add conv3d gemm forward alternative

theano/gpuarray/opt.py

@@ -1739,6 +1739,45 @@ def local_abstractconv3d_gemm(node):
     return [rval]
 
 
+@local_optimizer([AbstractConv3d])
+def local_abstractconv3d_alt(node):
+    if not isinstance(node.op, AbstractConv3d):
+        return None
+    img, kern = node.inputs
+    if (not isinstance(img.type, GpuArrayType) or
+            not isinstance(kern.type, GpuArrayType)):
+        return None
+    ctx = infer_context_name(img, kern)
+
+    border_mode = node.op.border_mode
+    subsample = node.op.subsample
+    filter_dilation = node.op.filter_dilation
+
+    if ((border_mode == 'full') and (subsample == (1, 1, 1))):
+        if not node.op.filter_flip:
+            kern = kern[:, :, ::-1, ::-1, ::-1]
+        kern = kern.dimshuffle(1, 0, 2, 3, 4)
+        rval = GpuCorr3dMM_gradInputs('valid',
+                                      subsample,
+                                      filter_dilation)(
+            gpu_contiguous(kern), gpu_contiguous(img))
+
+    elif(subsample == (1, 1, 1) and filter_dilation == (1, 1, 1) and
+         border_mode == 'valid'):
+        if node.op.filter_flip:
+            kern = kern[:, :, ::-1, ::-1, ::-1]
+        rval = GpuCorr3dMM_gradWeights(border_mode,
+                                       subsample,
+                                       filter_dilation)(
+            gpu_contiguous(img.dimshuffle(1, 0, 2, 3, 4)),
+            gpu_contiguous(kern.dimshuffle(1, 0, 2, 3, 4)))
+        rval = as_gpuarray_variable(rval.dimshuffle(1, 0, 2, 3, 4),
+                                    context_name=ctx)
+    else:
+        return None
+    return [rval]
+
+
 @local_optimizer([AbstractConv3d])
 def local_abstractconv3d2d(node):
     if not isinstance(node.op, AbstractConv3d):
@@ -2608,6 +2647,7 @@ conv_metaopt.register([local_abstractconv_gemm_gradweights_alt])
 conv_metaopt.register([local_abstractconv_gradinputs_gemm_alt])
 conv_metaopt.register([local_abstractconv_cudnn_alternative])
 conv_metaopt.register([local_abstractconv3d2d])
+conv_metaopt.register([local_abstractconv3d_alt])
 abstractconv_groupopt.register('conv_metaopt', conv_metaopt, 'conv_meta', position=0)
 
 # Register cuDNN batch normalization implementation