Baunsgaard commented on PR #1869:
URL: https://github.com/apache/systemds/pull/1869#issuecomment-1660196477
On So010, with a 1000 x 1000 matrix and 48 threads in use:
we get 40 GB input processed with MatrixVector multiplication with a peak
MemoryBandwith of 200 GB
```
WriteTest Repetitions: 1000 ConstMatrix ( Rows:1000, Cols:1000,
Spar:1.0, Unique: 32)
Sum, 0.401+- 0.045 ms, 19974765027+-
1877895561 Byte/s, 399488+- 37557 Byte/s
MV mult, 0.182+- 0.036 ms, 43905190543+-
8386160115 Byte/s, 44738539+- 8545335 Byte/s
Update&Apply Scheme, 3.529+- 0.186 ms, 2267055252+-
121334592 Byte/s, 401301962+- 21477999 Byte/s
Update&Apply Scheme Fused, 1.627+- 0.056 ms, 4916655958+-
173706513 Byte/s, 870320068+- 30748595 Byte/s
Apply Scheme, 1.836+- 0.052 ms, 4358519905+-
125944181 Byte/s, 771521818+- 22293964 Byte/s
Update&Apply from Empty, 1.467+- 0.086 ms, 5451830796+-
298396674 Byte/s, 27356767+- 1497326 Byte/s
Normal Compression, 19.648+- 2.180 ms, 407180261+-
47363150 Byte/s, 72076866+- 8383971 Byte/s
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
