This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Gcc 3.1 performance regressions with respect to 2.95.3
> I applied your cross jumping patch, did a full bootstrap and ran the
> regression test suite. There are no regressions.
>
> Here are the results for the stepanov test. Surprisingly, when the -O
> optimization gains the most speed from your patch, resulting in the
> fastest executable. Now, gcc 3.1 has about the same performance as
> gcc 2.95, but at the -O3 level. Why have the additional optimizations
> no net effect in comparison to -O? Do you have an idea what is going wrong?
I think it is the presence of dead code, like it were in 3.0 case.
The dataflow pass should kill it. I will try to find time to benchmark
it this weekend, but I am somewhat overloaded right now.
Honza
>
> Thank you for enhancing the performance. I strongly suggest your patch
> should go in.
>
> Hope this helps,
>
> Peter Schmid
>
>
>
> -O
> test absolute additions ratio with
> number time per second test0
>
> 0 0.45sec 111.11M 1.00
> 1 0.46sec 108.70M 1.02
> 2 0.46sec 108.70M 1.02
> 3 0.45sec 111.11M 1.00
> 4 0.48sec 104.17M 1.07
> 5 0.53sec 94.34M 1.18
> 6 0.50sec 100.00M 1.11
> 7 0.50sec 100.00M 1.11
> 8 0.51sec 98.04M 1.13
> 9 0.46sec 108.70M 1.02
> 10 0.45sec 111.11M 1.00
> 11 0.45sec 111.11M 1.00
> 12 0.48sec 104.17M 1.07
> mean: 0.47sec 105.33M 1.05
>
> Total absolute time: 6.18 sec
>
> Abstraction Penalty: 1.05 was 1.12 without your patch applied.
>
> -O2
> test absolute additions ratio with
> number time per second test0
>
> 0 0.44sec 113.64M 1.00
> 1 0.45sec 111.11M 1.02
> 2 0.47sec 106.38M 1.07
> 3 0.62sec 80.65M 1.41
> 4 0.60sec 83.33M 1.36
> 5 0.56sec 89.29M 1.27
> 6 0.57sec 87.72M 1.30
> 7 0.56sec 89.29M 1.27
> 8 0.56sec 89.29M 1.27
> 9 0.62sec 80.65M 1.41
> 10 0.60sec 83.33M 1.36
> 11 0.62sec 80.65M 1.41
> 12 0.60sec 83.33M 1.36
> mean: 0.56sec 90.01M 1.26
>
> Total absolute time: 7.27 sec
>
> Abstraction Penalty: 1.26 was 1.30.
>
> -O3
> test absolute additions ratio with
> number time per second test0
>
> 0 0.44sec 113.64M 1.00
> 1 0.48sec 104.17M 1.09
> 2 0.48sec 104.17M 1.09
> 3 0.61sec 81.97M 1.39
> 4 0.60sec 83.33M 1.36
> 5 0.56sec 89.29M 1.27
> 6 0.56sec 89.29M 1.27
> 7 0.56sec 89.29M 1.27
> 8 0.56sec 89.29M 1.27
> 9 0.60sec 83.33M 1.36
> 10 0.61sec 81.97M 1.39
> 11 0.60sec 83.33M 1.36
> 12 0.60sec 83.33M 1.36
> mean: 0.56sec 89.99M 1.26
>
> Total absolute time: 7.26 sec
>
> Abstraction Penalty: 1.26 was 1.28.
>
>
>