Re: improvement of index-calculation
- From: Eric Grange <egrangeNO@xxxxxxxxxxxxxxx>
- Date: Wed, 27 Jul 2005 14:24:04 +0200
> const x, y, z: integer
If you can pass them as a unique const vector structure (without having to fill a vector structure for that sole purpose) like
const xyz : TIntegerVector
you may be able to help the compiler generate better code by reducing register pressure at the call point (and pass everything via registers).
Since what this routine does is rather simple, call overhead is probably not negligible.
Eric .
- Follow-Ups:
- Re: improvement of index-calculation
- From: Guenther Wimpassinger
- Re: improvement of index-calculation
- References:
- improvement of index-calculation
- From: Guenther Wimpassinger
- improvement of index-calculation
- Prev by Date: Re: Fastcode SSE3 Support Implementation
- Next by Date: Re: NAFCQY [Not A FastCode Question, Yay!]: Int64 DivMod
- Previous by thread: Re: improvement of index-calculation
- Next by thread: Re: improvement of index-calculation
- Index(es):