Loop optimization

Project Infrastructure LLVM Dev List Archives
1

Hi,

I was wondering if there is an optimization pass in LLVM or Polly that can optimize this given loop:

int main(int argc, char* argv) {

char s = 0;

// Loop is setting s = argv[1][0];

for (int i=0; i<argv[1][0]; i++) {

s++;

}

printf(“s = %c\n”, s);

return 0;

}

into:

int main(int argc, char* argv) {

// Optimized loop

char s = argv[1][0];

printf(“s = %c\n”, s);

return 0;

}

Is there an optimization that can do this?

And if not, maybe someone knows what hinders the optimizer?!

Thanks,

Peter

2

Without testing it I'd say loop idiom recognition (Transforms/Scalar/LoopIdiomRecognize.h) is the one that should kick in and optimize this.

3

Thanks a lot for your help.

Indeed if I compile with -OZ the loop gets replaced with something like this:

  if ( *v3 < 0 )
    v4 = 0;

Is there a reason why it does not get applied with O3 ?
I mean this a huge improvement to the loop.

4

I also noticed that it only works when I compile from C to IR but when I try to compile IR to IR or use opt to optimize BC it fails.
Any idea why ?

5

How do you invoke opt and could you please attach the input file.