User Tools
opt:graphics_cards
Differences
This shows you the differences between two versions of the page.
| Both sides previous revision Previous revision | |||
|
opt:graphics_cards [2018/08/31 13:02] pklapetek |
opt:graphics_cards [2018/09/04 17:24] (current) pklapetek |
||
|---|---|---|---|
| Line 71: | Line 71: | ||
| will run the calculation on the Tesla K40m card on our system. | will run the calculation on the Tesla K40m card on our system. | ||
| - | Even if we try to do our best to have one to one correspondence for everything on CPU and GPU, it can happen that some of the recent algorithms are not yet implemented on GPU. The reference should always be the CPU implementation. We therefore strongly recommend, namely for setting up the new task, to test the task first on CPU, crosscheck it with GPU and then run the various repetitive calculations on GPU to speedup the solution. | + | Even if we try to do our best to have one to one correspondence for everything on CPU and GPU, it can happen that some of the recent algorithms are not yet implemented on GPU. The reference should always be the CPU implementation. We therefore strongly recommend, namely for setting up the new task, to test the task first on CPU, crosscheck it with GPU and then run the various repetitive calculations on GPU to speedup the solution. Some known issues with GPU are listed also later on this page. |
| As an example you can test the following parameter files, comparing performance of GSvit on different number of CPU cores and on a single GPU on the above mentioned system. The calculation took 8 minutes 37 seconds on a single core, 1 minute 43 seconds on eight cores, and 54 seconds on the first GPU. | As an example you can test the following parameter files, comparing performance of GSvit on different number of CPU cores and on a single GPU on the above mentioned system. The calculation took 8 minutes 37 seconds on a single core, 1 minute 43 seconds on eight cores, and 54 seconds on the first GPU. | ||
| Line 84: | Line 84: | ||
| Generally, the bigger is the calculation size and problem complexity, the better GPU works comparing to CPU as the effect of administration (data transfers) is smaller. Also, too many unnecessary transfers, e.g. for image outputs at every few steps, degrade the GPU performance. | Generally, the bigger is the calculation size and problem complexity, the better GPU works comparing to CPU as the effect of administration (data transfers) is smaller. Also, too many unnecessary transfers, e.g. for image outputs at every few steps, degrade the GPU performance. | ||
| + | |||
| + | === Known issues and missing algorithms === | ||
| + | |||
| + | At present the following algorithms are known to be not yet implemented on GPU or known to have some other troubles and not recommended for use in GPU: | ||
| + | * focused source (key TSFF) | ||
| + | * layered focused source (key LTSFF) | ||
| + | |||
| + | Note that this list might be not complete and we always recommend a cross-check with CPU results when starting a new type of simulation. | ||
opt/graphics_cards.1535713346.txt.gz ยท Last modified: 2018/08/31 13:02 by pklapetek
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Noncommercial 4.0 International
