[gpu] add flops and microarch to gpu information
1 unresolved thread
1 unresolved thread
Merge request reports
Activity
added 7 commits
-
2a3c46d2...2d31f7c7 - 2 commits from branch
master
- c7bd28e6 - [gpu] adding FLOPS and microarch for gpus
- 12f6c89b - [gpu] rename method to get numbers of cores
- 24b01830 - [gpu] add performance and microarch in gpu information
- 7b830249 - [gpu] updated json file with gpu performance and microarch
- 4d165440 - [gpu] add exception for chifflot and grafitti
Toggle commit list-
2a3c46d2...2d31f7c7 - 2 commits from branch
added 8 commits
-
4d165440...a77807cf - 3 commits from branch
master
- d0465b2a - [gpu] adding FLOPS and microarch for gpus
- 3e5e63aa - [gpu] rename method to get numbers of cores
- 9e357de5 - [gpu] add performance and microarch in gpu information
- 89c09ea3 - [gpu] updated json file with gpu performance and microarch
- 16f662ec - [gpu] add exception for chifflot and grafitti
Toggle commit list-
4d165440...a77807cf - 3 commits from branch
- Resolved by Lucas Nussbaum
assigned to @lnussbau
added 1 commit
- 1219b157 - [gpu] replace microarch by microarthitecture
added 9 commits
-
214f17d8...2d17a26b - 2 commits from branch
master
- a53d8400 - [gpu] adding FLOPS and microarch for gpus
- a97219e0 - [gpu] rename method to get numbers of cores
- fd64f5d4 - [gpu] add performance and microarch in gpu information
- 6df4290c - [gpu] updated json file with gpu performance and microarch
- eccfd8ce - [gpu] add exception for chifflot and grafitti
- a8c2f9b1 - [gpu] replace microarch by microarchitecture
- 06af1f43 - [gpu] regenerate json files with microarchitecture
Toggle commit list-
214f17d8...2d17a26b - 2 commits from branch
42 72 'alias' => 'v100-sxm2-32', 73 'microarchitecture' => 'Volta', 74 'performance' => { 75 'fp-16' => 28260000000000, 76 'fp-32' => 14130000000000, 77 'fp-64' => 7066000000000, 78 }, 43 79 }, 44 80 'Tesla M2075' => { 45 81 'cores' => 448, 46 82 'compute_capability' => '2.0', 47 83 'short_name' => 'Tesla M2075', 48 84 'alias' => 'm2075', 85 'microarchitecture' => 'Maxwell', 86 'performance' => { 87 'fp-16' => 0, En effet, pas de données constructeurs, je vois pas pourquoi on pourrait pas lancer le code en half par contre pour l'extrapolation c'est pas possible (cf : A40 ou GTX 1080 Ti).
Edited by POUILLOUX LaurentOk, du coup en gros la question c'est si on préfère mettre une valeur théorique extrapolée possiblement erronée/imprécise, ou si on garde explicitement 0 pour dire qu'on a pas la donnée constructeur ? @lnussbau tu as un avis là dessus ?
mentioned in commit fdd63d06
Please register or sign in to reply