmd(param_52, param_53, param_54);
cmd_alloc = param_52;
cmd_ref = param_53;
cmd_limit = param_54;
if (!_1981)
{
break;
}
linewidth = asfloat(_260.Load(di * 4 + 8));
Alloc param_52 = cmd_alloc;
CmdRef param_53 = cmd_ref;
Tile param_54 = tile_1;
float param_55 = linewidth;
write_fill(param_52, param_53, param_54, param_55);
cmd_ref = param_53;
cmd_rad.index = _1378.Load(dd_1 * 4 + 0);
cmd_rad.mat = asfloat(uint4(_260.Load((di + 1u) * 4 + 8), _260.Load((di + 2u) * 4 + 8), _260.Load((di + 3u) * 4 + 8), _260.Load((di + 4u) * 4 + 8)));
cmd_rad.xlat = asfloat(uint2(_260.Load((di + 5u) * 4 + 8), _260.Load((di + 6u) * 4 + 8)));
cmd_rad.c1 = asfloat(uint2(_260.Load((di + 7u) * 4 + 8), _260.Load((di + 8u) * 4 + 8)));
cmd_rad.ra = asfloat(_260.Load((di + 9u) * 4 + 8));
cmd_rad.roff = asfloat(_260.Load((di + 10u) * 4 + 8));
Alloc param_56 = cmd_alloc;
CmdRef param_57 = cmd_ref;
CmdRadGrad param_58 = cmd_rad;
Cmd_RadGrad_write(param_56, param_57, param_58);
linewidth = asfloat(_266.Load(di * 4 + 8));
Alloc param_55 = cmd_alloc;
CmdRef param_56 = cmd_ref;
Tile param_57 = tile_1;
float param_58 = linewidth;
write_fill(param_55, param_56, param_57, param_58);
cmd_ref = param_56;
cmd_rad.index = _1399.Load(dd_1 * 4 + 0);
cmd_rad.mat = asfloat(uint4(_266.Load((di + 1u) * 4 + 8), _266.Load((di + 2u) * 4 + 8), _266.Load((di + 3u) * 4 + 8), _266.Load((di + 4u) * 4 + 8)));
cmd_rad.xlat = asfloat(uint2(_266.Load((di + 5u) * 4 + 8), _266.Load((di + 6u) * 4 + 8)));
cmd_rad.c1 = asfloat(uint2(_266.Load((di + 7u) * 4 + 8), _266.Load((di + 8u) * 4 + 8)));
cmd_rad.ra = asfloat(_266.Load((di + 9u) * 4 + 8));
cmd_rad.roff = asfloat(_266.Load((di + 10u) * 4 + 8));
Alloc param_59 = cmd_alloc;
CmdRef param_60 = cmd_ref;
CmdRadGrad param_61 = cmd_rad;
Cmd_RadGrad_write(param_59, param_60, param_61);
cmd_ref.offset += 48u;
break;
}
case 72u:
{
linewidth = asfloat(_260.Load(di * 4 + 8));
Alloc param_59 = cmd_alloc;
CmdRef param_60 = cmd_ref;
uint param_61 = cmd_limit;
bool _2066 = alloc_cmd(param_59, param_60, param_61);
cmd_alloc = param_59;
cmd_ref = param_60;
cmd_limit = param_61;
if (!_2066)
linewidth = asfloat(_266.Load(di * 4 + 8));
Alloc param_62 = cmd_alloc;
CmdRef param_63 = cmd_ref;
uint param_64 = cmd_limit;
bool _2087 = alloc_cmd(param_62, param_63, param_64);
cmd_alloc = param_62;
cmd_ref = param_63;
cmd_limit = param_64;
if (!_2087)
{
break;
}
Alloc param_62 = cmd_alloc;
CmdRef param_63 = cmd_ref;
Tile param_64 = tile_1;
float param_65 = linewidth;
write_fill(param_62, param_63, param_64, param_65);
cmd_ref = param_63;
uint index = _1378.Load(dd_1 * 4 + 0);
uint raw1 = _1378.Load((dd_1 + 1u) * 4 + 0);
Alloc param_65 = cmd_alloc;
CmdRef param_66 = cmd_ref;
Tile param_67 = tile_1;
float param_68 = linewidth;
write_fill(param_65, param_66, param_67, param_68);
cmd_ref = param_66;
uint index = _1399.Load(dd_1 * 4 + 0);
uint raw1 = _1399.Load((dd_1 + 1u) * 4 + 0);
int2 offset_1 = int2(int(raw1 << uint(16)) >> 16, int(raw1) >> 16);
CmdImage _2105 = { index, offset_1 };
Alloc param_66 = cmd_alloc;
CmdRef param_67 = cmd_ref;
CmdImage param_68 = _2105;
Cmd_Image_write(param_66, param_67, param_68);
CmdImage _2126 = { index, offset_1 };
Alloc param_69 = cmd_alloc;
CmdRef param_70 = cmd_ref;
CmdImage param_71 = _2126;
Cmd_Image_write(param_69, param_70, param_71);
cmd_ref.offset += 12u;
break;
}
case 5u:
{
bool _2119 = tile_1.tile.offset == 0u;
bool _2125;
if (_2119)
bool _2140 = tile_1.tile.offset == 0u;
bool _2146;
if (_2140)
{
_2125 = tile_1.backdrop == 0;
_2146 = tile_1.backdrop == 0;
}
else
{
_2125 = _2119;
_2146 = _2140;
}
if (_2125)
if (_2146)
{
clip_zero_depth = clip_depth + 1u;
}
else
{
Alloc param_69 = cmd_alloc;
CmdRef param_70 = cmd_ref;
uint param_71 = cmd_limit;
bool _2137 = alloc_cmd(param_69, param_70, param_71);
cmd_alloc = param_69;
cmd_ref = param_70;
cmd_limit = param_71;
if (!_2137)
Alloc param_72 = cmd_alloc;
CmdRef param_73 = cmd_ref;
uint param_74 = cmd_limit;
bool _2158 = alloc_cmd(param_72, param_73, param_74);
cmd_alloc = param_72;
cmd_ref = param_73;
cmd_limit = param_74;
if (!_2158)
{
break;
}
Alloc param_72 = cmd_alloc;
CmdRef param_73 = cmd_ref;
Cmd_BeginClip_write(param_72, param_73);
Alloc param_75 = cmd_alloc;
CmdRef param_76 = cmd_ref;
Cmd_BeginClip_write(param_75, param_76);
cmd_ref.offset += 4u;
render_blend_depth++;
max_blend_depth = max(max_blend_depth, render_blend_depth);
@ -1147,29 +1159,29 @@ void comp_main()
case 37u:
{
clip_depth--;
Alloc param_74 = cmd_alloc;
CmdRef param_75 = cmd_ref;
uint param_76 = cmd_limit;
bool _2170 = alloc_cmd(param_74, param_75, param_76);
cmd_alloc = param_74;
cmd_ref = param_75;
cmd_limit = param_76;
if (!_2170)
Alloc param_77 = cmd_alloc;
CmdRef param_78 = cmd_ref;
uint param_79 = cmd_limit;
bool _2191 = alloc_cmd(param_77, param_78, param_79);
cmd_alloc = param_77;
cmd_ref = param_78;
cmd_limit = param_79;
if (!_2191)
{
break;
}
Alloc param_77 = cmd_alloc;
CmdRef param_78 = cmd_ref;
Tile param_79 = tile_1;
float param_80 = -1.0f;
write_fill(param_77, param_78, param_79, param_80);
cmd_ref = param_78;
uint blend_1 = _1378.Load(dd_1 * 4 + 0);
CmdEndClip _2193 = { blend_1 };
Alloc param_81 = cmd_alloc;
CmdRef param_82 = cmd_ref;
CmdEndClip param_83 = _2193;
Cmd_EndClip_write(param_81, param_82, param_83);
Alloc param_80 = cmd_alloc;
CmdRef param_81 = cmd_ref;
Tile param_82 = tile_1;
float param_83 = -1.0f;
write_fill(param_80, param_81, param_82, param_83);
cmd_ref = param_81;
uint blend_1 = _1399.Load(dd_1 * 4 + 0);
CmdEndClip _2214 = { blend_1 };
Alloc param_84 = cmd_alloc;
CmdRef param_85 = cmd_ref;
CmdEndClip param_86 = _2214;
Cmd_EndClip_write(param_84, param_85, param_86);
cmd_ref.offset += 8u;
render_blend_depth--;
break;
@ -1204,23 +1216,31 @@ void comp_main()
break;
}
}
bool _2242 = (bin_tile_x + tile_x) < _1005.Load(8);
bool _2251;
if (_2242)
bool _2263 = (bin_tile_x + tile_x) < _1020.Load(8);
bool _2272;
if (_2263)
{
_2251 = (bin_tile_y + tile_y) < _1005.Load(12);
_2272 = (bin_tile_y + tile_y) < _1020.Load(12);
}
else
{
_2251 = _2242;
_2272 = _2263;
}
if (_2251)
if (_2272)
{
Alloc param_84 = cmd_alloc;
CmdRef param_85 = cmd_ref;
Cmd_End_write(param_84, param_85);
Alloc param_87 = cmd_alloc;
CmdRef param_88 = cmd_ref;
Cmd_End_write(param_87, param_88);
if (max_blend_depth > 4u)
{
uint scratch_size = (((max_blend_depth * 16u) * 16u) * 1u) * 4u;
uint param_89 = scratch_size;
MallocResult _2293 = malloc(param_89);
MallocResult scratch = _2293;
Alloc param_90 = scratch_alloc;
uint param_91 = scratch_alloc.offset;
Alloc param_92 = scratch.alloc;
alloc_write(param_90, param_91, param_92);
}
}
}

639
piet-gpu/shader/gen/coarse.msl generated
View file

File diff suppressed because it is too large Load diff

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

Binary file not shown.

BIN
piet-gpu/shader/gen/kernel4.dxil generated
View file

Binary file not shown.

2
piet-gpu/shader/gen/kernel4.hlsl generated
View file

@ -1250,7 +1250,7 @@ void comp_main()
CmdRef param_41 = cmd_ref;
CmdEndClip end_clip = Cmd_EndClip_read(param_40, param_41);
clip_depth--;
if (clip_depth < 4u)
if (clip_depth >= 4u)
{
base_ix_1 = ((blend_offset >> uint(2)) + (((clip_depth - 4u) * 16u) * 16u)) + (8u * (gl_LocalInvocationID.x + (8u * gl_LocalInvocationID.y)));
}

2
piet-gpu/shader/gen/kernel4.msl generated
View file

@ -1304,7 +1304,7 @@ kernel void main0(device Memory& v_297 [[buffer(0)]], const device ConfigBuf& _1
CmdRef param_41 = cmd_ref;
CmdEndClip end_clip = Cmd_EndClip_read(param_40, param_41, v_297);
clip_depth--;
if (clip_depth < 4u)
if (clip_depth >= 4u)
{
base_ix_1 = ((blend_offset >> uint(2)) + (((clip_depth - 4u) * 16u) * 16u)) + (8u * (gl_LocalInvocationID.x + (8u * gl_LocalInvocationID.y)));
}

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

Binary file not shown.

BIN
piet-gpu/shader/gen/kernel4_gray.dxil generated
View file

Binary file not shown.

2
piet-gpu/shader/gen/kernel4_gray.hlsl generated
View file

@ -1250,7 +1250,7 @@ void comp_main()
CmdRef param_41 = cmd_ref;
CmdEndClip end_clip = Cmd_EndClip_read(param_40, param_41);
clip_depth--;
if (clip_depth < 4u)
if (clip_depth >= 4u)
{
base_ix_1 = ((blend_offset >> uint(2)) + (((clip_depth - 4u) * 16u) * 16u)) + (8u * (gl_LocalInvocationID.x + (8u * gl_LocalInvocationID.y)));
}

2
piet-gpu/shader/gen/kernel4_gray.msl generated
View file

@ -1304,7 +1304,7 @@ kernel void main0(device Memory& v_297 [[buffer(0)]], const device ConfigBuf& _1
CmdRef param_41 = cmd_ref;
CmdEndClip end_clip = Cmd_EndClip_read(param_40, param_41, v_297);
clip_depth--;
if (clip_depth < 4u)
if (clip_depth >= 4u)
{
base_ix_1 = ((blend_offset >> uint(2)) + (((clip_depth - 4u) * 16u) * 16u)) + (8u * (gl_LocalInvocationID.x + (8u * gl_LocalInvocationID.y)));
}

BIN
piet-gpu/shader/gen/kernel4_gray.spv generated
View file

Binary file not shown.

2
piet-gpu/shader/kernel4.comp
View file

@ -259,7 +259,7 @@ void main() {
CmdEndClip end_clip = Cmd_EndClip_read(cmd_alloc, cmd_ref);
clip_depth--;
uint base_ix;
if (clip_depth < BLEND_STACK_SPLIT) {
if (clip_depth >= BLEND_STACK_SPLIT) {
base_ix = (blend_offset >> 2) + (clip_depth - BLEND_STACK_SPLIT) * TILE_HEIGHT_PX * TILE_WIDTH_PX +
CHUNK * (gl_LocalInvocationID.x + CHUNK_DX * gl_LocalInvocationID.y);
}