Re: Optimization techniques and undefined behavior

From	Bart <bc@freeuk.com>
Newsgroups	comp.compilers
Subject	Re: Optimization techniques and undefined behavior
Date	2019-05-01 12:40 +0100
Organization	virginmedia.com
Message-ID	<19-05-003@comp.compilers> (permalink)
References	(3 earlier) <19-04-037@comp.compilers> <19-04-039@comp.compilers> <19-04-042@comp.compilers> <19-04-045@comp.compilers> <19-04-049@comp.compilers>

Show all headers | View raw

On 30/04/2019 14:48, David Brown wrote:
> On 29/04/2019 19:15, Bart wrote:
>> On 29/04/2019 16:08, David Brown wrote:
>>> On 29/04/2019 01:31, Bart wrote:
>>
>>>> then you don't want the compiler being clever about overflow.
>>>
>>> I /do/ want a result consistent with a single expression, or splitting
>>> up the expression.
>>
>> Then the choice is between both ways giving you 1500000000, or both
>> giving you -647483648.
>
> Let me repeat - I do not care what the results are here.  I don't care
> if they are consistent with each other.  I don't care if they change
> between runs of the compiler.  I don't care if the result is a pink
> umbrella.

So, there's a bug in the program, an inadvertent overflow. But rather
than help in discovering that bug (such as giving the wrong results)
gcc (and its clone Clang) conveniently pretend that such bugs cannot
exist, and use that to give their code an unfair advantage:

     #include <stdio.h>

     int main(void) {
         int x,y,z;

         x=1000000000;
         z=0;

         for (int i=0; i<1000000000; ++i) {
             y=x*3/3;
             z+=y;
             ++x;
         }

         printf("%d\n",x);
         printf("%d\n",y);
         printf("%d\n",z);
     }

Optimising compilers that don't take advantage of that undefined
behaviour give timings here of 1.25 seconds (msvc) and 1.9 seconds
(pelles c), with some taking much longer.

The two that do, give timings of 0.22 seconds (gcc) and 0.05 seconds
(clang).

However, there are a couple of problems: (1) they give different
results; (2) they cheated by not doing a billion multiplies and divides.

Note this example also includes UB due to z+=y line, but I'm only
interested in the bottom 32 bits, albeit signed, as a kind of checksum
to compare with other compilers.

For that purpose, I need the final z value to be -101627306, which will
match the same 32-bit arithmetic across languages** and in assembly.

I don't need it to be the mathematically correct 1499999999500000000,
which seems to me what you'd like it to be. gcc/clang-O3 give me
1565039360 which is neither one nor the other.

-------------------------------

(** This is the above program auto-translated from C to one of my
languages, which is sort of interesting, this being a compiler group.

Normally this is just for to help in viewing torturous C source code as
the C semantics are not translated. But tweaked with the int32() cast to
match C's intermediate calculations (usually 64-bit here), this actually
works:

     import clib

     global function main()int32 =
         var int32 x
         var int32 y
         var int32 z
         var int32 i

         x := 1000000000
         z := 0
         i := 0
         while i<1000000000 do
             y := int32(x*3)/3
             z +:= y
             ++x
             ++i
         od
         printf("%d\n",x)
         printf("%d\n",y)
         printf("%d\n",z)
         return 0
     end

     proc start =
         stop main()
     end

The results match those of the non-gcc/non-clang C compilers (apart from
speed which is poor).)

Back to comp.compilers | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

Re: Optimization techniques David Brown <david.brown@hesbynett.no> - 2019-04-25 21:58 +0200
  Re: Optimization techniques Kaz Kylheku <847-115-0292@kylheku.com> - 2019-04-26 00:18 +0000
    Re: Optimization techniques David Brown <david.brown@hesbynett.no> - 2019-04-28 23:49 +0200
      Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-04-29 00:31 +0100
        Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-04-29 17:08 +0200
          Re: Optimization techniques and undefined behavior Christian Gollwitzer <auriocus@gmx.de> - 2019-04-29 18:10 +0200
            Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-04-30 14:46 +0200
              Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-01 13:53 +0100
                Re: Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-02 11:29 +0100
                Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-03 00:48 +0100
                Re: Optimization techniques and undefined behavior Martin Ward <martin@gkc.org.uk> - 2019-05-03 10:52 +0100
                Re: Optimization techniques and undefined behavior George Neuner <gneuner2@comcast.net> - 2019-05-04 17:44 -0400
                Re: Bounds checking, Optimization techniques and undefined behavior George Neuner <gneuner2@comcast.net> - 2019-05-05 17:10 -0400
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-06 08:14 +0200
                Re: Optimization techniques and undefined behavior Gene Wirchenko <genew@telus.net> - 2019-05-11 22:25 -0700
                Re: not a lot of memory, was Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-03 12:45 +0100
                Re: Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-03 13:29 +0100
                Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-03 23:10 +0100
                Re: Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-04 10:45 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-05 11:14 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Hans-Peter Diettrich <DrDiettrich1@netscape.net> - 2019-05-05 20:44 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior Hans-Peter Diettrich <DrDiettrich1@netscape.net> - 2019-05-06 10:15 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 11:04 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior "Nuno Lopes" <nuno.lopes@ist.utl.pt> - 2019-05-07 22:38 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-08 01:14 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-08 09:55 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior "Derek M. Jones" <derek@_NOSPAM_knosof.co.uk> - 2019-05-08 19:08 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-08 01:42 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-08 10:16 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-09 01:15 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-09 21:56 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-08 10:03 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-09 09:19 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior Kaz Kylheku <847-115-0292@kylheku.com> - 2019-05-10 03:38 +0000
                Re: Bounds checking, Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-08 14:37 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Christopher F Clark <christopher.f.clark@compiler-resources.com> - 2019-05-06 05:05 -0400
                Re: Bounds checking, Optimization techniques and undefined behavior George Neuner <gneuner2@comcast.net> - 2019-05-05 17:38 -0400
                Re: Bounds checking, Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-06 13:07 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 14:01 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-06 01:15 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior Andy Walker <anw@cuboid.co.uk> - 2019-05-06 14:40 +0100
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 15:05 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-08 10:18 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior Jan Ziak <0xe2.0x9a.0x9b@gmail.com> - 2019-05-06 05:39 -0700
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 15:42 +0200
                Re: Bounds checking, Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-06 16:32 +0200
                Re: Optimization techniques and undefined behavior George Neuner <gneuner2@comcast.net> - 2019-05-04 17:59 -0400
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-02 16:51 +0200
                Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-02 20:04 +0100
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-03 17:23 +0200
                Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-03 21:10 +0100
                Re: Optimization techniques and undefined behavior Martin Ward <martin@gkc.org.uk> - 2019-05-06 13:25 +0100
                Re: Optimization techniques and undefined behavior "Derek M. Jones" <derek@_NOSPAM_knosof.co.uk> - 2019-05-06 16:32 +0100
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 16:03 +0200
                Re: Optimization techniques and undefined behavior Martin Ward <martin@gkc.org.uk> - 2019-05-08 13:16 +0100
                Re: Optimization techniques and undefined behavior George Neuner <gneuner2@comcast.net> - 2019-05-08 15:13 -0400
                Re: Optimization techniques and undefined behavior "Robin Vowels" <robin51@dodo.com.au> - 2019-05-07 01:22 +1000
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 16:05 +0200
                Re: Optimization techniques and undefined behavior Christian Gollwitzer <auriocus@gmx.de> - 2019-05-02 22:22 +0200
          Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-04-29 18:15 +0100
            Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-04-30 15:48 +0200
              Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-01 12:40 +0100
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-02 17:27 +0200
                Re: Optimization techniques and undefined behavior Bart <bc@freeuk.com> - 2019-05-02 18:59 +0100
                Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 16:16 +0200
              Re: Optimization techniques and undefined behavior Martin Ward <martin@gkc.org.uk> - 2019-05-02 14:54 +0100
      Re: Optimization techniques and runtime checks Hans-Peter Diettrich <DrDiettrich1@netscape.net> - 2019-04-29 22:36 +0200
        Re: Optimization techniques and runtime checks David Brown <david.brown@hesbynett.no> - 2019-05-07 16:29 +0200
          Re: Optimization techniques and runtime checks Hans-Peter Diettrich <DrDiettrich1@netscape.net> - 2019-05-08 02:27 +0200
            Re: Optimization techniques and runtime checks David Brown <david.brown@hesbynett.no> - 2019-05-08 10:31 +0200
              Re: Optimization techniques and runtime checks Hans-Peter Diettrich <DrDiettrich1@netscape.net> - 2019-05-08 22:50 +0200
              Re: Optimization techniques and runtime checks "Robin Vowels" <robin51@dodo.com.au> - 2019-05-11 19:26 +1000
              Re: Optimization techniques and runtime checks Gene Wirchenko <genew@telus.net> - 2019-05-11 22:43 -0700
                Re: Optimization techniques and runtime checks David Brown <david.brown@hesbynett.no> - 2019-05-12 20:17 +0200
          Re: Optimization techniques and runtime checks Bart <bc@freeuk.com> - 2019-05-08 14:58 +0100
            Re: Optimization techniques and runtime checks David Brown <david.brown@hesbynett.no> - 2019-05-08 23:02 +0200
              Re: Optimization techniques and runtime checks Bart <bc@freeuk.com> - 2019-05-09 18:28 +0100
                Re: Optimization techniques and runtime checks David Brown <david.brown@hesbynett.no> - 2019-05-09 22:07 +0200
      Re: Optimization techniques Gene Wirchenko <genew@telus.net> - 2019-04-30 18:24 -0700
        Re: Optimization techniques David Brown <david.brown@hesbynett.no> - 2019-05-01 09:20 +0200
          Re: Optimization techniques Kaz Kylheku <847-115-0292@kylheku.com> - 2019-05-02 17:40 +0000
          Re: Optimization techniques and error detection Gene Wirchenko <genew@telus.net> - 2019-05-03 10:16 -0700
          Re: Optimization techniques "Robin Vowels" <robin51@dodo.com.au> - 2019-05-07 01:42 +1000
  Re: Optimization techniques Kaz Kylheku <847-115-0292@kylheku.com> - 2019-04-26 02:26 +0000
    Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-04-29 00:12 +0200
      Re: Optimization techniques and undefined behavior Kaz Kylheku <847-115-0292@kylheku.com> - 2019-05-02 17:18 +0000
        Re: Optimization techniques and undefined behavior David Brown <david.brown@hesbynett.no> - 2019-05-07 16:38 +0200

csiph-web