Re: 01/01: gnu: fftw: Build SIMD codelets.

[Eric Bavier]

Hello Mark,

No problem.  I don't mind some parallel work that results in more confidence in 
the changes.

I saw your commit 69d5909, which looks good to me (you even untangled my nasty 
quoting, which I often can't seem to get right).

Sorry for the slow response, and thanks for finding the issue in the first 
place.

Eric Bavier, Scientific Libraries, Cray Inc.

________________________________________
From: Mark H Weaver <address@hidden>
Sent: Tuesday, May 8, 2018 18:29
To: Eric Bavier
Cc: address@hidden
Subject: Re: 01/01: gnu: fftw: Build SIMD codelets.

Hi Eric,

Eric Bavier <address@hidden> writes:

> Ok, if you feel confident, that's fine with me.
>
> Otherwise, I have a patch that I've tested on x86-64, and in the
> middle of testing for armhf that I can push in a few hours.  I could
> easily add a commit that remove the "--enable-armv7a-cntvct" flag.

Sorry for the duplicate work, but I already pushed my fixes in commit
69d5909e032e2fba57814ea9db52389d384d9341 to core-updates.  I tested them
on x86_64, but nowhere else.  It's not ideal, and I certainly wouldn't
do such a thing on 'master', but given the large number of affected
builds and my uncertainty about when you would fix it, I didn't want to
wait any longer.

FYI, I also removed "--enable-armv8-cntvct-el0" on 64-bit ARM, because
<https://github.com/FFTW/fftw3/blob/master/README-perfcnt.md> suggests
that as on 32-bit ARM, the relevant instruction is not normally
available from user mode.

I removed "--enable-mips-zbus-timer" on MIPS as well, because I suspect
that it's not supported on Loongson devices, although I'm not certain.

If you see any problems with my commit, please let me know.

     Thanks!
       Mark