TensorRT-LLMs/cpp/tensorrt_llm/kernels/decoderMaskedMultiheadAttention/decoderXQAImplCommon.cpp
2025-07-09 09:26:27 +08:00

61 lines
2.5 KiB
C++

/*
* Copyright (c) 2020-2023, 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.
*
* Common utils to be shared between Precompiled and JIT implementation.
*/
#include "tensorrt_llm/kernels/decoderMaskedMultiheadAttention/decoderXQAImplCommon.h"
namespace tensorrt_llm::kernels
{
uint32_t getKernelMTileSize(uint32_t headGrpSize, bool isSpecDec, uint32_t qSeqLen, bool isXqaJit)
{
if (!isSpecDec)
{
return headGrpSize;
}
if (isXqaJit)
{
return 64;
}
uint32_t const gemmM = qSeqLen * headGrpSize;
return gemmM < 16 ? 16 : 32;
}
XQAKernelRuntimeHashKey getRuntimeHashKeyFromXQAParams(XQAParams const& xqaParams, bool isXqaJit)
{
unsigned int head_size = xqaParams.head_size;
unsigned int num_q_heads = xqaParams.num_q_heads;
unsigned int num_kv_heads = xqaParams.num_kv_heads;
TLLM_CHECK_WITH_INFO(num_q_heads % num_kv_heads == 0, "numQHeads should be multiple of numKVHeads.");
unsigned int num_q_heads_over_kv = num_q_heads / num_kv_heads;
unsigned int beam_width = xqaParams.beam_width;
// Use mTileSize = 16 kernels when qSeqLen <= 16.
unsigned int qSeqLen = static_cast<unsigned int>(xqaParams.generation_input_length);
unsigned int mTileSize = qSeqLen <= 16 ? 16 : 32;
// MultiQueryToken kernels can support any num_q_heads_over_kv that is power of 2.
unsigned int kernel_num_q_heads_over_kv = xqaParams.multi_query_tokens ? 0 : num_q_heads_over_kv;
// MultiQueryToken kernels can handle either 16/32 for M direction per CTA.
unsigned int kernel_m_tilesize
= getKernelMTileSize(num_q_heads_over_kv, xqaParams.multi_query_tokens, qSeqLen, isXqaJit);
return {xqaParams.kv_cache_data_type, head_size, beam_width, kernel_num_q_heads_over_kv, kernel_m_tilesize,
xqaParams.paged_kv_cache ? static_cast<unsigned int>(xqaParams.tokens_per_block) : 0, xqaParams.paged_kv_cache,
xqaParams.multi_query_tokens, xqaParams.is_fp8_output};
}
} // namespace tensorrt_llm::kernels