REFERENCE - THE JANE SERIES
CHAPTER 2 - RECOVERY 101
Jane had been in the company for 4 years and had seen the disastrous results of poor pre-silicon verification. When she worked in the lab, her team had spent weeks trying every possible configuration of the THX1000 hoping that it wasn't completely DOA. Her claim to fame was having placed an inverter on the parity pin, combined with aligning the base memory address with 16 zeros, and disabling the external memories. When the lab team saw that all the data coming out of the DUT was corrupted, they were finally able to convince the management that it wasn't just a faulty board.
She returned to her desk and clicked on the my_resume.doc in her favorites category. She wondered if there was something worth updating since last week. After staring at her resume for a few minutes, she picked up the phone and called Josh. Josh was a college buddy of hers who she had discussed pre-silicon verification with, the previous weekend. He seemed to be having a great time working in pre-silicon, which seemed a bit strange to her as she had seen countless engineers go into the pre-sil team in her department over the years and very few to remain beyond one or two years.
Josh listened as she went over her boss's ideas, one by one. Then he told her "It looks like you're ready for the 'The verification system'.", he continued to describe a five part system which gets you out of the verification rut. He described the first steps you need to take in order to get management buy-in to the dramatic changes, and gave her pointers to some of the industry articles on the subjects he described.
Over the next few days Jane set out on a two-part plan. One part to build a work-plan which reflected the cost of continuing their current 'methodology', the other an alternative plan if they adopt 'The verification system'. The first part of the plan involved talking to all of the people involved in the pre-silicon of the previous projects. She set out to get the real cost for each of the major blunders of the existing methodology. She asked the engineers about the amount of time they had spent in upkeep of all the legacy environments, the cost of porting tests, the ROI (return on investment) in training 'borrowed workers' from other groups, and what would be required to bring a silicon back bug-free in the current way they were doing things. She continued to get details on the costs of all of the other requirements; Working without a spec, working with inexperienced engineers, minimal support from the RTL team, verifying without probing any internal signals, support of many test-writers, not using HVLs, relying on internal development only, integrating non-tested code. It was quite a lot of work but it was material well worth the effort.
When she came back to work
on Monday morning, she had the complete plan mapped out. Mitchell gasped
when he looked at the project file. It showed a tape-out in August of
2009. His face turned red, and then he started to look at the separate
parts of the plan. She had estimated 1654 days for porting legacy tests,
923 days supporting the environments, 167 days of training borrowed workers,
486 days of redoing work of borrowed workers, and 1670 days of reviewing,
fixing and redoing tests before tape-out. With the resources available
it ended up as an August tape-out with a moderate to low probability of
achieving the zero-bug goal. He went one by one through each estimate,
trying to understand the reasoning behind it. She had clear data from
her many interviews on how much time it took to port tests, upkeep environments,
do training and the amount of time it took to review and redo things to
be sure they really do all they need to do. When he was done asking all
the great manager questions, and getting firm answers he sat back and
started to think. Two minutes seemed enough, before he looked at her and
said "I need you to present an alternative plan by tomorrow morning,
We need to show a working product by May of next year if we want to keep
this department open. I need you to keep an open mind". Jane smiled,
he had taken the bait.
Design right. Verify right. Ace Verification
About Us | Careers | Contact Us
Copyright © 2006 Ace Verification Corp. All rights reserved.