RFR 8218615 [lworld] C1 intrinsic support for System.arraycopy
Frederic Parain
frederic.parain at oracle.com
Fri Mar 1 20:13:29 UTC 2019
Looks good to me.
Fred
> On Feb 28, 2019, at 23:33, Ioi Lam <ioi.lam at oracle.com> wrote:
>
> https://bugs.openjdk.java.net/browse/JDK-8218615
> http://cr.openjdk.java.net/~iklam/valhalla/8218615-c1-arraycopy-intrinsics.v01/
>
> + If at compile time, we know that at least one of src or dst is a flattened array,
> always take the slow path (normal method invocation of System.arraycopy).
>
> + If src and/or dst may be a flattened array, at runtime, we check if they are
> actually flattened. If so, take the slow path.
>
> Note: unlike C2, C1 never deoptimizes arraycopy. I updated TestArrays.java
> accordingly.
>
> Thanks
> - Ioi
>
More information about the valhalla-dev
mailing list