def on_node_end(self, sdfg: SDFG, state: SDFGState, node: nodes.AccessNode,
outer_stream: CodeIOStream, inner_stream: CodeIOStream,
global_stream: CodeIOStream):
from dace.codegen.dispatcher import DefinedType # Avoid import loop
if is_devicelevel_gpu(sdfg, state, node) or is_devicelevel_fpga(
sdfg, state, node):
# Only run on host code
return
desc = node.desc(sdfg)
# Obtain a pointer for arrays and scalars
ptrname = cpp.ptr(node.data, desc, sdfg, self.codegen)
defined_type, _ = self.codegen.dispatcher.defined_vars.get(ptrname)
if defined_type == DefinedType.Scalar:
ptrname = '&' + ptrname
# Create UUID
state_id = sdfg.node_id(state)
node_id = state.node_id(node)
uuid = f'{sdfg.sdfg_id}_{state_id}_{node_id}'
# Get optional pre/postamble for instrumenting device data
preamble, postamble = '', ''
if desc.storage == dtypes.StorageType.GPU_Global:
self._setup_gpu_runtime(sdfg, global_stream)
preamble, postamble, ptrname = self._generate_copy_to_host(
node, desc, ptrname)
# Encode runtime shape and strides
shape = ', '.join(cpp.sym2cpp(s) for s in desc.shape)
strides = ', '.join(cpp.sym2cpp(s) for s in desc.strides)
# Write code
inner_stream.write(preamble, sdfg, state_id, node_id)
inner_stream.write(
f'__state->serializer->save({ptrname}, {cpp.sym2cpp(desc.total_size - desc.start_offset)}, '
f'"{node.data}", "{uuid}", {shape}, {strides});\n', sdfg, state_id,
node_id)
inner_stream.write(postamble, sdfg, state_id, node_id)