bug-global
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: cost of popen()


From: Shigio Yamaguchi
Subject: Re: cost of popen()
Date: Mon, 04 Jul 2005 09:01:10 +0900

> > I think that there is no way besides switching the specification by configu
ration variables.

Reviewing your patch, I realized that it was difficult to keep the verbose
message rea-time in an appropriate cost.

> It turned out that that is not right.
> 
> This is a revised version.
> When option -v is specified, source files are processed individually.
> I think that this can satisfy your requirement.

I don't agree. The -v(verbose) option should not change the behaveior.
But I agree with you in the point of leaving the user the alternative.
How about adding new option --noxargs to gtags?

This is the explanation to the users.
o New gtags use xargs(1) to improve the performance of making standard
  format of GTAGS and GRTAGS. It is not used for compact format and GSYMS.
  (right?)
o Under the influence of using xargs, the verbose message is not necessarily
  real-time. If you dislike this behavior, please use the --noxargs option
  of gtags. With the option, gtags doesn't use xargs, and the verbose message
  is real-time.
---end of explanation--

The xargs solution brings a big improvement in a comparatively small
project, but in a large project, the improvement is not so big.
For example, in case of GLOBAL(small), there is about 74% improvement,
but in case of FreeBSD kernel(big), there is only about 7% improvement.
In large project, the output of the verbose message seems to stop.
If there is the alternative, the users will accept the change, I think.

What do you think?
--
Shigio YAMAGUCHI <address@hidden> - Tama Communications Corporation
PGP fingerprint: D1CB 0B89 B346 4AB6 5663  C4B6 3CA5 BBB3 57BE DDA3




reply via email to

[Prev in Thread] Current Thread [Next in Thread]