Closed SimonDanisch closed 7 years ago
Together with https://github.com/JuliaGPU/CUDAdrv.jl/pull/19 this passes tests on v0.5 locally
Did this fix the isempty(CUDArt.cuda_ptrs)
test failure (which is strange, because when it does fail the array only contains (WeakRef(nothing), 0)
? If it did, I messed something up...
I am going to merge this soon so that we can tag a new version. Any last reviews/changes?
Any last reviews/changes?
Not from my side. But then again, CUDArt's test coverage is pretty meager... @timholy: could any of this break your usage of CUDArt?
It's already broken, but that's my fault for not writing enough tests. At some point I'll try to figure out what needs to be done to fix it.
The most urgent thing is to not stand in the way of the truly awesome work going on here, so I enthusiastically endorse going forward with whatever changes you folks want and tagging new releases. Just make sure that when you do tag a new version, it's 0.3.0; very likely, I will later tag a 0.2.4 release that incorporates some of the nice things that @maleadt added before rebasing the package on CUDAdrv. Then when I have a little more time, I'll update my own code to work with the new version and figure out what tests need to be added to CUDArt to make sure it stays working.
Onwards and upwards!
Excellent. I will merge this then and let it sit for a bit on master, before tagging v0.3
I'd like to thank whichever saint(s) got so much of CUDArt working again for my purposes. With the (tiny) fixes in #88 and #89, I seem to be back in business again in my lab's internal code again on current versions of CUDArt.
tests pass a bit better!
Please especially review the copy changes... I just blindly changed the signatures until it worked :D That code seems to be in need of refactoring!