显卡RTX2080TI 安装了CUDA10.1 ,想用cusparse库解线性方程组，参考例程执行到cusparseScsrsv2_analysis时，返回值为4,对应
CUSPARSE_STATUS_ARCH_MISMATCH the device only supports compute capability 2.0
and above.
这个怎么解决呢，意思是我的显卡不兼容计算能力1.0以下所以不能用这个函数吗？
const cusparseOperation_t trans = CUSPARSE_OPERATION_NON_TRANSPOSE;
const cusparseSolvePolicy_t policy = CUSPARSE_SOLVE_POLICY_USE_LEVEL;
float alpha = 1.0;
status1 = cusparseCreate(&cuSphandle);
status2 = cusparseCreateCsrsv2Info(&csrsv2_info);
status3 = cusparseScsrsv2_bufferSize(cuSphandle, trans, rows, nnz, cuSpDescr,
(float*)ValDevPtr, RowPtrDevPtr, ColIdxDevPtr, csrsv2_info,&pBufferSize);
cudaMalloc((void**)&pBuffer, pBufferSize);
status4 = cusparseScsrsv2_analysis(cuSphandle, trans, rows, nnz, cuSpDescr,
(float*)ValDevPtr, RowPtrDevPtr, ColIdxDevPtr,
csrsv2_info, policy, pBuffer);
printf("buffer size:%d B\n", pBufferSize*4);
if (status1 || status2 || status3)
{
printf("cuSPARSE failed!\n");
}
cout << (int)status4 << endl;
return 0;
程序输出：
Csrsv2 satrt!
buffer size:148992 B
7
对于这个status，cusparse中有两个描述
4.22章节中对于cusparseStatus_t的描述
The function requires a feature absent from the device architecture
To correct: compile and run the application on a device with
appropriate compute capability
以及，10.2.9章节中Status Returned的描述
the device only supports compute capability 2.0 and above.
