ztao

R.I.P.

> The code was probably not compiled with openmp. Could you call `faiss::check_openmp()` somewhere in the code? `faiss::check_openmp()` returns true.

I read the code: in IndexIVFFlat.cpp, add_with_ids calls add_core, and the add actions executes in serial. in IndexIVF.cpp, add_with_ids can deal with omp in parallel. This means IndexIVFFlat can only...

Actually when faiss was built with openblasp (not default blas), the train and add precedure can eatup all the cores. Somewhat wired, but why?

> 请问cpu kernel中并行的kernel编译成功了吗？如果并行的kernel编译失败，会使用尝试单线程的kernel。 此外，并行kernel默认使用CPU逻辑核心数的一半。如果CPU逻辑核心数少或者获取逻辑核心数时出现bug，也可能导致这种情况。可以尝试再次调用quantize手动设置CPU核心数： `model = model.quantize(bits=4, kernel_file="xxxxxx/quantization_kernels_parallel.so")` 请问cpu kernel中并行的kernel编译成功了吗？ 这个怎么看呢？都是conda + pip 安装的

> 命令行运行时会输出编译信息 >>> import cpm_kernels Traceback (most recent call last): File "", line 1, in File "/Users/xxx/miniconda3/envs/pytorch-1.7.0/lib/python3.8/site-packages/cpm_kernels/__init__.py", line 1, in from . import library File "/Users/xxx/miniconda3/envs/pytorch-1.7.0/lib/python3.8/site-packages/cpm_kernels/library/__init__.py", line 1, in from...

但是是不是这个设置，safari代码显示错位

@Qinka 哈哈哈啊

@messi824 这个repo的维护貌似不太积极，老兄还是转道吧 [https://github.com/Tencent/libco](https://github.com/Tencent/libco)

代码内部使用了**APPLE**关键字了，同时也支持kqueue模拟epoll的行为，所以应该是支持BSD系列的。近两年不完BSD了，所以没有环境测试验证。