Binary Optimization and Layout Tool - BOLT optimizes placement of instructions in memory to reduce CPU execution time by 2 percent to 15 percent.
Binary Optimization and Layout Tool - BOLT optimizes placement of instructions in memory to reduce CPU execution time by 2 percent to 15 percent.
Most of this goes over my head; does anyone know what “all compilers” means? They mention Clang and LLVM, but what about, say, Python, Rust, .Net, or Go? I assume it’s not those, because otherwise the headline would be “Facebook breakthrough speeds up all computers by 3-8%” or something like that.
Most binaries fit in CPU cache. Very large binaries (such as those at facebook) do not, and are slower as a result.
By modifying a binary to optimize for code size instead of speed, more of the binary fits into cache and there are fewer trips to main memory.
This improves overall performance for binaries that don’t fit into cache, while reducing it for binaries which would have fit anyways.