RISC-V Support

[ChihweiLHBird]

Describe the problem

I know it's still far away from real usage, but any plan?

Describe the solution you'd like

No response

Additional context

No response

[AenBleidd]

RISC-V platform is not so mature yet, and doesn't really have a significant share on end-user devices.
I keep this is backlog and will review from time to time to decide what should we do with it.

[SETIguy]

Is there really anything to do beyond adding the platform names and strings to the official list (if there still is one)? It seems that the linux distros that include BOINC and support RISC-V already supply boinc-client and boinc-manager packages. No projects yet have applications for 'riscv64-unknown-linux-gnu' or 'riscv32-unknown-linux-gnu' so anyone wanting to crunch would need to use anonymous platform. I haven't checked if the BSDs do, but I would guess they also have ports.

…

On Fri, Jan 10, 2025 at 2:35 AM Vitalii Koshura ***@***.***> wrote: RISC-V platform is not so mature yet, and doesn't really have a significant share on end-user devices. I keep this is backlog and will review from time to time to decide what should we do with it. — Reply to this email directly, view it on GitHub <#5983 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACS5ZMWPJB2JG5L62J7HDZ32J6PA7AVCNFSM6AAAAABULPXFQWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKOBSGMZTCMBYGQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[AenBleidd]

Is there really anything to do beyond adding the platform names and strings to the official list (if there still is one)?

I'm not really into this particular type of CPUs, but it would be nice also to detect an available set of instructions (like we have now with the CPUID) if there are any differences between standards.
Besides that (including the platform detection like you already mentioned) I see no other changes that should be done on our side to support RISC-V platform.

[SETIguy]

Following the linux kernel tradition of never doing the same thing in the same way even once, /proc/cpuinfo contains an 'isa' and also a 'hart isa' field that would be the equivalent of the flags or features field on x86 cpus and it is underscore delimited. isa : rv64imafdch_zicbom_zicboz_zicntr_zicsr_zifencei_zihintntl_zihintpause_zihpm_zfa_zba_zbb_zbc_zbs_sstc hart isa : rv64imafdch_zicbom_zicboz_zicntr_zicsr_zifencei_zihintntl_zihintpause_zihpm_zfa_zba_zbb_zbc_zbs_sstc If I feel energetic I'll modify hostinfo_unix.cpp and also check what the BSDs are doing... Eventually you may see a pull request.

…

On Sat, Jan 11, 2025 at 5:45 PM Vitalii Koshura ***@***.***> wrote: Is there really anything to do beyond adding the platform names and strings to the official list (if there still is one)? I'm not really into this particular type of CPUs, but it would be nice also to detect an available set of instructions (like we have now with the CPUID) if there are any differences between standards. Besides that (including the platform detection like you already mentioned) I see no other changes that should be done on our side to support RISC-V platform. — Reply to this email directly, view it on GitHub <#5983 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACS5ZMTRDIZ2DOTPIVOOR5D2KHCKPAVCNFSM6AAAAABULPXFQWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKOBVGUZTQMZUG4> . You are receiving this because you commented.Message ID: ***@***.***>