This mostly asks, 'are we comfortable with the multithreading implimentation as it is now?'
The specific thing to change in the trunk is at core/globals.h:71.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
This mostly asks, 'are we comfortable with the multithreading implimentation as it is now?'
The specific thing to change in the trunk is at core/globals.h:71.
The default threadcount should not be made 0 (use max) until the following conditions are satisfied.
1. The validation code must produce the same outcomes for single and multithreaded runs.
2. The multithreaded runs must generally be faster, at least for larger models.
3. Multithreading must be work consistently on all supported platforms.
At the present time it satisfies none of these of conditions in trunk.
This cannot be done until all multithreading is done using MTI and all modules properly lock data access.