Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 3 Jan 2013 19:54:06 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: Rejecting hashes in valid() due to memory allocation failures?

On 3 Jan, 2013, at 19:34 , Lukas Odzioba <lukas.odzioba@...il.com> wrote:
> 2013/1/3 magnum <john.magnum@...hmail.com>:
>> It's a problem with large arrays, especially on GPU. But sure, for most CPU formats you could use mem_alloc_tiny. But then you could also use a static array. I think most formats that allocate dynamically (typically OpenMP aware formats) should eventually be converted to mem_alloc/MEM_FREE. But that is lower prio.
> 
> We've got mem_alloc_tiny for opencl_mscash2 format.

I consider that a bug. Soon, every format will get a done() in bleeding-jumbo and then we can review these things. And after that, finally, we'll be able to use --test for GPU builds and include all GPU formats.

> I did some minor changes mostly in opencl/cuda formats.
> I also added mem_calloc(size) to memory.c/h, not a big deal but helps
> clean tables for candidates with just one line of code.
> If our mem allocation functions were macros we could easily add
> __FILE__ and __FUNCTION__ to debug output when something goes wrong
> like check_mem_allocation(inbuffer,outbuffer) does... just loose idea.
> Of course it is not needed for devs, but helps a lot when user submits
> error logs.

Thanks, committed now. I will start making use of that macro right away, I recently added some callocs in Dhiru's pbkdf2 formats.

magnum

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.