Hello,

  Trying to debug performance problems with wapbox I looked into OProfile
output. And some numbers
look a bit high. E.g. isn't computer spending too much time in vfprintf() ?
I looked into source
code and is really fflush() after vfprintf() so needed? I have nothing
against losing few lines of
logs after computer crash.

  Also maybe gwlist_search() and octstr_compare() could be optimized
further? :)

  wapbox/bearerbox is from the latest cvs.

  Mindaugas

# opreport --long-filenames | head -20
CPU: AMD64 processors, speed 2390.33 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit
mask of 0x00 (No unit mask) count 100000
CPU_CLK_UNHALT...|
  samples|      %|
------------------
 58616878 45.1081 /usr/local/sbin/wapbox.140_cvs
 29076280 22.3754 /no-vmlinux
 24184077 18.6106 /lib64/tls/libc-2.3.4.so
  6544707  5.0364 /usr/local/sbin/bearerbox.140_cvs
  5342770  4.1115 /lib64/tls/libpthread-2.3.4.so
  5204879  4.0054 /usr/sbin/squid

# opreport -l /lib64/tls/libc-2.3.4.so | head -10
CPU: AMD64 processors, speed 2390.33 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit
mask of 0x00 (No unit mask) count 100000
samples  %        symbol name
7247305  29.9673  vfprintf
2045210   8.4568  _IO_default_xsputn
1975909   8.1703  memcmp
1786094   7.3854  memchr
1322906   5.4702  _int_malloc
880740    3.6418  __find_specmb
770008    3.1839  strchr

# opreport -l /usr/local/sbin/wapbox.140_cvs | head -20
CPU: AMD64 processors, speed 2390.33 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit
mask of 0x00 (No unit mask) count 100000
samples  %        symbol name
8585828  14.6474  gwlist_search
7424628  12.6664  anonymous symbol from section .plt
7144434  12.1884  octstr_compare
6179058  10.5414  gwlist_get
4110397   7.0123  wap_addr_same


Reply via email to