RFR: 8343056: C2: Micro-optimize Node lists grow

Aleksey Shipilev shade at openjdk.org
Fri Oct 25 11:18:39 UTC 2024


On Fri, 25 Oct 2024 09:44:23 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> Seen this in Leyden profiles, but it is a generic issue. The profiles show `Node::grow` and `VectorSet::grow` as hot methods. The methods are small, and since they are implementing doubling-the-array scheme, we often exit early without actually growing. Outlining the actual growing part from the capacity check would allow inlining the common case.
> 
> I'll put some performance data in the comments.
> 
> Additional testing:
>  - [x] Linux x86_64 server fastdebug, `tier1`

On various tests on x86_64, this gives me +3% faster runs in -Xcomp scenarios.


## -Xcomp -XX:-TieredCompilation Hello

# Before
  Time (mean ± σ):     617.0 ms ±   3.8 ms    [User: 584.8 ms, System: 31.0 ms]
  Range (min … max):   612.5 ms … 625.7 ms    20 runs
 
# After
  Time (mean ± σ):     596.0 ms ±   5.8 ms    [User: 562.3 ms, System: 32.5 ms]
  Range (min … max):   588.6 ms … 610.6 ms    20 runs

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

PR Comment: https://git.openjdk.org/jdk/pull/21707#issuecomment-2437366117


More information about the hotspot-dev mailing list