[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAKtfLcvqJocNkFRQpDQD5+6tCA3DbU2fDoMmcF8ECb0N4BY3RA@mail.gmail.com>
Date: Sat, 8 Dec 2012 20:48:19 -0500
From: Alain Espinosa <alainesp@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: GCN: indexed access to VGPRs
On 12/8/12, Solar Designer <solar@...nwall.com> wrote:
> On page 53 in AMD_Southern_Islands_Instruction_Set_Architecture.pdf
> (SHA-1: c64c24077eab93b070d291da3624a8d988d70d5d), there's a mention
> that VGPRs may be indexed - something I had thought was not possible.
I think Nvidia cards supports this too, but using indexed registers is
slower than using "static" registers. In some fast kernels i get a
~50% speedup changing to normal registers.
> I don't know if there's currently a way to access this functionality
> without going to a level lower than OpenCL or not, though.
In OpenCL when you use a private array the compiler may use indexed registers.
saludos,
alain
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.
