JackAKirk
JackAKirk
> @AidanBeltonS My logs were from a build of llvm from 3/19, and it looks like the multistream patch was committed on 5/17, but for sanity I'll try building from...
> Would this achieve asynchronous submissions? > Actually I think just from the observation that it fixes the test failures it must be effectively blocking future submissions (at least those...
I've now added blocking waits (using `queue.wait()`) to all places where they were missing. I think that this is required for generally correct behaviour: we should block all streams in...
I've found out that these cusolver functions are apparently asynchronous, even though the Nvidia documentations implies that they are synchronous: therefore I think that `depends_on` is behaving correctly. Also I've...
> I've found out that these cusolver functions are apparently asynchronous, even though the Nvidia documentations implies that they are synchronous: therefore I think that `depends_on` is behaving correctly. Also...
@AidanBeltonS could you check this is all OK? Thanks
By the way, I corrected a lot of typos that are still relevant for the new document here: https://github.com/intel/llvm/pull/6525/files Basically if you ignore the new sections I added there (####...
Looks pretty good overall. I just made some pretty simple suggestions.
> > > You have this inconsistency throughout the spec. In many places, you refer to the three matrices as a, b, and c. However, the use enum refers to...
LGTM. I just suggested some small changes.