* Copyright (c) 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.
*/
* Copyright (c) 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.
*/
#include <cstddef>
#include <iostream>
#include <stdexcept>
#include <tuple>
#include <type_traits>
#include <utility>
#include <vector>
#include "kernel_operator.h"
#include "utils/log.h"
#include "atvoss.h"
#include "utils/layout/shape.h"
static constexpr int32_t HEIGHT = 1;
static constexpr int32_t WIDTH = 32;
static constexpr int32_t MAX_DIM = 8;
template <typename T1, typename T2, typename T3>
struct RmsNormConfig {
using DtypeV1 = T1;
using DtypeV2 = T2;
using DtypeV3 = T3;
using TileShape = Atvoss::Shape<HEIGHT, WIDTH>;
struct RmsNormCompute {
template <template <typename> class Tensor>
__host_aicore__ constexpr auto Compute() const
{
auto in1 = Atvoss::PlaceHolder<1, Tensor<DtypeV1>, Atvoss::ParamUsage::IN>();
auto in2 = Atvoss::PlaceHolder<2, Tensor<DtypeV2>, Atvoss::ParamUsage::IN>();
auto in3 = Atvoss::PlaceHolder<3, float, Atvoss::ParamUsage::IN>();
auto out = Atvoss::PlaceHolder<4, Tensor<DtypeV3>, Atvoss::ParamUsage::OUT>();
auto temp = Atvoss::PlaceHolderTmpLike<1>(in1);
return (
temp = in1 * in1, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp,
temp = in1 * out, out = ReduceSum<Atvoss::Pattern::AR>(temp), out = Broadcast<Atvoss::Pattern::AB>(out),
temp = Divs<WIDTH>(out), temp = temp * in3, temp = in3 * temp, temp = in3 * temp, out = Sqrt(temp),
temp = in1 / out,
out = in2 * temp
);
}
};
static constexpr Atvoss::Ele::DefaultBlockPolicy<TileShape> blockPolicy{TileShape{}};
static constexpr Atvoss::Ele::DefaultKernelPolicy kernelPolicy{Atvoss::Ele::DefaultSegmentPolicy::UniformSegment};
using ArchTag = Atvoss::Arch::DAV_3510;
using BlockOp = Atvoss::Ele::BlockBuilder<RmsNormCompute, ArchTag, blockPolicy, Atvoss::Ele::DefaultBlockConfig>;
using KernelOp = Atvoss::Ele::KernelBuilder<BlockOp, kernelPolicy>;
using DeviceOp = Atvoss::DeviceAdapter<KernelOp>;
};
int main(int argc, char const* argv[])
{
uint64_t shapeArray[MAX_DIM] = {0};
std::vector<uint64_t> shape{32, 32};
std::copy(shape.begin(), shape.end(), shapeArray);
Atvoss::Tensor<float> t1(nullptr, shapeArray, shape.size());
Atvoss::Tensor<float> t2(nullptr, shapeArray, shape.size());
Atvoss::Tensor<float> t3(nullptr, shapeArray, shape.size());
float a = 1.0f;
auto arguments = Atvoss::ArgumentsBuilder{}.inputOutput(t1, t2, a, t3).build();
using DeviceOp = typename RmsNormConfig<float, float, float>::DeviceOp;
DeviceOp deviceOp;
deviceOp.Run(arguments, nullptr);
return 0;
}
