Public Member Functions | |
| uint32_t | setup (Runtime::Execution::PipelineExecutionContext &pipelineExecutionContext) override |
| Must be called only once per executable pipeline and initializes the pipeline execution context. e.g, creates the individual operator states -> window handler. More... | |
| ExecutionResult | execute (Runtime::TupleBuffer &buffer, Runtime::Execution::PipelineExecutionContext &ctx, Runtime::WorkerContext &wc) override |
| Is called once per input buffer and performs the computation of each operator. It can access the state in the PipelineExecutionContext and uns the WorkerContext to identify the current worker thread. More... | |
| uint32_t | stop (Runtime::Execution::PipelineExecutionContext &pipelineExecutionContext) override |
| Must be called exactly once per executable pipeline to remove operator state. More... | |
 Public Member Functions inherited from NES::Runtime::Execution::ExecutablePipelineStage | |
| virtual | ~ExecutablePipelineStage ()=default |
| ExecutablePipelineStage (PipelineStageArity arity=PipelineStageArity::Unary) | |
| PipelineStageArity | getArity () const |
| Returns the arity of this pipeline. More... | |
| virtual uint32_t | start (PipelineExecutionContext &pipelineExecutionContext) |
| Must be called only once per executable pipeline and starts the executable pipeline. e.g. starts the threads for the window handler. More... | |
| virtual uint32_t | open (PipelineExecutionContext &pipelineExecutionContext, WorkerContext &workerContext) |
| Must be called exactly once per worker thread and initializes worker local state. For instance a worker local aggregation state. More... | |
| virtual uint32_t | close (PipelineExecutionContext &pipelineExecutionContext, WorkerContext &workerContext) |
| Must be called exactly once per worker thread to remove worker local state. More... | |
| virtual std::string | getCodeAsString () |
| return the code of the pipeline More... | |
Public Attributes | |
| CUDAKernelWrapper< InputRecord, OutputRecord > | cudaKernelWrapper |
Member Function Documentation
◆ execute()
|
inlineoverridevirtual |
Is called once per input buffer and performs the computation of each operator. It can access the state in the PipelineExecutionContext and uns the WorkerContext to identify the current worker thread.
- Parameters
-
inputTupleBuffer pipelineExecutionContext workerContext
- Returns
- 0 if an error occurred.
Implements NES::Runtime::Execution::ExecutablePipelineStage.
References NES::Runtime::WorkerContext::allocateTupleBuffer(), NES::Runtime::Execution::PipelineExecutionContext::emitBuffer(), NES::Runtime::TupleBuffer::getBuffer(), NES::Runtime::TupleBuffer::getNumberOfTuples(), and NES::Ok.
◆ setup()
|
inlineoverridevirtual |
Must be called only once per executable pipeline and initializes the pipeline execution context. e.g, creates the individual operator states -> window handler.
- Parameters
-
pipelineExecutionContext
- Returns
- 0 if no error occurred.
Reimplemented from NES::Runtime::Execution::ExecutablePipelineStage.
References NUMBER_OF_TUPLE.
◆ stop()
|
inlineoverridevirtual |
Must be called exactly once per executable pipeline to remove operator state.
- Parameters
-
pipelineExecutionContext
- Returns
- 0 if no error occurred.
Reimplemented from NES::Runtime::Execution::ExecutablePipelineStage.
Member Data Documentation
◆ cudaKernelWrapper
| CUDAKernelWrapper<InputRecord, OutputRecord> MultifieldGPUPipelineStage::cudaKernelWrapper |
The documentation for this class was generated from the following file:
- /nebulastream/nes-coordinator/tests/UnitTests/Query/GPUPipelineQueryExecutionTest.cpp
