Re: One byte integer

[Jim Oberholtzer <midrangel@xxxxxxxxxxxxxxxxx>]

Dan,

Agreed and I would bet the performance tools would be unable to detect 
it unless there were a very significant number operations against that 
type of field.  That tooling is the best in the business, but that's 
just too small to capture without the overflow exceptions and/or wait 
states for some reason.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On 3/26/2013 10:37 AM, Dan Kimmel wrote:
> That generalization is highly dependent on the architecture of the processor. The generalization is mostly based on the behavior of intel 8086-family chips that only had 16-bit numeric registers. Doing eight-bit arithmetic with a 16 bit register required a couple extra ops to handle overflow. This was in a time when those extra ops took time. With current Power technology, all that is parallelized in multiple processor pipes optimized by advanced compilers and I doubt if there's any performance impact at all.
>
> -----Original Message-----
> From:midrange-l-bounces@xxxxxxxxxxxx  [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Tim Bronski
> Sent: Tuesday, March 26, 2013 10:13 AM
> To: Midrange Systems Technical Discussion
> Subject: Re: One byte integer
>
> Does anyone know what the performance considerations might be on using a char instead of an int? I know it's a generalization but usually int processing is faster than shorts and chars. Not sure if this has any application here, just curious.
>
> In 3/26/2013 3:27 PM, Luis Rodriguez wrote:
> > >  Rob,
> > >
> > >  I believe that the OP stated somewhere that he needed to store a
> > >  number between 0 and 100. An one-byte integer field could allow
> > >  between -127 and
> > >  128 if signed and between 0 and 255 if unsigned.
> > >
> > >  Regards,
> > >  Luis Rodriguez
> > >  IBM Certified Systems Expert - eServer i5 iSeries
> --