search

hdl / conda-eda

Conda recipes for FPGA EDA tools for simulation, synthesis, place and route and bitstream generation.
https://anaconda.org/LiteX-Hub
Apache License 2.0
95 stars 26 forks source link

Add Xyce simulator as a package #210

Open mithro opened 2 years ago

proppy commented 2 years ago

First we need to identify:

umarcor commented 2 years ago

@proppy see:

proppy commented 2 years ago

@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

umarcor commented 2 years ago

@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.

saicharan0112 commented 2 years ago

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

cbalint13 commented 2 years ago

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

saicharan0112 commented 1 year ago

Thanks, @cbalint13 .

cbalint13 commented 1 year ago

Thanks, @cbalint13 .

@saicharan0112 ,

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

saicharan0112 commented 1 year ago

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.

cbalint13 commented 1 year ago

@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).

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.

proppy commented 1 year ago

2022年12月10日(土) 15:59 Balint Cristian @.***>:

@saicharan0112 https://github.com/saicharan0112

This is very interesting and useful for Fedora/RHEL users, @cbalint13 https://github.com/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.

Small clarification:

conda-forge actually has a farm on azure (we rely on it for the magic package: https://github.com/conda-forge/magic-feedstock)

And this repo has some continuous integration running on GitHub actions that upload packages to a dedicated channel: https://anaconda.org/LiteX-Hub

But I agree the Q&A is way less rigorous than a distro based farm like you have with fedora.

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).

— Reply to this email directly, view it on GitHub https://github.com/hdl/conda-eda/issues/210#issuecomment-1345159162, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAABJFFHGVKEPOLY6DRCEE3WMQS57ANCNFSM5VJB73AA . You are receiving this because you were mentioned.Message ID: @.***>

cbalint13 commented 1 year ago

@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) ?

proppy commented 1 year ago

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.

saicharan0112 commented 1 year ago

@proppy If it is of any help while building conda package for xyce - https://github.com/Xyce/Xyce/discussions/4#discussioncomment-169255

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

proppy commented 1 year ago

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