Re: [musl] Linux distribution for using RISC-V musl?

[Nagakamira <nagakamira@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1749 bytes --]

lto on riscv64 is broken I will revert the commit when I will have free
time and make it work

пн, 12 апр. 2021 г., 13:30 ardi <ardillasdelmonte@gmail.com>:

> Ataraxia looks really interesting: LLVM+musl+Toybox is the kind of
> distribution I’d love to have. Is there any page explaining how to install
> Ataraxia on RISCV? The docs only cover x86 and PPC, and besides the latest
> RISCV release I found is from 2019 or so...
>
> Also, regarding LLVM LTO, what’s the ticket for following that bug? I
> searched for it and all what I found is.
>
> Thanks!!
>
>
> Nagakamira <nagakamira@gmail.com> wrote:
>
>> Try ataraxia gnu/linux! But llvm lto have to be fixed.
>>
>> вс, 11 апр. 2021 г., 00:17 ardi <ardillasdelmonte@gmail.com>:
>>
>>> Really useful!! Thanks a lot!
>>>
>>> On Sat, Apr 10, 2021 at 2:24 PM Szabolcs Nagy <nsz@port70.net> wrote:
>>> >
>>> > * ardi <ardillasdelmonte@gmail.com> [2021-04-10 14:08:31 +0200]:
>>> > > Also, I've read that QEMU is able to emulate multicore RISC-V CPUs.
>>> > > Does musl support multicore RISC-V in 64bit? I mean, if I configure a
>>> > > multicore RISC-V in QEMU and I try to build a parallel loop using
>>> > > OpenMP for example, will such loop be parallelized if I'm using musl?
>>> >
>>> > yes.
>>> >
>>> > this has not much to do with the libc though.
>>> > openmp uses pthreads, pthreads uses clone and
>>> > clone is implemented by the os in some way.
>>> > whether things are parallel is not in the hands
>>> > of the libc and the logic is not target dependent,
>>> > the libc works the same way on x86_64 as on riscv.
>>> > there can be target specific synchronization bugs
>>> > in the libc, but those are just bugs.
>>> >
>>>
>>

[-- Attachment #2: Type: text/html, Size: 2971 bytes --]