brpc调用openmp时cpu消耗不受控制

在brpc服务中调用openmp进行矩阵运算，在设置了环境变量`OMP_NUM_THREADS和OMP_DYNAMIC="FALSE"`后，cpu消耗依旧不受控制。
例：`OMP_NUM_THREADS=4 OMP_DYNAMIC="FALSE"`；单个client请求的情况下，cpu消耗达到在600%-1000%之间波动；多个client并发的情况下，cpu消耗更加不可控。
已验证使用pthread对上述openmp的计算模块进行多线程调用时cpu消耗正常；已验证在brpc服务中去除openmp的相关模块后正常。
请问上述情况和bthread有没有关系？或者有什么排查方向吗？非常感谢！

The report names no source file or test. Start by reproducing the matrix workload in a brpc service with OMP_NUM_THREADS=4 and OMP_DYNAMIC=FALSE, compare one and multiple clients with the pthread and no-OpenMP cases, and investigate whether bthread affects CPU usage. Done means establishing the cause and recording a verified resolution or diagnosis.

Written by the indexing model from the issue text.