eth-cscs / DLA-Future

DLA-Future
https://eth-cscs.github.io/DLA-Future/master/
BSD 3-Clause "New" or "Revised" License
64 stars 14 forks source link

Fix Cholesky priorities #999

Closed rasolca closed 1 year ago

rasolca commented 1 year ago

Fix priorities messed up in #371.

E.g. https://github.com/eth-cscs/DLA-Future/commit/89ca6a0297d59021354676a2c5a87cfa34ea8c25#diff-eb1edb073915301db270aa8e4583adde9d3d0eacf0b9c3cfed3b00ffc2e8ce87L123

rasolca commented 1 year ago

cscs-ci run

rasolca commented 1 year ago

4 nodes on eiger

master:

[0]
[0] 1.07523s 2662.96GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[1]
[1] 1.02061s 2805.5GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[2]
[2] 1.05352s 2717.85GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[3]
[3] 1.01913s 2809.56GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[4]
[4] 1.05299s 2719.22GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC

[0]
[0] 3.7225s 6153.53GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[1]
[1] 3.97347s 5764.86GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[2]
[2] 3.82834s 5983.4GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[3]
[3] 3.83611s 5971.28GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[4]
[4] 3.81851s 5998.8GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC

new:

[0]
[0] 0.962872s 2973.72GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[1]
[1] 0.984092s 2909.6GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[2]
[2] 0.98801s 2898.06GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[3]
[3] 0.969163s 2954.42GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC
[4]
[4] 0.968429s 2956.66GFlop/s dL (20480, 20480) (512, 512) (8, 4) 16 MC

[0]
[0] 3.71782s 6161.27GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[1]
[1] 3.67812s 6227.76GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[2]
[2] 3.6927s 6203.18GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[3]
[3] 3.76707s 6080.73GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC
[4]
[4] 3.66226s 6254.75GFlop/s dL (40960, 40960) (512, 512) (8, 4) 16 MC