[net.micro] summary of talk by Kottke of Apple at Stanford

croft%Safe%su-score@sri-unix.UUCP (02/04/84)

From:  Bill Croft <croft%Safe@su-score>

Some interesting tidbits that came up at the Apple Macintosh talk last
Wednesday at Terman:

The 128K byte memory is organized as 64K 16 bit words.  During
active video, one word can be fetched each microsecond by BOTH
the processor and memory.  Each microsecond the video takes half
the cycle (500 ns) and the processor can take the other.
Since only the RAM is shared with video, processor access to 
ROM can occur at twice this rate (2 megawords per second).  This
is an important consideration since ROM contains many of the heavily
CPU bound code, such as Quickdraw and the OS.  The 2 megaword
rate can also occur for processor RAM access, when the video
is inactive during vertical retrace.  It might be a worthwhile
option to allow the user to idle the video during lengthy computation
(such as a C compilation), if it results in twice the thruput.

The disk controller used is the "IWM", Integrated Wozniak Machine,
a custom VLSI version of the "classic" Woz design for the Apple II.
The disk controller is about the only thing in the system that
is not interrupt driven.  They were forced to do this because of
engineering constraints;  they couldn't find a proper FIFO to
operate with the IWM.  To get around the lack of interrupts, the
disk is "polled" during sector IO.  During disk IO all other
interrupts are turned off for the duration: 12 milliseconds.
Fortunately they also poll for serial input during this interval;
otherwise they would lose characters.

The video screen dot aspect ratio is "square", so that images
undergo no distortion in the mapping from memory to display.
Unfortunately this is not the case on the Lisa's:  the Quickdraw
software on those machines must "draw" an ellipse when a circle
is desired on the screen;  they admitted this was a mistake.

Sound is generated by simple pulse width modulation:  integration
of the digital waveform output results in an inexpensive D/A
converter.

	--Bill Croft, Stanford Medical Center