Re: new version of MAD

[Bjorn Eriksson <mdeans algonet se>]

On Mon, 30 Jul 2001, Steef Boerrigter wrote:

> >  Using realloc sounds nice. And a hash table would be really nice.
> > (Well, hash tables can't be resized efficiently but I don't think that
> > would pose a problem.)
> You mean using a hashing table would be nice to tackle the slooooowness
> of MAD when it goes over ~10000 handles?

 Exactly.

> > > - Support for SGI Indy (and probably more RISC and or MIPS platforms)
> > > which need 8 byte aligned blocks for double variables.
> >  Sounds useful.
> Mad crashes on the very first alloc. So useful is an understatement.
> "Vital" is more appropriate ;-)

 Uhm, yeah, perhaps. No, not really. MAD is only to be used by
developer, I'd expect the very first developer (worth his salt) to try
this on his shiny new SGI Indy MIPS-whathaveyou [I'm envoius since I
don't have one! :)] would jump in and add a bit of alignment to fix this
fallacy.

> >  All processors in the x8[68] family can read 16-bit WORDs (and larger
> > quantities) on an arbitrary offset but reading a 16-bit WORD that
> > straddles a BYTE boundary requires the processor to do two separate
> > reads. (And this hurts even though it actually reads a whole cacheline
> This is useful information. I conclude from it that even on the intel
> architecture alignment is desired if we want to improve the speed of our
> program being debugged.

 Yes indeed. But I think you're making to much of this; the code we know
as 'mc' doesn't care much. Again, 'I think' and IMHO, in other words -
I'm simply guessing.

> >  The 'diff -u' files would be usefull. Or if you could give us the exact
> > version you changed so everyone could do the diff herself.
> 
> Uh oh, didn't you get the mad-new.[ch] as attachments?

 Yes. I did get your attachments but I have no way of knowing their
originin short of checking out all previous versions and doing some
guess work. (I'd rather not.)
 
> In that case either my mailer is sick or the attachments are cut off by
> the mailing list handler.

 I've noticed the warnings that get appended to your messages but they
seem to make it almost intact to the list. I'm pretty sure the problem
is on your end Steef. The message you sent directly to me looked exactly
the same. Please test your setup.


//Björnen, cya all tomorrow, I need to sleep now.