* Copyright (c) Huawei Technologies Co., Ltd. 2025. All rights reserved.
* 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.
*/
#include <algorithm>
#include <mki/base/operation_base.h>
#include <mki_loader/op_register.h>
#include <mki/utils/assert/assert.h>
#include <mki/utils/const/op_const.h>
#include <mki/utils/checktensor/check_tensor.h>
#include <mki/utils/log/log.h>
#include "atbops/params/params.h"
namespace AtbOps {
const int32_t Q_HEADDIM = 576;
const int32_t SPLIT_SIZE_ONE = 512;
const int32_t SPLIT_SIZE_TWO = 64;
using namespace Mki;
class MlaPreprocessOperation : public OperationBase {
static constexpr int64_t NUM_INPUT_TENSORS = 25;
public:
explicit MlaPreprocessOperation(const std::string &opName) noexcept : OperationBase(opName) {}
Kernel *GetBestKernel(const LaunchParam &launchParam) const override
{
MKI_CHECK(launchParam.GetParam().Type() == typeid(OpParam::MlaPreprocess), "Invalid Op Param.", return nullptr);
auto inDtype = launchParam.GetInTensor(0).desc.dtype;
auto param = AnyCast<OpParam::MlaPreprocess>(launchParam.GetParam());
bool deqOnTheFly = inDtype == TENSOR_DTYPE_FLOAT16 &&
param.quantMode == OpParam::MlaPreprocess::QuantMode::PER_TENSOR_ASYMM_QUANT;
if (deqOnTheFly) {
return GetKernelByName("MLAPreprocessKernel");
}
return GetKernelByName("MLAPreprocessBF16Kernel");
}
int64_t GetInputNum(const Any &specificParam) const override
{
return NUM_INPUT_TENSORS;
}
int64_t GetOutputNum(const Any &specificParam) const override
{
return DIM_4;
}
Status InferShapeImpl(const LaunchParam &launchParam, SVector<Tensor> &outTensors) const override
{
auto &tensorQ = launchParam.GetInTensor(DIM_0);
auto &tensorKEY = launchParam.GetInTensor(16);
auto param = AnyCast<OpParam::MlaPreprocess>(launchParam.GetParam());
if (param.cacheMode == 0) {
outTensors[DIM_0].desc = tensorQ.desc;
outTensors[DIM_0].desc.dims = {param.N, param.headNum, Q_HEADDIM};
outTensors[DIM_1].desc = tensorKEY.desc;
} else {
outTensors[DIM_0].desc = tensorQ.desc;
outTensors[DIM_0].desc.dims = {param.N, param.headNum, SPLIT_SIZE_ONE};
outTensors[DIM_1].desc = tensorKEY.desc;
outTensors[DIM_1].desc.dims[DIM_3] = SPLIT_SIZE_ONE;
outTensors[DIM_2].desc = tensorQ.desc;
outTensors[DIM_2].desc.dims = {param.N, param.headNum, SPLIT_SIZE_TWO};
outTensors[DIM_3].desc = tensorKEY.desc;
outTensors[DIM_3].desc.dims[DIM_3] = SPLIT_SIZE_TWO;
}
return Mki::Status::OkStatus();
}
};
REG_OPERATION(MlaPreprocessOperation);
}