expose more of SKMatrix

[mgood7123]

Description of Change

Bugs Fixed

• Fixes #

API Changes

None.

Behavioral Changes

None.

Required skia PR

#2153

PR Checklist

• Has tests (if omitted, state reason in description)
• Rebased on top of main at time of PR
• Merged related skia PRs
• Changes adhere to coding standard
• Updated documentation

[sebllll]

shouldn't it read sk_matrix_pre_translate() here?

[mgood7123]

Not sure, havnt worked on it in ages

[mattleibow]

Thanks for this PR! However, I am not sure the current SKMatrix approach is the way we want to go for 3.0.

I have been investigating this from a perf perspective and right now SKMatrix has terrible performance compared to SKMatrix44. It is about 200% slower because it has to hop the interop, and it also is deceiving because using a struct is supposed to be light, however it does expensive things.

• See Remove the native interop in SKMatrix #2779

It may even be cheaper to use Matrix4x4 under the hood. In addition to the current interop, the managed struct does not have the matrix type mask which skia uses to take shortcuts. As a result, even just passing the struct between C++ and C# causes skia to have to recalculate it. It may be better to just re-implement the whole thing in C#.

[mgood7123]

alright

…

On Sun, 3 Mar 2024 at 06:15, Matthew Leibowitz ***@***.***> wrote: Thanks for this PR! However, I am not sure the current SKMatrix approach is the way we want to go for 3.0. I have been investigating this from a perf perspective and right now SKMatrix has terrible performance compared to SKMatrix44. It is about 200% slower because it has to hop the interop, and it also is deceiving because using a struct is supposed to be light, however it does expensive things. - See #2779 <#2779> It may even be cheaper to use Matrix4x4 under the hood. In addition to the current interop, the managed struct does not have the matrix type mask which skia uses to take shortcuts. As a result, even just passing the struct between C++ and C# causes skia to have to recalculate it. It may be better to just re-implement the whole thing in C#. — Reply to this email directly, view it on GitHub <#2153 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGLITH7XCBP7JBLEKMB3ZZTYWIXM5AVCNFSM525MXEBKU5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJXGQ4DSNZSGMZQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>