RFR(M): 8231561: [lworld] C2 generates inefficient code for acmp

Tobias Hartmann tobias.hartmann at oracle.com
Thu Oct 10 13:00:12 UTC 2019


Hi Roland,

thanks for looking at this!

On 10.10.19 13:17, Roland Westrelin wrote:
> 
>> http://cr.openjdk.java.net/~thartmann/8231561/webrev.00/
> 
> Don't use need to also change the cmp() implementation so it takes the
> _clear_prop_bits into account? Otherwise you would risk replacing a
> LoadKlass with _clear_prop_bits = true with one with _clear_prop_bits =
> false which could cause a crash. I suppose replacing one with
> _clear_prop_bits = false with one with _clear_prop_bits = true is ok
> though but too complicated to handle.

Good catch, I've added hash/cmp implementations to LoadKlassNode and LoadNKlassNode:
http://cr.openjdk.java.net/~thartmann/8231561/webrev.01/

Thanks,
Tobias


More information about the valhalla-dev mailing list