Re: Fastcode Rules
From: Dennis (marianndkc_at_home3.gvdnet.dk)
Date: 10/26/03
- Next message: Dennis: "Re: FastCode logo challenge"
- Previous message: Pierre le Riche: "Re: FastCode logo challenge"
- In reply to: John O'Harrow: "Fastcode Rules"
- Next in thread: John O'Harrow: "Re: Fastcode Rules"
- Reply: John O'Harrow: "Re: Fastcode Rules"
- Reply: Eric Grange: "Re: Fastcode Rules"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sun, 26 Oct 2003 14:03:02 +0100
Hi John
"John O'Harrow" <john@elmcrest.demon.co.uk> wrote in message
news:3f9b9ffd$1@newsgroups.borland.com...
> Hi Dennis and other FastCoders,
>
> I have just been reading the fastcode rules thoroughly for the first time
> (Yes, I know its a bit late).
Yes to late ;-)
> I believe the Global rule 'It is not allowed to have code that set
compiler
> switches in functions' is unnecessary. Many routines (not just our
fastcode
> functions) may need range checking turned off or (more often) Boolean
> shortcuts turned on to function correctly. ps - Many of our current
> functions may fail if Boolean shortcuts are turned off.
>
> I do not see a problem with a Pascal function setting the compiler
settings
> it requires to operate, as long as it restores the previous settings when
it
> finishes.
I think that the code is dirty if it will not run with these settings on,
except in a few cases maybe. More opinions needed.
> To be clear, with the rule 'It is not allowed to read past the end of
> strings', does reading the null terminator (which is always present but
not
> included in the length) count as reading past the end.
I think the rule should be "It is not allowed to read past the end of
strings, but it is allowed to read the zero terminator"
> Finally, didn't we agree on a challenge rule that, where two or more
> function perform equally well on a particular processor (+/-1%), then the
> winner would be the function with the smallest size or best blended score
(I
> can't remember which).
Yes it was on the page too but I can't see it now either. Not all functions
have a blended score - they do not qualify for this category if the use SSE
or SSE2 and the rule must be
"If two or more functions perform equally well in the categories P4, P3,
Athlon XP, Pascal and blended, then the winner is the function with the
smallest size "
> I think we need to get the rules clarified soon - the challenge deadlines
> are only a few weeks away.
Agree I have tried this for more than a year :-)
> regards,
> John.
>
Regards
Dennis
- Next message: Dennis: "Re: FastCode logo challenge"
- Previous message: Pierre le Riche: "Re: FastCode logo challenge"
- In reply to: John O'Harrow: "Fastcode Rules"
- Next in thread: John O'Harrow: "Re: Fastcode Rules"
- Reply: John O'Harrow: "Re: Fastcode Rules"
- Reply: Eric Grange: "Re: Fastcode Rules"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|