jlperkin@uokmax.UUCP (J Les Perkins) (12/10/87)
I have been told about an article appearing in MacUser a few months ago stating that there was a problem with the Nubus architecture. The Nubus is supposed to enable a single card to take over control of the MacII. According to thissource the rom placement makes this impossible because of the downward compatibility of the Mac II with the SE. Therefore it would take a complete motherboardchange to correct the problem. In other words, big bucks! It also means you can't even fathom doing parallel processing until this is fixed. But what I am here to find out is if this article is right. I haven't found it yet and no other magazines have mentioned it. Plus no one here has mentioned it. I would appreciate some input into this subject, as I am contemplating bying a MacII. J. Les Perkins -just an ignorant student trying to correct that-
zrm@eddie.MIT.EDU (Zigurd R. Mednieks) (12/11/87)
In article <952@uokmax.UUCP> jlperkin@uokmax.UUCP () writes: I have been told about an article appearing in MacUser a few months ago stating that there was a problem with the Nubus... .... According to thissource the rom placement makes this impossible because of the downward compatibility of the Mac II with the SE. Therefore it would take a complete motherboardchange to correct the problem. In other words, big bucks! It also means you ... > > >J. Les Perkins The Mac II motherboard is a virtual NuBus slot, but that should not stand in the way of an upgrade board "mapping in" the Mac II ROMs, or more likely, copying them into faster memory. Motherboard assets like the SCSI and seriial ports could also be mapped into any addressing scheme from any NuBus slot. The SE compatibility mode consists of making each NuBus slot a 1MB partition in a 16MB SE-like address space. Why they bothered with this, I don't really know, since no reasonable application really cares about anything outside the application heap, nor even where the application heap is. This makes life tough for high-end (e.g. 24-bit) graphics cards because there isn't enough address space in each slot under this mapping scheme. (Hey Apple, any good answers for this seemingly backward-thinking design decision? And what about future MacII OSs? Will they have full 32-bit NuBus addressing for Mac NuBus peripherals?) None of this, however, nixes the idea of a NuBus upgrade card for a 68030 or other future processor. Nor is a multiprocessor setup out of the question. In fact, a separate NuBus processor would not have to limit itself to the 1MB per card addressing scheme of the current Mac OS. -Zigurd -- ------------------------------------------------------------------------ Zigurd Mednieks MURSU Corporation (617)424-0146 25 Exeter Street Boston, MA 02116
ws0n+@andrew.cmu.edu (Walter Ray Smith) (12/11/87)
> The SE compatibility mode consists of making each NuBus slot a 1MB > partition in a 16MB SE-like address space. ... > (Hey Apple, any good answers for this > seemingly backward-thinking design decision? It is certainly a backward-thinking design decision--thinking back to all the Mac software that assumes the top byte of an address is irrelevant. The "SE compatibility mode" (never thought of it that way...) is the hardware maintaining that illusion. Also, as far as I know, the current memory manager still uses the top bytes of its master pointers for flags, so the top byte of an address *must be* irrelevant under the Mac OS. A/UX is another story entirely... - Walt -- Walter Smith, CS graduate student, Carnegie-Mellon University uucp: ...!seismo!cmucspt!wrs (?) ARPA: Walter.Smith@andrew.cmu.edu usps: 5706 Darlington Rd.; Pittsburgh, PA 15217