Navigation and service

Usage of LAPACK under Linux

The routines are available in single and double precision .
Using a program driver.f calling LAPACK routines with the appropriate BLAS looks like follows:

Using the ifort-compiler:

ifort driver.f -llapack_ifort /usr/local/lib/libmkl_def.a -lguide -lpthread -lmkl

Using the g77-compiler:

g77 driver.f -llapack -lf77blas_platform -latlas_platform

Using the pgf90-compiler:

pgf90 driver.f -llapack_pgf90 -lblas_pgf90

Using the NAGWare-compiler (f90, f95, nagf90, nagf95):

nagf95 driver.f -llapack_nagf95 -lblas_nagf95

For the following Linux processors ATLAS BLAS libraries (Automatically Tuned Linear Algebra Software) are compiled here at FZJ:
Intel Pentium III processor, 16K L1 cache, 512K L2 cache (version 3.4.1) Intel Pentium 4 processor, 512K L2 cache, with SSE2 (version 3.6) AMD Athlon processor, 64K L1 cache, 256K L2 cache (version 3.4.1) AMD Athlon processor, 64K L1 cache, 256K L2 cache, with SSE1 (version 3.4.1)
You can use these optimized BLAS to get a better performance. The file /proc/cpuinfo on your Linux PC contains the necessary informations concerning the cpu. The usage with the f77 interfaces is:

pgf90 -g77libs driver.f -llapack_pgf90 -lf77blas_platform -latlas_platform
You have to replace platform with PIII_C512 or PIIISSE1_C256 or PIIISSE1_C512 or ATHLON_C256 or ATHLONSSE1_C256 or P4SSE2_C512 .