This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Code Bloat g++
- To: pfeifer at dbai dot tuwien dot ac dot at
- Subject: Re: Code Bloat g++
- From: "Martin v. Loewis" <martin at loewis dot home dot cs dot tu-berlin dot de>
- Date: Sat, 19 Feb 2000 10:58:22 +0100
- CC: jbuck at possibly dot synopsys dot com, gcc at gcc dot gnu dot org
- References: <Pine.BSF.4.21.0002190301050.22252-100000@deneb.dbai.tuwien.ac.at>
> Hmm, I feel there must be some extraneous stuff here, unless I
> misunderstood:
>
> deneb[112]: wc -l generate.ii
> 29131
>
> deneb[114]: grep LLBB generate.s | wc -l
> 86672
>
> That is, we have three times as many lexical blocks as lines of
> preprocessed code.
I think there is a counting error here: Each LLBB symbol appears twice
- once before the instruction that starts the block, and once in the
debugging information.
Counting lexical blocks is perhaps also easier with
tr -cd '{' <generate.ii | wc -c
Of course, this probably still gives more blocks than appear in
source. Some are apparently redundant, however, there are also good
reasons for others. For example, I guess inline functions result in
new lexical blocks every time they are inlined.
Regards,
Martin