The h Bar

12:01 AM

@ChrisWhite yeah visiting places in person made a HUGE difference

I absolutely fell in love with Berkeley

you said you'd be starting at Berkeley in September right?

@FenderLesPaul Berkeley is nice, and bustling. For the place alone I would prefer Santa Barbara, but Berkeley has a big advantage on the "things happening all the time" front. If you care for that kind of thing. And more choices in exotic food, too.

skill patrol

12:21 AM

not to mention you'll be nearer to the capital of raider nation :P

skill patrol

12:33 AM

Oh, he's gone? That was a good example of a "drive by" comment :D

user54412

@dmckee Yes. Evil.

user54412

@BernardMeurer It makes beautifully fast code. But we do all our development on gcc. icc is slower to work in new standards (do they even fully support c++11 yet?), and has more subtle optimization bugs. It also lets you get away with more that it shouldn't -- "You called sqrt(), so you must surely have meant to include cmath, let me do that for you silently."

user54412

I still half suspect intel chips have a secret flag that slows down code not compiled with icc.

user54412

Except there are now other architectures where icc makes better code than the compiler made for that architecture.

user54412

@FenderLesPaul Yes, September 1 if all goes well.

skill patrol

12:54 AM

Did you @dmckee watch the video I linked?

Bernard Meurer

@ChrisWhite Have you seen what icc does with bad branch prediction? It's bloody magical

user54412

@BernardMeurer ?

Bernard Meurer

@ChrisWhite stackoverflow.com/a/11227902/2080712

Read the update in the end

user54412

1:09 AM

ha

user54412

that's only icc 11

user54412

icc 16 probably solves the halting problem

Bernard Meurer

LMAO

HA

@Obliv IT WORKS

My Euler calculator works :3

with arbitrary precision

user54412

SO: where you can get +1745 for answering an optimization question with unoptimized code, drawing all the wrong lessons from it

Bernard Meurer

@ChrisWhite Why is x&1 faster than x%2 for testing if something is even?

3075

1:14 AM

mit ocw has a new video lecture course on graduate level string theory - link

user54412

@BernardMeurer It's not. They'll generate the same machine code. If they don't, either your compiler sucks or x is some weird type where those two expressions are different in corner cases.

Bernard Meurer

@ChrisWhite Oh, nice

Also, earlier you were saying you are against boost and things like that. How would you deal with arbitrary precision using only the stdlib?

user54412

As a general rule, the first 1000 times you want to do bit twiddling for performance, the compiler has already thought of that.

2

Bernard Meurer

@ChrisWhite Makes sense, smart people wrote it after all

user54412

@BernardMeurer I... wouldn't :p But if forced to I would make a class that acted like numbers (define all the arithmetic operators and such) but internally had an array with on-the-fly memory allocation and such, and defined all operations on that data structure.

user54412

1:19 AM

Maybe bundling allocations into large chunks, creating a virtual heap that my own code micromanages for its purposes, if performance is an issue.

Bernard Meurer

:O

You're a c++ monstah though, I'm a little far from that point

is the -Ofast flag any good? is it pure magic?

user54412

what compiler?

Bernard Meurer

GCC

No money for Intel CPU ;-;

user54412

oh, that uses fast math -- things like the fast-inverse-square-root algorithm used in graphics

user54412

it's... a little dangerous

user54412

1:22 AM

your results will no longer be IEEE compliant, and in general will be wrong in the last few bits

user54412

on the one hand, if your floating point algorithm blows up because of inaccuracies, it's the algorithm's fault

Bernard Meurer

fast-inverse-square-root? That magic number crap?

user54412

on the other hand, in most real applications it's only a matter of time before things become too inaccurate, and fast math hastens your program's demise

user54412

to be honest, a lot of optimization lore is built around hardware traits that simply aren't true anymore

Bernard Meurer

How do you benchmark your code? Like I have my code for the euler function and someone else's, how can I accurately check which one is faster?

Also, would you have any recommendations for writing testcases for C++ code?

I feel like I'm question-avalanching you; I'm sorry

user54412

1:27 AM

vectorization gets you 2-8x speedup, pipelining is a factor of 5-50, and fundamentally better hardware is some additional factor -- the result is compared to 20 years ago, floating point operations are often free

user54412

what hasn't improved over the years is memory access speed -- they invented cache hierarchies, and that's about it

user54412

@BernardMeurer run them? :p

Bernard Meurer

@ChrisWhite Pff, but how do I chronometer it?

I know it in Python, but dunno in C/C++

user54412

@BernardMeurer google makes a decent unit testing framework I've used before

user54412

@BernardMeurer well if your code takes long enough you can just time it externally from the command line

Bernard Meurer

1:31 AM

@ChrisWhite It's blazingly fast, I just want to know if it's fast enough for me to win this bet :p

user54412

27

Q: Timing algorithm: clock() vs time() in C++

For timing an algorithm (approximately in ms), which of these two approaches is better: clock_t start = clock(); algorithm(); clock_t end = clock(); double time = (double) (end-start) / CLOCKS_PER_SEC * 1000.0; Or, time_t start = time(0); algorithm(); time_t end = time(0); double time = diff...

Bernard Meurer

@ChrisWhite If we go to DanielSank's wedding will you be my date?

<3

user54412

You've offered your love to half this chat room ;P

dmckee

Somewhere I have a link to a (now rather old) presentation about how compilers bit-twiddle. Money line "gcc is smarter than the video codex programmer on all platforms".

Bernard Meurer

@ChrisWhite Tsc, don't slutshame me

user54412

1:38 AM

On the off chance someone can help with html: I'm making a table in a github wiki, and it automatically alternates row colors. How can I turn this off given that I can't control the style sheet?

Bernard Meurer

I wish, I really do

dmckee

Link for above: linux-kongress.org/2009/slides/… It's from 2009, so it is quite ancient these days, but i keep it in mind for humility.

2

Bernard Meurer

"Warning: contains assembly language code." ::closes webpage::

skill patrol

indeed, assembly language keeps me humble

dmckee

1:55 AM

Another good lines in there "very few humans know this" about something a compiler does because it's faster on one manufacturer's x86_64 architecture and the same speed on others.

Bernard Meurer

@ChrisWhite Alright, it's blazingly fast

to my eyes at least

user54412

2:12 AM

@dmckee "Fancy data structures help on paper, but rarely in reality" and "If the data structure can’t be explained on a beer coaster, it’s too complex." These should be required learning for CS-students-turned-programmers.

Bernard Meurer

Holy crap C++ is just silly fast

Still reading that pdf, can't understand any of the assembly though

user54412

There's some pretty excellent stuff in there. Wish I could hear the talk that went with it.

skill patrol

yup

@dmckee thanks for sharing :-)

Bernard Meurer

3:53 AM

@DanielSank particlezoo.net/shop.html

Hari Prasad

5:33 AM

@BernardMeurer "From the Standard model of Particle Physics and Beyond!" :)

M1-3 Neutrinos

Hagen boson

Pomeron

The Dark Side

6:23 AM

23

A: Have I embarassed my supervisors by solving a problem that a PhD student in my group was working on without success?

Of course I can't say anything definitive based on the limited information in your question, but one possibility is that you may have inadvertently committed a faux pas. Your advisor may be upset that he/she didn't anticipate this possibility and head it off. One of the basic rules of the mathe...

^ ever seen an answer sitting at 23 = +43 -20 !!!

It is good for communities when people voice their disagreements openly, even when the author has 97k reputation.

Downvoted because the idea that an undergraduate can take any blame for taking an assignment, and performing it too well for diplomatic niceties, is beyond my ability to accept. — Daniel R. Collins Apr 13 at 4:46

when you started trying to prove the theorem, what did you think was going to happen to this graduate student if you succeeded? is holding the OP responsible for not thinking carefully. I strongly disagree with this sentence. What is the OP supposed to think? Why should he think what is going to happen? He just took the topic from his advisor and worked on it. What did he do wrong? What else can an undergrad do when he is given a thesis topic to work on? — scaaahu Apr 13 at 9:16

The posting explicitly says "the supervisor offered me a thesis on the topic". That being the case, it seems absurd to suggest a rule was violated. "you don't compete with graduate students by working on their thesis problems" seems like a rule for faculty. A student who is "offered" "a thesis on the topic" cannot be considered to be violating such a rule because he succeeds. — Michael Hardy Apr 13 at 5:11

(As I post that, I realize that 97k rep is the reason why this collected +43, considering that the premise is plain wrong.)

But still, one data point where not having a silent-downvote culture helps the community. Overall, this would be a thread that this community would be really proud of.

("Thinking out loud") :P

user54412

7:03 AM

@TheDarkSide Uh, why do you say that? That answer is completely right. It doesn't say the person was in the wrong, but that a faux pas was committed.

user54412

None of those commentators seem to know what they're talking about.

DanielSank

@BernardMeurer That's cute. Get your girlfriend one.

The Dark Side

@ChrisWhite See, it could be that we disagree here. With all due respect to the OP, this answer is incoherent at best. Contrast "I don't mean to suggest that you are primarily to blame", with (later in the para), "However, when you started trying to prove the theorem, what did you think was going to happen to this graduate student if you succeeded?".

And then, there's:

Bernard Meurer

@DanielSank She's not my girlfriend!

DanielSank

@BernardMeurer Ok.

Bernard Meurer

7:15 AM

:p

The Dark Side

"It comes across as callous to be more concerned with getting credit yourself than the repercussions for this graduate student."

Bernard Meurer

I always wanted to say that

DanielSank

Why?

Bernard Meurer

idk, sounds like a frustrated teenager thing, never had the opportunity to use it

The Dark Side

(contd.) Does one really expect the undergrad to question the adviser - why are you assigning me a problem that your grad student is already working on? It could be case of contrasting cultures, but I find that a little difficult to visualize, given the kind of power structures I'm familiar with.

So, it is really not the undergrad's fault that he's in this uncomfortable situation, and realistically, there's nothing he could've to prevent getting into this. Sure there's a faux pas, and blame predominantly goes to the adviser, but the discussion starting from "However, when you started trying to prove the theorem, what did you think was going to happen to this graduate student if you succeeded", seems to suggest that OP is accusing the undergrad to be eating off the grad student's pie.

I don't quite agree with that.

@ChrisWhite - ^ ( when you see this ) :)

@DanielSank - Congratulations on the upcoming wedding. :)

Bye guys. :)

DanielSank

7:23 AM

@TheDarkSide ...?

The Dark Side

@DanielSank From the chat logs:

2 days ago, by DanielSank

@BernardMeurer @ChrisWhite I'm getting married in July. If you want to come by we could probably split the cost of stuff.

Hence.

DanielSank

@TheDarkSide Oh right. I think @BernardMeurer didn't understand what I meant by that.

I meant that if he and @ChrisWhite wanted to get married we could split the cost of the venue, etc. ;)

The Dark Side

WHAT!

I should've read the logs properly. I just saw a bi-starred message right below it, and jumped the gun. Sorry. Enjoy single-hood.

:D

DanielSank

I am getting married.

Bernard Meurer

@DanielSank You're right I hadn't

DanielSank

7:33 AM

I'm just pointing out that @BernardMeurer misunderstood my joke.

Bernard Meurer

As usual :p

DanielSank

Well, it's probably because I don't deliver them well.

Bernard Meurer

I don't get jokes in english very well

it't not a funny language :p

DanielSank

7:49 AM

@BernardMeurer It's pretty good for humor.

Not as good as Russian.

kevin Tah N.

google.com/…

Feeds

9:02 AM

David Z has added Chris White to the list of this room's owners.

8

David Z has added John Rennie to the list of this room's owners.

user147690

9:21 AM

Congratulations @ChrisWhite @JohnRennie

Slereah

10:47 AM

Guys

Why are Hilbert spaces separable in QM

Despite the position and momentum basis being not countable

Oh it's because position basis isn't part of the Hilbert space

nvm

What's an example of a countable basis for a free particle in basic QM?

Bass

11:54 AM

@Slereah $\mathrm{sin}(nx)$ ?

Slereah

I s'ppose

What is it an eigenvector of?

Bass

@Slereah The differential operator $\partial_x^2$, i.e. $p^2$

Slereah

true dat

Faheem Mitha

Why is this room called "The h Bar"? A reference to Planck's constant, presumably. But why "Bar"?

Loong

H-bar

H-bar or h-bar can refer to: H with stroke, a Latin letter H with a doubled horizontal stroke (Ħ ħ) reduced Planck constant, which the above symbol represents as a mathematical symbol, ħ = h/(2π) Antihydrogen, an antimatter element represented by the symbol H Steyr AUG H-bar, the light machine gun version of the AUG assault rifle...

Faheem Mitha

12:09 PM

@Loong Ah, right. The scaled version. Forgot about that. Thanks.

Slereah

A bar is a place where people congregate

John Duffield

@FaheemMitha : and talk.

Faheem Mitha

12:31 PM

Yes, I got the play on words. You forgot the most important verb. Drink.

Slereah

We don't actually drink here, though

Well we can, but independantly of being there

Faheem Mitha

@Slereah Yes, I meant an actual bar.

Or pub, or tavern, or whatever other name one might choose.

Slereah

Feel free to drink while here, though

Might make for more fun conversations

skill patrol

12:55 PM

after you sleep it off you may find yourself banned

Faheem Mitha

1:10 PM

@Slereah I don't really drink, personally. I might if someone offered me good red wine for free. But that rarely happens.

ACuriousMind

1:47 PM

@Bass That's not a wavefunction of a free particle, as it is not square integrable.

@Slereah The countable bases for free particles are not different from countable bases for bound particles, as they share their Hilbert spaces.

Slereah

ah yes right

skill patrol

2:04 PM

@DavidZ may I nominate ACuriousMind to be added to the list of this room's owners?

Slereah

Oh @ACuriousMind

yesterday, by Slereah

Is the argument that perturbation theory doesn't work because 1) it's basically a polynomial expansion in $\lambda$, so it assumes that the propagator is analytic 2) the theory isn't defined for $\lambda < 0$ 3) same argument as compact support functions not being analytic

ACuriousMind

@skillpatrol No.

@Slereah 2)

If the perturbation series converged, it would have to have a radius of convergence

Slereah

Ah, and you can't because it is on $[0\infty)?$

Closed around 0

ACuriousMind

But any radius of convergence around zero coupling would imply that the series converges for a negative coupling.

Which...could happen, but contradicts all physical expectations.

It's really just a heuristic argument

Slereah