8 ms·
> Performance results on SPECint 2017 show that we can compile LLVM-IR 8--24x faster than LLVM -O0 while being on-par in terms of run-time performance Wait - i
by vlovich123 1y ago
> Performance results on SPECint 2017 show that we can compile LLVM-IR 8--24x faster than LLVM -O0 while being on-par in terms of run-time performance
Wait - it’s 8-24x faster than O0 while producing code on par with O3???
- ummonk 1y agoNo, the generated code is on par with LLVM -O0. It's slower than LLVM -O1, never mind LLVM -O3.
- wiz21c 1y agoI guess it doesn't include linking ? (which takes quite some time)
- andyferris 1y agoOne thing I never understood in this context here (fast JIT/debug builds/hot reloads/-O0) is why you would need much static linking. Generally your modules are going to have a DAG relationship. Even code inside a large compilation unit could potentially be factored out (automatically) into smaller modules. Could you not just generate a bunch of small dynamically linked libraries? Would the system dynamic loader become the speed bottleneck? Even if so, wouldn't reloading just a portion of the DAG in a hot-reload context be much faster than linking everything beforehand?