[Bug bootstrap/19223] New: --disable-checking doesn't fully disable checking

[Bug bootstrap/19223] New: --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

During a `make bootstrap', I still get:

  gcc -c   -pipe -DENABLE_CHECKING -DENABLE_ASSERT_CHECKING

during stage1 even though I configured with --disable-checking. This is because
gcc/Makefile.in unconditionally does this:

  STAGE1_CHECKING = -DENABLE_CHECKING -DENABLE_ASSERT_CHECKING

Sure, because it's a Makefile variable, it can be overriden on the command line
but I think it wrong for it do this uncondionally. Maybe it'll be switched off
once the 4.0 branch is cut? Dunno.

Thanks

-- 
           Summary: --disable-checking doesn't fully disable checking
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: bootstrap
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: gschafer at zip dot com dot au
                CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-02 02:16 -------
No this is not a bug, this was done on purpose to catch some bugs which we were not catching before.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

------- Additional Comments From gschafer at zip dot com dot au  2005-01-02 02:21 -------
Fairynuff.

As long as someone remembers to switch it off before release, because if not, it
will then clearly be a bug.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-02 02:23 -------
Why do you say it is a bug, because stage2 and stage3 are not compiled with checking.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

------- Additional Comments From gschafer at zip dot com dot au  2005-01-02 02:30 -------
Because it currently makes my bootstraps slow as molasses. Yes, stage 2 and 3
are fine. But I thought released compilers were meant to have checking switched
off completely. If you release the compiler like this then checking of stage1
will still happen, thus unneccessarily slowing down bootstraps.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-02 02:34 -------
(In reply to comment #4)
> Because it currently makes my bootstraps slow as molasses. Yes, stage 2 and 3
> are fine. But I thought released compilers were meant to have checking switched
> off completely. If you release the compiler like this then checking of stage1
> will still happen, thus unneccessarily slowing down bootstraps.

So, what if the boostrap is slowed down, who cares, we don't, we changed this so we can catch bugs 
and because some people don't build non releases this will help find some bugs which we will not find 
right away.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

------- Additional Comments From gschafer at zip dot com dot au  2005-01-02 02:41 -------
Like I said, fairynuff.

But users will definitely notice slower bootstraps, thus further contributing to
the perceived feeling that GCC is forever getting slower. It will be a shame IMHO.

If the extra asserts in stage 1 help catch bugs, great, but this certainly
deviates from past GCC releases where checking was switched off completely. Not
to worry.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2005-01-02 22:14 -------
(In reply to comment #6)

> But users will definitely notice slower bootstraps, thus further
> contributing to the perceived feeling that GCC is forever getting 
> slower. 

This is an urban legend. Please do timings and come back with numbers. 4.0 is 
30-40% faster than 3.4 for C++ code on average at -O0, for instance. And still 
much faster at -O2, and it's got something like 50 new optimizer passes.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

------- Additional Comments From gschafer at zip dot com dot au  2005-01-02 22:27 -------
> This is an urban legend.

I agree with the above comment, but you've completely missed my point.

With gcc-3.4.3, I can `make bootstrap' with --enable-languages=c and it takes
7:25. With gcc-4.0, I do the same and it takes 14:50, exactly double the time!
This is with --disable-checking. The only thing I haven't checked yet is how
much difference it makes if I override STAGE1_CHECKING to be "". I'll report
back on that when I get more time.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[gschafer at zip dot com dot au]

------- Additional Comments From gschafer at zip dot com dot au  2005-01-05 00:19 -------
Here are some timimgs from my test runs. This is all done with:

  --enable-languages=c
  --disable-checking
  make bootstrap

gcc-4.0    14:50
gcc-4.0    12.25    -> with STAGE1_CHECKING manually removed

as you can see, STAGE1_CHECKING results in a 19.5% slowdown!

Here it is again, but this time adding in --disable-mudflap:

gcc-4.0    14.07
gcc-4.0    12.04    -> with STAGE1_CHECKING manually removed

Quicker, but still a 17% slowdown.

Now for comparison purposes, here is gcc-3.4.3

gcc-3.4.3  7.25

Conclusions:

 - STAGE1_CHECKING most definitely slows down C only bootstraps by up to 19.5%
 - C only bootstrap times on GCC-4 have almost exactly doubled since 3.4.3

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

[Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[pinskia at physics dot uc dot edu]

------- Additional Comments From pinskia at physics dot uc dot edu  2005-01-05 00:22 -------
Subject: Re:  --disable-checking doesn't fully disable checking


On Jan 4, 2005, at 7:19 PM, gschafer at zip dot com dot au wrote:

>  - C only bootstrap times on GCC-4 have almost exactly doubled since 
> 3.4.3

Well considering there is almost double the amount of code in 4.0.0
I would not doubt that.  Also I can tell you that on ppc-darwin, we
are faster at bootstrapping because of speedups made to genattrtab
which cuts the time in more than half for that program alone.

-- Pinski



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19223

Re: [Bug bootstrap/19223] --disable-checking doesn't fully disable checking

[Andrew Pinski]

On Jan 4, 2005, at 7:19 PM, gschafer at zip dot com dot au wrote:

>  - C only bootstrap times on GCC-4 have almost exactly doubled since 
> 3.4.3

Well considering there is almost double the amount of code in 4.0.0
I would not doubt that.  Also I can tell you that on ppc-darwin, we
are faster at bootstrapping because of speedups made to genattrtab
which cuts the time in more than half for that program alone.

-- Pinski