Re: Fw: Questions on -ftest-coverage and gcov

[Tehila Meyzels <TEHILA at il dot ibm dot com>]

gcc-owner@gcc.gnu.org wrote on 10/12/2006 15:14:20:

>
>
> Hi Chris,
>
> Have you looked into gcov.v (and other gcov* files)?
> I think that can help you to figure out things...
>
> For example (quoted from gcov.c). Deeper look can answer your 4th
question:
> /* The bbg file is generated by -ftest-coverage option. The da file is
>    generated by a program compiled with -fprofile-arcs. Their formats
>    are documented in gcov-io.h.  */

In addition, you can see exactly what those flags invoke by grep for:
flag_test_coverage
profile_arc_flag
FYI, you can find the corresponding flags (in the code itself) in
common.opt.

HTH,
Tehila.


>
> Good luck,
> Tehila.
>
>
>
>
>              linuxkaffee@gmx.n
>              et
>              Sent by:
To
>              gcc-owner@gcc.gnu         gcc@gcc.gnu.org
>              .org
cc
>
>
Subject
>              10/12/06 14:50            Questions on -ftest-coverage and
>                                        gcov
>
>
>
>
>
>
>
>
>
>
> Hi,
>
> I've spent couple of hours reading the gcov manual and running
> it with some examples. There are some questions that came up:
>
> 1) What exactly happens when gcc is invoked with "-ftest-coverage"?
> The manual says that basically a program flow graph is constructed
> in order to enable the assignment of profiling information
> to the corresponding program parts. How does the flow graph looks
> like in the .gcno file? Is it a low-level representation of the
> program consisting of basic blocks combined via edges? If so, this
> means that gcc transforms the (here) C source code to an intermediate
> representation (I suppose RTL), right?
>
> 2) How are the line numbers mapped to the corresponding basic
> blocks in the .gcno file? Is gcc proceeding each C statement in the
> source code and generates the corresponding low-level (RTL)
> code and when it notices that a new basic block is required, it
> takes the source code line number and assigns it to the currently
> proceeding basic block?
>
> 3) Question on an concrete example:
> This is the analysed source code listed on the gcov man pages.
>
>         -:    0:Source:tmp.c
>         -:    0:Graph:tmp.gcno
>         -:    0:Data:tmp.gcda
>         -:    0:Runs:1
>         -:    0:Programs:1
>         -:    1:#include <stdio.h>
>         -:    2:
>         -:    3:int
>         -:    4:main (void)
> function main called 1 returned 100% blocks executed 75%
>         1:    5:{
>         1:    6:  int i, total;
>         -:    7:
>         1:    8:  total = 0;
>         -:    9:
>        11:   10:  for (i = 0;
>         1:   10-block  0
>        11:   10-block  1
>        10:   10-block  2
>         -:   11:      i < 10;
>         -:   12:      i++)
>        10:   13:    total += i;
>         -:   14:
>              ...
>         1:   19-block  0
>         -:   20:}
>
> I don't understand the annotations in line 10 that represent a for
> loop. (On purpose, I've written the for-loop statements in separate
> lines.) The statement "i=0" is executed once and belongs to the
> block that also contains code from line 6-9. That's OK. Then,
> the exit test "i<10" is executed which should be placed in a new
> basic block to be able to jump to it. (I've verified this by
> looking at the generated assembly code.) However, the corresponding
> annotations "block 1" are placed before line 11. Why? Same for the
> following two statements "i++" and "total += i" that are also
> placed in the a separate basic block (in the generated assembly file).
> Also here, the annotations "block 2" are placed before the corresponding
> source code lines 12 and 13. Can you explain this to me?
>
> 4) In general, where can I find the source code for work done
> when gcc is invoked with "-ftest-coverage" and "-fprofile-arcs".
> I've downloaded the gcc v4.1 source code from gcc.gnu.org and
> grep'ed for "ftest-coverage" but found merely gcc.c containing
> this string. However, in gcc.c the flow graph (mentioned in 1))
> is not reconstructed.
>
> Thank you very much for your help.
>
> Best regards,
> Chris
> --
> Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen!
> Ideal für Modem und ISDN: http://www.gmx.net/de/go/smartsurfer
>
>