Results in JMP® will be incorrect if you are running JMP on a machine with an Intel® Xeon® Sapphire Rapids processor and the "Enable Advanced Linear Algebra Routines" preference is chosen. This preference enables the use of BLAS (Basic Linear Algebra Subprograms) and LAPACK (Linear Algebra Package) routines for linear algebra calculations in various JMP platforms and JSL functions. You can circumvent any errors, by selecting "File > Preferences", going to the "General" group, and ensuring the preference is unchecked. Future JMP releases will include updated versions of the BLAS and LAPACK routines.
To determine if your machine is impacted by the problems with the Sapphire Rapids processor, open and run the BLAS Numerics Check script. JMP will provide messaging as to the status of your machine.
In JMP 17, BLAS/LAPACK is used by the following platforms and functions
Platforms:
-
Hierarchical Cluster (with Method=Hybrid Ward)
-
Robust PCA Outliers (if Randomized SVD option is selected)
-
Principal Components (with Estimation Method=Randomized SVD)
-
Structural Equation Models
JSL built-in functions:
-
Matrix Mult Blas()
-
Sym Matrix Mult Blas()
-
SVD Lapack()
-
Random SVD()
-
QR Lapack()
-
Low Rank Symmetric Update Blas()
-
Quadratic Form Blas()
-
Add Vectors Blas()
-
Inner Product Blas()
-
Eigen Blas()
To help determine if your JMP installation is affected, Intel provides a list of products formerly Sapphire Rapids here.