This comes with tests and docs changes.

This also updates GpuEye to use it so that we have a real-world example.

make gpuarray/test_fft compatible with numpy<1.10

Full 3D convolution for conv3d2d

Make the DLT mlp example learn in float16

Fix the few failing tests with floatX=float32 in gpuarray

Linting: Pep8 indentation, consistent quotes

Addition of DownsampleFactorMaxGradGrad.grad and its test

Move sandbox.cuda test to 2nd jenkins job

Method DownsampleFactorMaxGradGrad.grad and method test_DownsampleFactorMaxGradGrad_grad have been added respectively to pool.py and tests/test_pool.py.

Use shift modulo axis size to handle large shifts in theano.tensor.roll

Fix blas C code having wrong alpha and beta for float16.

Add the theano flags gpuarray.sched

Add the user stack trace printing in nanguardmode

Fix problem in DebugMode error message.

Local gpua careduce

add jenkins test script

Fix handling of errors in the Set case of GpuIncSubtensor.