Was there some discussion about this approach of explicitly disabling LTO this way? IMO the tradeoff of LTO isn't worth it to add this to a couple hundred packages...
Well, according to our experiments, we'll have to disable ~80 packages out of almost 3000 that we have in staging project. From a preliminary statistics, we can save about 5% of size for binaries and debug info will shrink by 20%. Moreover, the LTO binaries are faster and we can prove that for packages like Python, Firefox, or SPEC benchmarks prove that.
Note that this is just a starting point, we would like to first reach openSUSE:Factory and then we'll see how viable idea is that.
Was there some discussion about this approach of explicitly disabling LTO this way? IMO the tradeoff of LTO isn't worth it to add this to a couple hundred packages...
Well, according to our experiments, we'll have to disable ~80 packages out of almost 3000 that we have in staging project. From a preliminary statistics, we can save about 5% of size for binaries and debug info will shrink by 20%. Moreover, the LTO binaries are faster and we can prove that for packages like Python, Firefox, or SPEC benchmarks prove that. Note that this is just a starting point, we would like to first reach openSUSE:Factory and then we'll see how viable idea is that.
@Beineri, @Vogtinator, @adrianSuSE, @alarrosa, @dirkmueller, @llunak: review reminder