You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As described in the issue above, Firefox may indicate support for compressed formats that it actually emulates in drivers, decompressing or transcoding to another format. This emulation has much higher cost than just transcoding to the 'right' format in KTX2Loader, but we don't necessarily know what the right format is because Firefox doesn't say. We don't have this problem on ANGLE, but currently we are preserving a less-optimal choice of target transcoding format for ANGLE because of the Firefox issue.
Solution
Consider detecting the Firefox user agent and adjusting format selection as a special case, if we can do so reliably
For Linux+Firefox+RadeonSI we should prefer BCn formats over ETC2 as the target from ETC1S, when both are "supported"
For other platforms, is the reverse happening? Would Firefox report BCn support when only ETC2 is actually available?
Test major browsers on multiple platforms, check target format selection, and implement unit tests for transcoder target formats if possible (WASM in tests may be tricky...)
Consider preferring BC1/3 over BC7 for ETC1S transcoding
Description
Related:
As described in the issue above, Firefox may indicate support for compressed formats that it actually emulates in drivers, decompressing or transcoding to another format. This emulation has much higher cost than just transcoding to the 'right' format in KTX2Loader, but we don't necessarily know what the right format is because Firefox doesn't say. We don't have this problem on ANGLE, but currently we are preserving a less-optimal choice of target transcoding format for ANGLE because of the Firefox issue.
Solution
Alternatives
n/a
Additional context
No response
The text was updated successfully, but these errors were encountered: