#!/usr/bin/env python # coding=utf-8 """ Function: This file mainly involves the constant variables. Copyright Information: Huawei Technologies Co., Ltd. All Rights Reserved © 2020 """ import os import stat class Constant: """ The class for constant. """ # error code for user:success MS_AICERR_NONE_ERROR = 0 # error code for user: error MS_AICERR_INVALID_PARAM_ERROR = 1 MS_AICERR_INVALID_PATH_ERROR = 2 MS_AICERR_CONNECT_ERROR = 3 MS_AICERR_INVALID_DUMP_DATA_ERROR = 4 MS_AICERR_OPEN_FILE_ERROR = 5 MS_AICERR_EXECUTE_COMMAND_ERROR = 6 MS_AICERR_INVALID_CONFIG_DATA_ERROR = 7 MS_AICERR_INVALID_SLOG_DATA_ERROR = 8 MS_AICERR_FIND_DATA_ERROR = 9 MS_AICERR_SINGLE_OP_ERR = 101 # 检查到单算子运行错误 MS_AICERR_MEMORY_ALLOCATION_ERR = 102 # 检查到内存分配错误 MS_AICERR_HARDWARE_ERR = 103 # 检查到硬件错误 MS_AICERR_OPERATOR_INPUT_DATA_ERR = 104 # 检查到算子输入数据错误 MS_AICERR_FRAMEWORK_MEMSET_MISSING = 105 # 检查到框架未执行memset清零 MS_AICERR_OPERATOR_ARGS_OVERWRITTEN = 106 # 检查到算子args被踩 MS_AICERR_ATOMIC_OPERATOR_OVERFLOW = 107 # 检查到atomic算子遇到溢出数据导致错误 WRITE_FLAGS = os.O_WRONLY | os.O_CREAT WRITE_MODES = stat.S_IWUSR | stat.S_IRUSR DIRECTORY_MASK = 0o700 MAX_READ_FILE_BYTES = 1024 * 1024 # 1M MAX_TAR_SIZE = 1 * 1024 * 1024 * 1024 # 1G DIR_PLOG = 'plog' AIC_ERROR_TUPLE_LEN = 9 # dump_data_parser UINT64_SIZE = 8 TIME_LENGTH = 1000 UINT64_FMT = 'Q' STRUCT_FORMAT_KEY = 'struct_format' DTYPE = 'dtype' SIZE_OF_DTYPE = {"DT_FLOAT": 4, "DT_FLOAT16": 2, "DT_INT8": 1, "DT_INT16": 2, "DT_UINT16": 2, "DT_UINT8": 1, "DT_INT32": 4, "DT_INT64": 8, "DT_UINT32": 4, "DT_UINT64": 8, "DT_BOOL": 1, "DT_DOUBLE": 8, "DT_STRING, -1}, {DT_DUAL_SUB_INT8": 1, "DT_DUAL_SUB_UINT8": 1, "DT_COMPLEX64": 8, "DT_COMPLEX128, 16}, {DT_QINT8": 1, "DT_QINT16": 2, "DT_QINT32": 4, "DT_QUINT8": 1, "DT_QUINT16": 2, "DT_RESOURCE": -1, "DT_STRING_REF": -1, "DT_DUAL": 5, "DT_BFLOAT16": 2, "DT_BF16": 2} #aicore_error_parser OBJ_DUMP_FILE = "cce-objdump" NEW_DUMP_FILE = "llvm-objdump" GRAPH_FILE = 0 # collection EXCEPTION_PATTERN = r"TIME:(\d+-\d+-\d+-\d+:\d+:\d+\.\d+\.\d+)[ \S]+?" \ r"device_id=(\d+)[ \S]+?stream_id=(\d+)[ \S]+?task_id=" \ r"(\d+).+?AICORE_INFO_START: core_id=(\d+).+?AIC_ERROR=" \ r"(\S+).+?PC_START=(\S+)(.+?)CURRENT_PC=(\S+)" AIC_ERROR_INFO_DICT = { 175: "vec_err_parity_err, VEC 中发生奇偶校验错误", 174: "vec_valu_ill_issue, VALU 指令传输顺序违反 ISA 约束", 173: "vec_pb_read_no_resp, SU 从 VEC 接收 PB 读取请求后长时间不响应", 172: "vec_pb_ecc_mberr, SU 返回给 VEC 的 PB 数据包含 ECC 错误", 171: "vec_biu_resp_err, BIU 返回给 VEC 的数据错误", 170: "vec_ic_ecc_err, 从 VEC ICACHE 中获取的指令发生 ECC 错误", 169: "vec_ill_vga_vpd_order, VAG 和 VPD 命令的顺序违反 IAS 约束", 168: "vec_ill_instr_padding, VAG 和 VPD 的 PADDING 指令不是 VNOP", 167: "vec_st_num_exceed_limit, st 指令数量超过 ISA 中指定的最大值", 166: "vec_ld_num_exceed_limit, ld 指令数量超过 ISA 中指定的最大值", 165: "vec_ex_num_mismatch, ex 指令所在的代码段包含非 ex 指令", 164: "vec_st_num_mismatch, st 指令所在的代码段包含非 st 指令", 163: "vec_ld_num_mismatch, ld 指令所在的代码段包含非 ld 指令", 162: "vec_ill_vloop_sreg, 第 4 层的 VLOOP 循环次数全部为 0", 161: "vec_ill_vloop_op, VLOOP 指令中的操作码错误", 160: "vec_vci_idata_out_range, VCI 指令的输入数据超出范围", 159: "vec_valu_neg_sqrt, VALU squart 操作的输入数据是负数", 158: "vec_valu_neg_ln, VALU lN 操作的输入数据是负数", 157: "vec_div_by_zero, VEC 的除零错误", 156: "vec_idata_inf_nan, 指令操作的输入数据是 INF/NAN", 155: "vec_ub_ecc_mberr, 访问 UB 时发生多位 ECC 错误", 154: "vec_ub_addr_wrap_around, UB 的访问地址超出范围", 153: "vec_instr_ill_sqzn, sqzn 值无效", 152: "vec_instr_ill_mask, 掩码值无效", 151: "vec_instr_misalign, 指令访问 UB 地址未对齐", 150: "vec_instr_ill_cfg, VEC 指令配置非法", 149: "vec_instrs_undef, 指令在 ISA 中未定义", 148: "vec_instr_timeout, 指令运行超时", 147: "vec_data_excpt_vec, VECTOR 写入/读取时报告数据异常", 146: "vec_data_excpt_su, SU 写入/读取时报告数据异常", 145: "vec_data_excpt_mte, MTE 写入/读取时报告数据异常", 144: "mte_err_l0c_rdwr_cflt, FIXP 读取 L0C 时发生读写操作冲突", 143: "fixp_l0c_ecc, FIXP 指令错误: 读取 L0C 时的 ECC 验证失败", 142: "cube_err_pbuf_wrap_around, CUBE FIXP_BUFFER 发生循环错误", 141: "fixp_err_out_write_overflow, FIXP 写入溢出错误", 140: "ccu_cross_core_set_ovfl_err, 跨核心通信的标志计数器值超过最大值 15", 139: "ccu_lsu_atomic_err, 标量 LSU 执行原子操作失败", 138: "mte_aipp_ecc_err, MTE AIPP 中发生多位 ECC 错误", 137: "mte_stb_ecc_err, MTE STB 中发生多位 ECC 错误", 136: "biu_unsplit_err, BIU 上发生异常, 例如 tag_id 错误或 FIFO 溢出", 135: "ccu_ub_overflow_err, AIC 和 AIV 共享 UB 时发生 LS/ST 访问覆盖", 134: "ccu_ub_wr_cflt, Lite Only 错误, CCU 向 UB 写入数据时发生 UB 读/写冲突", 133: "ccu_ub_rd_cflt, Lite Only 错误, CCU 读取 UB 时发生 UB 读/写冲突", 132: "mte_ktable_rd_addr_overflow, MTE 为空时发生读地址冲突并报告异常", 131: "mte_ktable_wr_addr_overflow, MTE 满时发生写地址冲突并报告异常", 130: "mte_ub_wr_cflt, Lite Only 错误, MTE 写入 UB 时发生 UB 读/写冲突", 129: "mte_ub_rd_cflt, Lite Only 错误, MTE 读取 UB 时发生 UB 读/写冲突并报告异常", 128: "mte_timeout, Lite Only 错误, MTE 指令或数据超时", 127: "ccu_safety_crc_err, MTE CRC 错误", 126: "cube_illegal_instr, CUBE 指令配置非法", 125: "mte_ub_rd_ovflw, MTE 读取 UB 地址溢出", 124: "mte_ub_wr_ovflw, MTE 写入 UB 地址溢出", 123: "ccu_pb_ecc_err, 参数缓冲区出现 2 位 ECC 错误", 122: "ccu_lsu_err, 缓冲区启用时, 堆栈访问指令缓存缺失", 121: "ccu_mpu_err, MPU 地址访问无效", 120: "ccu_seq_err, SEQ 命令序列不正确", 119: "mte_err_waipp, WAIPP 指令配置非法", 118: "mte_err_hebce, HEBCE 指令配置非法", 117: "mte_err_hebcd, HEBCD 指令配置非法", 116: "mte_err_instr_illegal_cfg, MTE 指令配置非法", 115: "cube_err_hset_cnt_ovf, CUBE HSET 计数器上溢错误", 114: "cube_err_hset_cnt_unf, CUBE HSET 计数器下溢错误", 113: "mte_err_cache_ecc, MTE 内部 MVF 缓存故障", 112: "ccu_err_parity_err, 安全功能期间 SU 内部缓冲区出现奇偶校验错误", 111: "mte_err_waitset, HWATI/HSET 配置不正确", 110: "mte_err_fifo_parity, 从 MTE 内部 FIFO 读取的数据不正确", 109: "sc_reg_parity_err, 安全检查期间 nManager 寄存器出现奇偶校验错误", 108: "fixp_err_fbuf_read_ovflw, FBUF 的读取地址超出范围", 107: "fixp_err_fbuf_write_ovflw, FBUF 的写入地址超出范围", 106: "fixp_err_write_ub_ovflw, UB 的写入地址超出范围", 105: "fixp_err_write_l1_ovflw, L1 的写入地址超出范围", 104: "fixp_err_read_ub_ovflw, UB 的读取地址超出范围", 103: "fixp_err_read_l1_ovflw, L1 的读取地址超出范围", 102: "fixp_err_read_l0c_ovflw, L0C 的读取地址超出范围", 101: "fixp_err_illegal_cfg, FIXP 配置非法", 100: "fixp_err_instr_addr_misal, 读取 L0C、读取 L1 和写入 FIXP 缓冲区地址不对齐", 99: "cnt_sw_bus_err, 在慢速上下文切换期间, SC 通过 AXI 总线传输数据, AXI 返回错误", 98: "ccu_neg_sqrt_fp, FP SQRT 计算单元的输入为负数", 97: "ccu_div0_fp, FP32 DIV0 错误", 96: "ccu_dc_tag_ecc, 数据缓存标签 RAM 中出现 2 位 ECC 错误", 95: "ccu_dc_data_ecc, 数据缓存数据 RAM 中出现 2 位 ECC 错误", 94: "ccu_veciq_ecc, VEC 问题队列中出现 2 位 ECC 错误", 93: "mte_err_read_3d_overflow, MTE load3d指令的读取溢出", 92: "mte_err_illegal_smallk_cfg, MTE load3d指令的small K配置不正确", 91: "mte_err_illegal_k_m_start_pos, K_M_START_POS的值非法", 90: "mte_err_illegal_k_m_ext_step, K_M_EXT_STEP的值非法", 89: "mte_err_illegal_chn_size, CHN_SIZE的值非法", 88: "mte_err_illegal_h_size, WINOA fmap H的值非法", 87: "mte_err_illegal_w_size, WINOA fmap W的值非法", 86: "mte_err_illegal_h_cov_pad_ctl, WINOA H padding的值非法", 85: "mte_err_illegal_v_cov_pad_ctl, WINOA V padding的值非法", 84: "mte_err_wino_l0b_read_overflow, WINOB读取的L1地址溢出, 发生了循环", 83: "mte_err_wino_l0b_write_overflow, WINOB向L0B地址写入时发生溢出", 82: "mte_err_dw_fmap_h_illegal, DEPTHWISE FMAP上配置的H值小于3", 81: "mte_err_dw_pad_conf_err, DEPTHWISE PADDING配置错误", 80: "mte_err_addr_misalign, 指令地址不对齐(ADDR_MISALIGN)", 79: "ccu_bus_err, D-cache读写UB时, 总线返回的响应值为非零值", 78: "ccu_addr_err, CCU指令地址检查错误", 77: "vec_instr_undef, 当前VEC版本不支持的指令", 76: "vec_instr_illegal_cfg, VEC支持非法配置的指令", 75: "vec_instr_addr_misalign, VEC指令访问的UB地址未对齐", 74: "mte_atm_addr_misalg, MTE原子加地址未对齐(AIC、AIV1、AIV2)", 73: "mte_illegal_schn_cfg, small_channel使能标志有效, 但不满足small_channel的条件", 72: "ccu_inf_nan, CCU执行的浮点指令的输入为nan/inf", 71: "vec_col2img_illegal_k_size, col2img的值非法", 70: "vec_col2img_illegal_fetch_pos, col2img的值非法", 69: "vec_col2img_illegal_1st_win_pos, col2img的值非法", 68: "vec_col2img_illegal_stride, col2img的值非法", 67: "vec_col2img_illegal_fm_size, col2img的值非法", 66: "vec_col2img_rd_dfm_addr_ovfflow, col2img的值非法", 65: "vec_col2img_rd_fm_addr_ovflow, col2img的值非法", 64: "ccu_sbuf_ecc, 在CCU的scalar buffer中报了ECC错误", 63: """biu_dfx_err, BIU模块返回了一个综合型错误, 把所有的错误类型或在一起""", 62: """vec_ub_wrap_around, vector指令自动计算UB地址时发生越界, X或者VA寄存器已经越界的情况不报此错误""", 61: """vec_ub_self_rdwr_cflt, vector指令同时读写了UB的同一个地址, 仅针对lite平台""", 60: """vec_ub_ecc, ecc error, 硬件问题""", 59: """vec_same_blk_addr, vector指令在同一个repeat写了相同的block, 代码不会写出这样的指令, 一般为硬件问题； VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 58: """vec_neg_sqrt, VSQRT指令被开方数为负数, VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 57: """vec_neg_ln, VLN指令的操作数为负数, VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 56: """vec_l0c_rdwr_cflt, L0C读写同地址冲突, 比如MMAD指令写L0C地址与MOV指令读L0C地址相同, CUBE_ERR_INFO中的cube_err_addr记录了冲突的地址""", 55: """vec_l0c_ecc, ecc error, 硬件问题""", 54: """vec_inf_nan, vector指令的操作数为Inf或者NaN, VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 53: """vec_illegal_mask, vector指令MASK域段为全0, VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 52: """vec_div0, VREC指令发生除0错误, VEC_ERR_INFO中的vec_err_rcnt记录了出错的repeat""", 51: """vec_data_excp_vec, vector指令访问的UB地址, 超过了DATA_EXP寄存器设置的UB访问地址范围""", 50: """vec_data_excp_mte, mte指令访问的UB地址, 超过了DATA_EXP寄存器设置的UB访问地址范围""", 49: """vec_data_excp_ccu, scalar指令访问的UB地址, 超过了DATA_EXP寄存器设置的UB访问地址范围""", 48: """mte_write_overflow, LOAD 2D写地址越界""", 47: """mte_write_3d_overflow, LOAD 3D写地址越界""", 46: """mte_unzip, unzip指令异常, MTE_ERR_INFO中的mte_err_type记录了异常类型: 000 uzp_write_over_turn_err 001 uzp_blk_num_zero_err 010 uzp_index_noenough_err 011 uzp_index_err 100 uzp_decompress_err""", 45: """mte_ub_ecc, ecc error, 硬件问题""", 44: """mte_tlu_ecc, ecc error, 硬件问题""", 43: """mte_rob_ecc, ecc error, 硬件问题""", 42: """mte_read_overflow, LOAD 2D读地址越界""", 41: """mte_padding_cfg, LOAD 3D指令padding配置错误, 和padding有关的一些限制如下: 1. Feature_map_size_w(after padding)>Filter_size_w(after dilation) 2. Feature_map_size_h(after padding)>=Filter_size_h(after dilation) 3. 1st filter window position <= Feature map size (with padding) – Filter size (after dilation) 4. Stride size <= Feature map size (with padding) – Filter size (after dilation).This formula fit for the entire feature map in convolution. In terms of a single 3D Load instruction, the feature map may be partial. Then in w dimension, the formula is tenable; In H dimension, it should be ignored.""", 40: """mte_l1_ecc, ecc error, 硬件问题""", 39: """mte_l0b_rdwr_cflt, L0B读写同地址冲突, 比如L0AD 3D指令写L0B地址与MMAD指令读L0B地址相同, CUBE_ERR_INFO中的cube_err_addr记录了冲突的地址""", 38: """mte_l0a_rdwr_cflt, L0A读写同地址冲突, 比如L0AD 3D指令写L0A地址与MMAD指令读L0A地址相同, MTE_ERR_INFO中的mte_err_addr记录了冲突的地址""", 37: """mte_illegal_stride, L0AD 3D指令stride配置错误, 有关的一些限制如下: 1. Stride size, W dimension ∈[1, 63](Bit width[5:0]), H dimension ∈[1, 63](Bit width[5:0]) 2. Stride size <= Feature map size (with padding) – Filter size (after dilation).This formula fit for the entire feature map in convolution. In terms of a single 3D Load instruction, the feature map may be partial. Then in w dimension, the formula is tenable; In H dimension, it should be ignored.""", 36: """mte_illegal_l1_3d_size, L0AD 3D指令L1_3D_SIZE配置错误, 有关的一些限制如下: FeatureMapW * FeatureMapH * (CIndex +Cincr+1) <= L1_3D_size""", 35: """mte_illegal_fm_size, L0AD 3D指令feature map size配置错误, 有关的一些限制如下: 1. Feature_map_size_w(after padding)>Filter_size_w(after dilation) Feature_map_size_h(after padding)>=Filter_size_h(after dilation) 2. 1st filter window position <= Feature map size (with padding) – Filter size (after dilation). 3. Stride size <= Feature map size (with padding) – Filter size (after dilation).This formula fit for the entire feature map in convolution. In terms of a single 3D Load instruction, the feature map may be partial. Then in w dimension, the formula is tenable; In H dimension, it should be ignored. 4. Due to L1 Buffer size is 1MB, Feature map size (without padding) and C dimension index should meet: FeatureMapW * FeatureMapH * (CIndex +1) <= 32768 When repeat mode=0, the C index increment should be consider(Cincr denote the increment of C0 index during repeat): FeatureMapW * FeatureMapH * (CIndex +Cincr+1) <= 32768 When set L1 3D size is valid FeatureMapW * FeatureMapH * (CIndex +Cincr+1) <= L1_3D_size""", 34: """mte_comp, FMC指令异常, 当前index未处理完又下发新的index table, MTE_ERR_INFO中的mte_err_type记录了异常类型: 000 fmc_read_over_turn_err 001 fmc_blk_num_zero_err""", 33: """mte_gdma_write_overflow, DMA MOV指令写越界""", 32: """mte_gdma_read_overflow, DMA MOV指令读越界""", 31: """mte_gdma_illegal_burst_num, DMA MOV指令burst_num不合法, 不能为0""", 30: """mte_gdma_illegal_burst_len, DMA MOV指令burst_len不合法, 不能为0""", 29: """mte_fpos_larger_fsize, L0AD 3D指令fetch position in filter 大于 filter size""", 28: """mte_fmapwh_larger_l1size, L0AD 3D指令w*h*c大于L1 size""", 27: """mte_fmap_less_kernel, L0AD 3D指令feature map size小于kernel size""", 26: """mte_f1wpos_larger_fsize, L0AD 3D指令1st filter window position大于Feature map size – Filter size""", 25: """mte_decomp, FMD指令异常, load index entry的数量和它后面最近的一个load decompressed data 要解压的数据块数量不一致, MTE_ERR_INFO中的mte_err_type记录了异常类型: 000 fmd_write_over_turn_err 001 fmd_blk_num_zero_err 010 fmd_blk_num_noequal_err 011 fmd_header_err 100 fmd_decompress_err""", 24: """mte_cidx_overflow, L0AD 3D指令c0 index L1 overflow""", 23: """mte_biu_rdwr_resp, 通过BIU读写out数据错误, MTE_ERR_INFO中的mte_err_type记录了异常类型, mte_err_addr记录了发生错误的地址(是触发问题的另一面的地址, 比如out -> L1读错误, 记的是L1地址, 又如ub -> out写错误, 记的是ub地址)""", 22: """mte_bas_raddr_obound, L0AD 3D指令base read addr(Xn)越界""", 21: """mte_aipp_illegal_param, AIPP指令aipp_spr寄存器配置有误, MTE_ERR_INFO中的mte_err_type记录了异常类型: 3'b000: aipp_mte_ex_round : 表示访问外部存储绕回 3'b001: aipp_mte_l1_round : 表示访问L1 buffer绕回 3'b010: aipp_mte_inerr : 表示配置AIPP SPR相关的fp16为INF或者NAN""", 20: """ifu_bus_err, BIU返回bus error给IFU, 意味着从out取指令错误, IFU_ERR_INFO中的ifu_err_type记录了异常类型, ifu_err_addr记录了错误的地址""", 19: """cube_l0c_wrap_around, cube指令自动计算L0C地址时发生越界, X或者VA寄存器已经越界的情况不报此错误""", 18: """cube_l0c_self_rdwr_cflt, 前后两条指令的L0C地址相同, 可能同时出现对相同的L0C地址同时读写的场景, CUBE_ERR_INFO中的cube_err_addr记录了冲突的地址""", 17: """cube_l0c_rdwr_cflt, L0C读写同地址冲突, 比如MMAD指令写L0C地址与MOV指令读L0C地址相同, CUBE_ERR_INFO中的cube_err_addr记录了冲突的地址""", 16: """cube_l0c_ecc, ecc error, 硬件问题""", 15: """cube_l0b_wrap_around, cube指令自动计算L0B地址时发生越界, X或者VA寄存器已经越界的情况不报此错误""", 14: """cube_l0b_rdwr_cflt, L0B读写同地址冲突, 比如L0AD 3D指令写L0B地址与MMAD指令读L0B地址相同, CUBE_ERR_INFO中的cube_err_addr记录了冲突的地址""", 13: """cube_l0b_ecc, ecc error, 硬件问题""", 12: """cube_l0a_wrap_around, cube指令自动计算L0A地址时发生越界, X或者VA寄存器已经越界的情况不报此错误""", 11: """cube_l0a_rdwr_cflt, L0A读写同地址冲突, 比如L0AD 3D指令写L0A地址与MMAD指令读L0A地址相同, MTE_ERR_INFO中的mte_err_addr记录了冲突的地址""", 10: """cube_l0a_ecc, ecc error, 硬件问题""", 9: """cube_invld_input, cube指令的输入(L0A/L0B/L0C)数据含有NaN或者Inf, 包含L0C累加场景""", 8: """ccu_ub_ecc, ecc error, 硬件问题""", 7: """ccu_neg_sqrt, sqrt指令出现被开方数为负数""", 6: """ccu_loop_err, 有程序jump, 从其他地方跳到loop body中""", 5: """ccu_loop_cnt_err, loop循环次数设置为0""", 4: """ccu_illegal_instr, 非法执行: 1.指令的binary错误 2.指令地址非对齐""", 3: """ccu_div0, scalar出现除0错误""", 2: """ccu_call_depth_ovrflw, 函数调用深度超过设置值""", 1: """biu_l2_write_oob, L2访问写越界, BIU_ERR_INFO中的biu_err_addr记录了越界的地址""", 0: """trap_or_timeout, trap指令或者超时或者未捕获到任何错误""", } SOC_ERR_INFO_DICT = {"000": "read poison, 读到脏数据", "001": "read oob, 表示L2在作为buffer模式时, 读地址超过了配置的L2虚拟地址的size", "010": "read bus error, 读请求时数据异常, 例如安全访问请求访问了非安全的地址, " "非安全访问的请求访问了安全地址, 访问请求收不到response、atomic运算异常(1980)", "011": "read decode error, 读请求访问的目的地址不在各个模块的地址空间内, 也即越界", "101": "write oob, 表示L2在作为buffer模式时, 写地址超过了配置的L2虚拟地址的size", "110": "write bus error, 写请求时数据异常, 例如安全访问请求访问了非安全的地址, " "非安全访问的请求访问了安全地址, 访问请求收不到response、atomic运算异常(1980)", "111": "write decode error, 写请求访问的目的地址不在各个模块的地址空间内, 也即越界"} FMC_ERR_INFO_DICT = {"000": "fmc_read_over_turn_err", "001": "fmc_blk_num_zero_err"} FMD_ERR_INFO_DICT = {"000": "fmd_write_over_turn_err", "001": "fmd_blk_num_zero_err", "010": "fmd_blk_num_noequal_err", "011": "fmd_header_err", "100": "fmd_decompress_err"} UNZIP_ERR_INFO_DICT = {"000": "uzp_write_over_turn_err", "001": "uzp_blk_num_zero_err", "010": "uzp_index_noenough_err", "011": "uzp_index_err", "100": "uzp_decompress_err"} AIPP_ERR_INFO_DICT = {"000": "aipp_mte_ex_round : 表示访问外部存储绕回", "001": "aipp_mte_l1_round : 表示访问L1 buffer绕回", "010": "aipp_mte_inerr : 表示配置AIPP SPR相关的fp16为INF或者NAN"} IFU_KEY = "IFU_ERR_INFO" CCU_KEY = "CCU_ERR_INFO" BIU_KEY = "BIU_ERR_INFO" CUBE_KEY = "CUBE_ERR_INFO" MTE_KEY = "MTE_ERR_INFO" VEC_KEY = "VEC_ERR_INFO" @property def max_top_n(self: any) -> int: """ max top n """ return 100 @property def min_top_n(self: any) -> int: """ mix top n """ return 1