Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

yoga::resolveValue -> Length::resolve #41939

Closed
wants to merge 2 commits into from
Closed

yoga::resolveValue -> Length::resolve #41939

wants to merge 2 commits into from

Commits on Dec 19, 2023

  1. Make CompactValue internal detail of yoga::Style (facebook#41776) 

    Summary:
X-link: facebook/yoga#1492

# Summary

In preparation to replace `CompactValue`, this fully encapsulates it as an implementation detail of `yoga::Style`.

The internal API now always operates on `Style::Length`, converted to `YGValue` at the public API boundary.

In the next step, we can plug in a new representation within `Style`, which should enable 64 bit values, and lower memory usage.


# Test Plan

1. Existing tests (inc for style, invalidation, CompactValue) pass
2. Check that constexpr `yoga::isinf()` produces same assembly under Clang as `std::isinf()`
3. Fabric Android builds
4. Yoga benchmark does style reads

# Performance

Checking whether a style is defined, then reading after, is a hot path, and we are doubling any space style lengths take in the stack (but not long-term on the node). After a naive move, on one system, the Yoga benchmark creating, laying out, and destroying a tree, ran about 8-10%  slower in the "Huge nested flex" example. We are converting in many more cases instead of doing undefined check, but operating on accessed style values no longer needs to do the conversion multiple times.

I changed the `CompactValue` conversion to YGValue/StyleLength path to check for undefined as the common case (since we always convert, instead of calling `isUndefined` directly on CompactValue. That seemed to get the difference down to ~5-6% when I was playing with it then. We can optimistically make some of this up with ValuePool giving better locality, and fix this more holistically if we reduce edge and value resolution.

On another machine where I tested this, the new revision went the opposite direction, and was about 5% faster, so this isn't really a cut and dry regression, but we see different characteristics than before.

# Changelog
[Internal]

Reviewed By: rozele

Differential Revision: D51775346
    @NickGerleman @facebook-github-bot
    NickGerleman authored and facebook-github-bot committed Dec 19, 2023
    Configuration menu
    Copy the full SHA
    bc154bc View commit details
    Browse the repository at this point in the history

  2. yoga::resolveValue -> Length::resolve (facebook#41939) 

    Summary:

X-link: facebook/yoga#1520

This code originates as `YGValueResolve`, used to compute a YGValue to a length in points, using a reference for 100%.

This moves it to `Style::Length`, so we can encapsulate parts of it (for style value functions), and make the API more cohesive now that we can do C++ style OOP with it.

Changelog: [Internal]

Reviewed By: joevilches

Differential Revision: D51796973
    @NickGerleman @facebook-github-bot
    NickGerleman authored and facebook-github-bot committed Dec 19, 2023
    Configuration menu
    Copy the full SHA
    e228a6a View commit details
    Browse the repository at this point in the history