boxing + unboxing a VT is not a no-op ?
tobias.hartmann at oracle.com
Tue Jul 17 12:31:08 UTC 2018
I had a look at the C2 generated code for valuelist_intbox_innervalue_inlined_reduce and it is
essentially the same highly optimized code that is generated for valuelist_intbox_inlined_reduce.
I'm not able to measure a performance difference either. Were you able to reproduce this with the
latest version of the repo? Maybe some of our recent patches fixed this problem as well.
More information about the valhalla-dev