Issues building any of the stable models

[ahmedalalousi]

Hi,

Having issues building any of the published stable models. Tried the airport and torus models: same error (output of make attached). Seems that g_tw_npe has either been moved, redefined ...etc., since this error is consistent across all models.

Any help gratefully received.

Ahmed.
outfile.txt

[helq]

Hi Ahmed, yes, g_tw_npe was removed a while ago. Once upon a time, ROSS supported multithreaded execution and MPI. Nowadays, it only supports parallelization via MPI. That variable indicated how many PEs were spawn in a specific MPI rank. That value is now 1. So, you can replace it for 1 anywhere you find it.

As for g_tw_pe, it was a pointer to an array of PEs. It is now just a pointer. So, the brackets in tw_kp_onpe(kpid, g_tw_pe[0]) should be removed: tw_kp_onpe(kpid, g_tw_pe).

There might be a couple more issues with ROSS-Models, it was archived after all right at the time when the changes above were being implemented.

There are a couple of models that work out of the box with the current version of ROSS, if you wanna try them. They are meant to be very simple models too: https://github.com/helq/highlife-ross/ and https://github.com/sanjaychari/ROSS-coffeeshop.

I hope that helps, don't forget to check out the FAQ

[ahmedalalousi]

Many thanks for this. My interest in the torus model in particular, is that I'm researching characteristics of this topology, and ROSS looked promising for a model like that, out of the box - looking to fully understand latency and throughput characteristics, versus network diameter, size and routing protocol. It still wouldn't compile after I made the changes you suggested: now complaining about an implicit declaration of tw_getpe (arning: implicit declaration of function ‘tw_getpe’; did you mean ‘tw_getkp’? [-Wimplicit-function-declaration] 1056 | pe = tw_getpe(kpid % 1);) The only models which did build successfully are phold, phold_test and coffeeshop. Thanks again, Ahmed.

…

On Fri, 15 Mar 2024 at 18:46, helq ***@***.***> wrote: Hi Ahmed, yes, g_tw_npe was removed a while ago. Once upon a time, ROSS supported multithreaded execution and MPI. Nowadays, it only supports parallelization via MPI. That variable indicated how many PEs were spawn in a specific MPI rank. That value is now 1. So, you can replace it for 1 anywhere you find it. As for g_tw_pe, it was a pointer to an array of PEs. It is now just a pointer. So, the brackets in tw_kp_onpe(kpid, g_tw_pe[0]) should be removed: tw_kp_onpe(kpid, g_tw_pe). There might be a couple more issues with ROSS-Models, it was archived after all right at the time when the changes above were being implemented. There are a couple of models that work out of the box with the current version of ROSS, if you wanna try them. They are meant to be very simple models too: https://github.com/helq/highlife-ross/ and https://github.com/sanjaychari/ROSS-coffeeshop. I hope that helps, don't forget to check out the FAQ <https://ross-org.github.io/faq/usage-faq.html> — Reply to this email directly, view it on GitHub <#189 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA47TA5ONHHPIG7LSLEN6UDYYM6ZZAVCNFSM6AAAAABEWONLNWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMBQGI2TGNRRGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[helq]

tw_getpe was also removed. Try replacing that line with "pe = g_tw_pe".

If you wanna try a more complete, although unmaintained, torus network model, check out CODES. I suggest you run the script CODES-compile-script.sh to get a working CODES binary. Check our most recent wiki entry, which focuses on the dragonfly model but gives you a nice tar.gz with a couple of examples of how to run a dragonfly model. Once you have the binary and can run an experiment with the dragonfly model, you can try running a torus network by unpacking the example attached to this comment :).

I hope that helps

torus-64.tar.gz