home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.atari.st.tech
- Path: sparky!uunet!usc!hela.iti.org!cs.widener.edu!dsinc!spool.mu.edu!yale.edu!ira.uka.de!math.fu-berlin.de!unidui!rrz.uni-koeln.de!rrz.uni-koeln.de!aeg03
- From: aeg03@rrz.uni-koeln.de (Jan T. Kim)
- Subject: GFA and FPU support (was: Some old questions)
- Message-ID: <1992Nov09.205424.132216@rrz.uni-koeln.de>
- Reply-To: kim@vax.mpiz-koeln.mpg.dbp.de
- Organization: Regional Computing Center, University of Cologne, F. R. Germany
- References: <af49f08f@Kralizec.fido.zeta.org.au>
- Date: Mon, 09 Nov 92 20:54:24 GMT
- Lines: 67
-
- In <af49f08f@Kralizec.fido.zeta.org.au> michael.smith@f842.n800.z3.fido.zeta.org.au (michael smith) writes:
-
- >Original to: fsctc@lerc03.lerc.nasa.gov
- >In a message of <27 Oct 92 22:37:01>, fsctc@lerc03.lerc.nasa.gov (3:713/602)
- >writes:
-
- > f> 4. Is there a version of GFA BASIC that supports the math coprocessor
- > f> on an ST (as a peripheral device)? Is there one that supports the math
- > f> coprocessor on the TT? Will this run under the TT RAM?
-
- >At this point, no.
-
- I don't think that's correct. When the copy of GFA basic 3.60 TT
- arrived here, I did the following quick test:
-
- t%=TIMER
- FOR i&=1 TO 1000
- l=l+SIN(RAD(i&))*COS(RAD(i&))
- NEXT i&
- t%=TIMER-t%
- PRINT "ohne ~TT?: ";t%/200;" sec.",l
- ~TT?
- CLR l
- t%=TIMER
- FOR i&=1 TO 1000
- l=l+SIN(RAD(i&))*COS(RAD(i&))
- NEXT i&
- t%=TIMER-t%
- PRINT "mit ~TT?: ";t%/200;" sec.",l
-
- The loop takes .825 seconds without and .180 seconds with ~TT,
- showing that there definitely is a very significant difference in
- speed between the TT specific rountines and those that are
- designed to run on all 680x0 Ataris. I haven't analyzed the code
- and verified if the speedup is due to FPU usage, though.
- For those not familiar wiht GFA: TT is a boolean function that
- returns TRUE if the program runs on a TT. Programs compiled and
- linked carry two versions of some routines, one that runs on all
- Ataris and one that is optimized for the TT. Originally, the
- pointers to those routines are set to use the all-Atari versions.
- If the TT function is called and a TT is detected, the TT
- function changes those pointers to use the TT optimized versions.
- At the CeBIT 1992, they demonstrated this at the GFA stand, and
- claimed the speedup was in large part due to FPU usage.
-
- >Richter have recently acquired the rights to GFA basic
- >(ST only) from GFA themselves, and are undertaking a complete rewrite, with
- >the intention of completely legalising the package.
-
- >They may add this support - however I have no information either way.
-
- According to what Richter told me on the phone recently, the GFA
- version developed by them will run and be supported on all
- Ataris, including the Falcon, and it will run under MultiTOS.
- This is also what Richter announced at the Atari Messe.
- Oh yes, and the support will not be continued to be done in the
- clumsy way described above, they assured me at the Messe. So save
- your sneering remarks :-)
-
- Greetinx, Jan
-
- +- Jan Kim -- X.400: S=kim;OU=vax;O=mpiz-koeln;P=mpg;A=dbp;C=de -+
- | Internet: kim@vax.mpiz-koeln.mpg.dbp.de |
- | |
- *----=< hierarchical systems are for files, not for humans >=-----*
-
-
-