Faster compilation: use a deque instead of Queue.

Change datastructure to a deque that is faster. Queue is for multi-thread communication with automatic locking.

gpuarray: Do not pass config.nvcc.compiler-bindir to g++

Fix crash in Lop

Fix Theano deprecation warning

chmod +x buildbot_debug

The contributing message is annoying.

Fix an error. GpuAllocEmpty in the new back-end was inheriting from A…

Fix random segfault on exit when the new backend is in use.

split jenkins buildbot debug_mode into separate job

Half 3D convolution for conv3d2d

Fix an error. GpuAllocEmpty in the new back-end was inheriting from Alloc that have a __call__ method. We don't want that. Also refactor the code to lower code duplication.

The problem is that the old backend assumes it is alone in the
universe and even when not in use will interact with the gpu.  Most
notably in this case it will forcibly destroy any primary context on
exit.  The only problem is that we are also using the primary context
and our cleanup handlers run after those of the old backend.

This is a major problem because cublas uses the runtime api which will
create a context whenever it thinks it needs one (like for cudaFree).
However all the allocations are from the old context that the old
backend destroyed.  So when it tries to clean up its resources the
low-level handlers get confused and we end up in a segmentation fault.

As to why it doesn't always happen, I figure it's because sometimes we
get lucky and addresses line up or something.

Anyway, if we stop the old backend from calling cudaThreadExit() the
segfaults go away.  This may cause a very small leakage of resources
for the few seconds we are running until we exit, but I don't think
this will be a problem.

Fix Lop to comply with new requirement that theano.grad's known_grad argument be an OrderedDict.

Postpone Composite name creating

Doc update: pdf figures for pdftex doc and update metadoc buildbot info