larsberg

28 ms·

301.
by larsberg 15y ago
Unfortunately, I can't figure out how to make the point any clearer, so I've removed it. My apologies.
302.
by larsberg 15y ago
Yes, NUMA effects will really kill you, though how much depends on the particular quad-proc topology. I have some measurments for the interested in a small workshop paper I put together (I gathered the numbers in the context of tuning our g
303.
by larsberg 15y ago
Great question; I have no idea. We just get mail back from the editors that says things like, "your PDF breaks our scripts; regenerate it directly with LaTeX." Some, like Springer-Verlag, also have an editorial process where you have to sen
304.
by larsberg 15y ago
Not directly; only through index mutual funds. Anything else is illegal, as my wife works in finance (HFT). Unlike congress, employees of trading firms and their families are forbidden from trading any instrument related to what their firm
305.
by larsberg 15y ago
XeTeX-generated PDFs are not compatible with the toolchains of some academic publishers (Cambridge University Press, which publishes the Journal of Functional Programming comes to mind, but I seem to recall Springer-Verlag having issues as
306.
by larsberg 15y ago
Are you arguing not enough kids go into the hard sciences? That's certainly another debatable point, but different from the one I'm making. I'm pointing out that of the people who _already_ go into the hard sciences and even get a PhD, we c
307.
by larsberg 15y ago
I disagree. It's a lack of research job opportunities that pushes people from hard sciences and into, especially finance IT. At the risk of sounding like NDT, funding for research labs (and yes, NASA) would suck up the hard sciences people
308.
by larsberg 15y ago
Apart from the obvious Jane Street example, quite a few. F# is used pretty widely at this point in the Chicago area. Many of the larger banks either have their own private dialect of ML or a strict, ML-like subset of Haskell (Standard Chart
309.
by larsberg 15y ago
Thanks! Yes, the overloading issue requires changes to both the parsing rules and some fairly tricky formal semantics changes as well. I've talked with Dave MacQueen about it in the past, and it didn't sound like something he was terribly a
310.
by larsberg 15y ago
Standard ML! Of course, I do research on Manticore (parallel dialect of Standard ML) and am one of the primary developers left on the SML/NJ implementation. So I'm a bit biased. But trying to be objective, the module system is absolutely am
311.
by larsberg 15y ago
The older age demographic may work in their favor --- the "about to retire" crowd might be willing to work with them on pretty favorable terms, especially if they are flexibile on location and hours. It would be a heck of a lot more interes
312.
by larsberg 15y ago
When I worked at MSFT, we used to break up (senior) people according to which stage of the project they were better at leading, either as technical ICs or managers. If you chunk projects into four parts: early, build-up, build-out, and rele
313.
by larsberg 15y ago
If your only reason for using gcc is speed and you're on a Linux platform, you should really try out icc. The loop unrolling and vector op generation are much better than gcc's (at least for code we've tested on, YMMV) and can result in som
314.
by larsberg 15y ago
One of the things even the high-powered public speaking trainers tell you is that the only way not to be nervous is if you no longer care. Any time you have an audience you actually want to learn something or be convinced to take some actio
315.
by larsberg 15y ago
Yup, I connect to an external monitor. And worse, it doesn't seem to give that memory back without rebooting. Sorry --- that was an important detail that I should have included.
316.
by larsberg 15y ago
Agreed, particularly on my MacBook Air where ~1GB is reserved for the shared video memory. I have to close Firefox or Chrome if I want to do any development work (our research compiler is a bit of a memory hog) without having a process get
317.
by larsberg 15y ago
Curious point: > replace universities > heading down wrong path last couple decades. not fun for students or professors. Is that (fun part) generally true? I'm in my thirties, and I certainly had lots of fun at my previous job (MSFT
318.
by larsberg 15y ago
Not all samples have numbers in them, but I was somwhat disturbed that it was not until 2012 that the LISP publishers used a monospace font that clearly distinguished 0 "zero" and O "capital letter o".
319.
by larsberg 15y ago
Make sure to investigate the firm and, in particular, their enforcement of non-competes before you join to ensure you are comfortable with their terms. One of the three in your bullet (1) has a long history of enforcing (and backing up with
320.
by larsberg 15y ago
> deciding whether someone will fit in the Google clique That can be shockingly important. You can be a great developer, but still not a good fit for a company. As an example, I saw several employees top-out at relatively low points at
321.
by larsberg 15y ago
Yes, we're familiar with that work. One technical bit of interest is that we don't use the C stack at all (we have heap-allocated continuation records), so we're curious how that will play out with the calling convention even beyond the iss
322.
by larsberg 15y ago
You are correct --- PCC (AFAIK) has not come about in the sense of carrying along a proof object that a formal verifier can validate is both correct and corresponds to the code payload. As the other commenter pointed out, the SLAM tools are
323.
by larsberg 15y ago
That linked article is from 2007. Interestingly, the stuff that the author mentions as useless (proof carrying code) is now basically The Way that developers handle code that has security, reliability, or robustness requirements. - For exam
324.
by larsberg 15y ago
Not significantly. If you check out the sources for most functional language compilers (GHC, any of the ML family, scheme, etc.), code generation and micro-optimizations make up a tiny fraction of the code and developer time investment. LLV
325.
by larsberg 15y ago
> academic lanaguages don't succeed because the incentives aren't aligned A much harder problem is that compilers are shockingly time-consuming to fully implement and test, much less to integrate with a full set of libraries and tools t
326.
by larsberg 15y ago
Probably none, though I'm putting words into their mouths. Many seemed to be looking for the mentoring, "office next door to <X>," and team aspects as well. Modulo health insurance, many of these people probably didn't really need a
327.
by larsberg 15y ago
Totally untrue. I used to manage a team within the Visual Studio team, and we regularly had people come in looking at a 2x-4x pay cut (because they were going from "CTO" or "Chief Architect" titles to "Software Development Engineer" positio
328.
by larsberg 15y ago
In systems research, it seems to mean something slightly more specific: how do you architect systems that can cope with crazy amounts of generated data? Being in the middle of academic job talk season, we hear people talking about a petabyt
329.
by larsberg 15y ago
In our research group, almost everybody just uses emacs to write and aspell to check. One person also uses TeXLive, but he is the one non-emacs person for development as well (uses bbedit, I believe).
330.
by larsberg 15y ago
I'm sort of skeptical that any of these will be usable as a complete replacement for a very long time. Heck, I can't even use pdflatex today because the toolchains for several conferences and journals (I'm looking at you, Journal of Functio

More ›