Benjamin Brock
Benjamin Brock
Okay, silly me, `-D_GLIBCXX_USE_TBB_PAR_BACKEND=0` seems to fix my compilation. Does this mean recent versions of `libstdc++` will automatically bring in legacy TBB by default, breaking oneTBB, even if I don't...
My understanding is that in the current public release of DPC++, the `-D_GLIBCXX_USE_TBB_PAR_BACKEND=0` is always necessary when using DPC++ asynchronous algorithms. This does seem to have been fixed in the...
Never mind--I didn't have the DPDK modules loaded properly.
I'm still running into issues, although I'm pretty sure the kernel modules are loading correctly now. ```Bash ubuntu@ip-172-31-44-242:~$ sudo modprobe rte_kni ubuntu@ip-172-31-44-242:~$ sudo modprobe urdma ubuntu@ip-172-31-44-242:~$ lsmod | grep rte...
Thanks for creating this issue. This is a result of the design of reple, which currently prints _new lines of output_. This means that if new output appears without a...
There's a feature already in repl, line epilogues, that could potentially fix this by adding a newline at the end of each line. There's a question about whether this is...
Also, someone emailed me with a potential fix to nondeterministic output, which is related, and could potentially provide a full fix here. This would involve printing out tags before and...
The `gemm_usm` example included with production oneMKL also throws the same error on PVC. ```bash (base) bbrock@sdp125071:~/src/issues/oneMKL_gemm$ ./gemm_usm ######################################################################## # General Matrix-Matrix Multiplication using Unified Shared Memory Example: # #...
This is actually running fine on Borealis, so I think this might be a configuration issue with ORTCE. I will get in touch with the people who run the cluster.