software job advice

[Evan]

:?:

[Mike]

"news" is getting a link from out of nowhere. this happened a few months back... i forget the cause.

but it has to be in the middle of a sentence without any punctuation around. the following use of news will have a link, but i didn't put it there.

[HAULN-SS]

Welll..that's not the whole story. I do enjoy the work, but right after i started, we had a management change. That part has sucked. The guy there now is definitely drinking the corporate koolaid (this is a small project, and we have a massive budget underrun). There is no real NEED for us to save money, the money is already there, and they have been spending it frivolously to try fill the gap. Not spending it on the employees is a mistake, I think.

if I can move to make the same or better money to a different company, doing the same work, why not?

[Mike]

if I can move to make the same or better money to a different company, doing the same work, why not?

eventually changing jobs every year will catch up to you.

[HAULN-SS]

I'm pretty far from fickle. This job, for sure, is leaving here next year. Since I am pretty sure I am not going to NM, it probably doesnt hurt to shop around I would think. I've only ever changed jobs once since college, and i stayed within Lockheed, so to the outside world I think it probably looks ok.

[Apoc]

if I can move to make the same or better money to a different company, doing the same work, why not?

first in is usually first out in a bad economy but if you know the job is going away anyway...

[Evan]

I was at Booz for 7 months and it certainly didnt hurt my next job prospects. (I think it helped me, everyone I interviewed loved that I hated Booz )

2 years is my minimum target to stay at a job but as long as you can explain it away with "my job is moving and Im checking out other opportunities" then its no problem.

[stevegula]

Mike is right that you don't want to look like a job hopper.

Evan is right that recruiters won't even bat an eye if it's due to relocation of job.

[damnit458]

What kind of PDE's / which numerical methods do you usually see? I would assume most of it is FEM stuff? IMO coding that stuff can be a real pain in the ass sometimes, although I do think it's pretty interesting.

[HAULN-SS]

The code I am working on these days is using Poissons equation. I have solvers written to do Gauss Seidel with Successive Over Relaxation, Dynamic ADI (alternating direction implicit) , Conjugate Gradient and Boltzmann methods. We also do some other algorithms too, and the first project I worked on here was doing an out of core LU decomposition (out of core because most of our problem sizes might have 500k + unknowns).

And yes, it is a pain in the ass to code, and especially to code in a way with optimal memory access. Most of the problems are memory bound, not computationally bound.

[CaptainHenreh]

What do french fish have to do with math?

[HAULN-SS]

heh, what a disturbing picture

[damnit458]

It's to bad that your having issues with your job because it sounds like pretty interesting. I would imagine looking into the different equations is hard to do (or at least understanding them sufficiently), but pretty rewarding in and of itself. I'm doing research on a Finite Element Method approach to solving the Navier Stokes Equations, but in terms of the coding I catch a little bit of a break since a lot of times you can get away with using other packages like FREEFEM to do the coding, which simplifies things immensely.

[HAULN-SS]

It is really interesting. I remember taking numerical methods classes and just assuming that only PhD students cared about such things. You are correct that looking into the equations is hard. I spend a lot of time reading papers on applications of X equation to Y problem. Trying to understand them sometimes forces me to go back to simpler things and work my way forward (Looking at poissions equation for our application the first time was a little intimidating: and as a side note the book "Div, Grad, Curl and all that" is a good reference)

I have never actually heard of FREEFEM but I just looked at it briefly just now, and I suspect it runs like crap. Actually, it looks like something I might get, and have to work on to make run better as part of my job. Stuff like these arithmetic if statements within loops runs absolutely terrible!
// we remove the boundary link
for (int k=0;k<nt;k++)
for (int j=0;j<3;j++)
if (TonBoundary[k] & MaskEdge[j])
{
int s0,s1;
VerticesNumberOfEdge(triangles[k],j,s0,s1);
// cout << s0 << " " << s1 << " ld " << linkd[s0] << " " << linkd[s1] << " lg " << linkg[s0] << " " << linkg[s1] << " apres " ;
linkg[s0] = linkg[s0] != -1 ? -2 : -1;
linkg[s1] = linkg[s1] != -1 ? -2 : -1;

linkd[s1] = linkd[s1] != -1 ? -2 : -1;
linkd[s0] = linkd[s0] != -1 ? -2 : -1;
// cout << " ld " << linkd[s0] << " " << linkd[s1] << " lg" << linkg[s0] << " " << linkg[s1] << endl;

}

What are you doing these days? grad school I suppose? Where at? I am going to try to find a place where I can still do this kind of work, even if I quit this job, I think. It is a lot of fun getting some physics code to run 60% faster, and also staying sharp on any math skills I was starting to forget as well.

[damnit458]

FREEFEM is used less for efficiency and more for the short length of time it takes to get something running. Generally we use it to test new methods on small scale problems, solely because it eliminates a LOT of the coding you would have to do. For instance, what I'm doing specifically deals with creating two different meshes, and then solving equations on each mesh to get an approximation. To actually write down what I need in C for the two meshes takes a hell of a lot longer than what I need to write in FREEFEM. That said, you're right in that it is inefficient, so larger scale problems are coded in C.

I'm in the PhD program at Clemson. I actually came in here a few years ago thinking I had no desire to do much in terms of computational mathematics, although once I got here I obviously changed my mind.