RFR: 8346194: Improve G1 pre-barrier C2 cost estimate [v2]

[Roberto Castañeda Lozano]

On Thu, 6 Mar 2025 10:35:07 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:

>> Hi all,
>> 
>>   please review this change that modifies pre-barrier node costs for loop-unrolling to only consider the fast path. The reasoning is similar to zgc (and the new costs as well): only the part of the barrier inlined into the main code stream, as the slow path is laid out separately and does/should not directly affect performance (particularly if there is no marking going on).
>> 
>> There are no differences/impact in performance since the post barrier cost is still very large, which fill be fixed elsewhere.
>> 
>> Testing: gha, perf testing standalone (neither micros nor actual benchmarks give any difference outside of variance), testing with JDK-8342382
>> 
>> Hth,
>>   Thomas
>
> Thomas Schatzl has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Update src/hotspot/share/gc/g1/c2/g1BarrierSetC2.cpp
>   
>   Co-authored-by: Roberto Castañeda Lozano <robcasloz at users.noreply.github.com>

Looks good, thanks for addressing my feedback Thomas!

Reducing the estimated GC barrier size could lead to over-unrolling, which might increase code cache pressure and ultimately affect performance. In practice, it seems that the effect of estimated GC barrier size on total code size is limited though: I studied the impact on C2-generated code size using DaCapo 23 on x64 and aarch64 and it is basically unaffected by this change. In the extreme case of estimating GC barrier size to be 0, the overall code size increase is of about 1% for x64 and 0.5% for aarch64. If each GC barrier (pre and post) is estimated to correspond to 20 nodes, the code size increase is further reduced to only 0.5% for x64 and 0.01% for aarch64. Beyond that, the code size increase becomes statistically insignificant.

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

Marked as reviewed by rcastanedalo (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/23862#pullrequestreview-2666612989