RFR: 8277906: Incorrect type for IV phi of long counted loops after CCP

Wed Dec 1 13:01:22 UTC 2021

On Wed, 1 Dec 2021 10:35:32 GMT, Roland Westrelin <roland at openjdk.org> wrote:

> This failure occurs because the iv phi of a long counted loop has the
> wrong type after CCP. That happens because, during CCP, after the type
> of the limit of the counted loop is updated, the type of the iv phi is
> not recomputed. The fix is to apply to long counted loops the logic
> that already exists for int counted loop.

Looks good to me but what about this code?

https://github.com/openjdk/jdk/blob/37ff7f3b66eaa74d62d6a93f2f34ec744db21834/src/hotspot/share/opto/phaseX.cpp#L1573-L1575

-------------

PR: https://git.openjdk.java.net/jdk/pull/6632