Multi-GPU

- Parallelize batches among GPUs and tree-reduce the gradients
- The effective batch size scales with the number of devices
- Batch size is multiplied by the number of devices
- Split batches between GPUs, and tree-reduce the gradients
- Detect machine topology (twin-GPU boards, P2P connectivity)
- Track device in syncedmem (thanks @thatguymike)
- Insert a callback in the solver for minimal code change
- Accept list for gpu flag of caffe tool, e.g. '-gpu 0,1' or '-gpu all'.
  Run on default GPU if no ID given.
- Add multi-GPU solver test
- Deterministic architecture for reproducible runs

1 files changed, 3 insertions, 0 deletions

diff --git a/src/caffe/util/blocking_queue.cpp b/src/caffe/util/blocking_queue.cpp
index f7c53f22..d1d1fa86 100644
--- a/src/caffe/util/blocking_queue.cpp
+++ b/src/caffe/util/blocking_queue.cpp
@@ -3,6 +3,7 @@
 
 #include "caffe/data_layers.hpp"
 #include "caffe/data_reader.hpp"
+#include "caffe/parallel.hpp"
 #include "caffe/util/blocking_queue.hpp"
 
 namespace caffe {
@@ -89,5 +90,7 @@ template class BlockingQueue<Batch<float>*>;
 template class BlockingQueue<Batch<double>*>;
 template class BlockingQueue<Datum*>;
 template class BlockingQueue<shared_ptr<DataReader::QueuePair> >;
+template class BlockingQueue<P2PSync<float>*>;
+template class BlockingQueue<P2PSync<double>*>;
 
 }  // namespace caffe