Closed RealMegatron closed 5 years ago
The optimization phases applied and also associated flags to turn them on/off can be found in the LowTier, HighTier, MidTier classes in the org.graalvm.compiler.core.phases package. Note that those flags may change between versions.
There are 7 optimizations mentioned in the paper: 1) Escape Analysis with Atomic Operations; 2) Loop-Wide Lock Coarsening; 3) Atomic-Operation Coalescing; 4) Method-Handle Simplification; 5) Speculative Guard Motion; 6) Loop Vectorization; and 7) Dominance-Based Duplication Simulation.
I know that 6) Loop Vectorization is not available in CE. I'm wondering how about the others? Are there any options to easily turn on/off these optimizations?
Many thanks.