Encode path_data and draw_data directly as u32
#817
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We already assume these are word-aligned, and so this would catch a failure of that assumption much earller. A future change would be to encode the full scene encoding as u32. That would be useful for linebender#810, as the current code has an alignment hazard
| @@ -331,7 +334,8 @@ impl Encoding { | |||
| pub fn encode_color(&mut self, color: impl Into<DrawColor>) { | |||
| let color = color.into(); | |||
| self.draw_tags.push(DrawTag::COLOR); | |||
| self.draw_data.extend_from_slice(bytemuck::bytes_of(&color)); | |||
| let DrawColor { rgba } = color; | |||
| } | ||
| if let Some((x, y)) = self.pending_images[*index].xy { | ||
| let xy = (x << 16) | y; | ||
| data.extend_from_slice(bytemuck::bytes_of(&xy)); | ||
| pos = *draw_data_offset + 4; | ||
| pos = *draw_data_offset + 1; | ||
| } else { | ||
| // If we get here, we failed to allocate a slot for this image in the atlas. | ||
| // In this case, let's zero out the dimensions so we don't attempt to render | ||
| // anything. | ||
| // TODO: a better strategy: texture array? downsample large images? | ||
| data.extend_from_slice(&[0_u8; 8]); |
There was a problem hiding this comment.
This is slightly confusing then being u8 still and just required some extra thought.
There was a problem hiding this comment.
Yes... I don't think there's really a cleaner way to express this, because data is still byte-wise. Happy to change it if needed.
I don't expect this to have a measurable impact. It's largely about making the semantics of the encoding match the reality, which is that these values are required to a multiple of 4 bytes (if they're not, things would go extremely badly wrong on the GPU). Separately, this will also make reasoning about alignment easier, if we end up doing the follow-up.
Yes, exactly. The scene buffer is always interpreted as an array of 32 bit unsigned integers. |
tomcur
added a commit
that referenced
this pull request
Feb 20, 2025
(On top of #817 to prevent conflicts.) WGSL expects little-endian data, but Vello currently doesn't convert to little-endian, except for colors: `DrawColor` specifies its endianness explicitly, with conversion handled by the Color crate. On big-endian hosts, the other encodings would break. Whether big-endian hosts are a valid target for Vello, I do not know. This is opened in part as a place to have that discussion. If this is desirable, other places will need attention. I have not taken a look yet.
tomcur
added a commit
that referenced
this pull request
Feb 20, 2025
(On top of #817 to prevent conflicts.) WGSL expects little-endian data, but Vello currently doesn't convert to little-endian, except for colors: `DrawColor` specifies its endianness explicitly, with conversion handled by the Color crate. On big-endian hosts, the other encodings would break. Whether big-endian hosts are a valid target for Vello, I do not know. This is opened in part as a place to have that discussion. If this is desirable, other places will need attention. I have not taken a look yet.
tomcur
added a commit
that referenced
this pull request
Feb 20, 2025
(On top of #817 to prevent conflicts.) WGSL expects little-endian data, but Vello currently doesn't convert to little-endian, except for colors: `DrawColor` specifies its endianness explicitly, with conversion handled by the Color crate. On big-endian hosts, the other encodings would break. Whether big-endian hosts are a valid target for Vello, I do not know. This is opened in part as a place to have that discussion. If this is desirable, other places will need attention. I have not taken a look yet.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We already assume these are word-aligned, and so this would catch a failure of that assumption very early, if it came up.
A potential alternative is to store
path_dataas f32 instead, as all values stored are f32. I've not done that for pretty weak reasons, and it should be trivial to change.A future change would be to encode the full scene encoding as
u32. That would be useful for #810, as the current code has an alignment hazard.