[vector] java doc for iota()

Yang Zhang (Arm Technology China) Yang.Zhang at arm.com
Wed Nov 13 10:41:45 UTC 2019


Hi Sandhya

>There is mention of partial wrapping in java doc of fromValues in VectorShuffle.java:
>     *
>     * <p> For each shuffle lane, where {@code N} is the shuffle lane
>     * index, the {@code N}th index value is validated
>     * against the species {@code VLENGTH}, and (if invalid)
>     * is partially wrapped to an exceptional index in the
>     * range {@code [-VLENGTH..-1]}.
Why the exceptional index is in the range [-VLENGTH..-1] in partially wrapping? Is it meaningful?

In VectorShuffle, Internal representation reorder is byte data. But for AArch64 SVE, maximum length of byte vector is 256(2048/8 bytes).  If [-VLENGTH..-1] of reorder is unused, we can reinterpret them as [128, 255] in AArch64 backend.

Regards
Yang



More information about the panama-dev mailing list