Mary Jane Irwin Receives The Kaufman Award

Mary Jane Irwin just got back from a cruise around the Greek islands with her husband of 53 years to celebrate being the first woman to receive the Kaufman award. When I wrote my post The 2019 Kaufman Award Goes to Mary Jane Irwin about her receiving the awards last week, I mostly just used the boilerplate biographical information from the press release. But that’s rather dry, so I called her up last week to get more color.

Mary Jane grew up in Tennessee when she was young. Her father was a faculty member at what is now the University of Memphis. Her mother was a teacher. As Mary Jane told me:

My mother’s model was for me to be a teacher, but I did an internship in elementary school and I knew it wasn’t for me.

She married early and then went to college with her husband at Memphis State University where she got a BS in mathematics. Her husband landed a job in Indiana so she applied to the University of Illinois Urbana-Champaign. It was the perfect fit so they moved there for six years. She discovered what was involved in being a professor, not just teaching but also doing research. She loved working with colleagues. She ended up working with her PhD advisor Jim Robertson. He was a designer of the ILLIAC II supercomputer but was best known as one of the proposers of the SRT division algorithm (the algorithm is named for its three creators, Robertson obviously being the R). The algorithm would become even more famous in 1994 when Intel got one of the lookup tables wrong. She graduated with a PhD in computer science in 1977, still thinking of herself as a computer architect. As she told me, “I fell into EDA by accident and I’ll tell you how that happened.”

When she graduated, she and her husband didn’t want to go to a big city so she came to Penn State as a faculty member. She said she never really considered industry. She just wanted to be at a research university and not just do teaching.

She divides her career into four decades. The first decade, starting in 1977, she did what most junior faculty do: publish enough to get tenure. She was interested in design but they didn’t really have everything needed to do actual implementation. Bob Owen was one of her graduate students. He left and went to IBM but then came back and she and Bob got a collaborative program going. Unfortunately, Bob died early in his late 40s.

The second decade, roughly the 1990s, were the “stars aligned” period. CMOS was the new technology in town and they were teaching it in classes. MOSIS had by then come into existence so they could fab chips. They got funding for what she called a “building project,” being paid to actually create something real. This was not easy to come by, since faculty grants were usually just enough to pay salary, one grad student, and a couple of conference trips. They were building signal processing architectures, custom chips in CMOS. They had a team of grad students by then, using interesting arithmetic components for discrete Fourier transform (DFT) hardware. In that era, silicon area was precious so they really needed to minimize the number of transistors. So they started trying to do logic synthesis for multi-level logic.

We fell into developing design tools out of necessity.

They developed multi-level logic synthesis by exhaustive search, which obviously only works on tiny designs. Even then, it took a week on their VAX (roughly 1 MIPS), so they had to beg their colleagues to keep off the machine and hope it stayed up for long enough. They had to come up with a language as input for this (Verilog existed but wasn’t really out there yet). They built a layout tool to go with this using a gate-matrix type of design. They decided they should submit it somewhere, and the obvious choice was DAC. Of the four papers they submitted, two were accepted, so those were the first of many DAC papers.

So they were doing chip design, fabrication, test, board design, and algorithm development for the DFT boards. There were two board-level designs with very different architectures. This was all enabled by a large team of grad students. By the end of the 1990s it became clear that area was no longer the main design constraint, it was power.

So Mary Jane started a completely new project, thinking about power consumption in multi-core designs and tools to measure it. They created a tool, SimplePower, to measure the power on a 5-stage pipeline architecture, and that tool was widely distributed and downloaded about 500 times.

Her interest in power didn’t go away, but she started to look at new technologies such as emerging memory technologies and upcoming technologies such as deep learning.

She retired in 2017 although the team still exists. She’s no longer actively working on research. She told me it is a great honor to receive the Kaufman Award but she couldn’t have done much without “great colleagues and fantastic grad students.”

She continues in retirement to be involved in professional service. “Paying it back,” as she put it. She’s more active now in ways to encourage women to come into science and engineering through things like the DAC Women’s Workshop and other institutions. She’s also trying to pay it forward to the next generation and is as active as she can be in writing nominations for women (“and men sometimes”) for promotion, tenure, awards, and so on.

You can meet Mary Jane in November when she will be at the Kaufman Award Dinner, where she will formally receive the award.

The Kaufman Award Dinner will take place on November 7 from 6:30pm to 9:30pm (PSA: it always runs late) at The GlassHouse in San Jose (2 South Market Street), hosted by the ESD Alliance and IEEE CEDA. Registration for the dinner will soon be up on the ESD Alliance website now that the announcement is public. I’ll see you there, too.