Minutes of Weekly Meeting, 2011-05-16

Meeting called to order: 11:05 AM EDT

1. Roll Call

Eric Cormack
Harrison Miles
Patrick Au
Ian McIntosh
Tim Pender
Carl Walker
Brian Erickson
Brad Van Treuren (joined 11:08)
Peter Horwood (joined 11:09)
Heiko Ehrenberg (joined 11:40)

Adam Ley

2. Review and approve previous minutes:

05/09/2011 minutes:

  • Draft circulated on 05/10/2011.
  • Three corrections noted, all in the discussion, item 4a:
    • In comment from Harrison change 'ha' to 'the',
    • In comment from Harrison change 'mpi.org' to 'mipi.org' and remove comment,
    • In comment from Ian wrap 'A N Other' in quotes.
  • Key takeaways to be added:
    • [Brad] This is not a only a BSDL problem - circuit model too.
    • [Harrison] Object oriented modeling may be a key (abstraction).
    • [Ian] Impacts across multiple standards.
    • [Brad] Tooling needs 'exceptions' to allow for variations (abstraction).
  • Motion to approve with the above amendments by Eric, seconded by Patrick, no objections or abstentions.

3. Review old action items

  • Adam proposed we cover the following at the next meeting:
    • Establish consensus on goals and constraints
    • What are we trying to achieve?
    • What restrictions are we faced with?
  • All: do we feel SJTAG is requiring a new test language to obtain the information needed for diagnostics or is STAPL/SVF sufficient? see also Gunnar's presentation, in particular the new information he'd be looking for in a test language
  • Ian/Brad: Condense gateway comments and queries into a concise set of questions. - Ongoing
  • All: Forward text file to Ian containing keywords from review of meeting minutes. - Ongoing.
  • Carl/Brad: Get annotated keyword worksheets to Ian by Wednesday Close of Business. - Ongoing
  • All: Consider how a keyword can be used to define the chain configuration for a given test step, and what that keyword might be.
  • Adam: Prepare response to Brad's enquiry on identifying different cores in a device within dot7. - Ongoing

4. Discussion Topics

{The order of items does not follow the agenda in the Calling Notice}

  1. The Multiple BSDL Problem
    • [Ian] Unfortunately, we're missing Adam's contribution here today, but he hopes to circulate something prior to the next meeting.
    • [Harrison] It's good to learn of other perspectives on an issue. I think I need to note that none of the other Working Groups are looking at BSDL. In the February or March issue of the IEEE magazine there was an article that discussed the increase in functionality of devices while the package size is reducing. This is mainly coming from cellphones and tablets where the desire is to push up speed but with lower power, for battery life. The way they're doing this is to reduce the clocks but use multiple processors.
    • [Harrison] Quiescent power is becoming an issue. Growth is mainly in the mobile market. There are pressures on the BSDL file and how JTAG is used.
    • [Harrison] The major players are all different from each other in how they approach things and this makes testing difficult. My idea with the keyword was that the tester can highlight to the tool where something needs special attention.
    • [Harrison] I think this gives a fairly straightforward way to progress. You keep the BSDL as a flat file: It has the necessary information, but needs the human dimension, depending on what you want to test, and that's where the keyword came in.
    • [Harrison] Changing the structure of BSDL is something that's industry wide, so you need some kind of abstraction.
    • [Brad] That doesn't sound like abstraction as it's normally applied to object orientation.
    • [Harrison] I mean it as a tool for the tool person.
    • [Brad] In object orientation, you get the same behaviour but in a different implementation.
    • [Harrison] I'm thinking short run over long run here. A keyword seems easier to implement.
    • [Brad] Look at HSDL and the way it 'includes' things: There you can have the include of a HSDL device or the include of a HSDL module, where a device has more low level primitives than a module has.
    • [Ian] I'm not sure I'm really grasping how we actually use the keyword: I'm finding if difficult to put what's in my head in words. The main issue seems to be the need to switch BSDLs at points during test. I tend to agree with what Harrison said about not changing BSDL; it's been around a long time and there will be inertia to any change other than small incremental changes. But how do you get the keyword to instruct the tooling appropriately?
    • [Harrison] You don't want to get too tied into how any specific implementation will work.
    • [Ian] No, it's the premise of SJTAG that we remain agnostic to specific devices or tools that may be used to implement SJTAG solutions.
    • [Harrison] SJTAG needs to get leverage from the buy-in of JTAG. It's a way to get to the DFX of silicon. It's picked up on at the device level and at the board level, but not enough at the system level.
    • {Heiko joined, 11:40}
    • [Harrison] What dot7 did was largely driven by the need for performance in debugging, not test.
    • [Ian] We should keep in mind that what we propose should be for the benefit of the end users, not just to promote the use of JTAG for it's own sake.
    • [Harrison] Yes, but JTAG doesn't need functionality to do 'bare metal' testing. Other methods will need to use the vendor's drivers, but before you can even get to the drivers then you need to have millions of instructions. A lot has to work.
    • [Harrison] JTAG gives you Go/NoGo with very little function. Once you have that then you have lots of tools you can use to hammer on the design. And once you have fully instrumented silicon then maybe you don't need JTAG. But we're not there yet.
    • [Brad] Ian, I think the fundamental problem is that the execution model for JTAG needs to change. It uses things like SVF and STAPL which are static representations. We can't do what we're trying to do without changing the current flow.
    • [Brad] BSDL is utilizing the dot1 state machine without understanding what the state machine can do.
    • [Harrison] Probably slanted towards the more general cases. SERDES tuning, bring up of boards without putting devices in danger has gone beyond the simple intent of the state machine. People want to use JTAG to tune their SERDES interfaces but that's not what dot1 was meant to do.
    • [Brad] The way things are, it as if all the devices in the chain operate as a unified machine. It's not quite as if they were in 'lock step', though.
    • [Harrison] There's a whole bunch of stuff that was never managed before. To some extent we have a 90's standard based on 80's technology that we're trying to use to solve 21st century problems.
    • [Ian] OK, I need to cover a few other things before we close today, so I'll need to stop this discussion there.
  2. Email Problems
    • [Ian] A few people have commented in recent weeks that they weren't always getting the group emails. To check this I sent out two emails, one through my own Multiple Recipient Address and the other through Carl's reflector, and asked for replies.
    • [Ian] Obviously, if some people didn't get those emails, then they wouldn't know to reply, which is why I'm raising the subject here.
    • [Patrick] I got both emails OK.
    • [Eric] I only sent one reply, but I got both emails.
    • [Peter] Our mailserver wasn't parsing the 'envelope-to' entry, as you suggested.
    • [Ian] So that looks like it's fixed now?
    • [Peter] Yes, should be.
    • [Brian] I got the emails, so my problem looks OK now.
    • [Ian] I did get strange bounce messages for you several months ago Brian, but none recently.
    • [Brian] If you use that later email address I gave you, then it should be OK.
  3. May/Spring Newsletter
    • [Ian] The next newsletter is due at the end of the month. I had hoped to have a draft to share, but I didn't get the time to spend on it last week.
    • [Ian] I'll take an action to prepare something during this week for review prior to the next meeting. In the meantime, any suggestions for material to include are welcome.
    • [Brad] Some small item on the Multiple BSDL issue may be worth including.
    • [Ian] Yes, OK, and if we can include a diagram to illustrate it, maybe one of your examples, then that could help get the idea over.
    • [Brad] It seems that the way this subject increased the attendance on the call suggests that this is something people relate to.
  4. ITC WG Posters
    • [Ian] I got in touch with Bill Eklow earlier today to let him know that I expect to be able to attend ITC and so should be able to support the poster session again this year.
    • [Ian] Bill has indicated that they expect to have a fairly full programme of board and system level papers this year. That should be good for us since I don't think we really catch the interest of the more chip-centric attendees.
    • [Brad] Well, I was brought out of retirement to review a few submissions, so that would confirm it.
    • [Harrison] I've seen a few too.
    • [Ian] I asked who was likely to attend, but only Harrison responded.
    • [Heiko] I will be there.
    • [Brian] I will too.
    • [Peter] I won't be able to spare the time.

5. Key Takeaway for today's meeting

  • [Brad] Abstraction needed to get away from tool specifics.
  • [Brad] 'Execution Model' of JTAG may need to change.

6. Schedule next meeting

Next Meeting:
May 23rd (11:00 AM EDT, 4:00 PM BST) - Heiko unlikely to attend, Brad may have difficulty joining.
No meeting on 30th due to Memorial Day holiday in US.

Schedule for June 2011:
6th, 13th, 20th, 27th.

7. Any other business


8. Review new action items

Ian: Prepare draft of Newsletter for review at next meeting.

9. Adjourn

Eric moved to adjourn at 12:07 PM EDT, seconded by Peter.

Respectfully submitted,
Ian McIntosh