|
|
@@ -101,6 +101,12 @@ to use a version that takes the a stream parameter.
|
|
|
|
|
|
Unfortunately, some CUDA libraries do not have stream variants of
|
|
|
kernels. This will seriously lower the potential for overlapping.
|
|
|
+If some CUDA calls are made without specifying this local stream,
|
|
|
+synchronization needs to be explicited with cudaThreadSynchronize() around these
|
|
|
+calls, to make sure that they get properly synchronized with the calls using
|
|
|
+the local stream. Notably, \c cudaMemcpy() and \c cudaMemset() are actually
|
|
|
+asynchronous and need such explicit synchronization! Use cudaMemcpyAsync() and
|
|
|
+cudaMemsetAsync() instead.
|
|
|
|
|
|
Calling starpu_cublas_init() makes StarPU already do appropriate calls for the
|
|
|
CUBLAS library. Some libraries like Magma may however change the current stream of CUBLAS v1,
|
