Question M1 Mac/Arm Native DC Projects

[jdiddy]

Hey all, aside from Einstein at Home, is anyone aware of DC projects that run natively on M1 Macs - not through Rosetta.

Thanks!

 

[StefanR5R]

You could click through the (incomplete) list of projects at https://boinc.berkeley.edu/projects.php, as far as they show a Mac icon, and visit the apps.php page of each. But I suspect there aren't any besides E@h FGRP5.

On second thought, such a search can probably be scripted. Maybe later today.

 

[ericlp]

At the moment, F@H is native for CPU cores, not GPU. That's the real question, is how to get the GPU cores running.

https://www.reddit.com/r/Folding/comments/tv4x6k

 

[StefanR5R]

On second thought, such a search can probably be scripted. Maybe later today.

DENIS@home has got one in beta test:
https://denis.usj.es/denisathome/apps.php
(Might come out of beta soon, once the project is un-paused.)

Nothing else besides this and Einstein@home came up in my search through boinc.berkeley.edu's URLs and some others.

 

[Skillz]

Try looking at wuprop. They'll show anyone who's running Arm on their CPUs

 

[StefanR5R]

Try looking at wuprop. They'll show anyone who's running Arm on their CPUs

Wouldn't this require a "Mac OS running on ARM" application version of the WUProp client (which doesn't exist)? Or would an Intel Mac client show it as alternative platform of its host? — On the other hand, WUProp's result tables don't cover the Mac-on-ARM platform yet, only Intel Mac, PPC Mac, and Linux or Android on ARM.

 

[mmonnin03]

WUProp doesn't show some processor data thats not a standard AMD/Intel CPU.. None of my RPIs or Android phones show up as mine. There are some posts about missing data there.

Filtering down to computation time has an Apple CPU, not ARM, for PG Sophie Germain and TRP apps. Not even E@H

 

[jdiddy]

DENIS@home has got one in beta test:
https://denis.usj.es/denisathome/apps.php
(Might come out of beta soon, once the project is un-paused.)

Nothing else besides this and Einstein@home came up in my search through boinc.berkeley.edu's URLs and some others.

Thanks Stefan and everyone for their responses. Was hoping I missed something but I'll keep my M1 machines churning on Einsten for the moment. It's a shame because the M1 gets through CPU units in under 2 hours.

 

[StefanR5R]

M1 owners who have a development toolchain installed could take these steps: Active DC projects with (to them) interesting subject? --> Non-GPU, non-VirtualBox application? --> Mac Intel port already existing? --> Source code publicly available or available on request? ------------> Update the code to current toolchain and system libraries, test a native build with real workunits, offer the binary back to the project.

Usually, the main developers of the DC projects themselves don't have access to hardware and developer toolchains of all of the platforms which their project actually runs on, let alone could run on.

 

[StefanR5R]

PrimeGrid enabled cpuGFN16_mt cpuGFN17MEGA_mt…cpuGFN22_mt on the "Mac OS running on an ARM 64-bit CPU" platform now.

I suspect relative power efficiency of this could be pretty good at the smaller GFNs, but not as good as with the latest one or two NVidia GPU generations at the bigger GFNs.

 

[StefanR5R]

Correction and update: The native multithreaded CPU-only application is for GFN17…GFN22.
As of today, there is also a native GPU application version for for GFN16…DYFL, called opencl_Apple_M_GFN16…opencl_Apple_M_GFNEXTREME.