Re: FORTH plusminus, was Re: [f-cpu] Winograd DCT on my seul.org account

[Cedric BAIL <cedric.bail@free.fr>]

> > I think that you understand the problem. The only way to schedule an 
> > algorithm in a high level language is after you generate the assembler, 
> > and this assembler can't be rigt scheduled when it's generated, because 
> > the intermediate representation didn't reflect the asm language. 
  
> Hmm AFAIK about gcc it does scheduling as part of compilation and 
A i don't understand what you mean by "as part of compilation". If you mean 
that gcc do the schedule before sending the asm code to asm, I agree, but I 
don't think that he do it before during the creation of the IR. 
 
> it seems to work (well it not so good in all cases). 
 
> I think that scheduling almost done assembler is hard because you 
> nave not all the information. For example gcc allows you to mark 
> conditions as "likely" or "unlikely" to branch - this should influence 
> scheduling but it is not known at later stages .. 
Yes, of course you need the knowledge of the IR and the ASM to compute the 
schedule. 
 
A+ 
   Cedric 
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/