| 1 |
A very short log of what has been done during the last days. I am |
| 2 |
working on the PBLAS and ScaLAPACK benchmarks, which is a very |
| 3 |
challenging topic, because it is very difficult to debug such |
| 4 |
applications. |
| 5 |
|
| 6 |
* I changed some parts of the BTL framework, adapting it to the |
| 7 |
distributed memory benchmarks. This has required writing two new |
| 8 |
perfanalyzers -- one for the root process, one for the other (node) |
| 9 |
processes. The nodes do not perform any measurement, while the root |
| 10 |
process broadcasts the needed informations, measures the time and |
| 11 |
manages the output (both std{out,err} and resulting file). |
| 12 |
|
| 13 |
* I added a BLACS library that provides an useful interface which |
| 14 |
scatters and gathers matrices and vectors. I also added a PBLAS |
| 15 |
library that inherits the BLACS one and will support the most common |
| 16 |
operations (at the moment just the parallel matrix-vector |
| 17 |
multiplication). |
| 18 |
|
| 19 |
* I added an action for the parallel matrix-vector multiplication |
| 20 |
which makes use of the two described interfaces. |
| 21 |
|
| 22 |
The matrix-vector multiplication is a case study for now. If |
| 23 |
everything goes fine (and it seems so, now), then more actions will be |
| 24 |
provided, for both PBLAS and ScaLAPACL, which share the same concepts. |
| 25 |
I plan to have tomorrow a working (but incomplete) Python module for |
| 26 |
PBLAS, too. |
| 27 |
|
| 28 |
Milestones for the next week: |
| 29 |
* Having working PBLAS and ScaLAPACK modules |
| 30 |
* Do some benchmarks using these modules and publish the results |
| 31 |
* Start the implementation of the advanced FFTW benchmarks, as |
| 32 |
previously described |
| 33 |
|
| 34 |
Best regards |
| 35 |
Andrea Arteaga |
Archives