I interviewed Andrew & Yunsup as part of my RISCY BUSINESS series! The episode is available here:
Edit: Annotations are up for the interview, you can find them in the youtube description. They timestamp each question/response/etc. to make it easy to jump to parts you are interested in. I’ll post the raw text here as well:
0:01 Welcome and introduce Andrew Waterman & Yunsup Lee
1:07 A highlight reel of life stories before RISC-V
1:22 Andrew Waterman: Life before RISC-V
2:46 Yunsup Lee: Life before RISC-V
4:18 Why decide to create a new ISA?
4:33 Andrew: On creating a simple basis for performing research
6:23 Yunsup: On removing complexity in order to enable development by small teams
7:12 What drew you specifically to RISC?
7:34 Yunsup: Necessity
8:39 Andrew: On following RISC I and RISC II
9:16 What appealed to you about having it be open hardware?
9:32 Yunsup: Openness enables understanding and learning
10:41 Andrew: On finding encouragement in the competition of ideas in the open source community
12:57 Yunsup: On the potential for open hardware enabling schools to teach what the industry actually uses
13:36 What are you hoping to see happen in the world of RISC-V?
13:51 Andrew: A RISC ISA is suitable for many domains of computing, from the lowest to the highest performing computing
15:15 Yunsup: On the need for customisation with the end of Moore’s Law
16:01 Do you think RISC-V being in anything and everything is an attainable goal, and what markets may RISC-V most thrive in?
16:24 Andrew: Embedded systems contain all sorts of chips
17:21 Yunsup: New markets
18:23 Chat comment: "My question is, considering the huge investment that is currently happening in custom chips for AI, how do you think SiFive can attract investors in the field of AI to pay more attention to your platform?"
18:50 Yunsup: Customisable chips may well serve machine learning research
20:27 Andrew: Evaluation of neural networks will frequently be done in energy constrained environments
21:11 Can you tell us a bit about SiFive in terms of what they offer customers?
21:29 Yunsup: Democratising access to custom silicon
23:07 Andrew: On the expense of building custom chips using a proprietary ISA
24:46 What are the coolest HiFive1 projects you’ve seen?
25:01 Yunsup: Printer controller and Zephyr port
26:30 Andrew: On the community porting the realtime Zephyr operating system to the HiFive1
27:12 Will SiFive release products using Freedom Unleashed?
27:30 Yunsup: Why build the HiFive1 board
28:22 Andrew: Excitement for a Unix capable RISC-V system coming out relatively soon
28:41 Any resources to recommend for learning about RISC-V or open hardware in general?
28:53 Andrew: The RISC-V User-Level ISA Specification - https://riscv.org/specifications - and two text books coming out this year based on Computer Organization and Design - https://www.elsevier.com/books/computer-organization-and-design-mips-edition/patterson/978-0-12-407726-3 - and Computer Architecture: A Quantitative Approach: https://www.elsevier.com/books/computer-architecture/hennessy/978-0-12-383872-8
29:55 Yunsup: Graduate-level course at Berkeley
31:25 Andrew: Plug Berkeley’s use of RISC-V in the classroom
32:05 What advice do you have for those looking to get involved in the RISC-V community?
32:24 Yunsup: If you want to learn how computers work, RISC-V is a good resource
34:27 Andrew: There’s a lot of work to be done on compilers, operating systems, software, microarchitecture, etc
35:04 Chat comment: "Is there any plan for MOOC (massive online open courses)?"
35:45 Andrew: On Berkeley possibly teaching CS61C as an online course: http://inst.eecs.berkeley.edu/~cs61c/sp17/
35:57 Point out the SiFive Forums as a good resource: http://forums.sifive.com/
36:12 Andrew: Shout-out to Megan Wachs
36:35 Give it a minute to see if some questions pop up
37:38 Suppose that is everything, but wait…
37:59 Chat comment: "What is your answer to Linus Torvalds when he said ISA is not important and it’s the community around ISA that matters?"
38:15 Andrew: On ISAs not mattering, especially in high end computing
39:28 Yunsup: Why not RISC-V if instruction sets don’t matter?
39:50 Andrew: It takes a village, so if you have resources to help port to RISC-V, pitch in!
40:03 Chat comment: "I missed the start, but I was wondering: Do you personally have anything you’d like to see RISC-V users do that isn’t necessarily endorsed or endorsable by SiFive?"
40:28 Andrew: The more implementations there are, the better the project will be for everyone
41:43 Yunsup: It gets stronger the more people look at it
42:57 That’s the end for real
Annotated by Miblo - https://handmade.network/m/Miblo