High sample count crashes IBLBaker

[remcohuijser]

When you choose a high number of samples (anything higher than 512) IBLBaker stops. In debug there is no exception given.

[remcohuijser]

I tried to disconnect the sample count from the samples per frame so that the samples are spread out over a number of frames. This seems to work as IBLBaker takes a couple of frames to calculate the end result and you can actually see the intermediate result being updated.

I did notice something strange though: it appears that the result from the previous pass are not correctly combined with the next pass. As a result result becomes 'smeared' along the edges (see image below):

Because of this, the resulting map is not usable :(

[derkreature]

Yeah, I imagine there is a bug in offsetting the sample Id in regards to hammersley point generation. I’ll take a look.

[derkreature]

Are you just hitting TDR timeouts?

[remcohuijser]

I am not sure to be honest. Any way I can check?

[derkreature]

I’ll take a look at this over the holiday break.

[remcohuijser]

Much appreciated!

[derkreature]

I am getting to this.