ComputeContext (object)¶
Members¶
| ComputeContext_perThreadData[] | perThreadData | |
| Float32 | calibrationDummyResult | |
| Float32 | parallelExecuteOverhead | |
| SInt32 | executingKLDFGBindings | |
| String[40] | loggingIndentPerDepth | |
| AsyncTaskQueue | asyncTaskQueue | Define it here so it’s in a singleton |
Methods¶
| ComputeContext ( in ComputeContext other ) | |
| ComputeContext () | |
| ComputeContext | clone ? () |
| Boolean | hasExecutingKLDFGBinding ? () |
| Float32 | measureParallelEvalOverhead ! ( in Boolean printResult ) |
Methods in detail¶
ComputeContext ( in ComputeContext other )
copy constructor
ComputeContext ComputeContext.clone? ()
clone method
Boolean ComputeContext.hasExecutingKLDFGBinding? ()
Patch to detect onDirty norifs coming from runtime KL->DFGBinding changes
Float32 ComputeContext.measureParallelEvalOverhead! ( in Boolean printResult )
Empiritically measures the cost of multithreaded eval and prints the number in seconds. However, we found that this is extremely varying from run to run, probably as the OS is the boss when it comes to waking up threads.
