[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20120424231235.GA14185@openwall.com>
Date: Wed, 25 Apr 2012 03:12:35 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: GPU eager execution (was: New patch for OpenCL SHA-512)
On Thu, Mar 22, 2012 at 08:00:54AM +0100, magnum wrote:
> I believe both AMD's and Nvidia's compilers generate code for eager
> execution in many cases (that is, unless the conditional blocks are too
> large or too complex). I don't know much details but I suppose the real
> problems begin when they do not: Normally all threads in a workgroup
> execute in sync.
FWIW, this presentation on GCN:
http://developer.amd.com/afds/assets/presentations/2620_final.pdf
gives some generated code samples that show that there's some hardware
support for eager execution - "exec mask" (tells which of the work-items
execute for real?) and the s_cbranch_vccz and s_cbranch_execz
instructions (branch for real if the "exec mask" would remain full or if
it became zero, respectively?)
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.
