barbagroup / inexact-gmres

Paper: inexact GMRES with fast multipole method and low-p relaxation
9 stars 4 forks source link

Why are we not able to run a problem of the same size as the example in Liu's book? #6

Closed labarba closed 9 years ago

labarba commented 9 years ago

The example in Liu's book ran on a laptop with a 2.4 GHz Pentium IV CPU with 1 GB RAM, taking 730 minutes to complete the 49 iterations to converge to a tolerance of $10^{-4}$ using $p=15$, for 40 cells, 7500 panels each (300k total panels).

Why do you say that for 40 cells you need to use the coarser mesh of 2048 panels? I mean, he was running on a Pentium IV, surely we can match that on our machines??

slayton58 commented 9 years ago

Do you still have a pdf copy of the book? I’d like to take a look at what he says about the test

On Jun 2, 2015, at 7:24 AM, Lorena A. Barba notifications@github.com wrote:

The example in Liu's book ran on a laptop with a 2.4 GHz Pentium IV CPU with 1 GB RAM, taking 730 minutes to complete the 49 iterations to converge to a tolerance of $10^{-4}$ using $p=15$, for 40 cells, 7500 panels each (300k total panels).

Why do you say that for 40 cells you need to use the coarser mesh of 2048 panels? I mean, he was running on a Pentium IV, surely we can match that on our machines??

— Reply to this email directly or view it on GitHub https://github.com/barbagroup/inexact-gmres/issues/6.

slayton58 commented 9 years ago

I believe there were issues with the local preconditioner and / or sparse matrix approach for the near-field that caused the memory issues. That’s my guess at least

On Jun 2, 2015, at 8:39 AM, Simon Layton slayton58@gmail.com wrote:

Do you still have a pdf copy of the book? I’d like to take a look at what he says about the test

On Jun 2, 2015, at 7:24 AM, Lorena A. Barba <notifications@github.com mailto:notifications@github.com> wrote:

The example in Liu's book ran on a laptop with a 2.4 GHz Pentium IV CPU with 1 GB RAM, taking 730 minutes to complete the 49 iterations to converge to a tolerance of $10^{-4}$ using $p=15$, for 40 cells, 7500 panels each (300k total panels).

Why do you say that for 40 cells you need to use the coarser mesh of 2048 panels? I mean, he was running on a Pentium IV, surely we can match that on our machines??

— Reply to this email directly or view it on GitHub https://github.com/barbagroup/inexact-gmres/issues/6.

labarba commented 9 years ago

That's not very helpful. I see no option but to avoid making any comparisons with Liu's example.