failed to build the so file with comparable performance

[yxw]

I managed to build the library from source code, however, the performance is significantly (20x) slower than the provided one. (0.08 vs. 0.004)

the reason why I have to rebuild from source is that that we found there is serious memory leak when using this build version, and we were trying to track down the issue.

I've tried using "-Ofast", "-mavx2" and others, I also tried different versions of GCC/clang , but still couldn't achieve even similar performance.

Could you help provide the exact compile / link parameters, or any other optimisation techniques you used to generate this binary release ?

thank you

-xiaowei