LLVM Is On The Path To Getting Faster
LLVM contributor Mikhail Zolotukhin has issued a report about the continued evolution of LLVM and its impact on the compiler's performance.
He's found that in the past few months LLVM is getting faster, which is a good sign as the months prior to that LLVM was getting slower.
He explained in his mailing list report, "Many areas of LLVM have been examined and improved since then: InstCombine, SCEV, APInt implementation, and that resulted in almost 10% improvement compared to January compiler. I remeasured compile time data for CTMark tests and annotated the biggest changes."
There still though are open opportunities for making LLVM even faster with some of the areas to work on include a known bits cache, SCEV, forming LCSSA, unnecessary pass dependencies, and libcxx changes. Read more in his mailing list post.
He's found that in the past few months LLVM is getting faster, which is a good sign as the months prior to that LLVM was getting slower.
He explained in his mailing list report, "Many areas of LLVM have been examined and improved since then: InstCombine, SCEV, APInt implementation, and that resulted in almost 10% improvement compared to January compiler. I remeasured compile time data for CTMark tests and annotated the biggest changes."
There still though are open opportunities for making LLVM even faster with some of the areas to work on include a known bits cache, SCEV, forming LCSSA, unnecessary pass dependencies, and libcxx changes. Read more in his mailing list post.
4 Comments