rusty@crystal.ARPA (Russel Sandberg) (07/21/83)
We just got our sun up running Berkeley 4.1c with a Fujitsu 80 Mbyte disk. Here are the benchmark results: cc hello.c on SUN 10r 1.1u 2.2s 32% We also ran the benchmark on some of our vaxes. Here are those numbers: VAX/780 rp07 disk, dual memory controllers, 4.1bsd 4r 0.4u 1.4s 45% VAX/750 Fujitsu 160, SI 9900 controller, 4.1bsd 6r 1.0u 1.8s 46% VAX/750 Eagle, SI 9900 controller, 4.1c 6r 1.0u 2.4s 57% VAX/730 Fujitsu 160, SI 6600, 4.1c 12r 2.2u 6.2s 66% Below is another benchmark that we run on virtual memory systems. We tried it out on the National 16032 system at USENIX and it crashed! Here are some numbers from our machines: (only real time is given) 780 (4.1bsd): 0:20 750 (4.1bsd): 0:28 750 (4.1c): 0:30 730 (4.1c): 1:02 SUN (4.1c): 8:56 (slow!) Masscomp: 12:56 note: ran out of memory at 70 (not virtual memory) If people would run this test (time vmbench) on their own systems and mail me the results I will post them to the net. ***************************************************************** vmbench.c ***************************************************************** #include <signal.h> main() { int i, pid; char *a, *malloc(); pid = getpid(); if (fork()) { /* parent waits for last child signal (for time) */ pause(); } for (i=0; i<100; i++) { a = malloc(10000); if (i%10 == 0) printf("%d: a=%x\n", i, a); if (fork()) exit(0); } kill(pid, SIGINT); }