We had a user send us a link to a post on a different forum regarding a slowdown during profiling.
(No idea why the issue was not mentioned here...)
Anyway, it did have us scratching our heads for a bit!

It has to be said the profile was very strange, as it was a sequential profile, which would likely cause heat related issues, with grey and primary ramps in 10 bit of some 400 patch values each.
Out of interest, such a profile would permute into a cube of some 64,000,000 point!

Anyway, it turns out it was the primary RGB patches that were the issue.
They were trying to populate the Gamut Triangle dropdown with the 400+ different values.
Totally unmanageable, as it was soaking up memory resources.

So, we have restricted the Gamut Triangle option when RGB Primary patches exceed a viable value for the Gamut Triangle dropdown, based on the Permuted Cube Size limit, with a smaller decimated list being generated.

Steve

Only to explain as I have been the trigger.
I was on the process to open a thread on the subject. There were indeed to things that I have noted:

1) When running cubes with additional WRGBCMY patches, I have noted, using Prometheus, which is very fast, delays in CS execution time. I have replicated with virtual probe. The delay grows with the ordinal number of the executed patch. I understand it depends from my PC hw.

2) A CS newbie, using my generation tool, set a 14 cube with 30 additional WRGBCMY patches (that 's a value that in my focuses patch-set generation I usually set at 1), and complained some execution problems. Anyhow not so many patches (2774 cube patches, 2739 WRGBCMY only patches). When I run that with virtual probe, it took, on my PC, 7 hours.

My comment:
I have noted the the 7 hours patch-set, takes the usual time when displayed, so why, instead of setting a limit, you don't introduce a characterisation fast mode? I.e a mode only recording the stimulus, XYZ and setting only what is needed in the bcs header for further displaying/elaboration?
That would also help people with less powerful PCs to get the most from their hw

What used to last 7 hours, now lasts 1 hour.
But what puzzles me is the different execution time depending how data have been stored in CS.
I will open another thread to explain may concern