Commit graph

854 commits

Author SHA1 Message Date
Raph Levien
0cf370f9c7 Mostly working rendering
This exposes interfaces to render glyphs into a texture atlas. The main changes are:

* Methods to plumb raw Metal GPU resources (device, texture, etc) into piet-gpu-hal objects.

* A new glyph_render API specialized to rendering glyphs. This is basically the same as just painting to a canvas, but will allow better caching (and has more direct access to fonts, bypassing the Piet font type which is underdeveloped).

* Ability to render to A8 target in addition to RGBA.

WIP, there are some rough edges, not least of which is that the image format changes are only on mac and cause compile errors elsewhere.
2022-01-19 12:10:51 -08:00
Raph Levien
833d993a4e More progress exposing interface
Much of the surface area exists for rendering now.

WIP of course still
2022-01-18 18:41:28 -08:00
Raph Levien
5e221d2e91 Add capability to function as a guest in Metal
WIP
2022-01-13 17:36:08 -08:00
Raph Levien
41c7118e91
Merge pull request #138 from linebender/new_element
Beginnings of new element pipeline
2021-12-15 09:03:29 -08:00
Raph Levien
9601989152 Tweak linguist settings
Mark shader/gen subdirectories as autogenerated to avoid cluttering diffs.
2021-12-08 12:49:40 -08:00
Raph Levien
d948126c16 Adjust workgroup sizes
Make max workgroup size 256 and respect LG_WG_FACTOR.

Because the monoid scans only support a height of 2, this will reduce
the maximum scene complexity we can render. But it also increases
compatibility. Supporting larger scans is a TODO.
2021-12-08 11:48:38 -08:00
Raph Levien
75496f5e67 Fix draw test
We'll revert this later, but for now trying to keep tests green.
2021-12-08 10:38:46 -08:00
Raph Levien
49c3a3923b Restore gradients and clips
This changes gradients and clips to the new encoding. Lightly tested.
2021-12-07 18:39:33 -08:00
Raph Levien
5585b3563c Account path tags correctly 2021-12-06 15:59:32 -08:00
Raph Levien
55395fa533 Get text working
This wires up the text rendering. WIP though, as it tends to get stuck.
Not sure if that's text related or just being triggered by it.
2021-12-06 11:08:58 -08:00
Raph Levien
3ee81ad5a9 Fixes for mac
Fix incorrect workgroup sizes, and change strategy for assigning binding
numbers; ultimately we should get correct values for those from shader
compilation, but this works for now.
2021-12-03 20:40:15 -08:00
Raph Levien
395d400cd1 Remove debug printing 2021-12-03 16:35:57 -08:00
Raph Levien
7f3427420b Partial DX12 fixes
As of this patch, cli works in release mode, but hangs in debug. There
are some validation errors about incompatible resouce states.
2021-12-03 16:24:30 -08:00
Raph Levien
c503ff28b0 Make shaders cross-platform
Translate all piet-gpu shaders into DXIL and MSL; move generated files
into the shader/gen directory.
2021-12-03 15:49:58 -08:00
Raph Levien
44327fe49f Beginnings of new element pipeline
This successfully renders the tiger; fills and strokes are supported.
Other parts of the imaging model, not yet.

Progress toward #119
2021-12-03 15:33:01 -08:00
Raph Levien
22b86072f2
Merge pull request #137 from linebender/draw_element
Add draw object stage
2021-12-03 15:32:51 -08:00
Raph Levien
875c8badf4 Add draw object stage
This is one of the stages in the new element pipeline. It's a simple
one, just a prefix sum of a couple counts, and some of it will probably
get merged with a downstream stage, but we'll do it separately for now
for convenience.

This patch also contains an update to Vulkan tools 1.2.198, which
accounts for the large diff of translated shaders.
2021-12-02 13:37:16 -08:00
Raph Levien
50153a7797
Merge pull request #136 from linebender/path_element
Path stream processing
2021-12-02 13:36:34 -08:00
Raph Levien
70723bf076 Very minor reorg
Move transform stage into its own file, cargo fmt.
2021-12-01 15:46:36 -08:00
Raph Levien
1d1801c1aa Cross-platform path stage shaders 2021-12-01 08:42:06 -08:00
Raph Levien
8af4707525 Fix uninitialized variable 2021-12-01 08:34:41 -08:00
Raph Levien
178761dcb3 Path stream processing
This patch contains the core of the path stream processing, though some
integration bits are missing. The core logic is tested, though
combinations of path types, transforms, and line widths are not (yet).

Progress towards #119
2021-12-01 07:33:24 -08:00
Raph Levien
5ea5c4bb9a
Merge pull request #135 from linebender/bufwrite
Access buffer data through mapping
2021-12-01 07:06:29 -08:00
Raph Levien
a7a5b84c86 Clean up stray files 2021-11-30 10:34:42 -08:00
Raph Levien
3039a2ac39 Merge branch 'master' into bufwrite 2021-11-30 10:31:16 -08:00
Raph Levien
0880975c5b
Merge pull request #134 from linebender/new_element
Start work on new element pipeline
2021-11-30 10:20:56 -08:00
Raph Levien
97bc4c4471 Tweak BufWrite::push
Make it pass the value, not a reference, to more closely match Vec's
behavior.  It's not a big difference because the type is `Copy`, but
still better.
2021-11-26 07:50:45 -08:00
Raph Levien
f1d7560b3c Tweak extend implementation
The one that takes T is more useful than the one that takes references
to T. When specialization lands, we will be able to have both under the
`extend` name.
2021-11-25 22:02:04 -08:00
Raph Levien
ac0fb228c1 Fix Metal port 2021-11-25 21:34:06 -08:00
Raph Levien
9fb2ae91eb Access buffer data through mapping
This patch includes a number of changes to encourage reading and writing
buffers through mapping rather than copying data as before.

This includes a new `BufWrite` abstraction which is designed for filling
buffers. It behaves much like a Vec<u8>, but with fixed capacity.
2021-11-25 21:27:08 -08:00
Raph Levien
8f7ed161a6 Tune transform test parameters
Previous threshold was seeing occasional failures, and also fairly wide
variance in the error. This seems to be reliable, but hasn't been
validated extremely rigorously.
2021-11-24 09:40:52 -08:00
Raph Levien
5af8aa5e52
Merge pull request #133 from linebender/bytemuck
Use bytemuck
2021-11-24 08:02:22 -08:00
Raph Levien
47f8812e2f Start work on new element pipeline
There's a bit of reorganizing as well. Shader stages are made available
from piet-gpu to the test rig, config is now a proper structure
(marshaled with bytemuck).

This commit just has the transform stage, which is a simple monoid scan
of affine transforms.

Progress toward #119
2021-11-24 08:01:43 -08:00
Raph Levien
a8103a4c20 cargo fmt
I really need to make this automated. There are some small challenges though.
2021-11-23 08:49:06 -08:00
Raph Levien
abe2a6ceef Fix tests to use bytemuck 2021-11-23 08:48:14 -08:00
Raph Levien
2ebdd942cf Use bytemuck
Get rid of `PlainData` trait and use `Pod` from bytemuck instead.
2021-11-23 08:24:16 -08:00
Raph Levien
ecdd7fd817
Merge pull request #132 from linebender/dx12_desc
Better DX12 descriptor management
2021-11-23 07:34:31 -08:00
Raph Levien
0762cc763c Implement clear_buffers on Metal
Since clearing functionality is not built-in, use a compute shader.

Simplify tests greatly; they don't need the workaround.
2021-11-20 22:24:52 -08:00
Raph Levien
657f219ce8 Better DX12 descriptor management
Reduce allocation of descriptor heaps. This change also enables clearing
of buffers, as the handles are needed at command dispatch time.

Also updates the tests to use clear_buffers on DX12. Looking forward to
being able to get rid of the compute shader workaround on Metal.

This is a followup on #125, and progress toward #95
2021-11-20 16:36:43 -08:00
Raph Levien
95d356c08f
Merge pull request #128 from linebender/prefix_war
Fix write-after-read in prefix test
2021-11-16 08:10:20 -08:00
Raph Levien
80b6150bd7
Merge pull request #129 from nilsmartel/patch-1
Use proper hyperlink in README
2021-11-15 07:16:23 -08:00
Nils Martel
96f39af14b
Use proper hyperlink in README 2021-11-15 11:25:17 +01:00
Raph Levien
8015eb25a1 Also fix write-after-read in elements.com
On further testing, this resolves a hard lockup on Intel 630 on the
mmark stress test, so is worth getting into the repo.
2021-11-14 08:23:37 -08:00
Raph Levien
69b6632085 Fix write-after-read in prefix test
Thanks to Jeff Bolz for spotting the write-after-read hazard on the
sh_flag accesses. This fixes observed failures on Nvidia Turing and
Ampere on DX12.
2021-11-14 07:13:15 -08:00
Raph Levien
76a6f1fec8
Merge pull request #127 from linebender/atomic
Increase testing of atomics
2021-11-14 07:10:39 -08:00
Raph Levien
f32f2d7f95 Add linked list DXIL
Not sure why it wasn't in the previous commit.
2021-11-12 16:01:27 -08:00
Raph Levien
d66f67fa09 Actually add README 2021-11-12 15:27:47 -08:00
Raph Levien
4c42da2d46
Merge pull request #126 from linebender/dx12
Improve DX12 portability
2021-11-12 15:24:24 -08:00
Raph Levien
27bedd9ef1 Add README
Also verify linked list results.
2021-11-12 15:23:31 -08:00
Raph Levien
c6965de557 Add linked list test
Measure bandwidth of building linked lists with atomics.
2021-11-12 10:23:31 -08:00