TensorRT-LLMs/cpp/tensorrt_llm/kernels/lruKernel.h
Yihan Wang 9df4dad3b6
[None][fix] Introduce inline namespace to avoid symbol collision (#9541)
Signed-off-by: Yihan Wang <yihwang@nvidia.com>
2025-12-12 23:32:15 +08:00

62 lines
1.8 KiB
C++

/*
* Copyright (c) 2020-2024, NVIDIA CORPORATION. All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#pragma once
#include "tensorrt_llm/common/assert.h"
#include "tensorrt_llm/common/config.h"
TRTLLM_NAMESPACE_BEGIN
namespace kernels
{
struct lruParams
{
int batch, width;
int max_seqlen; // only valid for padded input.
int block_size; // used for the cases that enable fused gate
bool remove_padding;
// Common data pointers.
void* __restrict__ A_ptr;
void* __restrict__ x_ptr;
void* __restrict__ y_ptr;
void* __restrict__ y_bias_ptr;
void* __restrict__ gate_ptr;
void* __restrict__ gate_bias_ptr;
void* __restrict__ gate_x_ptr;
void* __restrict__ gate_x_bias_ptr;
void* __restrict__ gate_a_ptr;
void* __restrict__ gate_a_bias_ptr;
void* __restrict__ state_ptr;
void* __restrict__ out_ptr;
int const* __restrict__ last_token_ids_ptr;
int const* __restrict__ slot_mapping_ptr;
};
////////////////////////////////////////////////////////////////////////////////////////////////////
template <typename T>
void invokeRGLRU(lruParams& params, cudaStream_t stream);
template <typename T>
void invokeRGLRUUpdate(lruParams& params, cudaStream_t stream);
} // namespace kernels
TRTLLM_NAMESPACE_END