IRC channel logs

<rekado>civodul: thank you, I’ll give it a try

<PurpleSym>Is there a reason we’re packaging an OpenCL ICD, which disappeared years ago from the internet, and not the official Khronos Group one?

<civodul>PurpleSym: you mean mesa-opencl-icd? no idea

<civodul>BTW, congrats on getting rocm in place!

<civodul>we should try to build pytorch with it

<rekado>…did I package this? If so: I didn’t (and still don’t) understand how this OpenCL stuff works.

*rekado checks the git log

<rekado>hah, yes, it was me.

<rekado>so… no reason! Just plain old ignorance.

<rekado>FYI: I have picked up work on this on-demand reproducible AWS VMs with RStudio or Jupyter Notebooks, all running in Guix environment containers.

<rekado>we will be testing this internally for a while and then begin publishing some of the components.

<rekado>there are *so* many quirks when AWS is involved…

<PurpleSym>civodul: No, I mean ocl-icd.

<PurpleSym>My guess would be that Khronos’ ICD did not exist back then.

<PurpleSym>rekado: It’s pretty simple: You link against a generic libOpenCL.so, which then forwards your API calls to the actual implementation (and provides an API to query them).

<rekado>R 4.1.1 is out. I’ll prepare the upgrade tonight or tomorrow and build a dedicated R branch on ci.guix.gnu.org

<drakonis>bit of a silly question, guix-hpc intersects with something such as guix-science, yeah?

<rekado>yes

<rekado>because HPC users are often guix-science users

<drakonis>i see

<drakonis>i did not mean the channel

<drakonis>but rather another interest group

<drakonis>but i understand

<drakonis>the intersection remains

<drakonis>a guix interest group for general science related packages

<rekado>yes, that’s how I understood your question, but my answer repeated the ambiguity, because it is the same either way

<drakonis>aye

IRC channel logs

2021-08-10.log