Article 163051 of comp.os.vms:
(I hate having to follow up my own posts - one day I'll learn to proofread
BEFORE I post!)

greg@indiana.edu (Gregory Travis) writes:

>I can't remember what a divide-by-zero did on the CPU hardware.  I don't
>recall there being any mechanism (until the later XJ ins

OOps.  I forgot to revise that.  My feeble memory tells me that a
divide-by-zero on the 6600 simply caused the CPU to halt.  PPU 0 would
detect that the CPU had halted and would actually look at the instructions
and operands in the vicinity of the halt PC to determine if a divide by
zero (as opposed to a program halt) caused the halt.

>Most 6600 instructions typically took either 3 or 4 major cycles.

Should be "3 or 4 MINOR cycles" (i.e. 300 to 400ns)

  Branches
>took a little longer (branches were very expensive on the 6600).  Thus
>the 6600 was nominally about 3 million instructions per second (I earlier
>wrote 1 mips).

Typical real-world results with good functional parallelism were somewhere
around 4-5 mips.

greg
--

greg		greg@indiana.edu	http://gtravis.ucs.indiana.edu/