def run(self): dep_signal = drvapi.hsa_signal_t(dependent_signal._id) roc.hsa_amd_memory_async_copy( host_dst.device_pointer.value, cpu_ctx._agent._id, local_memory.device_pointer.value, gpu_ctx._agent._id, kSize, 1, ctypes.byref(dep_signal), completion_signal) h2l_start.release() # signal h2l to start
def run(self): h2l_start.acquire() # to wait until l2h thread has started roc.hsa_amd_memory_async_copy(local_memory.device_pointer.value, gpu_ctx._agent._id, host_src.device_pointer.value, cpu_ctx._agent._id, kSize, 0, None, dependent_signal)
def run(self): dep_signal = drvapi.hsa_signal_t(dependent_signal._id) roc.hsa_amd_memory_async_copy(host_dst.device_pointer.value, cpu_ctx._agent._id, local_memory.device_pointer.value, gpu_ctx._agent._id, kSize, 1, ctypes.byref(dep_signal), completion_signal) h2l_start.release() # signal h2l to start
def run(self): h2l_start.acquire() # to wait until l2h thread has started roc.hsa_amd_memory_async_copy( local_memory.device_pointer.value, gpu_ctx._agent._id, host_src.device_pointer.value, cpu_ctx._agent._id, kSize, 0, None, dependent_signal)