milazzo@RICE.EDU (Paul Milazzo) (12/22/85)
In <382@pedsgd.UUCP>, Bob Halloran <bobh@pedsgd.UUCP> ventures: "Granted, the graphics chips relieve the [Amiga's] 68000 from dealing with the display, as it must in the Mac, but the 68000 and the various co-processors must still contend for access to memory. YES, I/O can proceed independently, etc., but only ONE of the processors can use memory at a time. This HAS to impact the throughput of the 68000." This turns out not to be the case. First, the Amiga's coprocessors can *only* access the first 512k of RAM. Any additional RAM resides on a completely separate bus to which only the 68000 has access. The loader allows a program developer to specify that those program segments containing graphics structures should reside in the shared RAM, and the remainder of the program in "fast" RAM, if available. Second, the shared bus is run at twice the CPU speed, and the CPU and coprocessors use alternate bus cycles. Thus, under "normal" conditions the CPU sees no bus contention whatsoever. Only when the coprocessor bus cycles are inadequate for a particular task do the coprocessors begin to preempt the CPU's bus cycles. Most commonly, this happens whenever a 640-pixel-wide display with more than 3 (I think) bit planes must be displayed. Default text windows use the 640-wide mode with two bit planes, and so do not slow the CPU. N.B. I own neither of the machines being compared. Much of the above information comes from INFO-AMIGA postings by Amiga technical personnel, and any inaccuracies therein are no doubt due to my faulty recollection of those messages. I welcome corrections. Paul G. Milazzo Dept. of Computer Science Rice University, Houston, TX Domain: milazzo@rice.EDU ARPA: milazzo@rice.ARPA BITNET: milazzo@ricecsvm UUCP: {cbosgd,convex,hp-pcd,shell,sun,ut-sally,waltz}!rice!milazzo