HALCoGen 4.02.00 and Errata PBIST#4, Continued

Charles,

Since before the holiday we've actually been scrubbing our errata descriptions to try to remove vague terms; 'much shorter' sounds like it might fit that category. My favorite though is 'small pulse' in the NHET errata :)

I'll see if I can get a preview of the new version of this errata - as hopefully it will be more precise. If not you've given us a good example of something to improve.

Thanks, Anthony.
I hope the new errata doesn't insist on using the errata_PBIST_4() HALCoGen API.
Charlie

Hi Charlie,

Ok, it may insist on the API -- if so why do you think it's overkill? Would be good to know so I can discuss that w. the folks working on the errata.
(if it actually isn't considered overkill by them - then maybe we are not clear enough in the errata. if they agree it's overkill - could be good feedback for the HalCoGen team.)

Charlie,

The part's datasheet includes a table showing the PBIST memory groups along with the expected number of clock cycles for each individual memory being tested. This is what is meant by the "normal" execution time in the erratum description.

The issue is that when you kick off a PBIST run it could quit without actually running the memory self-test. This will result in a "significantly shorter than normal" execution time, and this is the condition that can be trapped using the software loop (or RTI/PMU timestamps).

Anthony,
If I understand the API correctly, it is running undocumented tests many times and measuring the elapsed time. This seems to be overkill as running and timing the original HALCoGen PBIST self check would be sufficient. It runs PBIST on ROM expecting an error. All that's needed is the successful test time, plus the maximum failure time or an equation for "significantly shorter than normal".
Using the API is not just a case of extra execution time; I have to document all our code, so I would have to get more detail on the undocumented tests.

Sunil,
The datasheet does not have an entry for RAM Group 1 (PBIST ROM) using the march13n algorithm. This is an abnormal case that is only run to generate a PBIST error. I could measure this time, but would still need the failure criteria.

Thanks, Charlie

Hi Anthony,
I just noticed you sent me a friend request on 2/11 to discuss this. It has since expired. Please send it again.
Thanks, Charlie

Hi Anthony,
Any luck getting a preview copy of the errata for Silicon Rev C (SPNZ195E)?
Thanks, Charlie