[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150824053819.GA21430@openwall.com>
Date: Mon, 24 Aug 2015 08:38:19 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Formats using non-SIMD SHA2 implementations
On Mon, Aug 24, 2015 at 11:43:50AM +0800, Lei Zhang wrote:
> BTW, I encountered an issue testing with OpenMP. I used some stack arrays as vector buffers, but gcc-5 cannot align them properly when OpenMP is enabled (thus giving me segfaults). This seems to be a known issue in some older gcc: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=16660
This isn't OpenMP-related, and it isn't an issue in some older gcc -
even though (I am surprised) they're keeping the "bug" open - I guess
they're still considering enhancing gcc to deal even with that.
The stack is only guaranteed to be 16-byte aligned (and on 32-bit x86
even this wasn't the case until gcc 2.95, and some systems didn't
provide this alignment even later). When you use AVX2, you need 32-byte
alignment, but you can't reliably have that on the stack (unless you do
it on your own: over-allocate space and align the pointer as necessary).
> I also used stack arrays as vector buffers in 7z, but somehow it just works fine.
Luck. Arguably, that's bad luck (hides a real problem from you).
> If this cannot be solved, I could switch to global arrays anyway.
This can be worked around in a number of ways. But you really shouldn't
expect the system to just do it for you.
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.
