msobjdump样例【免费下载链接】asc-toolsAscend C Tools仓是CANN基于Ascend C编程语言推出的配套调试工具仓。项目地址: https://gitcode.com/cann/asc-tools概述本样例基于MatmulLeakyRelu算子演示融合编译场景下msobjdump工具的使用方式。样例通过编译matmul_leakyrelu.asc生成融合编译产物再对生成的ELF文件执行解析。msobjdump工具的详细说明请参考msobjdump工具。支持的产品Ascend 950PR/Ascend 950DTAtlas A3 训练系列产品/Atlas A3 推理系列产品Atlas A2 训练系列产品/Atlas A2 推理系列产品目录结构介绍├── 04_msobjdump │ ├── CMakeLists.txt // 编译工程文件 │ ├── data_utils.h // 数据读入写出函数 │ ├── matmul_leakyrelu.asc // Ascend C算子实现 调用样例 │ └── scripts │ ├── gen_data.py // 输入数据和真值数据生成脚本文件 │ └── verify_result.py // 真值对比文件算子描述样例功能MatmulLeakyRelu的计算公式为C A * B Bias C C 0 ? C : C * 0.001样例参数M 1024K 256N 640样例规格如下表所示样例类型(OpType)MatmulLeakyRelu样例输入nameshapedata typeformatA[M, K]float16NDB[K, N]float16NDBias[N]floatND样例输出C[M, N]floatND核函数名matmul_leakyrelu_custom样例实现实现流程通过GenerateTiling实现host侧的Tiling计算通过CalcGMOffset完成分核计算通过Iterate接口完成矩阵乘计算通过LeakyRelu实现激活函数计算调用实现使用内核调用符调用核函数。编译运行在本样例根目录下执行如下步骤编译并执行算子。配置环境变量请根据当前环境上CANN开发套件包的安装方式选择对应配置环境变量的命令。默认路径root用户安装CANN软件包source /usr/local/Ascend/cann/set_env.sh默认路径非root用户安装CANN软件包source $HOME/Ascend/cann/set_env.sh指定路径install_path安装CANN软件包source ${install_path}/cann/set_env.sh执行如下命令若能正常显示-h信息则表示工具环境正常功能可正常使用。msobjdump -h样例执行mkdir -p build cd build; # 创建并进入build目录 cmake ..;make -j; # 编译工程 python3 ../scripts/gen_data.py # 生成测试输入数据 ./demo # 执行编译生成的可执行程序执行样例 python3 ../scripts/verify_result.py output/output.bin output/golden.bin # 验证输出结果是否正确确认算法逻辑正确执行结果如下说明精度对比成功。test pass!调用msobjdump工具解析样例中的demo为融合编译生成的ELF文件。若该ELF中包含.aicore_binary段msobjdump会自动提取该段内容并继续解析无需手工拆分中间文件。解析融合编译产物msobjdump --dump-elf ./demo本样例实际输出如下.ascend.meta META INFO RUNTIME_IMPLICIT_INFO: DOUBLE_PAGE_TABLE_ADDR RUNTIME_IMPLICIT_INFO: FFTS_ADDR VERSION: 1 RUNTIME_IMPLICIT_INFO: DOUBLE_PAGE_TABLE_ADDR RUNTIME_IMPLICIT_INFO: FFTS_ADDR VERSION: 1 RUNTIME_IMPLICIT_INFO: SIMD_TRACE_SPACE .ascend.meta. [0]: _Z23matmul_leakyrelu_customPhS_S_S_S_N7AscendC6tiling11TCubeTilingE_mix_aic KERNEL_TYPE: MIX_AIC_MAIN CROSS_CORE_SYNC: USE_SYNC MIX_TASK_RATION: [1:2] .ascend.meta. [0]: _Z23matmul_leakyrelu_customPhS_S_S_S_N7AscendC6tiling11TCubeTilingE_mix_aiv KERNEL_TYPE: MIX_AIC_MAIN CROSS_CORE_SYNC: USE_SYN全量打印融合编译产物中的device信息msobjdump --dump-elf ./demo --verbose本样例实际输出如下.ascend.meta META INFO RUNTIME_IMPLICIT_INFO: DOUBLE_PAGE_TABLE_ADDR RUNTIME_IMPLICIT_INFO: FFTS_ADDR VERSION: 1 RUNTIME_IMPLICIT_INFO: DOUBLE_PAGE_TABLE_ADDR RUNTIME_IMPLICIT_INFO: FFTS_ADDR VERSION: 1 RUNTIME_IMPLICIT_INFO: SIMD_TRACE_SPACE .ascend.meta. [0]: _Z23matmul_leakyrelu_customPhS_S_S_S_N7AscendC6tiling11TCubeTilingE_mix_aic KERNEL_TYPE: MIX_AIC_MAIN CROSS_CORE_SYNC: USE_SYNC MIX_TASK_RATION: [1:2] .ascend.meta. [0]: _Z23matmul_leakyrelu_customPhS_S_S_S_N7AscendC6tiling11TCubeTilingE_mix_aiv KERNEL_TYPE: MIX_AIC_MAIN CROSS_CORE_SYNC: USE_SYNC MIX_TASK_RATION: [1:2] [elf heard infos] ELF Header: Magic: 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 Class: ELF64 Data: 2s complement, little endian Version: 1 (current) OS/ABI: UNIX - System V ABI Version: 0 Type: EXEC (Executable file) Machine: unknown: 0x1029 Version: 0x1 Entry point address: 0x0 Start of program headers: 64 (bytes into file) Start of section headers: 33504 (bytes into file) Flags: 0x940000 Size of this header: 64 (bytes) Size of program headers: 56 (bytes) Number of program headers: 3 Size of section headers: 64 (bytes) Number of section headers: 16 Section header string table index: 14 Section Headers: [Nr] Name Type Address Off Size ES Flg Lk Inf Al [ 0] NULL 0000000000000000 000000 000000 00 0 0 0 [ 1] .text PROGBITS 0000000000000000 0000e8 006c94 00 AX 0 0 4 ...................................................................................... [15] .strtab STRTAB 0000000000000000 007ce0 0005fc 00 0 0 1 Key to Flags: W (write), A (alloc), X (execute), M (merge), S (strings), I (info), L (link order), O (extra OS processing required), G (group), T (TLS), C (compressed), x (unknown), o (OS specific), E (exclude), D (mbind), p (processor specific) There are no section groups in this file. Program Headers: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align LOAD 0x0000e8 0x0000000000000000 0x0000000000000000 0x006ca7 0x006ca7 R E 0x1000 LOAD 0x0070e8 0x0000000000007000 0x0000000000007000 0x000210 0x000210 RW 0x1000 GNU_STACK 0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000 RW 0 ......获取融合编译产物中的ELF文件列表msobjdump --list-elf ./demo对当前样例产物终端提示如下ELF file 0: demo.aicore.o解压融合编译产物中的ELF文件mkdir -p objdump_out msobjdump --extract-elf ./demo执行上述命令默认在当前执行路径下落盘demo.aicore.o文件若需指定路径可通过--out-dir进行设置。【免费下载链接】asc-toolsAscend C Tools仓是CANN基于Ascend C编程语言推出的配套调试工具仓。项目地址: https://gitcode.com/cann/asc-tools创作声明:本文部分内容由AI辅助生成(AIGC),仅供参考