* Copyright (c) 2025-2026 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
* \file device_launcher_binding.cpp
* \brief
*/
#include "machine/runtime/launcher/device_launcher_binding.h"
#include "adapter/api/runtime_define.h"
#include "machine/runtime/launcher/device_launcher.h"
#include "machine/runtime/context/device_launcher_context.h"
namespace npu::tile_fwk::dynamic {
int ExportedOperatorDeviceLaunchOnceWithDeviceTensorData(
ExportedOperator* op, const std::vector<DeviceTensorData>& inputList,
const std::vector<DeviceTensorData>& outputList,
DeviceStream aicoreStream, bool streamSynchronize, uint8_t* devCtrlCache, const DeviceLauncherConfig& config)
{
RtStream aicoreStreamValue = reinterpret_cast<RtStream>(aicoreStream);
return DeviceLauncher::DeviceLaunchOnceWithDeviceTensorData(
op->GetFunction(), inputList, outputList, aicoreStreamValue,
streamSynchronize, op, reinterpret_cast<DevControlFlowCache*>(devCtrlCache), config);
}
int DeviceSynchronize(DeviceStream aicpuStream)
{
RtStream aicpuStreamValue = reinterpret_cast<RtStream>(aicpuStream);
return DeviceLauncher::DeviceSynchronize(aicpuStreamValue, GetContextAiCoreStream());
}
int DeviceRunOnce(Function* function, uint8_t* hostCtrlCache, const DeviceLauncherConfig& config)
{
return DeviceLauncher::DeviceRunOnce(function, reinterpret_cast<DevControlFlowCache*>(hostCtrlCache), config);
}
int HasInplaceArgs(Function* function) { return DeviceLauncher::HasInplaceArgs(function); }
void DeviceLauncherInit() { DeviceLauncherContext::Get().Initialize(); }
void DeviceLauncherFini() { DeviceLauncherContext::Get().Finalize(); }
void CopyDevToHost(const DeviceTensorData& devTensor, DeviceTensorData& hostTensor)
{
DeviceMemoryUtils().CopyFromDev(
(uint8_t*)hostTensor.GetAddr(), (uint8_t*)devTensor.GetAddr(), devTensor.GetDataSize());
}
void CopyHostToDev(const DeviceTensorData& devTensor, DeviceTensorData& hostTensor)
{
DeviceMemoryUtils().CopyToDev(
(uint8_t*)devTensor.GetAddr(), (uint8_t*)hostTensor.GetAddr(), devTensor.GetDataSize());
}
uint8_t* CopyHostToDev(uint8_t* data, uint64_t size)
{
return DeviceMemoryUtils(false).CopyToDev((uint8_t*)data, size, nullptr);
}
void ChangeCaptureModeRelax() { ExchangeCaptureModeRelax(); }
void ChangeCaptureModeGlobal() { ExchangeCaptureModeGlobal(); }
void GetCaptureInfo(RtStream aicoreStream, AclMdlRI& rtModel, bool& isCapture)
{
(void)GetStreamCaptureInfo(aicoreStream, rtModel, isCapture);
}
void* RegisterKernelBinary(const std::vector<uint8_t>& kernelBinary)
{
return RegisterKernelBin(kernelBinary);
}
void UnregisterKernelBinary(void* hdl)
{
int ret = RuntimeDevBinaryUnRegister(hdl);
if (ret != RT_SUCCESS) {
MACHINE_LOGE(RtErr::RT_REGISTER_FAILED, "unregister kernel failed, ret: %d", ret);
}
}
static std::unordered_map<ExportedOperator*, std::shared_ptr<ExportedOperator>> exportedOperatorDict;
ExportedOperator* ExportedOperatorBegin()
{
std::shared_ptr<ExportedOperator> op = std::make_shared<ExportedOperator>();
exportedOperatorDict[op.get()] = op;
return op.get();
}
void ExportedOperatorEnd(ExportedOperator* op) { op->ResetFunction(Program::GetInstance().GetLastFunction()); }
}