On Mon, Feb 2, 2009 at 11:30 PM, Peter Duniho wrote:
> On Feb 1, 2009, at 8:37 PM, Michael Ash wrote:
>
>> Thanks to both you and Rob for repeating the experiment. Interesting
>> that it's so variable. On my computer it would occasionally require
>> several million to crash, but most of the time i
On 03/02/2009, at 2:30 PM, Peter Duniho wrote:
On my computer, the one time I saw it crash, it was an
NSInvalidArgumentException, complaining that in "-[MyOperation
start]" the "receiver has already started or finished". I'm
wondering if this is the exception everyone sees, or if the actu
On Feb 1, 2009, at 8:37 PM, Michael Ash wrote:
Thanks to both you and Rob for repeating the experiment. Interesting
that it's so variable. On my computer it would occasionally require
several million to crash, but most of the time it explodes within the
first few hundred thousand.
I've been te
On Sun, Feb 1, 2009 at 8:04 PM, Peter N Lewis wrote:
> At 16:27 -0500 1/2/09, Michael Ash wrote:
>>
>> It crashes reliably on my Mac Pro. Note that it uses only one queue.
>> Note that it only enqueues operations from the main thread. Note that
>> it uses a very simple custom NSOperation subclass,
At 16:27 -0500 1/2/09, Michael Ash wrote:
It crashes reliably on my Mac Pro. Note that it uses only one queue.
Note that it only enqueues operations from the main thread. Note that
it uses a very simple custom NSOperation subclass, and doesn't use
NSInvocationOperation at all. Note that it does n
On 02/02/2009, at 7:27 AM, Michael Ash wrote:
It crashes reliably on my Mac Pro. Note that it uses only one queue.
Note that it only enqueues operations from the main thread. Note that
it uses a very simple custom NSOperation subclass, and doesn't use
NSInvocationOperation at all. Note that it
On Sat, Jan 31, 2009 at 10:15 PM, Robert Marini wrote:
> Easily reproduced doesn't always translate into guaranteed to occur. In my
> experience, using a single queue in your application is a sufficient
> safeguard as no system framework I've encountered causes an issue.
Well, it turns out that
If your original algorithm is scalar, you can get upto 4x increase by
moving to SIMD (SSE2/3) if your algorithms can be paralleled. Add
that to the 7 extra cores and you can get upto 32x speed up. In the
real world you're very (very) unlikely to reach max throughput because
of data depend
On 31 Jan 09, at 19:15, Robert Marini wrote:
Easily reproduced doesn't always translate into guaranteed to
occur. In my experience, using a single queue in your application
is a sufficient safeguard as no system framework I've encountered
causes an issue. That isn't to say that the API is
On Sat, Jan 31, 2009 at 10:15 PM, Robert Marini wrote:
> Easily reproduced doesn't always translate into guaranteed to occur. In my
> experience, using a single queue in your application is a sufficient
> safeguard as no system framework I've encountered causes an issue. That
> isn't to say that
Easily reproduced doesn't always translate into guaranteed to occur.
In my experience, using a single queue in your application is a
sufficient safeguard as no system framework I've encountered causes an
issue. That isn't to say that the API is without quirks but they can
usually be adjus
On 31 Jan 09, at 15:11, Chris Hanson wrote:
On Jan 31, 2009, at 2:04 PM, jurin...@eecs.utk.edu wrote:
There IS a know bug with the NSInvocationQueue method on intels using
10.5.6 which I have read will be fixed on 10.6.
Please do not say things like this without citing a specific
source. Ot
On Jan 31, 2009, at 2:04 PM, jurin...@eecs.utk.edu wrote:
There IS a know bug with the NSInvocationQueue method on intels using
10.5.6 which I have read will be fixed on 10.6.
Please do not say things like this without citing a specific source.
Otherwise you are spreading rumors.
In gener
On Sat, Jan 31, 2009 at 5:04 PM, wrote:
> I have a bullet-proof scientific app I developed using cocoa. I just
> purchased a new MacPro with the dual quad processors.
>
> Earlier posts attempting to determine MAXIMUM theoretical speedup have
> gotten bogged down with semantic differencea between
At the risk of stating the obvious: it seems like your limiting factor
in speedup will depend a lot more on your algorithm design than
NSInvocationQueue, the number of cores/processors, etc.
In other words, this isn't really an Apple-specific question. It's a
matter of parallel algorithm de
15 matches
Mail list logo
