mirror of
https://github.com/vllm-project/vllm-ascend.git
synced 2025-10-20 21:53:54 +08:00
bcc313e8f2
### What this PR does / why we need it? - Adds the `mla_preprocess` custom kernel to provide an optimized pre-processing operator for Multi-head Latent Attention (MLA) on Ascend NPUs. - Wires the new kernel into the C++ extension pipeline so vLLM can invoke it directly, cutting Python-side tensor shuffling and memory copies that previously bottlenecked MLA compilation paths. ### Does this PR introduce any user-facing change? - No. The change only introduces a low-level kernel; public APIs and inference behavior remain unchanged. ### How was this patch tested? - Dedicated Ascend kernels are not covered by our CI yet, so no extra automated tests were added. Future MLA-focused regression runs will cover this path. - vLLM version: v0.11.0 Signed-off-by: Chen Chen <0109chenchen@gmail.com>
37 lines
1.5 KiB
C++
37 lines
1.5 KiB
C++
/* Adapted from
|
|
* https://gitee.com/ascend/ascend-transformer-boost.git
|
|
*
|
|
* Copyright (c) 2024 Huawei Technologies Co., Ltd.
|
|
* This file is a part of the CANN Open Software.
|
|
* Licensed under CANN Open Software License Agreement Version 1.0 (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.
|
|
*/
|
|
#ifndef INCLUDE_HARDWARE_H
|
|
#define INCLUDE_HARDWARE_H
|
|
|
|
enum class ArchType { ASCEND_V220, ASCEND_V200, ASCEND_M200 };
|
|
|
|
template <ArchType ArchTag>
|
|
struct HardwareInfo {
|
|
static uint32_t const l2BW = 5;
|
|
static uint32_t const hbmBW = 1;
|
|
static uint32_t const supportMix = 0;
|
|
static uint32_t const l1Size = 512 * 1024;
|
|
static uint32_t const l0ASize = 64 * 1024;
|
|
static uint32_t const l0BSize = 64 * 1024;
|
|
static uint32_t const l0CSize = 128 * 1024;
|
|
static uint32_t const l2Size = 192 * 1024 * 1024;
|
|
static uint32_t const biasSize = 1024;
|
|
static uint32_t const fixBufSize = 7 * 1024;
|
|
static uint32_t const ubSize = 192 * 1024;
|
|
static uint32_t const fractalSize = 512;
|
|
static uint32_t const l1l0BlockSize = 32;
|
|
static uint32_t const btBlockSize = 64;
|
|
static uint32_t const fbBlockSize = 128;
|
|
};
|
|
|
|
#endif
|