Google To Work On Better PGO Support For LLVM
Some of Google's compiler developers will be working on upstream improvements to LLVM for better supporting Profile Guided Optimizations.
Profile Guided Optimizations (PGO) is the process of using profiling data by the compiler to tune the program's runtime performance, but can mean needing to compile the program twice. PGO can be very effective at improving the performance of software with modern compilers and is well supported by GCC.
LLVM/Clang has been working on PGO support for some time while now Google developers are hoping to do more for improving PGO in LLVM.
For those interested in all of the technical details how Google is looking to improving the LLVM PGO support, see this mailing list post. However, the improvements will likely not come immediately as this is a longer-term work item for them, but they mentioned to Google that PGO is very important. "Internally, we have a large dependency on PGO to get peak performance in generated code."
Profile Guided Optimizations (PGO) is the process of using profiling data by the compiler to tune the program's runtime performance, but can mean needing to compile the program twice. PGO can be very effective at improving the performance of software with modern compilers and is well supported by GCC.
LLVM/Clang has been working on PGO support for some time while now Google developers are hoping to do more for improving PGO in LLVM.
For those interested in all of the technical details how Google is looking to improving the LLVM PGO support, see this mailing list post. However, the improvements will likely not come immediately as this is a longer-term work item for them, but they mentioned to Google that PGO is very important. "Internally, we have a large dependency on PGO to get peak performance in generated code."
3 Comments