BillW%SRI-KL@sri-unix.UUCP (08/08/83)
From: William "Chops" Westfield <BillW @ SRI-KL> Just how much of a "virtual machine" is provided by Visi-On? Normally, when someone mentions a virtual machine, I envision something like Smalltalk, where object code is transporable from machine to machine, and the underlying "engine" ends up interpreting byte codes or some such. However, from the general tone of your message, I got the impression that Visi-On provides more of an operating system type environment - You have a set of system calls that are callable from many languages, and are present on all machines. Sort of like Unix, which is the only major operating system that I can think of that runs on a bunch of different hardware. Bill W
BRACKENRIDGE%USC-ISIB@sri-unix.UUCP (08/14/83)
From: Billy <BRACKENRIDGE@USC-ISIB> I suppose "virtual machine" is a misnomer as it has come to mean so many things as to be "virtually meaningless". I believe the preferred terms are Visi Host for the low level stuff and Visi Machine for the higher level code. The system calls are not callable from many languages -- Visi C is the only language. Visi C is a fairly restricted subset of C designed for maximum portability and to be independant of word size while currently optimised for 16 bit words. I think in the Visi On system we see something fundamentally different from any portable operating system like Unix or CP/M. The Visi On system incorporates a model of a machine with a large virtual address space, a bit map display, and a mouse. Currently a hardware implementation of such a machine is not competitive in the office marketplace. The best compromise was to take a cost effective machine like the PC and write a lot of assembly language code to do raster-ops and memory management in software. Some of the UCSD Pascal people were at the Visi seminar. Their intent is to write Pascal to C translators that could port the UCSD base of programs into various C environments. I haven't looked in to the details, but I suspect some built in Pascal functions could be directly mapped to Visi On machine calls. Many of the low levels of the Visi On system may be made available in silicon. I expect many of the PC compatible companies will make super set PC machines to exploit the Visi On model. Currently, for example, the Compaq has a high resolution screen mode that is a superset of the IBM-PC graphics capability. Some of the people from the Intel display processor chip design team were at the seminar. The current Intel display processor (and the NEC equivalent) are selling about as well as stand alone word processors or dumb video terminals. Texas Instruments has already announced their chip entries in the raster-op race. They have a dual ported RAM that has a built in shift register and a raster-op engine based on the TMS320 architecture (fast multiply, built in barrel shifter, etc.) The Intel 286 processor is code compatible with the current 8086 but some dozen times faster. An upgrade to this chip would speed up memory management and provide memory protection while requiring no changes in user or Visi Machine level code. I don't think MS-DOS or applications that run directly under MS-DOS will be able to utilize the 286 without extensive recoding and a basic change in memory management and process structure.