Re: [PATCH] Add 2-pass mode to gtags

[Shigio YAMAGUCHI]

> Four incremental patch files are attached. They are used as follows. 

Thank you very much.

I made the comparison report.

Program        |Output of time(2)                             |difference with 
current
---------------+----------------------------------------------+-----------------------
current cvs    |128.84 real        84.77 user        16.75 sys| 100% real%
builtin-parser | 97.45 real        63.85 user         6.73 sys|  75% real%      
-25%
external-writer|100.93 real        65.07 user         6.98 sys|  78% real%      
+3%
2pass-parsing  | 75.60 real        44.29 user         4.99 sys|  58% real%      
-17%
1pass-parsing  | 58.58 real        27.79 user         4.17 sys|  45% real%      
-13%
---------------+----------------------------------------------+-----------------------
(Target source code: linux-2.6.21.3)

The builtin-parser, 2pass-pasing and 1pass-parsing are great improvements.

I was not able to recognize significant of external-writer. I tested the 
1pass-parsing
without external-writer too. But the result was same.

58.58 real        27.79 user         4.17 sys   1pass-parsing)
57.64 real        26.45 user         3.88 sys   (1pass-parsing without 
external-writer)

I do not think that external-writer is significant, because:
o It does not improve efficiency too much.
o It requires external programs. It complicates the program itself, and makes 
it difficult
  to measure the performance of the program.

Though it might be effective in a multi-processing environment, other 
processors should
be left for other programs if there is no great reason, I think.

Anyway, these are great improvements.
Thank you again.

--
Shigio YAMAGUCHI <address@hidden>
PGP fingerprint: D1CB 0B89 B346 4AB6 5663  C4B6 3CA5 BBB3 57BE DDA3