Robert Swain

@freechelmi - are you aware of someone working on such a decoder/encoder implementation? I would say that we should set the requirement as low as possible.

I added the necessary parts to the C API for ice connection state I believe though the aggregation between components would need to be checked in case it is specified...

@stefanalund - perhaps you know how to fix this? Can it be just a small change in the API or application options somewhere?

I’ll give this another look over soon. I would like to have reviewed this and other things already but life is what it is. I appreciate your patience.

I definitely appreciate the need that you put forth in the description and you clarify your approach well. A couple of things come to mind: - When bind groups are...

In my opinion we should remove the watch for changes method. It’s racey and so unpredictable. Better to use the resource from the beginning, in my opinion.

I find it difficult to get an overview of whether the benchmarks are good or bad. There are so many and they go either way. One approach I've seen used...

I think these sound like great improvements and welcome them. At the same time, I still feel like we’re limiting performance artificially based on an assumption of thread contention and...

This looks controversial. I hear the reasoning. I guess ideally it would be fixed in mesa but even then it will take time for new drivers to roll out. Still,...

I did test this PR with positive/negative uniform scales, and also non-uniform scales with all/some positive components, and all negative components. At least empirically, this looks correct. And unless I...