Re: CFD: the future of ports on powerpc64/12 and powerpc64/11

On 5/28/21 14:33, [email protected] [email protected] wrote:
> [attempting re-send from a properly subscribed-to address.]
> 
> For a long time I have been attempting to keep ports building on
> powerpc64/12.  This, along with mips*/12, is still stuck on having
> GCC4.2.1 in base.  (Of course, in 13/14, we are based on clang.)
> 
> (I have not even looked at the state of ports on 11 in over a year.)
> 
> Although most of the individual problems are not that hard to fix or
> work around, the fact is that I have become overwhelmed by the number
> of them.  This is both for existing ports where updates switch to taking
> advantage of c11 or c++11 (or later) features, but, most notably, for
> the number of new ports added every week.
> 
> The problems noticed in the last 1-2 months are:
> 
>   math/openblas (I am told there is an upstream fix)
>   math/mpdecimal (affects lang/python* but it can be worked around)
>   math/clp
>   devel/indi
>   devel/py-gobject3 (also affects python)
>   print/libraqm
>

    While I do applaud your efforts and am thankful that FreeBSD even
exists on the ppc64 ( big endian? ) platform I suspect there are very
few actual users of 12.x in the wild. If any. I guess traffic out of
ports may reveal some statistics. I have tried to use FreeBSD 13 and of
course CURRENT on an old PowerMac quad PPC970MP.  The results a very
underwhelming given that there is no way to use more than a single CPU
core. That has been discussed and explained in depth here before. I do
not see that ever being fixed as the much more popular POWER8 and POWER9
based systems are a far better target to focus on.  In short, dropping
support on powerpc64/12 seems perfectly reasonable. Unless of course
someone with a spare million dollars shows up at the door and really
wants these things to work.


-- 
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken
GreyBeard and suspenders optional