Fedora 33 Moving Closer To LTO-Optimizing Packages
Going back to last year Fedora has been working to enable link-time optimizations by default for their packages. That goal wasn't achieved for Fedora 32 but for Fedora 33 this autumn they still have chances of marking that feature off their TODO list.
LTO'ing the Fedora package set can offer not only performance advantages but in some cases smaller binaries as well. This is all about applying the compiler optimizations at link-time on the binary as a whole for yielding often sizable performance benefits and other optimizations not otherwise possible. LTO is great as we often show in benchmarks, especially in the latest GCC and LLVM Clang compilers.
Red Hat's Jeff Law has published a report on the latest Fedora 33 mass rebuild attempt when using LTO. Some bugs were uncovered to little surprise, but progress is being made on isolating those bugs for what may be issues with the code themselves or otherwise compiler problems. Law commented, "the immediate actions are to find a near term resolution for the LTO issues which I've assigned to myself in [Bugzilla]. There aren't that many and I'm confident we'll be able to close them out in a reasonable timeframe."
Current F33 package failures are outlined here.
So with some luck, Fedora 33 will be able to ship this fall with the vast majority of packages being LTO-optimized. The Fedora LTO plans remain outlined on this Wiki page. Hopefully we will be seeing more Linux distributions employ LTO and other optimizations by default moving forward considering the pretty good shape these days for LTO in GCC and Clang.
LTO'ing the Fedora package set can offer not only performance advantages but in some cases smaller binaries as well. This is all about applying the compiler optimizations at link-time on the binary as a whole for yielding often sizable performance benefits and other optimizations not otherwise possible. LTO is great as we often show in benchmarks, especially in the latest GCC and LLVM Clang compilers.
Red Hat's Jeff Law has published a report on the latest Fedora 33 mass rebuild attempt when using LTO. Some bugs were uncovered to little surprise, but progress is being made on isolating those bugs for what may be issues with the code themselves or otherwise compiler problems. Law commented, "the immediate actions are to find a near term resolution for the LTO issues which I've assigned to myself in [Bugzilla]. There aren't that many and I'm confident we'll be able to close them out in a reasonable timeframe."
Current F33 package failures are outlined here.
So with some luck, Fedora 33 will be able to ship this fall with the vast majority of packages being LTO-optimized. The Fedora LTO plans remain outlined on this Wiki page. Hopefully we will be seeing more Linux distributions employ LTO and other optimizations by default moving forward considering the pretty good shape these days for LTO in GCC and Clang.
12 Comments