Searched refs:ParallelExecute (Results 1 – 4 of 4) sorted by relevance
/external/tensorflow/tensorflow/compiler/mlir/tensorflow/ir/ |
D | tf_device_ops.td | 133 ParallelExecute op concurrently executes variadic number of regions. Regions 136 ParallelExecute ops can be used instead. As so, regions within 137 ParallelExecute op must not have control/data dependencies. 139 While explicit dependencies between regions are disallowed, ParallelExecute 141 communication via send/recvs). In this case, users of ParallelExecute op 143 correctness. Regions in ParallelExecute may include Resource ops. 146 the users of the ParallelExecute op must provide mechanism (via send/recvs 149 ops before/after ParallelExecute ops are guaranteed. That is, execution of 150 regions inside ParallelExecute op is blocked until all inputs to all regions 151 are materialized and ops following ParallelExecute op are blocked until all
|
/external/autotest/utils/ |
D | parallel.py | 11 class ParallelExecute(object): class
|
D | unittest_suite.py | 245 pe = parallel.ParallelExecute(functions, **dargs)
|
/external/tensorflow/tensorflow/core/kernels/ |
D | scatter_functor.h | 140 Index ParallelExecute(OpKernelContext* c, const Device& d, 228 return ParallelExecute(c, d, params, updates, indices);
|