CUDA tutorial/fr: Difference between revisions

Jump to navigation Jump to search
Created page with "= Facteurs de performance de base = == Memory transfer == * PCI-e est extrêmement lent (4-6Go/s) en comparaison à la mémoire hôte et la mémoire de la carte graphique * Mi..."
(Created page with "= Avantages de la mémoire partagée= Jusqu'ici, tous les transferts en mémoire dans le ''kernel'' ont été via la mémoire régulière (globale) du GPU, ce qui est relative...")
(Created page with "= Facteurs de performance de base = == Memory transfer == * PCI-e est extrêmement lent (4-6Go/s) en comparaison à la mémoire hôte et la mémoire de la carte graphique * Mi...")
Line 166: Line 166:
</syntaxhighlight>
</syntaxhighlight>


= Basic performance considerations =
= Facteurs de performance de base =
== Memory transfer ==
== Memory transfer ==
* PCI-e is extremely slow (4-6 GB/s) compared to both host and device memories
* PCI-e est extrêmement lent (4-6Go/s) en comparaison à la mémoire hôte et la mémoire de la carte graphique
* Minimize host-to-device and device-to-host memory copies
* Minimisez les copies de mémoire dans les deux directions.
* Keep data on the device as long as possible
* Gardez les données sur la carte graphique le plus longtemps possible.
* Sometimes it is not effificient to make the host (CPU) do non-optimal jobs; executing it on the GPU may still be faster than copying to CPU, executing, and copying back
* Sometimes it is not effificient to make the host (CPU) do non-optimal jobs; executing it on the GPU may still be faster than copying to CPU, executing, and copying back
* Use memcpy times to analyse the execution times
* Use memcpy times to analyse the execution times
rsnt_translations
56,430

edits

Navigation menu