Pekka Jääskeläinen

Do our Arm maintainer @jianbinfang have an idea here?

Does it work with older LLVM 11 (I doubt) or older PoCL? basic device isn't tested as much as pthread, so this is not surprising. We should test it more,...

@isuruf @jrprice any ideas? This is the last "regression" we know of (pocl 1.16 + LLVM 11 vs. pocl 1.15 + LLVM 10).

Shall we just mark this as a known regression for the new release?

Very nice! Thanks for your willingness to contribute back to the open source. I'm looking forward for finding the time to review this.

@Oblomov I agree. This is what is done between basic and several other drivers (at least pthread), the ugly C style of "inheritance".

One option would be to rename 'pthread' to 'multithread'/'mt'/'multicore'/mc' and just implement the TBB as an alternative code path for its thread scheduler in it. If the TBB works, I...

Looks much better now! We'll take a closer look in the group and test it out. * Congrats on the IWOCL paper! Can you add a link to it in...

> I've seen different copyright claim styles: "pocl developers", a person, an institution, or both. What is the preferred way of claiming copyright at the beginning of each file? For...

Otherwise this now looks good and I started testing it (with Michal's fixes on top). I noticed this breaks build when there are no host devices enabled. I found it...