Hello: I am an applied math phd student looking for the relative merits of learning assembly for speedy computation as compared to C or Fortran, which seem to be the standard. Particularly, are there forums for algorithm development in medical imaging, or modeling stochastic PDE's, routines akin to libraries like LAPAC, BLAS, Intel MKL, etc.?

Honestly, I doubt that there are any forums that are dedicated to mere noise functions.

However there ARE forums dedicated to algorithmic advancement, and to specific algorithm families.

However there ARE forums dedicated to algorithmic advancement, and to specific algorithm families.

the relative merits of learning assembly for speedy computation as compared to C or Fortran

Assembly will always have the edge over those languages when it comes to number crunching

__based on the same algorithm being used__. I've looked at enough disassembled code written in C to realize that a lot of time is wasted by those languages transferring data from memory to registers and back to memory for each computation, even if that same data must be handled immediately again.

This may not matter if you are considering computations which would take only a few seconds followed by wait times. However, when computations must be performed continuously, more will be done if written and optimized in assembly.

Honestly, I doubt that there are any forums that are dedicated to mere noise functions.

However there ARE forums dedicated to algorithmic advancement, and to specific algorithm families.

OK, this would be a good start I guess. Suppose I am trying to first work with a vector/matrix library which has been optimized for intel based processors (please note that I am very new to this so I don't even know what is current or even available), and would like to use the vector/matrix library for things such as fast linear algebraic solutions to tridiagonal systems (extremely common in solutions to implicit schemes in PDE's). Are people doing work on this? Numerical stability is key here, and fast implementations would have widespread utility for computational scientists. Any help is good help

Hell yes, people are doing work in this area - in particular, physics simulation springs to mind, something I've worked with a lot lately, I needed to diagonalize inertia tensors stored as 3x3 matrices, and had to learn all about Jacobian rotation and other solutions.... then I implemented it all in asm, but I did not optimize the code as much as I could have.

Anyway, just an example.

Anyway, just an example.