[Coco] About the tylg (type language) ident report

Gene Heskett gheskett at wdtv.com
Tue Oct 23 20:45:41 EDT 2012


On Tuesday 23 October 2012 20:19:33 Robert Gault did opine:

> I've just realized that our current use of ident to report a 6809 vs
> 6309 module is essentially useless for the NitrOS-9 project. The reason
> for this is that the asm files used to build the project do not include
> a variable for language. The result is that a module such as IOMAN
> which can be built for both 6809 and 6309 systems will always give an
> ident report as a 6809 module.
> 
> Does anyone care about this? It could be corrected but would require a
> large amount of work on the NitrOS-9 project and the results could
> still be confusing.
> 
> If a correction were to be made and I am not volunteering, each .asm in
> the project would need to be checked for the presence of 6309 code and
> when present the following added to the asm file.
> 
> 	ifne H6309
> tylg	set	Systm+Obj6309
> 	else
> tylg	set	Systm+Objct
> 	endif
> 
> You should not do this when 6309 code is not present or all programs in
> a 6309 package would be labeled 6309 even if no such code were present.

I believe this is probably less of a problem than it could be, primarily 
because the huge majority of the modules that go to make up the kernel do 
not actually have any 6309 only code in them.  If that module builds on a 
6809 premise setting, without the above ifne verse, then it likely does not 
actually have 6309 code in it.  Or this assembler (referring to the nitros9 
assembler, not mamou) would be squawking about it.

So far, ioman and the level3 only nitro are the two modules needing fixing 
that we have found.  I think that tells us that there are lots of the 
kernels separated out calls that probably could be further optimized for 
the 6309. The 'hot' paths however can IMO wait until we have stability 
again.

This somewhat resembles the "if it not broken, don't fix it" method.  Fixes 
here to make it conditionally build optimized for each cpu, are generally 
of the .1% speedup variety except for certain bit shift situations where 
the first 8 bits each way can be done in 3 or 4 instructions, about 20x 
faster than the usual loop till done construct, and which none of the so-
called optimizers we have know a thing about, so it will take many such 
fixes to get another 2% worth of real throughput.

> --
> Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco


Cheers, Gene
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
My web page: <http://coyoteden.dyndns-free.com:85/gene> is up!
Am I in GRADUATE SCHOOL yet?



More information about the Coco mailing list