On Tue, 19 Apr 2022 08:41:50 GMT, Jie Fu <ji...@openjdk.org> wrote: > Hi all, > > The Current Vector API doc for `LSHR` is > > Produce a>>>(n&(ESIZE*8-1)). Integral only. > > > This is misleading which may lead to bugs for Java developers. > This is because for negative byte/short elements, the results computed by > `LSHR` will be different from that of `>>>`. > For more details, please see > https://github.com/openjdk/jdk/pull/8276#issue-1206391831 . > > After the patch, the doc for `LSHR` is > > Produce zero-extended right shift of a by (n&(ESIZE*8-1)) bits. Integral only. > > > Thanks. > Best regards, > Jie
I need to think a little more about this. The specification is not accurate and likely requires a CSR. My initial thoughts are I would prefer the operation to retain reference to the succinct definition using the logical right shift operator but we add additional specification explaining the cases for `byte` and `short`, namely that the result is widened to an `int` as if by `(a & ((1 << ESIZE) - 1))` and the result narrowed. That's commonly (at least for the widening part) how `>>>` is used with `byte` and `short`, and i think would be clearer to be explicit in that regard. ------------- PR: https://git.openjdk.java.net/jdk/pull/8291
