Re: regarding optimization options in phase ordering

[Richard Guenther <richard.guenther@gmail.com>]

On Fri, Aug 7, 2009 at 1:50 PM, pms<pmshiva@gmail.com> wrote:
>
> Hi,
>     We've a problem here. we were trying to use cc1 with & without -O
> option to verify the optimizations happening in our sample code. these r the
> list of outputs after each compilation
> without -O
> pms@Shiva:~/Desktop/Compilers/GCC/build/test$ ls
> 1.c                1.c.011t.ehopt        1.c.038t.release_ssa
> 1.c.001t.tu        1.c.012t.eh           1.c.123t.optimized
> 1.c.003t.original  1.c.013t.cfg          1.c.125t.blocks
> 1.c.004t.gimple    1.c.014t.cplxlower0   1.c.126t.final_cleanup
> 1.c.006t.vcg       1.c.015t.veclower     1.c.203t.statistics
> 1.c.007t.useless   1.c.021t.cleanup_cfg  1.s
> 1.c.010t.lower     1.c.023t.ssa          a.out
>
> with -O
>
> 1.c                           1.c.051t.ccp2         1.c.085t.sink
> 1.c.001t.tu                   1.c.052t.forwprop2    1.c.086t.loop
> 1.c.003t.original             1.c.054t.alias        1.c.087t.loopinit
> 1.c.004t.gimple               1.c.055t.retslot      1.c.088t.copyprop4
> 1.c.006t.vcg                  1.c.056t.phiprop      1.c.089t.dceloop1
> 1.c.007t.useless              1.c.057t.fre          1.c.090t.lim
> 1.c.010t.lower                1.c.058t.copyprop2    1.c.093t.sccp
> 1.c.011t.ehopt                1.c.059t.mergephi2    1.c.094t.empty
> 1.c.012t.eh                   1.c.061t.dce1         1.c.099t.ivcanon
> 1.c.013t.cfg                  1.c.062t.cselim       1.c.104t.cunroll
> 1.c.015t.veclower             1.c.063t.ifcombine    1.c.107t.ivopts
> 1.c.021t.cleanup_cfg          1.c.064t.phiopt1      1.c.108t.loopdone
> 1.c.023t.ssa                  1.c.066t.ch           1.c.111t.reassoc2
> 1.c.024t.early_optimizations  1.c.068t.cplxlower    1.c.113t.dom2
> 1.c.025t.einline2             1.c.069t.sra          1.c.114t.phicprop2
> 1.c.026t.copyrename1          1.c.070t.copyrename3  1.c.115t.cddce2
> 1.c.027t.ccp1                 1.c.071t.dom1         1.c.117t.dse2
> 1.c.028t.forwprop1            1.c.072t.phicprop1    1.c.118t.forwprop4
> 1.c.029t.addressables1        1.c.073t.dse1         1.c.119t.phiopt3
> 1.c.030t.esra                 1.c.074t.reassoc1     1.c.121t.copyrename4
> 1.c.031t.copyprop1            1.c.075t.dce2         1.c.122t.uncprop
> 1.c.032t.mergephi1            1.c.076t.forwprop3    1.c.123t.optimized
> 1.c.033t.cddce1               1.c.077t.phiopt2      1.c.124t.nrv
> 1.c.034t.sdse                 1.c.078t.objsz        1.c.125t.blocks
> 1.c.036t.switchconv           1.c.079t.ccp3         1.c.126t.final_cleanup
> 1.c.037t.profile              1.c.080t.copyprop3    1.c.203t.statistics
> 1.c.038t.release_ssa          1.c.081t.fab          1.s
> 1.c.048t.addressables2        1.c.082t.sincos       a.out
> 1.c.049t.copyrename2          1.c.083t.crited
>
> But here, we tried to see the differences, until 1.c.027t.ccp1, the output
> for the following source 1.c
> #include
> int main()
>
> {
>        int a=5;
>        int b;
>        b=a;
>        printf("the number is :%d",b);
> }
>
> was 1.c.026t.copyrename1
> ;; Function main (main)
>
> main ()
> {
>  int b;
>  int a;
>
> :
>  a_2 = b_1(D);
>  return;
>
> }
> but in 1.c.027t.ccp1, the output doesnot contain the actual assignment b=a.
> ;; Function main (main)
>
> main ()
> {
>  int b;
>  int a;
>
> :
>  return;
>
> }
>
> We want to know, without b=a, how is it generating the final code for b=a

Nothing.  Because it's a dead statement.

Richard.

> Kindly help
>
> --
> View this message in context: http://www.nabble.com/regarding-optimization-options-in-phase-ordering-tp24863416p24863416.html
> Sent from the gcc - Dev mailing list archive at Nabble.com.
>
>