Re: Determining row of pixels among a group that has the max number

homer123123123@[EMAIL PROTECTED]
 wrote:
> Thanks for the info so far. Sorry about the confusion regarding the
> actual number of pixels. It is  90 of 10x1. The raw data is coming
> from temperature sensors. We are trying to figure out the hotspots in
> a flow. I am trying to think of simpler ways to pinpoint the bank of
> sensors which has the hot spot.

> Therefore I was thinking of creative ways to analyze the data instead
> of just looking at it as a list of temperature points.

Why not just treat it as a vectorisable problem and use the SSE2 
extentions on a Pentium to check 16 bytes at a time for find maximum?

You might need to tweak your hardware to measure 304x300 or null pad the 
unmeasured values to take best advantage of the SIMD instructions.

Thesholding would actually be slower since there is no single 
instruction to threshold and count the number of hits.

I am still puzzled why this has to be done so quickly.

Regards,
Martin Brown
** Posted from http://www.teranews.com
**