Use GpuAllocEmpty for Gemm16 output.

theano/sandbox/gpuarray/nerv.py

@@ -9,7 +9,7 @@ from theano.scalar import as_scalar, constant
 
 from . import opt
 from .basic_ops import (as_gpuarray_variable, gpu_alloc, gpu_from_host,
-                        host_from_gpu)
+                        host_from_gpu, GpuAllocEmpty)
 from .opt_util import alpha_merge, output_merge
 from .pycuda_helper import ensure_pycuda_context
 
@@ -100,7 +100,7 @@ def local_dot_to_gemm16(node):
         fgraph = node.inputs[0].fgraph
         A = gpu_from_host(node.inputs[0])
         B = gpu_from_host(node.inputs[1])
-        C = gpu_alloc(numpy.asarray(0, dtype='float16'),
+        C = GpuAllocEmpty(dtype='float16')(
             shape_i(A, 0, fgraph), shape_i(B, 1, fgraph))
         return [host_from_gpu(Gemm16()(C, 1.0, A, B, 0.0))]
 
@@ -121,7 +121,13 @@ def local_gemm16_output_merge(node, *inputs):
 def local_gemm16_inplace(node):
     if type(node.op) != Gemm16 or node.op.inplace:
         return
-    return [Gemm16(relu=node.op.relu, inplace=True)(*node.inputs)]
+    inputs = list(node.inputs)
+    C = inputs[0]
+    if (C.owner and
+            isinstance(C.owner.op, GpuAllocEmpty) and
+            len(C.clients) > 1):
+        inputs[0] = C.owner.op(*C.owner.inputs)
+    return [Gemm16(relu=node.op.relu, inplace=True)(*inputs)]
 
 optdb.register('local_gemm16_inplace',
                tensor.opt.in2out(local_gemm16_inplace,