1) Lmbench (by Larry McVoy)

compile: gmake get results (to build and run)

run: gmake rerun (for reruns) output: to a file (~Results/<OS>/<hostname>

source: http://btp1da.phy.uni-bayreuth.de/ftp/pub/alpha/lmbench http://reality.sgi.com/employees/lm/lmbench/get_lmbench.html#S-3

note: lmbench measures only a system's ability to transfer data between processor, cache, memory, network, and disk. It does not measure other parts of the system, such as the graphics subsystem, nor is it a MIPS, MFLOPS, throughput, saturation, stress, graphics, or multiprocessor test suite. It is frequently run on multiprocessor (MP) systems to compare their performance against uniprocessor performance, but it does not takeadvantage of any multiprocessor features.

To uncompress lmbench.tar.Z, uncompress lmbench.tar tar -xf lmbench.tar findings: When the program was compiled as described above on pc10 and pc0, "mhz.c" wasn't compiled properly as the author wished it to be. It was a known bug by LM. This bug would cause in wrong recognition of the processor speed and wrong results for memory latency. To fix this, mhz.c needs to be compiled separately using gcc and move the execuatble mhz to /bin dir.

2) Linpack (by Al Alburto)

compile: gmake

run: all the executables output: to a file (Linapackc.res)

source: ftp://ftp.nosc.mil/pub/aburto/linpackc/

measure: Mflops

note: Results here are sensitive to cache effects and memory speed.

3) HINT

compile: gcc -DDOUBLE -DIINT hint.c hkernel.c -o hint

run: hint output: final diagnosis on screen and details in a file

source: ftp://ftp.scl.ameslab.gov/pub/HINT/serial/

measure: "QUIPS" (QUality Improvement Per Second) as a function of time. "Net QUIPS" is a single number that summarizes the QUIPS over time.

4) Dhrystone

compile: follow instruction in dhry.bat

run: DITTO output: to a file (dhry.res)

source: ftp://ftp.nosc.mil/pub/aburto/dhrystone/

measure: Dhrystones/sec and VAX MIPS rating

note: A synthetic integer program which provides a MIPS rating based upon a 'typical' instruction mix. Very sensitive to optimization and results will appear erratic unless degree of optimization is carefully tracked.

5) Whetstones

compile: g++ -DUNIX whets.c -o whets

run: whets output: to a file(whets.res)

source: ftp://ftp.nosc.mil/pub/aburto/whetstone/

measure: MWIPS, Mflops1, Mflops2, Mflops3, Cosmops, Expmops, Fixpmops, Ifmops, Eqmops