On Mon, 20 May 2024 19:21:44 GMT, Paul Sandoz <psan...@openjdk.org> wrote:
>> Hi, >> Can you help to review this simple patch? >> Some index check in Byte/ShortVector.fromArray/fromArray0Template seems not >> necessary, could be removed. >> Thanks > > That does not look correct and will only check a prefix indexes. A > `ByteVector` with a shape of 256 bits has 32 lanes, whereas an `IntVector` of > the same shape has 8 lanes. The `mapOffset` array will hold 32 indexes that > need checking, so we need to loop through `mapOffset` array four times. > Thanks @PaulSandoz for comment. I just re-ran the vector api tests with this > patch on x64 and riscv64, but seems no failures triggered. Let me check > further, either I missed something or maybe there is some gap in test to be > filled. More likely a gap in the tests, not sufficiently checking for out of bounds access across the range in the mapOffset array. ------------- PR Comment: https://git.openjdk.org/jdk/pull/18977#issuecomment-2122917109
