Hi,
I just joined this group. For the record, I built a 2.2.4 kernel with SMP
enabled and it started fine
on my dual PPro box [TYAN S1668 motherboard, 440FX PCI-ISA]. Since there
are no "official" benchmarks
to verify dual processor speed, I wrote the following applet based on the
"Parallel programming mini-howto".
This applet calculates the value of PI given the number of interations (say
100000) and the number of threads.
It would be interesting for others to build and run this, then report some
results for comparison. In particular, I'm interested in BX chipsets and
how shared memory configurations differ. Building the applet:
gcc -O pi.c -o pi -D_REENTRANT -lpthread -lm
./pi 100000 2 // times 2 threads calculating PI using 100000
iterations
Source code:
(See attached file: pi.c)
Here are my timings: (dual PPRO 150 MHz)
Process: 0.029640 secs.
Thread 0: 0.028207 secs.
Thread 1: 0.028114 secs
Estimation of pi is 3.141593
The total process took 29.6 ms. which consists of 2 threads running in
parallel (thread 0 = 28.2 ms, thread 1 = 28.1 ms).
On another note, I am building a dual Celeron PCI card (custom design) and
would like to combine Real-time Linux with SMP into an embedded, bootable
kernel. Anyone ever work on this?
Regards,
Terry Arden
pi.c
