Merge pull request #5469 from abergeron/gpu_advsub1

theano/gpuarray/subtensor.py

@@ -19,7 +19,7 @@ except ImportError:
 
 from .type import GpuArrayType, gpu_context_type
 from .basic_ops import (as_gpuarray_variable, HideC, GpuKernelBase, Kernel,
-                        infer_context_name)
+                        infer_context_name, gpu_contiguous)
 
 
 iadd_reg = {}
@@ -413,7 +413,7 @@ class GpuAdvancedSubtensor1(HideC, tensor.AdvancedSubtensor1):
         if ilist__.type.dtype != 'int64':
             ilist__ = tensor.cast(ilist__, 'int64')
 
-        ilist_ = as_gpuarray_variable(ilist__, ctx_name)
+        ilist_ = gpu_contiguous(as_gpuarray_variable(ilist__, ctx_name))
 
         if ilist_.type.dtype != 'int64':
             raise TypeError('index must be int64')

theano/gpuarray/tests/test_subtensor.py

@@ -7,7 +7,7 @@ from theano import tensor
 from theano.compile import DeepCopyOp
 from theano.tensor.tests import test_subtensor
 
-from ..basic_ops import HostFromGpu, GpuFromHost
+from ..basic_ops import HostFromGpu, GpuFromHost, GpuContiguous
 from ..elemwise import GpuDimShuffle
 from ..subtensor import (GpuIncSubtensor, GpuSubtensor,
                          GpuAdvancedSubtensor1,
@@ -37,7 +37,7 @@ class G_subtensor(test_subtensor.T_subtensor):
             # avoid errors with limited devices
             dtype='float32',
             ignore_topo=(HostFromGpu, GpuFromHost,
-                         DeepCopyOp))
+                         DeepCopyOp, GpuContiguous))
         # GPU opt can't run in fast_compile only.
         self.fast_compile = False
         assert self.sub == GpuSubtensor