C++ universal "maybeowned" adapting container used everywhere instead of std::vector #3794
Comments
4 tasks
|
When looking into this we should keep in mind |
4 tasks
|
Next step of this is also to support strides |
|
Talking to a user, not only strides would be useful but also byte offsets. Example: Struct like this:
We want it to be convenient to log arrays of this struct both as points and colors! |
This was referenced Nov 14, 2023
Wumpf
added a commit
that referenced
this issue
Nov 15, 2023
…ted constructs) (#4236) ### What Not much else has changed yet except some doc massaging. First step towards solving * #3794 `rerun::Collection` will evolve from here on and be the corner piece of our zero copy & component adaptability strategy (just like ComponentBatch is already!) _This was originally opened on the wrong branch by accident:_ * #4225 ### Checklist * [x] I have read and agree to [Contributor Guide](https://github.com/rerun-io/rerun/blob/main/CONTRIBUTING.md) and the [Code of Conduct](https://github.com/rerun-io/rerun/blob/main/CODE_OF_CONDUCT.md) * [x] I've included a screenshot or gif (if applicable) * [x] I have tested [demo.rerun.io](https://demo.rerun.io/pr/4225) (if applicable) * [x] The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG - [PR Build Summary](https://build.rerun.io/pr/4225) - [Docs preview](https://rerun.io/preview/3de46c8a5d294eb5ad68e5beb35a64d0610e992a/docs) <!--DOCS-PREVIEW--> - [Examples preview](https://rerun.io/preview/3de46c8a5d294eb5ad68e5beb35a64d0610e992a/examples) <!--EXAMPLES-PREVIEW--> - [Recent benchmark results](https://ref.rerun.io/dev/bench/) - [Wasm size tracking](https://ref.rerun.io/dev/sizes/)
4 tasks
Wumpf
added a commit
that referenced
this issue
Nov 17, 2023
…re (#4247) * Part of the series of #3794 ### What ... and various improvements to `rerun::Collection` in general to be up for the task. Added copy assignment/construction, added vector conversion, removed serialization method (handled at callsites now) Also, I overhauled and generalized our builtin `rerun::CollectionAdapters` further. They give now more exact guarantees on when they copy & move data. To support the considerably increased significance and abilitites of this data structure I added a lot more tests, in particular checking on the aforementioned move/copy guarantees. Huge thanks to @AnkeAnke for pointing out to me how we can support std compatible contains generically in a trivial way (`rerun::type_traits::value_type_t`!). This made it possible that this PR requires almost no changes in user/test/example code since most types just snap in the same as before. As expected this change improves the performance on the image logging benchmark (which logs 4 16k rgba images). Timing on M1 Max went from ~2.9s to ~2.4s and on the AMD windows box I use from ~3.0s to ~2.5s (did a bunch of runs, values are excluding prepare and process setup, AMD fluctuates a lot) DRAFT until I'm sure this works on more compilers. ### Checklist * [x] I have read and agree to [Contributor Guide](https://github.com/rerun-io/rerun/blob/main/CONTRIBUTING.md) and the [Code of Conduct](https://github.com/rerun-io/rerun/blob/main/CODE_OF_CONDUCT.md) * [x] I've included a screenshot or gif (if applicable) * [x] I have tested [demo.rerun.io](https://demo.rerun.io/pr/4247) (if applicable) * [x] The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG - [PR Build Summary](https://build.rerun.io/pr/4247) - [Docs preview](https://rerun.io/preview/825bf7905ccaef0f23d87a4ffe0ae406018edffa/docs) <!--DOCS-PREVIEW--> - [Examples preview](https://rerun.io/preview/825bf7905ccaef0f23d87a4ffe0ae406018edffa/examples) <!--EXAMPLES-PREVIEW--> - [Recent benchmark results](https://build.rerun.io/graphs/crates.html) - [Wasm size tracking](https://build.rerun.io/graphs/sizes.html)
5 tasks
|
The original issue was solved. Split all the other things that came up into different issues and list them in a tracking issue here |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The ComponentBatch class introduced in #3791 is not actually specific to components at all. We should build it out to a "universal" container that we use in all codegen wherever we use std::vector right now.
It is also what all serialize methods should work with.
rerun::ComponentBatchtorerun::Collection(and related constructs) #4236rerun::Collectionalmost everywhere we'd usestd::vectorbefore #4247The text was updated successfully, but these errors were encountered: