[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20141218162454.BPX17.19903.imail@eastrmwml114>
Date: Thu, 18 Dec 2014 16:24:54 -0500
From: <jfoug@....net>
To: john-dev@...ts.openwall.com
Subject: Re: ARM NEON
---- Solar Designer <solar@...nwall.com> wrote:
> Yes, or we may detect NEON in some other way, e.g.:
>
> https://mailman.videolan.org/pipermail/vlc-commits/2011-July/007663.html
>
I do not believe there is any probe there. There simply is a flag where the user can list --enable-NEON or something like that. At this time, we have auto detection probing that actually builds code, and runs it on the system to make sure that it will build and run. The code used is preposterously bad, BUT it works well enough to return a 0 error status if everything builds and runs, and will either fail to build, or core, if the machine can not handle it.
Yes, the other plumbing listed in the above post will have to be there (to relay the information that this is a NEON build to the rest of the build system). But the test magnum is referring to is an active probe.
Here is the probe code for AVX2 usability (with -mavx2 set on the CC command line)
#include <immintrin.h>
#include <stdio.h>
extern void exit(int);
int main(){
__m256i t, t1;
*((long long*)&t)=1;
t1=t;
t=_mm256_mul_epi32(t1,t);
if((*(long long*)&t)==88)printf(".");
exit(0);
}
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
