[13] RFR (S): 8059241: C2: Excessive RemoveUseless passes during incremental inlining

Aleksey Shipilev shade at redhat.com
Sat Jan 26 10:08:53 UTC 2019


On 1/26/19 12:34 AM, Vladimir Ivanov wrote:
> I did some measurements with earlier version [1] and didn't notice any regressions:
>                           BEFORE                 AFTER
> Box2DBench            50.439 ±  5.644       43.549 ±  2.520  ms/op
> CryptoBench            6.972 ±  0.137        7.019 ±  0.163  ms/op
> DeltaBlueBench       630.607 ±  6.784      644.160 ± 11.296  us/op
> EarleyBoyerBench      20.284 ±  0.415       20.735 ±  0.378  ms/op
> GbemuBench            58.731 ±  2.852       59.245 ±  3.936  ms/op
> NavierStokesBench      6.606 ±  0.042        6.628 ±  0.057  ms/op
> PdfJSBench            83.063 ±  4.480       78.909 ±  2.268  ms/op
> RaytraceBench       3583.622 ± 40.646     3578.562 ± 45.962  us/op
> RegexpBench           70.060 ±  2.538       72.189 ±  2.651  ms/op
> RichardsBench        233.414 ±  2.990      233.116 ±  5.641  us/op
> SplayBench           956.145 ± 37.547      917.133 ± 37.245  us/op
> 
> Let me know about your findings.

Oh no, if you did Octane/JMH runs, this is already okay. I would struggle with plain Octane runner
separately.

-Aleksey

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20190126/65965d8c/signature.asc>


More information about the hotspot-compiler-dev mailing list