Author Topic: Downscaling or Computational Fluid Flow anyone?  (Read 1167 times)

0 Members and 1 Guest are viewing this topic.

Offline fortran

  • Member
  • *
  • Posts: 7
Downscaling or Computational Fluid Flow anyone?
« on: March 02, 2019, 08:37:57 PM »
Hello from Dawson Creek, BC, Canada (where the Alaska Highway begins).

Having a working weather station gets my foot in the door, so to speak.  I eventually want to have many weather related sensors on the farm.  I have some frost problems, so I am thinking about putting together a LWIR camera and suspend it form a balloon to try and image how frost flows down the landscape.

I am about 5 miles downwind of a 130+ MW wind farm.  So, part of my goals are to model how the wind flows across the landscape.  I downloaded the source to WindNinja more than 1 year ago, but not yet played with it.  I also have one or tow other packages which might work on that.

I also want to see if I can relate weather in my region to Global Circulation Model results.  So, I have a bit more than a PC here.  I have about 46 amd64 cores for CPU, a bit more than 10 TFLOPs of GPU, and about 10 TB of RAID-10 for storage.  My M.Eng. thesis way back when was number crunching.

Whether I can build hardware with big fingers and eyesight getting worse, is to be determined.  :-)

Offline Mattk

  • Forecaster
  • *****
  • Posts: 2136
Re: Downscaling or Computational Fluid Flow anyone?
« Reply #1 on: March 02, 2019, 08:44:51 PM »
So you mean you might be a F77 man as well :)

Offline fortran

  • Member
  • *
  • Posts: 7
Re: Downscaling or Computational Fluid Flow anyone?
« Reply #2 on: March 02, 2019, 09:48:30 PM »
I started with FORTRAN-IV.  For a while we had WATFOR and WATFIV at the university (on an Amdahl).  Then we had FORTRAN-VS (which is F77).  Off to graduate school.  Lots of UN*X machines, so I had real f77.  But the machine I ran my model on was a VAX 11/785.  My model was a dynamical system, where the number of equations decreased randomly with time.  And so I had doubly linked lists and manual garbage collection to keep my data space compact.  After that, I had some exposure to M$ writing a FORTRAN-77 compiler in Pascal (it caused me grief).  I have yet to write anything in f90 or newer.  But, I once read that someone did not know what the specific name of the number crunching language of the future would be, but it would be some kind of FORTRAN.  (I butchered that, but you probably get the sentiments).

I don't know how I am going to do things here.  I have been running BOINC to exercise my GPUs and learn some things.  BOINC is about as decoupled as a person can get.  I have OpenMPI installed, but I haven't even run test programs for that.  I am closely following the Mesa-3D/clover type stuff, hoping that there is an OpenSource path to good numerics (probably using OpenCL).

I have time to upgrade hardware and learn stuff in the winter.  In summer, I often don't.  In the future, I may have livestock on the farm, which can cut into programming and learning at any time of the year.  But, if I do have livestock, it won't be a lot.

But, I became sort of infamous at my undergraduate university because of F77.  The Petroleum Engineering students were not very good at computers, but they used the computers the most in my department.  And they had a particular assignment every year, which required them to write a fixed format FORTRAN data file (I think they were calling some IMSL library).  I had no time for interruptions in my 4th year.  The first Petroleum student who came to bug me when this assignment came up, I sat him down and I wrote a 20 questions program in F77 to write this dumb fixed format data file.  He didn't know what all the fields were about, but I did the best I could.  I permitted it world, left no documentation in it and left for another university.  Three years later, I came back to visit, and noticed someone doing maintenance on my old program.  I learned that program was the biggest time saver the department had ever seen.  :-)