withers@nitmoi.enet.dec.com (George A. Withers) (04/09/91)
In some previous article gone by the wayside in the great
Assembly/C Battle, Matt Dillon mentioned that Dice had
Amiga-isms such as the "chip" type qualifier to force a
single variable into chip RAM like SAS/C AND Aztec C. I
only recently upgraded to v5.0d and I didn't see mention
of this in the manual so does such a qualfier exist and, if
so, what is it?
Thanks,
George
------------------------------------------------------------------------------
George A. Withers, Jr. | "There is no life I know to compare
Digital Equipment Corp., 97 Piper Road | with pure imagination. Living
Acton, MA 01720 AT&T: 508.264.2339 | there you'll be free .. if you
Addr: withers@nitmoi.enet.dec.com | truly wish to be." - W. Wonka
------------------------------------------------------------------------------
DISCLAIMER: "Don't look at me! I didn't do it!" (Krusty the Clown)bj@cbmvax.commodore.com (Brian Jackson) (04/11/91)
In article <21925@shlump.nac.dec.com> withers@nitmoi.enet.dec.com (George A. Withers) writes: > >In some previous article gone by the wayside in the great >Assembly/C Battle, Matt Dillon mentioned that Dice had >Amiga-isms such as the "chip" type qualifier to force a >single variable into chip RAM like SAS/C AND Aztec C. I >only recently upgraded to v5.0d and I didn't see mention >of this in the manual so does such a qualfier exist and, if >so, what is it? I am unaware of this existing in the current Aztec compiler. According to Aztec, however, the next major release of the Manx C compiler will have the CHIP keyword. bj >Thanks, George ----------------------------------------------------------------------- | Brian Jackson Software Engineer, Commodore-Amiga Inc. GEnie: B.J. | | bj@cbmvax.cbm.commodore.com or ...{uunet|rutgers}!cbmvax!bj | | "We defy augury" | -----------------------------------------------------------------------
dillon@overload.Berkeley.CA.US (Matthew Dillon) (04/11/91)
In article <21925@shlump.nac.dec.com> withers@nitmoi.enet.dec.com (George A. Withers) writes: > >In some previous article gone by the wayside in the great >Assembly/C Battle, Matt Dillon mentioned that Dice had >Amiga-isms such as the "chip" type qualifier to force a >single variable into chip RAM like SAS/C AND Aztec C. I >only recently upgraded to v5.0d and I didn't see mention >of this in the manual so does such a qualfier exist and, if >so, what is it? > >Thanks, >George Well, I *assumed* that Aztec had added it, I don't have 5.0d myself. If they haven't, well, there are no excuses left for them. -Matt > ------------------------------------------------------------------------------ > George A. Withers, Jr. | "There is no life I know to compare > Digital Equipment Corp., 97 Piper Road | with pure imagination. Living > Acton, MA 01720 AT&T: 508.264.2339 | there you'll be free .. if you > Addr: withers@nitmoi.enet.dec.com | truly wish to be." - W. Wonka > ------------------------------------------------------------------------------ > DISCLAIMER: "Don't look at me! I didn't do it!" (Krusty the Clown) -- Matthew Dillon dillon@Overload.Berkeley.CA.US 891 Regal Rd. uunet.uu.net!overload!dillon Berkeley, Ca. 94708 USA
plav@cup.portal.com (Rick M Plavnicky) (04/14/91)
dillon@overload.Berkeley.CA.US (Matthew Dillon) recently wrote: >In article <21925@shlump.nac.dec.com> withers@nitmoi.enet.dec.com (George A. W > >>In some previous article gone by the wayside in the great >>Assembly/C Battle, Matt Dillon mentioned that Dice had >>Amiga-isms such as the "chip" type qualifier to force a >>single variable into chip RAM like SAS/C AND Aztec C. I >>only recently upgraded to v5.0d and I didn't see mention >>of this in the manual so does such a qualfier exist and, if >>so, what is it? >> >>Thanks, >>George > > Well, I *assumed* that Aztec had added it, I don't have 5.0d myself. > If they haven't, well, there are no excuses left for them. > > -Matt > >> --------------------------------------------------------------------------- >> George A. Withers, Jr. | "There is no life I know to compa >> Digital Equipment Corp., 97 Piper Road | with pure imagination. Living >> Acton, MA 01720 AT&T: 508.264.2339 | there you'll be free .. if you >> Addr: withers@nitmoi.enet.dec.com | truly wish to be." - W. Wonka >> --------------------------------------------------------------------------- > DISCLAIMER: "Don't look at me! I didn't do it!" (Krusty the Clown) > >-- > > Matthew Dillon dillon@Overload.Berkeley.CA.US > 891 Regal Rd. uunet.uu.net!overload!dillon > Berkeley, Ca. 94708 > USA This comes up every so often. Manx currently does NOT offer the ``chip'' keyword in their product. (Current product is referred to as 5.0e, but signs on as 5.0d dated Sep 30 1990) However, there WAS some mention made on their support BBS recently: <begin included material> MESSAGE: 5492 DATE/TIME: 04-04-91 6:44pm FROM : JEFF DAVIS (SYSOP) RECEIVED : NO TO : PAT HEUVEL PRIVATE : NO SUBJECT: Compiler Suggestion THREAD : YES FOLDER : D, "Amiga Folder" We are going to add the chip keyword, probably for the release after the 5.2 release. - Jeff Davis <end included material> Anyway, Matt's right: there's no excuse. Disclaimer: I bought and use Manx's compiler, but it sure does have its shortcomings. /* Rick Plavnicky ...!sun!cup.portal.com!plav -or- plav@cup.portal.com */
jupiter@olympus.mlb.fl.us (Kurt R. Hawkes) (04/14/91)
In article <41242@cup.portal.com>, Rick M Plavnicky writes: > dillon@overload.Berkeley.CA.US (Matthew Dillon) recently wrote: > > >In article <21925@shlump.nac.dec.com> withers@nitmoi.enet.dec.com (George A. W > > > >>In some previous article gone by the wayside in the great > >>Assembly/C Battle, Matt Dillon mentioned that Dice had > >>Amiga-isms such as the "chip" type qualifier to force a > >>single variable into chip RAM like SAS/C AND Aztec C. I > >>only recently upgraded to v5.0d and I didn't see mention > >>of this in the manual so does such a qualfier exist and, if > >>so, what is it? > >> > >>Thanks, > >>George > > > > Well, I *assumed* that Aztec had added it, I don't have 5.0d myself. > > If they haven't, well, there are no excuses left for them. > > > > -Matt > > > > This comes up every so often. Manx currently does NOT offer the ``chip'' > keyword in their product. (Current product is referred to as 5.0e, but > signs on as 5.0d dated Sep 30 1990) > > However, there WAS some mention made on their support BBS recently: > > <begin included material> > > MESSAGE: 5492 DATE/TIME: 04-04-91 6:44pm > FROM : JEFF DAVIS (SYSOP) RECEIVED : NO > TO : PAT HEUVEL PRIVATE : NO > SUBJECT: Compiler Suggestion THREAD : YES > FOLDER : D, "Amiga Folder" > > We are going to add the chip keyword, probably for the release after > the 5.2 release. > - Jeff Davis > > <end included material> > > > Anyway, Matt's right: there's no excuse. Well, I've come across this same problem. I discovered that the linker has two options which allows one to specify sections of the code to be loaded into chip or fast ram. Look on page 5-15 of the Aztec C Reference Manual. Here is a quick summary. +c - load into chip memory c - executable code +f - load into fast memory d - initialized data b - uninitialized data (bss) So for example: ln +cdb +fc ... would cause the executable code to be loaded into fast memory while data resides in chip memory. Generally, I use ln +cd ... which specifies that the initialized data to be loaded into chip while uninitialized data and code be loaded normally (fast ram in my cast beacuse of FastMemFirst). I hope this helps... -- Kurt Hawkes Amiga 2000 (Hey, it works! Works well!) // (Just cover up the C word!) Warning: // Choose: Dangerous // (a) jupiter@olympus.mlb.fl.us \ Thinker \\ // (b) jupiter@cis.ufl.edu \ On Board! \X/ or (c) khawkes@gnu.ai.mit.edu / \
laughlin@fornax.UUCP (Bob Laughlin) (04/15/91)
In article <41242@cup.portal.com> plav@cup.portal.com (Rick M Plavnicky) writes: > ......... >This comes up every so often. Manx currently does NOT offer the ``chip'' >keyword in their product. (Current product is referred to as 5.0e, but >signs on as 5.0d dated Sep 30 1990) > ......... One way to force screen data into CHIP ram is to use the +cd linker option with Manx 5.0d. This puts all initialized data into CHIP memory. If the initialized data in your program is largely image data then this is a reasonable alternative and is preferable to AllocMem'ing CHIP mem and copying it over. Manx 5.0d does need the chip keyword though. -- Bob Laughlin laughlin@cs.sfu.ca
dillon@overload.Berkeley.CA.US (Matthew Dillon) (04/16/91)
In article <2492@fornax.UUCP> laughlin@fornax.UUCP (Bob Laughlin) writes: >In article <41242@cup.portal.com> plav@cup.portal.com (Rick M Plavnicky) writes: >> ......... >>This comes up every so often. Manx currently does NOT offer the ``chip'' >>keyword in their product. (Current product is referred to as 5.0e, but >>signs on as 5.0d dated Sep 30 1990) >> ......... > >One way to force screen data into CHIP ram is to use the +cd linker >option with Manx 5.0d. This puts all initialized data into CHIP memory. >If the initialized data in your program is largely image data then this >is a reasonable alternative and is preferable to AllocMem'ing CHIP mem >and copying it over. Manx 5.0d does need the chip keyword though. >-- > Bob Laughlin laughlin@cs.sfu.ca That's unreasonable, CHIP memory is at a premium these days. Only the actual *bitmaps* need to be in CHIP. -Matt -- Matthew Dillon dillon@Overload.Berkeley.CA.US 891 Regal Rd. uunet.uu.net!overload!dillon Berkeley, Ca. 94708 USA
waggoner@omews1.intel.com (Mark Waggoner) (04/17/91)
In article <dillon.6541@overload.Berkeley.CA.US> dillon@overload.Berkeley.CA.US (Matthew Dillon) writes: >In article <2492@fornax.UUCP> laughlin@fornax.UUCP (Bob Laughlin) writes: >> >>One way to force screen data into CHIP ram is to use the +cd linker >>option with Manx 5.0d. This puts all initialized data into CHIP memory. >>If the initialized data in your program is largely image data then this >>is a reasonable alternative and is preferable to AllocMem'ing CHIP mem >>and copying it over. Manx 5.0d does need the chip keyword though. >>-- >> Bob Laughlin laughlin@cs.sfu.ca > That's unreasonable, CHIP memory is at a premium these days. Only > the actual *bitmaps* need to be in CHIP. > -Matt You don't have to put ALL initialized data into CHIP memory. If you put the data you need in CHIP in a seperate file you can use the +cd option for that module and then switch it back to normal allocation for other modules. I don't remember the appropriate options for doing that. It is admittedly not as convenient as a _chip keyword, but you can do what you want to do. Mark Mark Waggoner waggoner@ichips.intel.com (503) 696-4590 No, I don't speak for intel.
stephane@Chucla.CAM.ORG (Stephane Laroche) (04/20/91)
In article <dillon.6541@overload.Berkeley.CA.US> (Matthew Dillon) writes: >In article <2492@fornax.UUCP> laughlin@fornax.UUCP (Bob Laughlin) writes: >> >>One way to force screen data into CHIP ram is to use the +cd linker >>option with Manx 5.0d. This puts all initialized data into CHIP memory. >>If the initialized data in your program is largely image data then this >>is a reasonable alternative and is preferable to AllocMem'ing CHIP mem >>and copying it over. Manx 5.0d does need the chip keyword though. >>-- >> Bob Laughlin laughlin@cs.sfu.ca > > That's unreasonable, CHIP memory is at a premium these days. Only > the actual *bitmaps* need to be in CHIP. It's unreasonable unless you put all your image data in the same file (or in multiple files) and only use the +cd option on those files. This is equivalent to the _chip keyword as long as you don't mix non-chip data with chip data. > -Matt Stephane Laroche | stephane@Chucla.CAM.ORG | Montreal, Qc. CANADA +1 514 277-8605 | laroche@ee470.ee.mcgill.ca | McGill University
peter@sugar.hackercorp.com (Peter da Silva) (04/22/91)
In article <1991Apr17.161732.27842@omews63.intel.com> waggoner@ichips.intel.com (Mark Waggoner) writes: > You don't have to put ALL initialized data into CHIP memory. If you put the > data you need in CHIP in a seperate file you can use the +cd option for that > module and then switch it back to normal allocation for other modules. I don't believe the linker allows this. I wrote browser with the expectation that they would eventually do that, with the pointer imagery in a separate file, but I'm still waiting. -- Peter da Silva. `-_-' <peter@sugar.hackercorp.com>.
rzv30@cccvm.ccc.amdahl.com (Rene Vega) (04/24/91)
The manx linker will not split data segments into chip and fast memory. I developed a program that postprocesses the load module and selectively sets the chip memory attribute in sepcific hunks. The only requirement is that the files containing the data to be moved into chip memory be compiled with the -mdm option. If you're interested, I can post the source code. Rene' A. Vega Computer & Systems Architecture