Extremely slooooow __sys_ftruncate?
Steve Kargl
sgk at troutmask.apl.washington.edu
Thu Mar 20 20:46:15 PDT 2008
On Thu, Mar 20, 2008 at 08:02:54PM -0700, Steve Kargl wrote:
> In the process of helping to debug a problem with gcc-4.4.0
> (actually a gfortran problem), I run gprof on the executable.
> The profile shows that __sys_ftruncate is extremely slow.
>
> % cumulative self self total
> time seconds seconds calls ms/call ms/call name
> 85.6 6.05 6.05 51830 0.12 0.12 __sys_ftruncate [2]
> 5.6 6.44 0.40 0 100.00% .mcount (101)
> 1.7 6.56 0.12 51872 0.00 0.00 _lseek [5]
> 1.6 6.67 0.11 52055 0.00 0.00 sigprocmask [6]
> 0.8 6.73 0.06 103687 0.00 0.00 memset [14]
> 0.4 6.76 0.03 488 0.06 0.06 __sys_write [18]
> 0.4 6.79 0.03 0 100.00% formatted_transfer_scalar
>
> time ./z
> 184.21 real 0.98 user 6.57 sys
>
> This program should finish well under 184 seconds. The same program
> and exact same gcc/gfortran source on linux shows
> real 0m0.555s user 0m0.103s sys 0m0.452s
>
> Is __sys_ftruncate known to have performance problems?
>
Well, I tried to kludge together a C example based on the
the Fortran test code and gfortran runtime library.
cc -o z -O -pg -static a.c
time ./z
239.96 real 0.08 user 8.12 sys
% cumulative self self total
time seconds seconds calls ms/call ms/call name
81.4 6.58 6.58 51830 0.13 0.13 _ftruncate [3]
17.1 7.96 1.38 52056 0.03 0.03 _write [4]
0.7 8.01 0.05 0 100.00% .mcount (46)
0.6 8.05 0.05 51830 0.00 0.00 __sys_lseek [5]
0.1 8.06 0.01 1 7.51 7.51 fstat [9]
0.1 8.07 0.01 1 7.51 8027.03 main [1]
0.1 8.07 0.00 51830 0.00 0.00 lseek [13]
Hmmm, ftruncate() mapped to _ftruncate :-/
Here's the program
#include <sys/stat.h>
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
int
main(void) {
int fd, off, i, j;
fd = open("junk", O_RDWR | O_CREAT, S_IRUSR | S_IWUSR);
off = 0;
for (i = 0; i < 5183; i++) {
if (i%23 == 0) printf("i = %d\n", i);
for (j = 0; j < 10; j++) {
write(fd, &off, sizeof(off));
off++;
lseek(fd, off, SEEK_SET);
ftruncate(fd, off);
}
}
return 0;
}
--
Steve
More information about the freebsd-current
mailing list