Path: chuka.playstation.co.uk!news
From: sosman@terratron.com (Steven Osman)
Newsgroups: scee.yaroze.freetalk.english
Subject: Re: Daft Question [long]
Date: Mon, 03 Sep 2001 03:56:22 GMT
Organization: PlayStation Net Yaroze (SCEE)
Lines: 27
Message-ID: <3b93ff49.33320963@www.netyaroze-europe.com>
References: <9m1d9c$b2i7@www.netyaroze-europe.com> <01c12ceb$fba64880$9709e4d5@pal-s-omnibook> <9m9c88$ioj5@www.netyaroze-europe.com> <01c12edc$cc2bf440$991fe4d5@pal-s-omnibook> <9me1g1$1sb13@www.netyaroze-europe.com> <9meajl$1sb17@www.netyaroze-europe.com> <01c13018$85386ae0$122e1bd4@pal-s-omnibook> <9mjelr$dra1@www.netyaroze-europe.com> <9mrotl$mfo3@www.netyaroze-europe.com> <3b9250e6.158216893@www.netyaroze-europe.com> <01c133fe$0a5a89e0$b00ae4d5@pal-s-omnibook>
NNTP-Posting-Host: dsl092-099-074.nyc2.dsl.speakeasy.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Newsreader: Forte Agent 1.5/32.452

...here's where the compiler's optimization could really come in
handy, then, right?

On 2 Sep 2001 22:28:11 GMT, "pal" <pal_news@hotmail.com> wrote:

>Steven Osman <sosman@terratron.com> wrote in article
><3b9250e6.158216893@www.netyaroze-europe.com>...
>><snip>Remember that the processor can take UP TO 32 cycles...  presumably,
>>that would be if it has to do a tonne of additions<snip>
>
>Indeed according to mips docs it processes 1bit per cycle; a whole mul/div
>always takes 32 cycles (although I don't know whether mul/divs can be done
>less-than-32bit-ly) or a bit more depending on instruction variants. What
>is meant by "up to" is that, as mul/divs are processed in parallel in
>something like a second dedicated alu, other instructions can be executed
>until the result of the mul/div is actually used.
>Ie if you have something like:
>a *= b
>x += y
>x += a
>the first line will be executed in parallel with the second, so that those
>first two lines will take 32 cycles.
>In the following order however the first two lines would take 1+32 cycles:
>x += y
>a *= b
>x += a