>>225
4.10R(i386)でやってみた。(Opteron240x2)
ここらあたりと比べるといまいちだが、gccでも健闘しているとも言えるかな。
ttp://benchmarks.jp/bench_test/himeno.html
やっぱPGIコンパイラがないのが困るよね。(PGIってLinuxlatorで動くの？？)

$ mpif77 -O3 himenobmtxpr.f
$mpirun -np 2 ./a.out
Sequential version array size
mimax= 257 mjmax= 129 mkmax= 129
Parallel version array size
mimax= 257 mjmax= 129 mkmax= 67
imax= 256 jmax= 128 kmax= 65
I-decomp= 1 J-decomp= 1 K-decomp= 2

Start rehearsal measurement process.
Measure the performance in 3 times.
MFLOPS: 968.11987 time(s): 0.42486 0.00169377949
Now, start the actual measurement process.
The loop will be excuted in 423 times.
This will take about one minute.
Wait for a while.
Loop executed for 423 times
Gosa : 0.00104661949
MFLOPS: 960.013135 time(s): 60.411124
Score based on Pentium III 600MHz : 11.5887632