本文主要是介绍OrangePi AIpro 香橙派 昇腾 Ascend C 算子开发 与 调用 - Tiling实现，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
OrangePi AIpro 香橙派 昇腾 Ascend C 算子开发 与 调用 - Tiling实现
flyfish
前置知识
基于Kernel直调工程的算子开发流程图
其中有一个Tiling实现
什么是Tiling、Tiling实现
计算API，包括标量计算API、向量计算API、矩阵计算API，分别实现调用Scalar计算单元、Vector计算单元、Cube计算单元执行计算的功能。
数据搬运API，计算API基于Local Memory数据进行计算，所以数据需要先从Global Memory搬运至Local Memory，再使用计算API完成计算，最后从Local Memory搬出至Global Memory。执行搬运过程的接口称之为数据搬移API，比如DataCopy接口。
大多数情况下，Local Memory的存储，无法完整的容纳算子的输入与输出，需要每次搬运一部分输入进行计算然后搬出，再搬运下一部分输入进行计算，直到得到完整的最终结果，这个数据切分、分块计算的过程称之为Tiling。根据算子的shape等信息来确定数据切分算法相关参数（比如每次搬运的块大小，以及总共循环多少次）的计算程序，称之为Tiling实现。
昇腾AI处理器在进行数据搬运和Vector计算时，对于搬运的数据长度和UB首地址都有必须32B对齐的要求。
当需要从Global拷贝11个half数值到Local时，使用DataCopy将拷贝16个half(32B)数据到Local上，Local[11]~Local[15]被写成无效数据-1。
非对齐搬入内存
当需要从Local拷贝11个half数值到Global时，使用DataCopy将拷贝16个half(32B)数据到Global上，Global[11]~Global[15]被覆写成-1。
非对齐搬出内存
Tiling实现完成后，获取到的Tiling切分算法相关参数，会传递给kernel侧，用于指导并行数据的切分。由于Tiling实现中完成的均为标量计算，AI Core并不擅长，所以我们将其独立出来放在host CPU上执行。
tiling实现
TilingData参数设计，TilingData参数本质上是和并行数据切分相关的参数，本示例算子使用了2个tiling参数：totalLength、tileNum。totalLength是指需要计算的数据量大小，tileNum是指每个核上总计算数据分块个数。比如，totalLength这个参数传递到kernel侧后，可以通过除以参与计算的核数，得到每个核上的计算量，这样就完成了多核数据的切分。tiling实现代码中通过上下文获取输入输出的shape信息，并对应设置TilingData。
原始的
// 实现核函数
extern "C" __global__ __aicore__ void add_custom(GM_ADDR x, GM_ADDR y, GM_ADDR z)
{// 初始化算子类，算子类提供算子初始化和核心处理等方法KernelAdd op;// 初始化函数，获取该核函数需要处理的输入输出地址，同时完成必要的内存初始化工作op.Init(x, y, z);// 核心处理函数，完成算子的数据搬运与计算等核心逻辑op.Process();
}// 调用核函数
void add_custom_do(uint32_t blockDim, void* l2ctrl, void* stream, uint8_t* x, uint8_t* y, uint8_t* z)
{add_custom<<<blockDim, l2ctrl, stream>>>(x, y, z);
}
tiling实现
add_custom_tiling.h
#ifndef ADD_CUSTOM_TILING_H
#define ADD_CUSTOM_TILING_H
#include <cstdint>struct AddCustomTilingData {uint32_t totalLength;uint32_t tileNum;
};
#endif
add_custom.cpp
#include "add_custom_tiling.h"
#include "kernel_operator.h"constexpr int32_t BUFFER_NUM = 2; // tensor num for each queueclass KernelAdd {
public:__aicore__ inline KernelAdd() {}__aicore__ inline void Init(GM_ADDR x, GM_ADDR y, GM_ADDR z, uint32_t totalLength, uint32_t tileNum){this->blockLength = totalLength / AscendC::GetBlockNum();this->tileNum = tileNum;this->tileLength = this->blockLength / tileNum / BUFFER_NUM;xGm.SetGlobalBuffer((__gm__ half *)x + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);yGm.SetGlobalBuffer((__gm__ half *)y + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);zGm.SetGlobalBuffer((__gm__ half *)z + this->blockLength * AscendC::GetBlockIdx(), this->blockLength);pipe.InitBuffer(inQueueX, BUFFER_NUM, this->tileLength * sizeof(half));pipe.InitBuffer(inQueueY, BUFFER_NUM, this->tileLength * sizeof(half));pipe.InitBuffer(outQueueZ, BUFFER_NUM, this->tileLength * sizeof(half));}__aicore__ inline void Process(){int32_t loopCount = this->tileNum * BUFFER_NUM;for (int32_t i = 0; i < loopCount; i++) {CopyIn(i);Compute(i);CopyOut(i);}}private:__aicore__ inline void CopyIn(int32_t progress){AscendC::LocalTensor<half> xLocal = inQueueX.AllocTensor<half>();AscendC::LocalTensor<half> yLocal = inQueueY.AllocTensor<half>();AscendC::DataCopy(xLocal, xGm[progress * this->tileLength], this->tileLength);AscendC::DataCopy(yLocal, yGm[progress * this->tileLength], this->tileLength);inQueueX.EnQue(xLocal);inQueueY.EnQue(yLocal);}__aicore__ inline void Compute(int32_t progress){AscendC::LocalTensor<half> xLocal = inQueueX.DeQue<half>();AscendC::LocalTensor<half> yLocal = inQueueY.DeQue<half>();AscendC::LocalTensor<half> zLocal = outQueueZ.AllocTensor<half>();AscendC::Add(zLocal, xLocal, yLocal, this->tileLength);outQueueZ.EnQue<half>(zLocal);inQueueX.FreeTensor(xLocal);inQueueY.FreeTensor(yLocal);}__aicore__ inline void CopyOut(int32_t progress){AscendC::LocalTensor<half> zLocal = outQueueZ.DeQue<half>();AscendC::DataCopy(zGm[progress * this->tileLength], zLocal, this->tileLength);outQueueZ.FreeTensor(zLocal);}private:AscendC::TPipe pipe;AscendC::TQue<AscendC::QuePosition::VECIN, BUFFER_NUM> inQueueX, inQueueY;AscendC::TQue<AscendC::QuePosition::VECOUT, BUFFER_NUM> outQueueZ;AscendC::GlobalTensor<half> xGm;AscendC::GlobalTensor<half> yGm;AscendC::GlobalTensor<half> zGm;uint32_t blockLength;uint32_t tileNum;uint32_t tileLength;
};extern "C" __global__ __aicore__ void add_custom(GM_ADDR x, GM_ADDR y, GM_ADDR z, AddCustomTilingData tiling)
{KernelAdd op;op.Init(x, y, z, tiling.totalLength, tiling.tileNum);op.Process();
}
main.cpp
#include "add_custom_tiling.h"
#include "data_utils.h"
#ifndef ASCENDC_CPU_DEBUG
#include "acl/acl.h"
#include "aclrtlaunch_add_custom.h"
#else
#include "tikicpulib.h"
extern "C" __global__ __aicore__ void add_custom(GM_ADDR x, GM_ADDR y, GM_ADDR z, AddCustomTilingData tiling);
#endifint32_t main(int32_t argc, char *argv[])
{uint32_t blockDim = 8;size_t tilingSize = 2 * sizeof(uint32_t);size_t inputByteSize = 8 * 2048 * sizeof(uint16_t);size_t outputByteSize = 8 * 2048 * sizeof(uint16_t);#ifdef ASCENDC_CPU_DEBUGuint8_t *tiling = (uint8_t *)AscendC::GmAlloc(tilingSize);ReadFile("./input/input_tiling.bin", tilingSize, tiling, tilingSize);uint8_t *x = (uint8_t *)AscendC::GmAlloc(inputByteSize);uint8_t *y = (uint8_t *)AscendC::GmAlloc(inputByteSize);uint8_t *z = (uint8_t *)AscendC::GmAlloc(outputByteSize);ReadFile("./input/input_x.bin", inputByteSize, x, inputByteSize);ReadFile("./input/input_y.bin", inputByteSize, y, inputByteSize);AscendC::SetKernelMode(KernelMode::AIV_MODE);ICPU_RUN_KF(add_custom, blockDim, x, y, z,*reinterpret_cast<AddCustomTilingData *>(tiling)); // use this macro for cpu debugWriteFile("./output/output_z.bin", z, outputByteSize);AscendC::GmFree((void *)x);AscendC::GmFree((void *)y);AscendC::GmFree((void *)z);AscendC::GmFree((void *)tiling);
#elseCHECK_ACL(aclInit(nullptr));int32_t deviceId = 0;CHECK_ACL(aclrtSetDevice(deviceId));aclrtStream stream = nullptr;CHECK_ACL(aclrtCreateStream(&stream));AddCustomTilingData *tiling;uint8_t *xHost, *yHost, *zHost;uint8_t *xDevice, *yDevice, *zDevice;CHECK_ACL(aclrtMallocHost((void **)(&tiling), tilingSize));ReadFile("./input/input_tiling.bin", tilingSize, tiling, tilingSize);CHECK_ACL(aclrtMallocHost((void **)(&xHost), inputByteSize));CHECK_ACL(aclrtMallocHost((void **)(&yHost), inputByteSize));CHECK_ACL(aclrtMallocHost((void **)(&zHost), outputByteSize));CHECK_ACL(aclrtMalloc((void **)&xDevice, inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));CHECK_ACL(aclrtMalloc((void **)&yDevice, inputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));CHECK_ACL(aclrtMalloc((void **)&zDevice, outputByteSize, ACL_MEM_MALLOC_HUGE_FIRST));ReadFile("./input/input_x.bin", inputByteSize, xHost, inputByteSize);ReadFile("./input/input_y.bin", inputByteSize, yHost, inputByteSize);CHECK_ACL(aclrtMemcpy(xDevice, inputByteSize, xHost, inputByteSize, ACL_MEMCPY_HOST_TO_DEVICE));CHECK_ACL(aclrtMemcpy(yDevice, inputByteSize, yHost, inputByteSize, ACL_MEMCPY_HOST_TO_DEVICE));ACLRT_LAUNCH_KERNEL(add_custom)(blockDim, stream, xDevice, yDevice, zDevice, tiling);CHECK_ACL(aclrtSynchronizeStream(stream));CHECK_ACL(aclrtMemcpy(zHost, outputByteSize, zDevice, outputByteSize, ACL_MEMCPY_DEVICE_TO_HOST));WriteFile("./output/output_z.bin", zHost, outputByteSize);CHECK_ACL(aclrtFree(xDevice));CHECK_ACL(aclrtFree(yDevice));CHECK_ACL(aclrtFree(zDevice));CHECK_ACL(aclrtFreeHost(xHost));CHECK_ACL(aclrtFreeHost(yHost));CHECK_ACL(aclrtFreeHost(zHost));CHECK_ACL(aclrtFreeHost(tiling));CHECK_ACL(aclrtDestroyStream(stream));CHECK_ACL(aclrtResetDevice(deviceId));CHECK_ACL(aclFinalize());
#endifreturn 0;
}
这篇关于OrangePi AIpro 香橙派 昇腾 Ascend C 算子开发 与 调用 - Tiling实现的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
