First we need to identify:

• upstream
• build documentation
• dependencies

@proppy see:

• https://github.com/hdl/containers/blob/main/debian-bullseye/xyce.dockerfile
• Xyce cannot be fetched from sandia.gov containers#56
• Get Xyce and the PDK playing nice together google/skywater-pdk#186
• https://github.com/Xyce/Xyce/blob/master/INSTALL.md

@umarcor thanks for providing all of those pointers.

It seems like one of the bigger deps "trilinos" is already present in conda-forge:
https://anaconda.org/conda-forge/trilinos

@proppy, note that you need to use version 12.12.1 of Trilinos (https://github.com/hdl/containers/blob/main/debian-bullseye/xyce.dockerfile#L54). Fortunately, it is available: https://anaconda.org/conda-forge/trilinos/labels.

Any progress on this? Is there something I can help here?

Any progress on this? Is there something I can help here?

@saicharan0112

If you want it for fedora/rhel you can have latest (including xyce):
https://copr.fedorainfracloud.org/coprs/rezso/HDL

Also the repo can be invoked within Docker for convenience, a short demo:
https://gist.github.com/cbalint13/1de29fbe77e8dedddd3861b5f49a1835

Thanks, @cbalint13 .

Thanks, @cbalint13 .

@saicharan0112 ,

In meanwhile, Xyce part comes with complete openmpi & mpich flavours for distributed simulations:

• https://copr.fedorainfracloud.org/coprs/rezso/HDL/build/5123262/
• https://copr.fedorainfracloud.org/coprs/rezso/HDL/build/5122252/

This is very interesting and useful for Fedora/RHEL users, @cbalint13 . I wish we had something similar for most used distros (ubuntu or debian in general). I am seeing conda packages as more like a platform independent way for installations. Not sure if the conda packaging mechanism keeps track on dependencies like its done with Fedora and the conda-forge.

@saicharan0112

This is very interesting and useful for Fedora/RHEL users, @cbalint13 . I wish we had something similar for most used distros (ubuntu or debian in general).

• Native packages are doable but there is no build-farm like COPR (free) that can target ubuntu / debian world.
• The same is with conda (no farm), so the burden of recipes is in charge of the users having their compile times.

I am seeing conda packages as more like a platform independent way for installations. Not sure if the conda ?packaging mechanism keeps track on dependencies like its done with Fedora and the conda-forge.

• There is a very rigorous QA assurance of such native packages (.rpm/.deb) are distro compliant on their own.
• One can easily compose his own docker images out of these rhel/fedora packages (without any compile times).

@proppy

Small clarification: conda-forge actually has a farm on azurehttps://github.com/conda-forge/magic-feedstock)

That is interesting. The very builders are kind of (re)purposed azure, etc. instances ?

(we rely on it for the magic package
And this repo has some continuous integration running on GitHub actions that upload packages to a dedicated channel: https://anaconda.org/LiteX-Hub

Magic / Litex (also) are small ones, many HDL/EDA build space reaches >100 Gbyte and some >12h of build times.
What resources would be on these instances for regular builds of bigger (i.e. Xyces / Trilinos, two good examples) ?

That is interesting. The very builders are kind of (re)purposed azure, etc. instances ?

I think they are using https://azure.microsoft.com/en-us/products/devops/pipelines/.

What resources would be on these instances for regular builds of bigger (i.e. Xyces / Trilinos, two good examples) ?

For conda-eda specifically, we discussed seting up custom github runner in #184.

@proppy If it is of any help while building conda package for xyce - Xyce/Xyce#4 (comment)

This works very smooth. I did this on 3 machines.

@saicharan0112 Thanks! I think we're doing something similar here already https://github.com/hdl/conda-eda/pull/263/files#diff-131443e74262f22bf247f65d935f723616400829a221a5437dac94721f4d1b40