Closed Nanotwerp closed 1 year ago
@Nanotwerp Can you give some explanations for the changes you made?
The basic changes are just refactoring so that the patch can apply. The LTO flag changes were made because I thought that they would just be redundant to keep, but I'm doubting that upon further review.
If this page is up to date on defaults, GCC's default LTO partitioning value of -flto-partition=balanced
compiles units in parallel (similar to Clang's ThinLTO), with the downside of having less max potential performance due to having less cross-module optimizations. `-flto-partition=one, on the other hand, operates like Clang's Full LTO.
-fdevirtualize-at-ltrans
seems to still be a good option for aggressive optimization too, with its only downside seeming to be that it can "significantly increase the size of the streamed data", whatever that means!
You added an unnecessary context line, you don't need more than 3. And I think we should keep the compiler flags, unless we have a good reason to remove them. It doesn't take too much time to link wineserver on my system.
You added an unnecessary context line, you don't need more than 3. And I think we should keep the compiler flags, unless we have a good reason to remove them. It doesn't take too much time to link wineserver on my system.
I agree! Is this new commit better?
@Nanotwerp now the commit is essentially empty
@Nanotwerp now the commit is essentially empty
Well...
This is very embarrassing! I think that I was previously using another patch that conflicted with the LTO patch, and blamed the patches failing on the wrong patch! Sorry for wasting your time!
@Nanotwerp Can you give some explanations for the changes you made?