A 64bit-no-graphics executable of carma was built using intel's icc-mkl v.11.0 using

icc -fast -parallel -D_FILE_OFFSET_BITS=64 Color_carma.c -L/opt/intel/Compiler64/11.0/074/mkl/lib/em64t/ -lmkl_lapack

The executable is here carma64.bz2.

Timings for a cartesian PCA problem with 570 selected atoms over 15000 frames …

Executable	Timings (from time)
64bit, repeat 1	69.500u 7.116s 1:20.60 95.0%
64bit, repeat 2	69.696u 7.240s 1:21.70 94.1%
64bit, repeat 3	69.712u 7.208s 1:20.88 95.0%
64bit, repeat 4	69.876u 7.128s 1:21.42 94.5%
32bit, repeat 1	151.825u 6.960s 2:51.46 92.6%
32bit, repeat 2	151.785u 7.192s 2:50.84 93.0%
32bit, repeat 3	152.041u 7.024s 2:52.37 92.2%
32bit, repeat 4	151.925u 7.120s 2:52.06 92.4%

… and for 1065 selected atoms:

Executable	Timings (from time)
64bit, repeat 1	281.345u 8.200s 4:32.68 106.1%
64bit, repeat 2	282.013u 8.508s 4:33.92 106.0%
32bit, repeat 1	589.660u 7.912s 10:41.24 93.1%
32bit, repeat 2	589.640u 7.892s 10:41.93 93.0%