Accelerating the phase-linking step

[sheyu921]

When I firstly installed the miaplpy and tested the phase-linking setp, it was real slow. I tried to chang the version of numpy and scipy and this is not helpful. After many tests, I found that using the mkl to accelerate the setp is working. It accelerated about 10 times than before. However, it remain ran very slow for a large area with Ubuntu Linux. Then, I limit the number of BLAS/OpenBLAS/MKL threads to 1 to avoid excessive "process × thread" nesting. This accelerated about 30 times than the last step.

For Ubuntu 24.04, after installed the miaplpy, install the mkl by

mamba install "libblas=*=*mkl"

Before run the phasing-linking, run the code below first

export OMP_NUM_THREADS=1 MKL_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 

For macos, after installed the miaplpy, run the code below

mamba install "libblas=*=*accelerate" "liblapack=*=*accelerate"

I hope this is helpful for people who found the same problem.