RFR: 8203352: Improve java implementation of Integer/Long.numberOfLeadingZeros

[Claes Redestad]

On 2018-05-18 16:35, Martin Buchholz wrote:
>
>
> On Fri, May 18, 2018 at 3:51 AM, Claes Redestad 
> <claes.redestad at oracle.com <mailto:claes.redestad at oracle.com>> wrote:
>
>
>     Correctness is checked by existing tests, mainly
>     test/jdk/java/lang/Integer|Long/BitTwiddle.java
>
>
> Do java library tests get run regularly with intrinsics disabled?
> (Perhaps as a result of testing with -Xint?)

Not sure but we might be running these with a few different stress flags 
in later tiers, but in the
standard mode the BitTwiddle tests are more likely to test -Xint and 
C1-compiled modes than the
C2 intrinsic.. I've also manually run exhaustive tests to ensure the new 
implementation matches
the result of the intrinsic version for all possible inputs (rather than 
corner cases + 1000 random
inputs like BitTwiddle does)

/Claes