The following dynamic instruction counts for each of the SPEC CPU2006 workloads were obtained on a Intel Pentium4 processor.
The measurements were done using hardware performance counters which were accessed using the papiex tool (website).
The benchmarks were compiled using the Intel icc,icpc and ifort compilers, version 9.1 (website). The optimization flags were set to -O2.
WARNING!!!
Recent experiments have shown that instruction counts using hardware performance counters (as with the perfex tool) are not correct. Because of some strange implementation issue of the Pentium 4 processor, too much instructions are counted. For example: the mesa benchmark from the SPEC CPU2000 suite has about 291,680 million instructions (counted with PIN and HPCs on AMD machines), but perfex counts 298,575 million instructions (i.e. a difference of 8 billion instructions !). Any solutions to this issue are highly appreciated.
Any comments or suggestions are welcome... (see contact info)
400.perlbench (average: 772.580.351.014, total: 2.317.741.053.042)
401.bzip2 (average: 386.199.758.551, total: 2.317.198.551.306)
403.gcc (average: 133.537.033.442, total: 1.201.833.300.984)
410.bwaves
416.gamess (average: 1.873.306.471.946, total: 5.619.919.415.838)
429.mcf
433.milc
434.zeusmp
435.gromacs
436.cactusADM
437.leslie3d
444.namd
445.gobmk (average: 340.772.604.707, total: 1.703.863.023.538)
447.dealII (1 input)
450.soplex (average: 416.392.801.433, total: 832.785.602.867)
453.povray
454.calculix (1 input)
456.hmmer (average: 1.589.536.815.371, total: 3.179.073.630.743)
458.sjeng
459.GemsFDTD (1 input)
462.libquantum
464.h264ref (average: 1.410.299.866.321, total: 4.230.899.598.964)
465.tonto
470.lbm
471.omnetpp
473.astar (average: 676.604.316.068, total: 1.353.208.632.137)
481.wrf
482.sphinx3
483.xalancbmk