2021 AM Turing Award Recipient Jack Dongarra Turing Lecture: "A Not So Simple Matter of Software"

Jack J. Dongarra received the 2021 ACM A.M. Turing Award for pioneering contributions to numerical algorithms and libraries that enabled high performance computational software to keep pace with exponential hardware improvements for over four decades. Dongarra is a University Distinguished Professor of Computer Science in the Electrical Engineering and Computer Science Department at the University of Tennessee. He also holds appointments with Oak Ridge National Laboratory and the University of Manchester.
Dongarra delivered his Turing Lecture, "A Not So Simple Matter of Software," at SC22 (sc22.supercomp...) on Tuesday, November 15, 2022. In his lecture, he examines how high-performance computing has changed over the last 40 years, looks toward future trends, and discusses how a new generation of software libraries and algorithms is needed to use dynamic, distributed, and parallel environments effectively.
amturing.acm.o....

[@SelenaTheory]

Sitting in these lectures transports me back in time. What an amazing legacy.

[@hedrickwetshaves1997]

39:53 Nice laptop comparison. I'd love to have a graph of several laptops and smartphone chips of recent years with comparison of 1990s and 2000s supercomputers of their equivalent measurement. Maybe PowerPoint presentation of what they each look like, so you can brag to friends using this true apples to apples comparison. Any help with this would be most appreciated. 😉

[@hedrickwetshaves1997]

48:29 Is extremely informative and eye opening. Thank you!

[@hedrickwetshaves1997]

Main question: How to appropriately compare All/Total compute power of a Smartphone's CPU, AI, GPU, etc with a Supercomputer from the 1990s or 2000s. [Taking note that Supercomputers have traditionally been benchmarked with Vector double precision FP64 Floating Point Operations per second.] This might have to be theoretical peak performance, as I'm not sure if there are currently any programs which fully utilize all Tensor Cores and AI processing performance at once. Alternatively: How to compare a GPU today with a Supercomputer of the 2000s. I'm trying to understand FP16 INT8. etc in the comparisons which are promoted lately by the big names in chip processing. Any help is greatly appreciated.