Correctly tag for gpu scan inplace opt for GPU.

theano/sandbox/cuda/__init__.py

@@ -493,6 +493,8 @@ def use(device,
                        'fast_run')
         optdb.add_tags('gpu_after_fusion',
                        'fast_run')
+        optdb.add_tags('gpu_scanOp_make_inplace',
+                       'fast_run')
 
     if force:
         try:

theano/sandbox/cuda/opt.py

@@ -2535,12 +2535,13 @@ def local_gpu_allocempty(node):
 def typeInfer(node):
     return typeConstructor
 
+# Do not register in fast_run or fast_compile.
+# It will be added to fast_run if the GPU is enabled.
 optdb.register('gpu_scanOp_make_inplace',
                scan_opt.ScanInplaceOptimizer(typeInfer=typeInfer,
                                              gpu_flag=True),
                75,
                'gpu',
-               'fast_run',
                'inplace',
                'scan')
 

theano/sandbox/gpuarray/__init__.py

@@ -78,6 +78,7 @@ if pygpu:
             import theano.compile
             theano.compile.shared_constructor(gpuarray_shared_constructor)
             optdb.add_tags('gpuarray_opt', 'fast_run', 'fast_compile')
+            optdb.add_tags('gpua_scanOp_make_inplace', 'fast_run')
         elif (config.init_gpu_device.startswith('cuda') or
               config.init_gpu_device.startswith('opencl')):
             if config.device != 'cpu':
@@ -91,6 +92,7 @@ if pygpu:
             import theano.compile
             theano.compile.shared_constructor(gpuarray_shared_constructor)
             optdb.add_tags('gpuarray_opt', 'fast_run', 'fast_compile')
+            optdb.add_tags('gpua_scanOp_make_inplace', 'fast_run')
 
         from .basic_ops import (GpuAlloc, GpuAllocEmpty, GpuContiguous, GpuEye,
                                 GpuFromHost, GpuJoin, GpuReshape, GpuSplit,

theano/sandbox/gpuarray/opt.py

@@ -977,11 +977,12 @@ def _scan_type_infer(node):
                             context_name=context_name)
     return typebuild
 
+# Do not register in fast_run or fast_compile.
+# It will be added to fast_run if the GPU is enabled.
 optdb.register('gpua_scanOp_make_inplace',
                scan_opt.ScanInplaceOptimizer(typeInfer=_scan_type_infer,
                                              gpua_flag=True),
                75,
                'gpuarray',
-               'fast_run',
                'inplace',
                'scan')