Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200822130653.GA12682@openwall.com>
Date: Sat, 22 Aug 2020 15:06:54 +0200
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: OpenCl lib not detected

On Sat, Aug 22, 2020 at 12:13:07PM +0000, freeroute wrote:
> I think its solved.

Thank you for letting us all know what worked for you.

> as written in /john/doc/INSTALL-OPENCL file.

I guess you mean README-OPENCL.

> "The new autoconf (./configure) should find your OpenCL installation and enable it.  If it doesn't, you may need to pass some parameters about where it's located..."
> 
> So I used these parameters:
> "./configure LDFLAGS=-L/usr/local/cuda-11.0/targets/x86_64-linux/lib CFLAGS=-I/usr/local/cuda-11.0/targets/x86_64-linux/include
>     make -sj4"
> 
> and now I have GPU support.

Great, but unfortunately overriding CFLAGS drops our usual content from
there, which means you get a build without compiler optimizations.  Such
build will use CPUs extremely inefficiently.  (I've just confirmed that
this is the case.)

I recommend that you use CPPFLAGS instead of CFLAGS there.  And yes, I
see README-OPENCL gives an example with CFLAGS - we need to fix that.

In fact, given your posting with the full configure output, it appears
that you only need the LDFLAGS there.

> I think the autoconf did not find the OpenCL lib.

It looks so.  Per the full configure output, it looks like the OpenCL
headers you have in /usr/include/CL are working OK.

As an alternative to passing explicit LDFLAGS, you could ensure the
library path is specified somewhere under /etc/ld.so.conf.d, and run
"ldconfig" to ensure the paths specified in there are in effect.

Alexander

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.