RFR: 8265262: CITime - 'other' incorrectly calculated
Tobias Hartmann
thartmann at openjdk.java.net
Tue May 18 14:34:39 UTC 2021
On Mon, 17 May 2021 16:36:26 GMT, Nils Eliasson <neliasso at openjdk.org> wrote:
> This CR fixes a few issues with the CITIme output for C2:
>
> 1) The other category for _t_optimize is not removing time spent in _t_vector
>
> 2) Some of the _t_incrInline sub counters is called from different contexts - calculating 'other' from total time spent in _t_incrInline expects that the counter usage is strictly hierarchical.
>
> 3) I've placed the non-hierarchical counters in braces.
>
> 4) Code Installation is a part of Code Emission (_t_output). Indentation fixed.
>
> 5) Moved "renumber live" after "Vector" so that they appear in order.
>
> 6) Added sub counters "shorten branches" and "fill buffer" to "Code Emission" phase, and added an other category. Before more than 50% of time in Code Emission was unaccounted for, now it's less than 25%.
>
> Please review,
> Best regards,
> Nils Eliasson
Looks good to me.
-------------
Marked as reviewed by thartmann (Reviewer).
PR: https://git.openjdk.java.net/jdk/pull/4065
More information about the hotspot-compiler-dev
mailing list