Nicolas Miller

> lxqt-panel 0.9.0-1 > liblxqt 0.9.0-2 Are these the packages in the `community` repository ? If yes, you should try with the `*-git` packages from the aur as the 0.9...

It does work. There are multiple problems though. First it's only temporary, as it's not stored in lxqt config files, and at least openbox doesn't update its configuration accordingly, so...

Hello @dnhsieh-intel, so I've had a look at this and I could reproduce the issues, I'm not getting segfaults but I am getting the incorrect results. I don't think it's...

Okay so I have figured out what is going on, with the way `interop_task` and `host_task` are currently defined they are required to be synchronous. Meaning that starting asynchronous work...

> In my view there are two issues here: (1) functional failures (wrong answers) and (2) segfaults. For now I am only focused on (1). @andrewtbarker do you still see...

I see, that's interesting, it could have to do with the context stuff in https://github.com/oneapi-src/oneMKL/pull/136 still. However I'm still unable to reproduce it locally with the latest DPC++ compiler so...

Sure, I'm doing the following: ``` export PATH=/path/to/llvm/build/bin:$PATH export LD_LIBRARY_PATH=/path/to/llvm/build/lib cmake -GNinja .. -DENABLE_CUBLAS_BACKEND=True -DENABLE_MKLCPU_BACKEND=False -DENABLE_MKLGPU_BACKEND=False -DREF_BLAS_ROOT=/usr -DTARGET_DOMAINS=blas ninja ctest ``` And everything passes: ``` 100% tests passed, 0 tests...

I've rebuilt with the revision you provided (`cae91b2b13af3915f4b6f990216146adde973903`), and I'm still not getting any segfault, not sure if there's something wrong in my setup or if I'm just being lucky....

> @npmiller Which CUDA version are you using? 11.x? We have 10.2 installed. Do you think that would make a difference? I am running with 11.6, I've tried to switch...

That's pretty strange, I still tried the reproducer and your instructions but I'm still not getting any segfault. You might be right that it's some sort of local configuration issue