search for: narchiv

...st more code which causes live ranges to expand and increases spills. Nothing really scary here, and in many ways validates the core premise -- this does actually remove barriers to the mid-level optimizer. I've also found one case where stack coloring appears to just fail for no reason: "NArchive::N7z::GetStringForSizeValue(unsigned int)" in 7zHandler.cpp has exactly the same sequence of instructions after the change but 3x larger stack. But this appears to be rare and mostly relating to functions with already small stack sizes. I've filed PR28821 to track this. Anyways, while cl...

Chris added alloca merging in the inliner a looooong time ago, 2009. The reason he added it was because at the time we didn't do stack coloring and without it we had serious stack size problems in LLVM. Since then, a few rather important things have changed: - We have reasonably powerful stack coloring support in the backend based on lifetime markers - Clang (the primary frontend I'm