jhs@MITRE-BEDFORD.ARPA (01/28/86)
Has anybody out there noted the news about the 65816 16-bit upgrade chip for the 6502? If what I read was correct, this will be a plug-in replacement for the 6502 which will extend the instruction set to handle a lot of 16-bit operations, sort of a la the 6800 or 8088 families. It is rumored that Apple is considering using this to upgrade their model II series, and it has occurred to me that it would be neat to drop one in my Atari 800XL. If it is truly upward compatible at the machine code level, that would open up a new world of enhanced performance and flexibility, such as (I speculate) 16-bit relative addressing, no more scrounging for zero-page addresses, etc. Any comments from the truly knowledgeable? -John Sangster jhs at mitre-bedford.arpa
info-atari@ucbvax.UUCP (01/28/86)
The 65816 book will soon be published by Brady. The author is David Eyes. If you really want one badly, give them a call, you might be able to bum a review copy if it is not yet on the shelves. Western Design Center did the design. Unfortunately their app-notes and other documentation are (last I looked) inadequte. The chip, on the other hand, is very real and very fast.
Fleysher.wbst@XEROX.COM (Dan Fleysher) (01/28/86)
re: dropping a 65816 into your Atari 800XL in place of the 6502, to upgrade it to 16 bits: Note that ANTIC (the screen display chip) is also an 8 bitter, so you would have to confine your display RAM to within 64K addresses. That's a real bummer - because one of the real advantages of lots of memory would be to have several alternate screens in memory at once and switch between them for animation, etc. Also, having usable RAM above the display area violates Atari 8-bit software architecture, which always puts the display area at the top of available RAM - the extra memory would not be usable by most 8 bitter programs (but you could make RAM disk(s) out of it). Then there's the problem that none of the existing memory decoding handles more than 8 bits. Also, you'd have to find a place to put the additional memory chips without destroying the RF shield - a packaging nightmare. Sounds like an awful lot of work to me, and when you finished up you would still not be able to run ST programs because of instruction set differences between the 65816 and ANTIC and their ST equivalents. Unless you want the experience of tearing apart your 8-bitter and learning something, I think you'd be further ahead to buy a new ST. (Of course, those of us who use our 8 bitters for word processing, game playing and bulletin-board-calling don't really need 16 bits anyway...) Dan -----------------------------------
oyster@uwmacc.UUCP (Vicious Oyster) (01/29/86)
In an article, Fleysher.wbst@XEROX.COM (Dan Fleysher) writes: >re: dropping a 65816 into your Atari 800XL in place of the 6502, to >upgrade it to 16 bits: Note that ANTIC (the screen display chip) is >also an 8 bitter, so you would have to confine your display RAM to >within 64K addresses.... >having usable RAM above the display area violates Atari 8-bit software >architecture, which always puts the display area at the top of available >RAM - the extra memory would not be usable by most 8 bitter programs >(but you could make RAM disk(s) out of it). > Unless I'm mistaken, that's merely a convention. The display memory can be anywhere (subject of course, to the usual alignment restrictions). >Also, you'd have to find a place to put the >additional memory chips without destroying the RF shield - a packaging >nightmare. Granted, a problem; but there are a *lot* of 800XL's out there with 256K and 512K already. >Sounds like an awful lot of work to me, and when you finished up you >would still not be able to run ST programs because of instruction set >differences between the 65816 and ANTIC and their ST equivalents. >Unless you want the experience of tearing apart your 8-bitter and >learning something, I think you'd be further ahead to buy a new ST. (Of >course, those of us who use our 8 bitters for word processing, game >playing and bulletin-board-calling don't really need 16 bits anyway...) I think you missed the point. The chip would be faster if only you take into consideration the fact that it could do memory transfers and arithmetic in 16 bits. I once attempted to alter the public domain Small-C compiler to generate 6502 code, and I gave up when I got sick of all the high-byte/low-byte garbage-- the compiler would have turned out twice as big and twice as slow. And who wants to run ST programs, anyway? (:-) - Joel ({allegra,ihnp4,seismo}!uwvax!uwmacc!oyster) P.S. I'd like to change my "split the group" vote from a no to a yes, just so the obnoxious amiga/mac/ham-radio cross(and mis)-postings go away.
jhs@MITRE-BEDFORD.ARPA (01/30/86)
Re: where to put the memory... --- The 65816 sounds like a very nifty idea to me, but certainly no substitute for buying an ST if you desire one. As an upgrade for the 800XL, it would be desirable to keep everything as compatible as possible. Since Atari has already devised a scheme for banked memory in the 130XE, and CDY and others(?) have successfully and compatibly extended that scheme to at least 262K, it would make sense to keep the current banking scheme. See Mapping The Atari, 1985 edition, Appendix 16, page 238, for a full description of the 130XE memory management scheme. Basically, they put the extra memory into the 65K memory map by replacing the original 16K segment between $4000 and $7FFF (decimal 16384 to 32767). Location $D301 (decimal 54017) controls the bank switching. Two bits there are used to control which of four possibilities occur with regard to whether the CPU or the ANTIC chips stay with standard memory or use the selected 16-K bank of extended memory. If you add more than 131 K but stay with their conventions you can always switch to XL/XE compatibility mode to run "old" software. Note that with this arrangement, both the low memory area and the high ROM area are unaffected. This is why everything still works (can be made to work easily). Re: why to do the conversion --- I concur that the reason for the conversion is to get more addressing flexibility within the 65K memory space. A whole lot of assembly language programs will be easier to write and more compact with the extended instruction set. If the 65816 in fact is truly object-code upward compatible with the 6502, I for one am CERTAINLY going to want to drop one into my 800XL even though I bought it basically as a toy computer and for use in ham radio applications and general electronic fooling around. This will in no way affect my interest in adding an ST to the collection for more serious work. --- By the way, does anybody know when this chip will really be available; where; and how much? I would appreciate solid info. John Sangster jhs at mitre-bedford.arpa