PEs    Trial 1     Trial 2 (times in microseconds)
2         7.28        7.54
3        11.63    13010.92
4        15.01      268
5     21073.14       18.11
6        19.63       18.97
7        17.83    22698.7
8        23.82     4099.2
9        24.72     9334.79
10    39859.54     1591.71
11    39586.55    38599.86
12    30229.56    39799.5
13    28860.67     5684.71
14    38742.38    39956.55
15    43117.49    39952.83
16    26762.15    26230.78
17    39458.78    59911.38
18    46232.86    45327.22
19    45462.67    54762.76
20    59744.57    39657.49
21    54023.2     72209.07
22    78213.27    49108.57
23    69221.4     79534.28
24    71599.46    81920.28
25    45443.61    76059.46
26    50649.56    72402.17
27    39936.37    66906.83
28    64324.64    44803.16
29    59913.22    53071.9
30    44569.65    70844.15
31    55428.5     47399.51
32    54708.82    64102.45
33    34413.99    52384.35
34    66390.67    97134.79
35    68839.46    62719.52
36    43243.98    59132.83 (59.1 ms)

Each entry is an average of 100 iterations. Benchmark used is from OSU MPI microbenchmarks (osu_barrier.c).

These are the configure options I used to cross-compile and build MPICH for TILE-Gx:

mkdir build && cd build
PREFIXROOT=/opt/tilegx
../configure CC=tile-gcc CXX=tile-g++ FC=tile-gfortran F77=tile-gfortran \
    --prefix=$PREFIXROOT/mpich --host=tile \
    --disable-fortran \
    --enable-fast=yes --with-thread-package=posix

Are there any build options that I'm not using that I should be? I'm trying to focus this build specifically for SMP performance.

Thanks,

Bryant Lam