https://open-source-silicon.dev logo
#openlane
Title
# openlane
r

Rob Taylor

02/13/2021, 3:29 PM
@User @User @User its coming togther… robtaylor/openlane@f84c4b0 (github.com)
🙌 2
Also I’ve given you three access to my repo, so you can hack there.
We should set up the docker hub tokens on the efabless repo then, it should be more or less working and ready for patch cleanup and merge
@Anton Blanchard I’m having problems with klayout on ppc64le - its currently at 5 hours and still building o_o
a

Amr Gouhar

02/15/2021, 5:38 PM
@Rob Taylor (cc: @Ahmed Ghazy, @Anton Blanchard): I've been working on converting the rest of the CI into github actions on the last days of the past week and I think it's done https://github.com/agorararmard/openlane/blob/actions-test-branch/.github/workflows/openlane_ci.yml (any input is very appreciated). I won't PR that now to efabless/openlane because we still need to figure out a way to secure some reliable self-hosted runners and we're still paying travis CI, so :--). Ideally, we'd just need to replace the docker-build job with what @Rob Taylor developed so far and everything would be working perfectly. P.S: we already have the docker hub secret tokens in the efabless/openlane repo as
DOCKERHUB_USER
and
DOCKERHUB_PASSWORD
.
r

Rob Taylor

02/15/2021, 7:59 PM
@Amr Gouhar Ok, i’ve updated my branch to use those secrets. Lets see if it all works!
@Amr Gouhar why do you want self-hosted runners?
@Amr Gouhar Also, its probably worth you running
make docker-cache CACHE_ID=shapebuild DOCKER_ID=efabless
👍 1
that will punt the artifacts over and speed things up a fair bit
a

Amr Gouhar

02/15/2021, 8:01 PM
@Rob Taylor: Github secrets will not be available to external PRs. So, it won't work (be reflected in efabless's dockerhub until we merge this PR into some branch).
r

Rob Taylor

02/15/2021, 8:01 PM
ah of course, makes sense
a

Amr Gouhar

02/15/2021, 8:03 PM
@Rob Taylor: We want the self-hosted runners for running the test cases themselves, because they take a lot of time if not parallelized properly and some cases require a lot of RAM. It's not blocking but it's nice to have (and we're working towards it).
r

Rob Taylor

02/15/2021, 8:04 PM
ahha, yes makes sense. builds are also very slow with only two cores.
It seems not too hard to hookup GCP servers.
or azure for that matter . I guess its just down to who’ll give the project the most free credits ;)
a

Amr Gouhar

02/15/2021, 8:07 PM
Yeah, we're going for GCP servers since the MPW project and all :--).
r

Rob Taylor

02/15/2021, 8:08 PM
I can get you contact with the MS for Startups team
a

Amr Gouhar

02/15/2021, 8:12 PM
Thanks! But, let's see how the GCP servers turn out first since they are already pseudo-available.
r

Rob Taylor

02/15/2021, 8:12 PM
nod let me know 🙂
@Amr Gouhar might make sense though to also have a AWS graviton instance for building/testing ARM , and maybe an IBM hosted ppc64le instance as well…
a

Amr Gouhar

02/16/2021, 1:19 PM
@Rob Taylor: Makes sense to me. But, I'm not the one calling the shots in this one, so I'll ask and tell you what they decide to do.
r

Rob Taylor

02/16/2021, 2:00 PM
@Amr Gouhar understood. Let them know that @Anton Blanchard is going to see if IBM could make a machine available to us
👍 1