NO

Author Topic: optimization bug?  (Read 2891 times)

icube

  • Guest
optimization bug?
« on: October 17, 2013, 02:10:31 PM »
Code: [Select]
#include <stdio.h>
#include <stdlib.h>

int main(int argc, char** argv) {
    int svc = 0;
    int dsvc = 0;
    //int clc = 0;
    int buffer[] = {0, 1, 2, 2, 2, 3, 4, 6, 1};
    for(int i = 0; i < 9; i++) {
            int sv, dsv;
            sv = buffer[i];
            dsv = sv - svc;
            int ddsv = dsv - dsvc;
            //printf("  %d %d %d \n", svc, sv, ddsv);
            if(ddsv > -1 && ddsv < 1) {
                printf("  %d\n", ddsv);
                //clc ++;
            } /*else {
                if(clc > 0) {
                    printf("[%d]\n", clc);
                }
                clc = 0;
            }*/
            svc = sv;
            dsvc = dsv;
    }
    return 0;
}

-std:C11 -Tx86-coff -Ob1 -fp:precise -W1 -Gd
output:
Code: [Select]
0
0
0
0

-std:C11 -Tx86-coff -Ot -Ob1 -fp:precise -W1 -Gd
output:
Code: [Select]
0
0
-1
0
0


Pelles C 7.00.350 win64

RSRC0009.DLL: Version 7.00.1
SUPPORT64.DLL: Version 7.00.0
CFORMAT64.DLL: Version 7.00.12
FDIFF64.DLL: Version 7.00.2
PORC64.DLL: Version 7.00.15
POBR64.DLL: Version 7.00.1
SQLITE364.DLL: Version 3071100
POCC.EXE: Version 7.00.18
POASM.EXE: Version 7.00.3
POLINK.EXE: Version 7.00.3
IDESPAWN64.EXE: Version 7.00.1

neo313

  • Guest
Re: optimization bug?
« Reply #1 on: October 17, 2013, 05:01:27 PM »
I get the first result (0,0,0,0) for any optimisation .
Pelles C 7.00.355 , Windows 7 32 bit

RSRC0009.DLL: Version 7.00.1
SUPPORT.DLL: Version 7.00.0
CFORMAT.DLL: Version 7.00.12
FDIFF.DLL: Version 7.00.3
PORC.DLL: Version 7.00.15
POBR.DLL: Version 7.00.1
SQLITE3.DLL: Version 3071100
POCC.EXE: Version 7.00.25
POASM.EXE: Version 7.00.3
POLINK.EXE: Version 7.00.3
IDESPAWN.EXE: Version 7.00.1


Interestingly i got a similar compiler bug as well today. It only appears when using any kind of optimization and is similar to a bug i previously posted, this time it is a global pointer to a struct.
Unfortunately my code is too complex to replicate, and smaller examples didn't show the bug.

Does anyone have any suggestion what should i do in this case? I really like this IDE and I have learned a Lot with it, but i really don't think coding around compiler bugs is optimal. Is it time to switch PellesC for something else?

It turns out i had a bug in MY code. (⌒_⌒;) My code was fixed, bug is still here; now I'm 100% it's the optimization bug.
« Last Edit: October 18, 2013, 01:56:58 AM by neo313 »

Offline frankie

  • Global Moderator
  • Member
  • *****
  • Posts: 2113
Re: optimization bug?
« Reply #2 on: October 18, 2013, 03:43:50 PM »
Yes it is definetly an optimization bug.
For some reason the compiler simplifies the comparison on line:
Code: [Select]
if((ddsv > -1) && (ddsv < 1)) {generating a single and simple jump above (ja) statement  :'(.
As a workaround declare volatile the variable ddsv
Code: [Select]
            int volatile ddsv = dsv - dsvc;

As general rule always try to compile your code without any optimization, if it works there could be an optimization bug.
You can then try to remove it disabling optimization for a specific function. You can do this by enclosing function in #pragmas
Code: [Select]
            #pragma optimize(none)    //Disable optimization
            int myfunc(....)    //function
            #pragma optimize()    //Restore command line optimizations
« Last Edit: October 18, 2013, 03:50:34 PM by frankie »
"It is better to be hated for what you are than to be loved for what you are not." - Andre Gide