Hi
Regarding the free Arm Performance Libraries...
"Arm Performance Libraries provides optimized standard core math libraries for high-performance computing applications on Arm processors. This free version of the libraries provides optimized libraries for Arm® Neoverse N1-based Armv8 AArch64 implementations that are compatible with various versions of GCC. You do not require a license for this version of the libraries."
Does the above mean the libraries work exclusively with Arm® Neoverse N1-based Armv8 AArch64 processors?
I've successfully built HPL Linpack with the free Arm Performance Libraries for Ubuntu 16.04+ with GCC-9.3 on my Raspberry Pi 4.
However, I get...
"mpirun noticed that process rank 0 with PID 0 on node node9 exited on signal 4 (Illegal instruction)."
I'm guessing the libraries use Armv8.2-A instructions, not implements on my Armv8-A Cortex-A72 cores? Or, maybe I've missed some compile/link options?
My HPL Make.armpl file references armpl as follows...
# ---------------------------------------------------------------------- # - Linear Algebra library (BLAS or VSIPL) ----------------------------- # ---------------------------------------------------------------------- # LAinc tells the C compiler where to find the Linear Algebra library # header files, LAlib is defined to be the name of the library to be # used. The variable LAdir is only used for defining LAinc and LAlib. # LAdir = /opt/arm/armpl_20.2_gcc-9.3 LAinc = LAlib = -L$(LAdir)/lib -larmpl -lgfortran -lamath -lm
And after I "export LD_LIBRARY_PATH=/opt/arm/armpl_20.2_gcc-9.3/lib" on a single node, all of the libraries are found, but during running with the illegal instruction.
Best wishes
John
Cortex A72 included?
Yes, Cortex-A72 is supported.
Great! Thanks!
Hi Chris
That is fabulous news!
Thank you