The AMD Core Math Library (ACML) is a set of numerical routines tuned specifically for AMD64 platform processors (including Opteron™ and Athlon™64 ). The routines, which are available via both FORTRAN 77 and C interfaces, include BLAS, LAPACK, FFT, RND, and many others.
Availability & Restrictions
OSC supports use of ACML on the Glenn cluster.
To configure your environment for use of ACML load the “acml” module appropriate for your compiler:
|Portland Group, Multi-Core||
This step is required for both building and running ACML applications. The default version is 4.0.1.
Building With ACML
The Glenn ACML modules will automatically configure your environment to locate appropriate include files and libraries. When the ACML modules are loaded the following environment variables will be set:
|$OMP_NUM_THREADS||Number of OpenMP threads to be used. Set to 1 in all ACML modules except acml-pgimp, which sets it to 4. Should be set by the user AFTER loading the module if a different value is needed.|
|$ACML_CFLAGS||Compiler flags: include path to be used|
|$ACML||Linker flags: libraries to be used|
Usage with the Portland Group Compiler, C code
To build a sample C-code that uses the ACML library with the Portland Group Compiler on Glenn, follow the example steps below after logging into the system:
$ cp ~support/examples/ACML/example.c . $ module load acml-pgi $ pgcc $ACML_CFLAGS $ACML -lm -lpgftnrtl –lrt example.c $ ./a.out
Usage with the Portland Group Compiler, FORTRAN code
To build a sample FORTRAN-code that uses the ACML library with the Portland Group Compiler on Glenn, follow the example steps below after logging into the system:
$ cp ~support/examples/ACML/example.f . $ module load acml-pgi $ pgf77 example.f $ACML $ ./a.out
- Intel MKL
- Armstrong Libraries and Compilers Group