Sun May 24 18:16:00 GMT 2015
I have done some experiments with optimization flags enabled by
different O flag levels.
I took optimizations enabled by O[1..3] and tried to enable them all at
once without calling
O flag levels. I have found out that improvements I got by using all
flags used by the
particular O flag level are not even close to what I would get by
explicitly using O flag levels.
What is the reason for not reaching the same level of performance of O
I did the following tests:
1. Enable optimizations specified in the Optimize Options manual of the
version of gcc I'm using on my system.
2. Enabled optimizations that were named to be enabled by "-Q -O[1..3]
In both cases, I have reached much slower performance compared to the
one reached by calling O[1..3].
In my tests I have reached speedup of below 2%, while equivalent O flag
would give me 40% speedup.
More information about the Gcc-help