Project Home
Project Home
Documents
Documents
Wiki
Wiki
Discussion Forums
Discussions
Project Information
Project Info
Forum Topic - 6.4.1 and SMP: (6 Items)
   
6.4.1 and SMP  
Will 6.4.1 on x86 support more then 8 CPU?  
Re: 6.4.1 and SMP  
No.  Please send me your >8 way machine - stat! ;-)

Colin

Mario Charest wrote:
> Will 6.4.1 on x86 support more then 8 CPU?  
> 
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post28313
> 

-- 
cburgess@qnx.com
RE: 6.4.1 and SMP  
You can't get an >8 cpu machine???  IBM was ready to lends us a 16cpu machine to test with.  We declined ;-)

>8 cpu machine are money wise in a different class, but that will change in June with the arrival of AMD Istanbul six- 
core processor, making 12 cores machines (2 sockets) "affordable"


> -----Original Message-----
> From: Colin Burgess [mailto:community-noreply@qnx.com]
> Sent: April-29-09 8:45 AM
> To: ostech-core_os
> Subject: Re: 6.4.1 and SMP
> 
> No.  Please send me your >8 way machine - stat! ;-)
> 
> Colin
> 
> Mario Charest wrote:
> > Will 6.4.1 on x86 support more then 8 CPU?
> >
> > _______________________________________________
> > OSTech
> > http://community.qnx.com/sf/go/post28313
> >
> 
> --
> cburgess@qnx.com
> 
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post28314
> 
Re: 6.4.1 and SMP  
On Wed, Apr 29, 2009 at 08:26:02AM -0400, Mario Charest wrote:
> Will 6.4.1 on x86 support more then 8 CPU?  

The shipped binary won't. By changing the value of PROCESSORS_MAX (see
services/system/smpswitch.h), you can recompile to support up to 32 
without other source changes.

-- 
Brian Stecher (bstecher@qnx.com)        QNX Software Systems
phone: +1 (613) 591-0931 (voice)        175 Terence Matthews Cr.
       +1 (613) 591-3579 (fax)          Kanata, Ontario, Canada K2M 1W8
RE: 6.4.1 and SMP  

> -----Original Message-----
> From: Brian Stecher [mailto:community-noreply@qnx.com]
> Sent: April-29-09 9:12 AM
> To: ostech-core_os
> Subject: Re: 6.4.1 and SMP
> 
> On Wed, Apr 29, 2009 at 08:26:02AM -0400, Mario Charest wrote:
> > Will 6.4.1 on x86 support more then 8 CPU?
> 
> The shipped binary won't. By changing the value of PROCESSORS_MAX (see
> services/system/smpswitch.h), you can recompile to support up to 32
> without other source changes.

Brian, you mentioned that a while ago when I asked the same question about 6.4.  My comment was that there was no way I 
would ship a kernel I build myself without having tools to validate it thoroughly.  If it's suppose to be that simple 
and QNX isn't doing it, I cannot safely assume that is as simple as you make it sound ;-)  Notice how I cleverly avoided
 talking about marketing, management and business decision.


> 
> --
> Brian Stecher (bstecher@qnx.com)        QNX Software Systems
> phone: +1 (613) 591-0931 (voice)        175 Terence Matthews Cr.
>        +1 (613) 591-3579 (fax)          Kanata, Ontario, Canada K2M 1W8
> 
> _______________________________________________
> OSTech
> http://community.qnx.com/sf/go/post28321
> 
Re: 6.4.1 and SMP  
On Wed, Apr 29, 2009 at 09:24:16AM -0400, Mario Charest wrote:
> 
> 
> > -----Original Message-----
> > From: Brian Stecher [mailto:community-noreply@qnx.com]
> > Sent: April-29-09 9:12 AM
> > To: ostech-core_os
> > Subject: Re: 6.4.1 and SMP
> > 
> > On Wed, Apr 29, 2009 at 08:26:02AM -0400, Mario Charest wrote:
> > > Will 6.4.1 on x86 support more then 8 CPU?
> > 
> > The shipped binary won't. By changing the value of PROCESSORS_MAX (see
> > services/system/smpswitch.h), you can recompile to support up to 32
> > without other source changes.
> 
> Brian, you mentioned that a while ago when I asked the same question about 6.4.  My comment was that there was no way 
I would ship a kernel I build myself without having tools to validate it thoroughly.  If it's suppose to be that simple 
and QNX isn't doing it, I cannot safely assume that is as simple as you make it sound ;-)  Notice how I cleverly avoided
 talking about marketing, management and business decision.
> 

It's more a testing and memory cost issue. We don't currently have the 
machines here to validate the >8 configuration. Also, there are a number of
statically allocated arrays based on the value of PROCESSORS_MAX, so bumping
the value increases the footprint for everybody. It's long been my fond
dream to get rid of the constant entirely. I have a few glimerings on how
to do it - it's a matter of finding the time.

-- 
Brian Stecher (bstecher@qnx.com)        QNX Software Systems
phone: +1 (613) 591-0931 (voice)        175 Terence Matthews Cr.
       +1 (613) 591-3579 (fax)          Kanata, Ontario, Canada K2M 1W8