算子开发时：

op_host侧：rnnMatmul1.SetDim(rnnParams.sysAicCoreNum);

matmul初始化方式：inputMM.Init(inputMMTiling, &pipe);

计算：if ASCEND_IS_AIC { processInputMM();}

processInputMM(){

if (GetBlockIdx() < inputMMTiling.usedCoreNum) {

    // printf("inputOffsets.AOffset: %d\n", inputOffsets.AOffset);

    // printf("inputOffsets.BOffset: %d\n", inputOffsets.BOffset);

    // printf("inputOffsets.COffset: %d\n", inputOffsets.COffset);

    inputMM.SetTensorA(xGm[inputOffsets.AOffset]);

    // DumpTensor(xGm[inputOffsets.AOffset], __LINE__, 16);

    inputMM.SetTensorB(weightGm[inputOffsets.BOffset]);

    // DumpTensor(weightGm[inputOffsets.BOffset], __LINE__, 16);

    inputMM.SetBias(biasGm[inputOffsets.BOffset]);

    // DumpTensor(biasGm[inputOffsets.BOffset], __LINE__, 16);

    if (inputTail.nCoreIndx == inputTail.notTailNCoreCount && inputTail.mCoreIndx == inputTail.notTailMCoreCount) {

      inputMM.SetTail(inputTail.tailSingleCoreM, inputTail.tailSingleCoreN);

    } else if (inputTail.nCoreIndx == inputTail.notTailNCoreCount) {

      inputMM.SetTail(inputMMTiling.singleCoreM, inputTail.tailSingleCoreN);

    } else if (inputTail.mCoreIndx == inputTail.notTailMCoreCount) {

      inputMM.SetTail(inputTail.tailSingleCoreM, inputMMTiling.singleCoreN);

    }

    inputMM.IterateAll(outputGm.workspace[inputOffsets.COffset], false);

    // DumpTensor(outputGm.workspace[inputOffsets.COffset], __LINE__, 16);

  }

}

DumpTensor不开启时，执行算子报错信息如下：

 

EZ9999: Inner Error! 

EZ9999: [PID: 37099] 2024-10-17-03:04:47.202.186 The error from device(chipId:1, dieId:0), serial number is 36, there is an fftsplus aivector error exception, core id is 0, error code = 0, dump info: pc start: 0x12c0c00d146c, current: 0x12c0c00d19a8, vec error info: 0xa10114de18, mte error info: 0xfbf65c754f, ifu error info: 0x212c136c11000, ccu error info: 0xe46730102300005c, cube error info: 0, biu error info: 0, aic error mask: 0x6500020bd00028c, para base: 0x12c1000d0080.[FUNC:ProcessStarsCoreErrorInfo][FILE:device_error_proc.cc][LINE:1218] 

        TraceBack (most recent call last): 

       The extend info: errcode:(0, 0x200000000000000, 0) errorStr: The MPU address access is invalid. fixp_error0 info: 0x65c754f, fixp_error1 info: 0xfb fsmId:0, tslot:0, thread:0, ctxid:0, blk:9, sublk:1, subErrType:4.[FUNC:ProcessStarsCoreErrorInfo][FILE:device_error_proc.cc][LINE:1230] 

       Kernel task happen error, retCode=0x26, [aicore exception].[FUNC:PreCheckTaskErr][FILE:davinic_kernel_task.cc][LINE:1221] 

       AICORE Kernel task happen error, retCode=0x26.[FUNC:GetError][FILE:stream.cc][LINE:1062] 

       Aicore kernel execute failed, device_id=9, stream_id=2, report_stream_id=2, task_id=0, flip_num=0, fault kernel_name=SelfDynamicRNN_161fc099eb0c35971672981fb05c101e_10000006_mix_aic, fault kernel info ext=none, program id=0, hash=14702801425534950479.[FUNC:GetError][FILE:stream.cc][LINE:1062] 

       [AIC_INFO] after execute:args print end[FUNC:GetError][FILE:stream.cc][LINE:1062] 

       rtStreamSynchronize execute failed, reason=[aicore exception][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53] 

       synchronize stream failed, runtime result = 507015[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:161] 

 

aclrtSynchronizeStream failed. ERROR: 507015 

DumpTensor开启：

inputOffsets.AOffset，inputOffsets.BOffset，inputOffsets.COffset，TensorA,TensorB,Bias数据打印均符合预期，DumpTensor(outputGm.workspace[inputOffsets.COffset], __LINE__, 16)打印值均为零

请问这个问题如何解决？？