Opened 12 years ago
Closed 10 years ago
#2040 closed enhancement (fixed)
Simcode scales badly for large linear systems
Reported by: | cschubert | Owned by: | wbraun |
---|---|---|---|
Priority: | normal | Milestone: | 1.9.1 |
Component: | Backend | Version: | trunk |
Keywords: | Cc: |
Description
Hi,
the time taken in Simcode scales at least quadratic with the size of the linear systems inside.
Maybe, someone can take a look at that.
I used my Pendulum_N model with Tearing deactivated, which you find attached.
Here are some numbers:
bodies time taken/s
10 1.18552
20 3.3209
30 6.8185
40 11.5792
50 16.9876
60 23.6133
70 31.6687
80 42.3227
90 52.981
At 90 bodies Simcode takes up more than 50% of the total translation time.
Attachments (1)
Change History (4)
Changed 12 years ago by cschubert
comment:1 Changed 11 years ago by lochel
- Owner changed from jfrenkel to wbraun
- Status changed from new to assigned
comment:2 Changed 11 years ago by sjoelund.se
- Milestone changed from 1.9.0 to 1.9.1
comment:3 Changed 10 years ago by wbraun
- Resolution set to fixed
- Status changed from assigned to closed
Seems to be fixed in meantime, since for N=80:
omc NPendulum.mos +d=backenddaeinfo true "" true "" record SimulationResult resultFile = "NPendulum_res.mat", simulationOptions = "startTime = 0.0, stopTime = 1.0, numberOfIntervals = 500, tolerance = 1e-06, method = 'dassl', fileNamePrefix = 'NPendulum', options = '', outputFormat = 'mat', variableFilter = '.*', cflags = '', simflags = ''", messages = "", timeFrontend = 13.248213017, timeBackend = 11.742481002, timeSimCode = 4.24461807, timeTemplates = 7.091320538, timeCompile = 3.974631141, timeSimulation = 6.239920618, timeTotal = 46.541280557 end SimulationResult;
Postponed until 1.9.1