Re: Interrupting EDP...
> On 29 Jan 2016, at 16:15, Michel Hack <mhack@xxxxxxx> wrote:
>
> Before this discussion is ruled out-of-order...
>
> On Fri, 29 Jan 2016 11:53:16 +0100, Ulrich Kulisch wrote:
>> On a modern processor that performs 10 to the power of 15 operations in
>> a second these can be done in really a very tiny portion of a second.
>
> I’d like one of those if it fits on my desk!
Nah. Your desk would collapse, or the room would rapidly over-heat!
(weight ~10 tonnes, power ~ 1MW)
> But seriously, the interrupt issue is real in any NON-DEDICATED machine,
> because a general-purpose OS -- and even some special-purpose ones --
> would not know that an EDP is in progress.
>
> Besides -- interrupting *one* EDP wasn't even the issue. The issue was
> computing a large number of independent EDPs, and THAT computation would
> surely be interruptable unless the machine was indeed dedicated to this
> calculation.
Still, being serious for a moment, I’d like to take the opportunity
to thank Ulrich and the members for this discussion, and to say that
I can see a way to get most of what Ulrich wants (an interruptible
EDP operation using a small number of stock instructions) using adaptions
to the ARM-NEON SIMD co-processor.
There might be potential export control restrictions, but it’d be fun to try..
Dave
>
> Michel.
> ---Sent: 2016-01-29 16:23:06 UTC