From 32a64b8af10b973c1355347ae45dfde38bcb8fcf Mon Sep 17 00:00:00 2001 From: "c.diddens" Date: Tue, 1 Sep 2026 08:49:59 +0200 Subject: [PATCH] Update the installation instructions for 0.2.0 Python 3.10+ is the new baseline, and the Mac arm64 section now describes both workarounds for the missing MKL Pardiso solver (Rosetta 2, or PETSc/SLEPc with MUMPS). Claude-Session: https://claude.ai/code/session_01S6QpmY6eWUhGegwWdzACex --- installation.html | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-) diff --git a/installation.html b/installation.html index 2744cbd..1a156f1 100644 --- a/installation.html +++ b/installation.html @@ -2,12 +2,12 @@

Installation

-

When you have Python 3.9 to 3.14 installed (CPython, 64-bit ,x86_64),

+

When you have Python 3.10 or higher installed (CPython, 64-bit),

python -m pip install pyoomph

-

should install the basic framework. On Macs with Apple silicon (arm64, e.g. M1-M4) processor systems, please execute it in a Rosetta2 terminal (see below). +

should install the recent release. For maximum performance and system-specific information, please refer to the sections below.

If you cannot manage to install it, refer to our tutorial. If this cannot help, you can ask for help (c.diddens@utwente.nl)

@@ -31,19 +31,26 @@

On Mac

-

If you have a recent Mac with an Apple silicon (arm64) chip, you must run all commands in a Rosetta 2 terminal, see here how to set it up (note that recent systems must be handled as described here). Also, please make sure to not upgrade your mkl package, i.e. enforce it to version 2021.4.0. Alternatively, you can compile it directly on arm64, but with the downside of not being able to use the fast Pardiso solver. See here for details. Also, you might have to use a less recent version of python. See here for details.

- - +

If you have a recent Mac with an Apple silicon (arm64) chip, mind the following: + Python cannot install the fast MKL Pardiso solver on arm64 natively and will use Accelerate by default, which becomes unacceptably slow for larger problems. There are two workarounds, either install pyoomph in a Rosetta 2 terminal to use MKL Pardiso or install PETSc/SLEPc and use MUMPS as fast linear solver instead. + + For the first option, you must run the installation command above in a Rosetta 2 terminal, see here how to set it up (note that recent systems must be handled as described here). Also, please make sure to not upgrade your mkl package, i.e. enforce it to version 2021.4.0.

+ +

python3 -m pip install mkl==2021.4.0

+ +

See here for further details. Also, you might have to use a less recent version of python, see here for details.

+ +

For the second option (PETSc with MUMPS), please refer to the installation instructions.

+ -

Make sure to have the XCode developer tools, e.g., by installing them via

+

In any case, make sure to have the XCode developer tools, e.g., by installing them via

xcode-select --install

- -

and test pyoomph via

+

(in Rosetta2 for the first option) and test pyoomph via

python -m pyoomph check all