HomeSort by relevance Sort by last modified time
    Searched refs:GDR (Results 1 - 2 of 2) sorted by null

  /external/tensorflow/tensorflow/contrib/verbs/
patch_notes_verbs_with_0_copies.md 5 Following HKUST research on the use of GPU direct, and their [GDR implementation](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/gdr/README.md), we wish to adopt the 0 copies approach and apply it to the current verbs implementation, while keeping the current implementation advantages, such as configurability and the use of RDMA for control messages.
9 Compared with the current GRPC, verbs and GDR implementation, the result implementation gave the best performance for every model, with any number of nodes. For VGG16 on 8 nodes with 4 P100 GPUs each, the prototype beat the second place by over 15%.
README.md 30 3. Following HKUST research on the use of GPU direct, and their [GDR implementation](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/gdr/README.md), there is a smart way to benefit from the TensorFlow allocation theme which is mostly pool based, i.e allocators pre-allocate a large memory block, and allocate the tensors from there. By attaching a custom Visitor to relevant allocators, we can do a single registration of the entire memory block, which zeros the registration overhead. Once the block is registered, each new tensor allocated will be at a registered address, which will allow us to do direct RDMA writes to it.
55 The tensor transfer process is initiated when the receiver requests a tensor. In code it is done by calling **Rendezvous::Recv()** or **Rendezvous::RecvAsync()**. The TensorFlow base implementation handles the case where the requested tensor is located on the same node. The more interesting case where the requested tensor is located on a remote node (receiver != sender) is to be handled in a derivation of the pure virtual **BaseRemoteRendezvous::RecvFromRemoteAsync()**. TensorFlow provides a default GRPC based implementation which comes in the vanilla version but suffers in scalability when running large models. Our RDMA based implementation presumes to be more scalable. HKUST's contrib GDR implementation is more scalable than GRPC, and less scalable than ours, only because we did our evolution based on it.

Completed in 60 milliseconds