Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 27 Mar 2010 17:35:25 +0100
From: "Magnum, P.I." <>
Subject: Re: bench.c fix for extreme c/s figures

On 03/27/2010 01:27 PM, Solar Designer wrote:
> On Sat, Mar 27, 2010 at 01:12:26PM +0100, Magnum, P.I. wrote:
>> The Jumbo patches use long long's anyway so I now fixed bench.c without
>> the half-assed try to keep using the int64. I just uploaded fullmpi-6 to
>> the wiki. Works fine on Linux 32 and 64, I hope it works for W.A too.
> We'll see it it works for W.A. or not.  Meanwhile, I notice that it
> will likely fail on 32-bit builds, because neither "count" nor "clk_tck"
> is cast to a 64-bit type before you multiply them.  You will likely want
> to add a cast to (unsigned long long) in two places.  Also, I suggest
> that you use "unsigned long long", not just "long long".
> Thanks,
> Alexander

Ah, yes. I never pushed my 32-bit tests towards that high numbers.
There are most likely more of such ignorance in my patches. It works 
fine for W.A though.

I'll add the casts and have a look for other similar misses. Some MPI 
environments don't support reduction for unsigned long longs so I learnt 
to avoid them, but in this case no reduction is used so I guess it's a 
good idea too.


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.