I am curious to see how wu-blast would perform on a niagara. As a very basic test, I ran the cooltst program on one of my intel blades that is currently running wu-blast.
cooltst.ksh v2.3.2
System configs
Host Name: XXXXXXXXX
Chip Arch: P15
OS: Linux 2.6.11.4-21.9-smp, 64-bit
[Virtual]Processors: 4
processor 0 on-line 3600 MHz
processor 1 on-line 3600 MHz
processor 2 on-line 3600 MHz
processor 3 on-line 3600 MHz
=> Running fpstat… (08/23/06 20:43:02)
=> Runtime (fpstat) ended… (08/23/06 20:43:15)
=> Running procls.sh… (08/23/06 20:43:15)
=> Completed data collection… (08/23/06 20:43:16)
=> Analysis of collected data… (08/23/06 20:43:16)
PROCESSES FOR ANALYSIS:
Process Summary:
125 Active Processes, 224 LWPs
PID NLWP CMD
7786 1 blastp (119.0%)
7731 1 blastp (103.0%)
7754 1 blastx (99.6%)
7791 1 perl (58.0%)
7742 6 run_search (11.1%)
7788 1 SystemCommandPr (6.0%)
11239 31 java (5.2%)
22999 31 java (5.2%)
7783 1 SystemCommandPr (5.0%)
23559 30 java (4.7%)
#
Huh ? According to the test, more than 100% of operations are floating point. This is in contrast to what Lawrence finds with ncbi-blast…
Technorati Tags: niagara, t1000, t2000, sun, blast, ncbi, wu-blast,niagaracmt