[cvsnt] Win32 performance

[Tony Hoyle]

On Fri, 17 Jan 2003 13:49:30 +1300, "Martin Rothbaum"
<martin.rothbaum at nzqa.govt.nz> wrote:

For what it's worth this is gprof output for cvsnt running in local
mode on Linux.. there are no obvious bottlenecks that I can see, so
I'd expect the preformance to be comparable:

Flat profile:

Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls   s/call   s/call  name
 46.93      2.14     2.14     7322     0.00     0.00
linevector_delete
 44.30      4.16     2.02     7325     0.00     0.00  linevector_add
  1.54      4.23     0.07      596     0.00     0.00
rcsbuf_valpolish_internal
  1.10      4.28     0.05    76020     0.00     0.00  rcsbuf_getkey
  0.66      4.31     0.03       35     0.00     0.12
apply_rcs_changes
  0.66      4.34     0.03        9     0.00     0.00  linevector_copy
  0.44      4.36     0.02    36062     0.00     0.00  getnode
  0.44      4.38     0.02    32761     0.00     0.00  hashp
  0.44      4.40     0.02     5055     0.00     0.00  dellist
  0.44      4.42     0.02     3811     0.00     0.00  getstr
  0.44      4.44     0.02        9     0.00     0.00  linevector_free
  0.22      4.45     0.01    11402     0.00     0.00  getdelta
  0.22      4.46     0.01     8089     0.00     0.00  findnode_fn
  0.22      4.47     0.01     4901     0.00     0.00  getlist
  0.22      4.48     0.01     2910     0.00     0.00  translate_symtag
  0.22      4.49     0.01     2425     0.00     0.00  fgetentent
  0.22      4.50     0.01     2200     0.00     0.00  AddEntryNode
  0.22      4.51     0.01     1999     0.00     0.00
RCS_reparsercsfile
  0.22      4.52     0.01     1999     0.00     0.00  do_file_proc
  0.22      4.53     0.01     1998     0.00     0.00
free_rcsnode_contents
  0.22      4.54     0.01      571     0.00     0.00  open_file
  0.22      4.55     0.01      158     0.00     0.00  Lock_Cleanup
  0.22      4.56     0.01        3     0.00     1.42  RCS_deltas
  0.00      4.56     0.00   197441     0.00     0.00  xmalloc
  0.00      4.56     0.00    60624     0.00     0.00  xstrdup

(Everything below here was 0.00% of runtime - insignificant)

The linevector_add/delete calls look significant but each iteration
accounts for <0.001% of runtime.  This is the function that is called
to read in the deltas from RCS, so for a large checkout that's going
to be called a lot (although it would be interesting to work out if it
was possible to get the iteration count down).

Tony