I think those tests cover all interesting corner cases, so the only way
I see it can be improved is to make it more exhaustive (say generate a
large random sample of tests every run). Do you feel that is needed?
/Claes
On 2019-04-09 01:35, Joseph D. Darcy wrote:
Should any additional cases be added to
test/jdk/java/lang/Math/DivModTests.java to cover the new implementation?
Thanks,
-Joe
On 4/5/2019 10:21 AM, Claes Redestad wrote:
On 2019-04-05 17:41, Andrew Haley wrote:
On 4/5/19 2:44 PM, Claes Redestad wrote:
Testing: tier1-2, all Math tests run locally, -prof perfasm
verification
on the provided microbenchmark.
Looks good.
Thanks!
I've kicked the tyres on AArch64, and it looks like a useful
optimization. The
gains when the divisor is constant (a common case) are modest but
worthwhile.
Thanks for trying it out and glad to hear it helps on AArch64 as well.
/Claes
