From d542a61d00a68fd1ca7ee38b523831b982612db9 Mon Sep 17 00:00:00 2001 From: msrocka Date: Fri, 21 Jul 2023 14:39:58 +0200 Subject: [PATCH] remove todos --- README.md | 27 +++++++-------------------- 1 file changed, 7 insertions(+), 20 deletions(-) diff --git a/README.md b/README.md index 61a51d1..4cb31ea 100644 --- a/README.md +++ b/README.md @@ -6,13 +6,17 @@ This is an experimental project that links the Intel Math Kernel (MKL) as calcul ## Building -The build scripts are written in Dart so that you need to have a current version of Dart (3.x) installed as well as the `archive` -library. To install the library, run: +The build scripts are written in Dart so that you need to have a current version +of Dart (3.x) installed as well as the `archive` library. To install the +library, run: + ```bash dart pub add archive ``` -On macOS with Apple M1/M2, one will need to set the Rust toolchain to `stable-x86_64-apple-darwin`: +On macOS with Apple M1/M2, one will need to set the Rust toolchain to +`stable-x86_64-apple-darwin`: + ```bash rustup install stable-x86_64-apple-darwin rustup default stable-x86_64-apple-darwin @@ -30,20 +34,3 @@ This will download the MKL Python package and its dependencies from pypi.org and The Java part has an `MKL` class with the native method-bindings and methods for loading the libraries from a folder. The method `MKL.loadFrom(DIR)` will load the libraries from the folder `{DIR}/olca-mkl-x64_v{VERSION}`. The current version is `1` (the next version would be `2`, then `3` etc.) and it is not the version of the MKL but the version of the openLCA JNI bindings for the MKL. The method `MKL.loadFromDefault()` will try to load the libraries from the openLCA default workspace `~/openLCA-data-1.4/olca-mkl-x64_v{VERSION}`. With the methods `MKL.isLibraryDir` and `MKL.isDefaultLibraryDir` you can also test if a directory contains the MKL library folder with the required libraries (useful for the integration in openLCA later). **Note** that there is only support for `x64` CPUs, e.g. on macOS M1/2, you need to run a `x64` JVM using the compatibility layer. - -## Python scripts - -On macOS Apple M1/M2, one needs to use `python3-intel64`: - -```bash -python3-intel64 -m pip install scipy numpy -python3-intel64 scripts/pardiso_example.py -``` - -## TODO - -* test performance on AMD processors (we added a method `mkl_serv_intel_cpu_true`, see https://danieldk.eu/mkl-amd-zen/) -* test if and how it works on macOS M1/M2 using the x64 compatibility layer -* move Java part to olca-core (`mkl` branch first) + tests -* port library download/packaging to the openLCA packager script (maybe with an `--mkl` flag) -* test in gdt-server containers