Consider using existing environments?

[mithro]

There are a number of other projects to distribute tooling;

• Conda - used by LiteX-BuildEnv - https://github.com/timvideos/conda-hdmi2usb-packages
• Apio - used by icestudio - https://github.com/FPGAwars/apio
• BitBake / OpenEmbedded - https://github.com/nathanrossi/meta-hdl
• Nix - https://git.m-labs.hk/m-labs/nix-scripts
• Edalize / FuseSoC - https://github.com/olofk/fusesoc

Is there a way to collaborate and/or reuse stuff here?

[mballance]

Funny you should mention this, since I just became of Conda last night. Definitely something I’ll investigate to see what can be leveraged!

[mithro]

@mballance - Happy to chat with you while at LatchUp about the advantages / disadvantages of each.

[krupan]

I just started developing my own tool similar to edapack. Then I found edapack and figured I should probably just use it. Then I saw this discussion about other existing tools. @mithro should I look into one of the others instead? I'm most interested in building and maintaining my own clones of many of these projects (icarus, verilator, ghdl, etc.) that have my own patches applied (at least until they get accepted upstream). I need to build the EDA tools for redhat/centos and ubuntu (various releases of each). I started looking at conda and it looks promising, except I'm not sure if I need any of the python stuff it seems to include

If this is not the right place for this discussion feel free to email me directly: bmurdock@gmail.com, or suggest some other forum/chat/mailing-list

[mballance]

Hi,
Thanks for contacting me about this! I started EDAPack a while before I was aware of Conda, but have since decided that Conda fills ~95% of what I need an EDA toolchain manager to do. The couple things that Conda doesn't do well (IMO) that an EDA toolchain manager could be called on to do are to easily manage multiple versions of a given tool (eg I need to use different synthesis-tool versions for different projects), and to easily manage 3rd-party tools (eg Quartus, Vivado, etc).
Consequently, I'd highly recommend Conda for your work here.
You raise the point of Conda including Python, and that is a bit of an overhead. Check out 'minconda' for a smaller-footprint install. Another point, here, is that Conda might be available pre-installed with continuous-integration services. Azure, at least, provides a very simple way activate a pre-installed Conda (have a look at one of my pipelines file here: https://github.com/mballance/fwrisc/blob/master/azure-pipelines.yml).
Conda packages definitely support custom patches applied, so you would be covered there. @mithro has a GitHub repository here (https://github.com/litex-hub/litex-conda-eda) with package recipes for a set of tools (eg Icarus, Verilator, Yosys, etc). The broader Conda community also offers a very large set of packages.
Conda bundles enough 'stuff' with it that you shouldn't need to create separate builds for different Linux platforms. One build should be sufficient, which is a very nice thing!
Hope the info above is helpful, and would be very happy to have another contributor to making OSS EDA tools easily accessible!

Best Regards,
Matthew

[krupan]

Thanks for replying! That's good to hear about conda, I'll definitely give it a look. By the way, I'm doing this as part of my new business selling and supporting open source EDA tools: https://millenniumsoftwaredesign.com/

…

On Wed, Feb 5, 2020, 6:26 PM Matthew Ballance ***@***.***> wrote: Hi, Thanks for contacting me about this! I started EDAPack a while before I was aware of Conda, but have since decided that Conda fills ~95% of what I need an EDA toolchain manager to do. The couple things that Conda doesn't do well (IMO) that an EDA toolchain manager could be called on to do are to easily manage multiple versions of a given tool (eg I need to use different synthesis-tool versions for different projects), and to easily manage 3rd-party tools (eg Quartus, Vivado, etc). Consequently, I'd highly recommend Conda for your work here. You raise the point of Conda including Python, and that is a bit of an overhead. Check out 'minconda' for a smaller-footprint install. Another point, here, is that Conda might be available pre-installed with continuous-integration services. Azure, at least, provides a very simple way activate a pre-installed Conda (have a look at one of my pipelines file here: https://github.com/mballance/fwrisc/blob/master/azure-pipelines.yml ). Conda packages definitely support custom patches applied, so you would be covered there. @mithro has a GitHub repository here (https://github.com/litex-hub/litex-conda-eda) with package recipes for a set of tools (eg Icarus, Verilator, Yosys, etc). The broader Conda community also offers a very large set of packages. Conda bundles enough 'stuff' with it that you shouldn't need to create separate builds for different Linux platforms. One build should be sufficient, which is a *very* nice thing! Hope the info above is helpful, and would be very happy to have another contributor to making OSS EDA tools easily accessible! Best Regards, Matthew — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AAPBETWJC6X3L6LROGHDXE3RBNRNFA5CNFSM4HKZ24WKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEK5S4NI#issuecomment-582692405>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAPBETV43EJ32D6ORBDO3ADRBNRNFANCNFSM4HKZ24WA> .

[mithro]

FYI - @mballance - Conda supports multiple "environments" and each environment can have different versions of the tools installed.