On the previous episode of "Solving in Parallel"... >From Nist HQ, Dr. Guyer reported his Pysparse / Trilinos comparison results...
> ...running examples/phase/anisotropy.py (*with* viewers) > for 10 steps > > solver 100x100 500x500 1000x1000 > > pysparse 5.75 65.8 287 > Trilinos w 1 proc 12.6 176 844 > Trilinos w 2 proc 13.7 134 710 Meanwhile, here at the Chicken of the Sea Think Tank: Machine Info: o CPUs (/proc/cpuinfo helpfully reports): 4-processors -- Dual Core Opteron Processor 280s, running at 2411.111 MHz, . o kernel (/proc/version says): Linux version 2.6.26-2-amd64... o memory: 8,200,116k Eddie brought up a point about the compiler: we compiled trilinos, OpenMPI, and all the support libs that didn't come with the OS (basically a Debian release), specifically many of the numeric libs, with GCC 4.3.2. Continuing with Dr. Guyer's anisotropy example, here are our times for his 3 mesh sizes (same 10 steps, but no viewer) solver 100x100 500x500 1000x1000 ---------------------------------------------------------------------------- pysparse 2.4 43 171 seconds trilinos 1 proc 4.5 103 480 trilinos 2 proc 2.9 54 263 trilinos 4 proc 3.3 37 192 Note the near equivalence of the trilinos-1 to pysparse ratio, between Guyer's result and the above, in the case of the 1000 by 1000 mesh: 2.9 and 2.8. It is our results for both 2 and 4 processors working on the 500x500 mesh that seem "in- consistent" in this little table. As Alice once noted, "Curiousier and curiousier"...
