Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Academics, we need to talk

matt-welsh.blogspot.com

Post reply on HN
Print view View original

121–130 of 170 posts

Re: Academics, we need to talk

#121

Post by mizzao »

forgottenpass wrote: post #9

Now, I don't think all academic research has to be relevant to industry. I'm getting kind of sick of this tactic, where this fake concession is made before going on about how academia isn't designed well enough to deliver to industry. If authors of articles like this didn't care about the efficiency of academia to deliver them research, what's the big deal if there are a bunch of people somewhere running around in ci…

Matt Welsh used to be a tenured systems professor at Harvard. He probably knows more about academia than 99.9% of people in industry.

Re: Academics, we need to talk

#122

Post by gluggymug »

nickpsecurity wrote: post #113
gluggymug wrote: post #111

Earlier quoted context omitted.

The flow is not good IMO. They haven't followed engineering practices which is one of the issues mdwelsh was talking about. If they've synthesized to 45nm and 28nm, where's all their synthesis stuff - constraints etc.? They have no back end stuff, very little docs, almost no tests, almost no verification infrastructure.

Hmm. Im clearly not an ASIC guy so I appreciate the tip on this. News to me. Ill try to look into it. Any link you have where people mention these and any other issues?

Maybe I was a bit harsh with the "almost no tests". They have some tests.

Someone name fmarch asked on https://news.ycombinator.com/item?id=10831601 about verification against the ISA model.

It possibly can be done via a torture tester apparently, https://github.com/ucb-bar/riscv-torture , but taking a quick look I don't think it handles loops, interrupts, floating point instructions etc.

Re: Academics, we need to talk

#123

Post by imh »

mdwelsh wrote: post #12
forgottenpass wrote: post #9

Now, I don't think all academic research has to be relevant to industry. I'm getting kind of sick of this tactic, where this fake concession is made before going on about how academia isn't designed well enough to deliver to industry. If authors of articles like this didn't care about the efficiency of academia to deliver them research, what's the big deal if there are a bunch of people somewhere running around in ci…

(I'm the author of the original blog post.) Ad hominem attacks aside, I do think it's a big deal if a bunch of academics are spending time working on the wrong things, if for no better reason than maximizing efficiency: Don't forget that most academic research is funded by the government. Since I also happen to help decide which research proposals Google funds, I also care that academics are well-aligned with the pro…

How do you think about pure mathematics research then? Or fundamental physics? Or art/literature/history's place in publicly funded universities? There's so much more to all of this than pushing forward in industry.

Re: Academics, we need to talk

#124

Post by jldugger »

nickpsecurity wrote: post #71
pklausler wrote: post #52

Earlier quoted context omitted.

Google patents a lot of cool stuff, actually. But the ideas tend to not get noticed in academia; when was the last time you saw a patent in a bibliography?

Academia patents tons of stuff. They just put that in patent applications and technology portfolios instead of bibliographies.

The point is that patents don't appear to contribute to the sum total knowledge that academics build on top of. They're by and large intellectual dead ends, if profitable ones.

Re: Academics, we need to talk

#125

Post by nickpsecurity »

jldugger wrote: post #124
nickpsecurity wrote: post #71

Earlier quoted context omitted.

Academia patents tons of stuff. They just put that in patent applications and technology portfolios instead of bibliographies.

The point is that patents don't appear to contribute to the sum total knowledge that academics build on top of. They're by and large intellectual dead ends, if profitable ones.

Oh I mostly agree. Some organizations will only fund you if they get IP out of it. Many significant tech in that area. But it's mostly an after the fact make money thing. Or just bullshit altogether.

Re: Academics, we need to talk

#126

Post by bashinator »

hackbinary wrote: post #93
forgottenpass wrote: post #9

Now, I don't think all academic research has to be relevant to industry. I'm getting kind of sick of this tactic, where this fake concession is made before going on about how academia isn't designed well enough to deliver to industry. If authors of articles like this didn't care about the efficiency of academia to deliver them research, what's the big deal if there are a bunch of people somewhere running around in ci…

I'm getting kind of sick of this tactic, where this fake concession is made before going on about how academia isn't designed well enough to deliver to industry. This +1, er +1,000,000. I would firmly state that academia by definition is not supposed to 'deliver to industry', and not even to society. The whole point of academia is to give the very smartest people (whatever that means) the freedom to explore ideas, so…

If I recall correctly, Maxwell was once asked by a minor noble-person what value his research would lend to industry.

Re: Academics, we need to talk

#127

Post by nickpsecurity »

gluggymug wrote: post #122
nickpsecurity wrote: post #113

Earlier quoted context omitted.

Hmm. Im clearly not an ASIC guy so I appreciate the tip on this. News to me. Ill try to look into it. Any link you have where people mention these and any other issues?

Maybe I was a bit harsh with the "almost no tests". They have some tests. Someone name fmarch asked on https://news.ycombinator.com/item?id=10831601 about verification against the ISA model. It possibly can be done via a torture tester apparently, https://github.com/ucb-bar/riscv-torture , but taking a quick look I don't think it handles loops, interrupts, floating point instructions etc.

There didn't seem to be a lot in there but I don't know Scala. I wish it was scripted in Lua or something with the Scala doing execution and analysis. Make it easier for others to follow.

Doesn't seem nearly as thorough as what I've read in ASIC papers on verification. They did (co-simulation?), equivalence, gate-level testing, all kinds of stuff. Plus, you did it for a living so I take your word there. I do hope they have some other stuff somewhere if they're doing tapeouts at 28nm. Hard to imagine unless they just really trust the synthesis and formal verification tools.

The flow is here:

http://www.cs.berkeley.edu/~yunsup/papers/riscv-esscirc2014....

Are those tools and techniques good enough to get first pass if the Chisel output was good enough to start with? Would it work in normal cases until it hits corner cases or has physical failures?

Re: Academics, we need to talk

#128

Post by mcguire »

knughit wrote: post #42
cushychicken wrote: post #36

Earlier quoted context omitted.

>it's a big deal if a bunch of academics are spending time working on the wrong things At the risk of sounding a little prickly with my comment - isn't it a little presumptuous to write off a certain amount of research as "wrong"? There are plenty of examples of research that didn't have a clear purpose leading to breakthroughs - safety glass, microwave ovens, and Teflon all spring to mind. Also, focusing on efficien…

Matt is talking about academic who publish papers showing that System X is Y% better than System Z, under completely nonsensical conditions. They aren't advancing the state of the art, they are mapping the terrain of the compost heap, but don't realize it.

Is he? Maybe he should have said that.

Re: Academics, we need to talk

#129

Post by ddingus »

andrepd wrote: post #119
ddingus wrote: post #100

Earlier quoted context omitted.

As would others vote to continue it. Relevance is a very broad term, with industry a subset.

>Relevance is a very broad term, with industry a subset. That's very insightful. It's easy to lose sight of things and be reduced to a particular worldview. The Higgs boson doesn't have to have commercialization potential for us to justify spending a few billion and a few decades discovering it. :) It's an end in itself.

That's how I see it. Pure research is necessary and possible.

Re: Academics, we need to talk

#130

Post by mcguire »

nickpsecurity wrote: post #69
js8 wrote: post #37

Earlier quoted context omitted.

I think I agree with samth - it's not the academia, but the industry who needs to open up. Especially Google has a reputation of being very secretive.

You and samth seem to be missing the point: academic research is often built using methods that will ensure no real-world success while aiming to achieve real-world success. Factoring in real-world constraints, best practices, and existing work-arounds will let academics achieve better results on average. Baseline of practicality goes up. And again, these are all academic projects that aim at being practical. The poi…

You have read the hair shirt paper, right?
Post reply on HN