[jdk17] RFR: JDK-8263321: Regression 8% in javadoc-steady in 17-b11

[Hannes Wallnöfer]

On Mon, 14 Jun 2021 17:51:59 GMT, Vicente Romero <vromero at openjdk.org> wrote:

> just curious, what was the performance improvement after this change?

Running the JavadocSteadyCompilerBench benchmark on my laptop (with other apps running) I get close to 5% improvement, which is similar to what I see in the regression. Before/After change results below.

Result "execute":
  294,600 ±(99.9%) 5,860 ms/op [Average]
  (min, avg, max) = (284,487, 294,600, 336,963), stdev = 10,416
  CI (99.9%): [288,740, 300,460] (assumes normal distribution)

Result "execute":
  281,528 ±(99.9%) 2,316 ms/op [Average]
  (min, avg, max) = (275,671, 281,528, 292,972), stdev = 4,117
  CI (99.9%): [279,212, 283,844] (assumes normal distribution)

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

PR: https://git.openjdk.java.net/jdk17/pull/42