AMBER 8 PMEMD Benchmark Done During RECOMB04 Meeting
If you are looking for current AMBER 8 PMEMD benchmark, please visit here
I always want to do some AMBER 8 benchmark on Apple's G5 machines, but apparently there is no G5 nearby, *sigh*.
Recently I got a chance to access powermac G5 (sponsored by Apple) in the RECOMB04. I wasn't able to run sander on those G5 email console because my old binaries still need IBM's dynamic libraries. Later I finally found the way to build static binary (Yes, I already tried "-Wl,-static".) with xlf 8.1beta on my powerbook. Here is how:
1. mkdir /opt/ibmcmp/lib_save/ 2. mv /opt/ibmcmp/lib/*.dylib /opt/ibmcmp/lib_save/ 3. echo "int mjhsiehkickme(){}" > test.c 4. xlc -c test.c -o test.o 5. ar rvs test.a test.o; ranlib test.a 6. mv test.a /opt/ibmcmp/lib/libxlomp_ser.a 7. Then you now can make a static binary with xlf. By the way, remember to add this to your shell's environment variables: setenv XLFRTEOPTS "namelist=old:xrf_messages=no"I only tested it on jac and gb_mb because there were too many people eager to use these consoles and I still haven't gotten a chance to use it again.
============================================================================= "jac" == Joint Amber/Charrm DHFR benchmark. This is the protein DHFR, solvated with TIP3 water, in a periodic box. There are 23,558 total atoms, and PME used with a direct space cutoff of 9 Ang. This is the benchmark in benchmarks/jac subdirectory of the Amber 8 distribution. ------------------------------------------------------------------------------- name date CPU OS compiler npcu ps per day ------------------------------------------------------------------------------- RECOMB04 3/31 1.8 Ghz G5 MacOS X 10.3 IBM XLF 8.1b 1 138 (sander) mjhsieh 3/18 400 Mhz G4 MacOS X 10.3 IBM XLF 8.1b 1 25 (sander) mjhsieh 3/18 400 Mhz G4 MacOS X 10.3 IBM XLF 8.1b 1 26 (pmemd) ============================================================================ "gb_mb" == Generalized Born myoglobin simulation. This protein has 2492 atoms, and is run with a 20 Ang. cutoff and a salt concentration of 0.2 M, with nrespa=4 (long range forces computed every 4 steps.) This is the test case in the benchmarks/gb_mb subdirectory of the Amber 8 distribution. ------------------------------------------------------------------------------- node-name CPU OS compiler npcu ps-per-day ------------------------------------------------------------------------------- RECOMB04 3/31 1.8 Ghz G5 MacOS X 10.3 IBM XLF 8.1b 1 184 mjhsieh 3/18 400 Mhz G4 MacOS X 10.3 IBM XLF 8.1b 1 23 ============================================================================ "fix": This is a factor_ix benchmark, from Bob Duke, also a protein solvated with TIP3 water, in a periodic box. There are 90,906 total atoms, and PME used with a direct space cutoff of 8 Ang. This test is in amber8/benchmarks/factor_ix. This uses dt=0.0015, so the conversion is ps-per-day = 129.6/(time-per-step) ------------------------------------------------------------------------------- name date CPU OS compiler npcu ps-per-day ------------------------------------------------------------------------------- mjhsieh 3/18 400 Mhz G4 MacOS X 10.3 IBM XLF 8.1b 1 10 ============================================================================ "hb": This is a hemoglobin benchmark, also a protein solvated with TIP3 water, in a periodic box. There are 44,247 total atoms, and PME used with a direct space cutoff of 8 Ang. This 500-step test is in amber8/benchmarks/hb; it uses a truncted octahedral box, and nrespa=2. ------------------------------------------------------------------------------- name date CPU OS compiler npcu ps-per-day ------------------------------------------------------------------------------- mjhsieh 3/18 400 Mhz G4 MacOS X 10.3 IBM XLF 8.1 1 16