Add Buffer::byte_stride and use it on Android
#315
Conversation
madsmtm
added
enhancement
madsmtm
force-pushed
the
madsmtm/stride
branch
from
1240fbb to
54a8954
Compare
| /// or failing that, make sure to chunk rows by the stride instead of the width. | ||
| /// | ||
| /// This is guaranteed to be `>= width * 4`, and is guaranteed to be a multiple of 4. | ||
| pub fn byte_stride(&self) -> NonZeroU32 { |
There was a problem hiding this comment.
I considered exposing fn stride instead, and have the buffer size be stride * height * (pixel_format.bit_depth() / 8), but I don't think that works with 24-bit formats like PixelFormat::Rgb8 that are packed into memory as [R0, G0, B0, R1, G1, B1, P0, P1] (two padding bytes at the end)?
There was a problem hiding this comment.
But perhaps there's a nicer way to define it that I'm not seeing?
There was a problem hiding this comment.
|
This currently seems to be worst-of-both-worlds: exposing the extra stride to the user while not taking advantage of it on Android by removing its backing buffer? Perhaps we should wait for the pixelformat / type API before landing this, which should finally enable that. |
True, though I do plan to add
We should be able to do it after #289. I'll fix it in whichever of these PRs land last. |
madsmtm
force-pushed
the
madsmtm/remove-deref
branch
from
2944dc3 to
5cba2e5
Compare
To help with debugging incorrect stride/width calculations, it helps that there are more common backends (than Android) that have `stride != width * 4`.
madsmtm
force-pushed
the
madsmtm/stride
branch
from
54a8954 to
c543d19
Compare
Fixes #291 and allows using
IOSurfaceon macOS in the future a la #95.This is also important for supporting pixel formats with a different bit depth than
32, see #98.Users should prefer to use
pixel_rowsorpixels_iteradded in #312 to stride correctly. In case they don't, I've changed the macOS and Wayland backends to require astride % 16whencfg!(debug_assertions), this should make it easier to see when one is not following this requirement.