From root Mon Dec 10 12:57:22 1984
>From hpfcll Mon Dec 10 12:05:31 1984 remote from hpfcla
>From mear Mon Dec 10 12:05:46 1984 remote from hpfcll
To: hpfcla!rob
Subject: Memory diagnostics
Status: R

Rob,

I will be receiving  some  information on the Indigo memory  diagnostics
sometime  this  month.  I will  have to  modifiy  these  diagnostics  to
account for differences in the memory controllers.

Do you have or know of someone  in the  facility  who has an ERS for the
Indigo (and Cheetah) memory controllers?

I would like to develop an outline for the memory diagnostics which may
be specific to Firefox.  (The standard memory  tests/diagnostics  should
port from Indigo  directly.)  Below is first stab at the outline.  Since
you and the MC chip designer are experts on the internal  workings, your
review and additions to this outline would be helpful.

Test for detection and correction and reporting of single bit errors.

	Force  a bad bit in  every  bit  location  in the  data  and the
	syndrome  bits.  What are good data patterns to use?  (A's, 5's)
	Make sure that data read is the same as data written.  Make sure
	that error is reported as DERR_MAP.

Test for detection of double bit errors (no errors mapped out).

	Force two bad bits in selective  bit  locations  in the data and
	syndrome  bits.  Check  for  report  or  DERR_HARD  error.  Must
	install proper HPMC interrupt handler.

Test mapping hardware

	Force one bad bit in the  data/syndrome.  Read the location  and
	map  out the bad  bit.  Force  the  same  error  to see if it is
	corrected  without being  reported.  Force a double bit error to
	see if it is corrected and reported.  Try and map for both banks
	of the 2MB board.

How can the disable error correction  option be tested or used?  Perhaps
it can be used in the above  tests to insure that a data error is indeed
present.





Initialize memory



					Charlie Mear
					hpfcla!mear



