On Mon, 22 Jan 2024 17:23:22 GMT, Raffaello Giulietti <rgiulie...@openjdk.org> wrote:
> AFAIK, IntrinsicCandidate methods are only relevant in JIT compiled code. A > test that checks correctness might not reach the compilation stage, and > execute only in the bytecode interpreter. > > But IMO using the result of `divWord()` as suggested is simple enough and > avoids hidden dependencies. @rgiulietti Perhaps I was not clear enough. I'm not intended to use `Long.divideUnsigned(long, long)` and `Long.remainderUnsigned(long, long)` in the tests, but in the algorithm implementation, instead of using an hard-coded implementation, to use directly the assembly code, although this implies to calculate the approximations of `q` and `r` two times instead of one. ------------- PR Comment: https://git.openjdk.org/jdk/pull/17291#issuecomment-1904488957
