From shade at openjdk.org Thu Sep 8 09:36:15 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 09:36:15 GMT Subject: RFR: CODETOOLS-7903290: jcstress: Inline Counter.record(R) method Message-ID: Currently, jcstress uses a "convenience" method on a hot path, which delegates to the real one. This relies on JIT compilers to inline it through. We don't have to rely on this, and can call the full method directly. ------------- Commit messages: - Fix Changes: https://git.openjdk.org/jcstress/pull/120/files Webrev: https://webrevs.openjdk.org/?repo=jcstress&pr=120&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903290 Stats: 18 lines in 3 files changed: 0 ins; 10 del; 8 mod Patch: https://git.openjdk.org/jcstress/pull/120.diff Fetch: git fetch https://git.openjdk.org/jcstress pull/120/head:pull/120 PR: https://git.openjdk.org/jcstress/pull/120 From shade at openjdk.org Thu Sep 8 10:09:22 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 10:09:22 GMT Subject: RFR: CODETOOLS-7903291: jcstress: Optimize jcstress trap handling Message-ID: <1U5ORFLSIy6gtRQgJR1TGv3DF1EwgDj-61yALaR6D6U=.d47a12ac-1867-409c-9b00-dfcd61ca0eda@github.com> Currently, jcstress does `jcstress_sink(...)` method to let early NP-checks for field accesses persist. But this is not required: we can just store the value in the local variable, and the NP-checking effects would persist. This saves cycles calling the method in interpeter mode. ------------- Commit messages: - Fix Changes: https://git.openjdk.org/jcstress/pull/121/files Webrev: https://webrevs.openjdk.org/?repo=jcstress&pr=121&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903291 Stats: 14 lines in 1 file changed: 0 ins; 6 del; 8 mod Patch: https://git.openjdk.org/jcstress/pull/121.diff Fetch: git fetch https://git.openjdk.org/jcstress pull/121/head:pull/121 PR: https://git.openjdk.org/jcstress/pull/121 From shade at openjdk.org Thu Sep 8 10:20:27 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 10:20:27 GMT Subject: RFR: CODETOOLS-7903292: jcstress: Allocation profiling is not available on older JDKs Message-ID: [CODETOOLS-7903216](https://bugs.openjdk.org/browse/CODETOOLS-7903216) introduced a regression where allocation profiling was ruled to be unavailable on older JDKs. ------------- Commit messages: - Fix Changes: https://git.openjdk.org/jcstress/pull/122/files Webrev: https://webrevs.openjdk.org/?repo=jcstress&pr=122&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903292 Stats: 1 line in 1 file changed: 0 ins; 0 del; 1 mod Patch: https://git.openjdk.org/jcstress/pull/122.diff Fetch: git fetch https://git.openjdk.org/jcstress pull/122/head:pull/122 PR: https://git.openjdk.org/jcstress/pull/122 From shade at openjdk.org Thu Sep 8 14:20:08 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 14:20:08 GMT Subject: Withdrawn: CODETOOLS-7903292: jcstress: Allocation profiling is not available on older JDKs In-Reply-To: References: Message-ID: <5HPx-oEv7piwZZfuv9fhGRKGmcWNbouS5--XcEUjFOI=.e09d98c3-47b6-4e11-8fdc-ed0ed9aef54c@github.com> On Thu, 8 Sep 2022 10:14:08 GMT, Aleksey Shipilev wrote: > [CODETOOLS-7903216](https://bugs.openjdk.org/browse/CODETOOLS-7903216) introduced a regression where allocation profiling was ruled to be unavailable on older JDKs. This pull request has been closed without being integrated. ------------- PR: https://git.openjdk.org/jcstress/pull/122 From duke at openjdk.org Thu Sep 8 14:20:15 2022 From: duke at openjdk.org (duke) Date: Thu, 8 Sep 2022 14:20:15 GMT Subject: git: openjdk/jcstress: 3 new changesets Message-ID: <51694c51-0ea8-4b67-b20a-52363647fdf8@openjdk.org> Changeset: 5e7ed13e Author: Aleksey Shipil?v Committer: GitHub Date: 2022-09-08 16:15:31 +0000 URL: https://git.openjdk.org/jcstress/commit/5e7ed13e918b4afbe947ba8333f39809064845cf 7903290: jcstress: Inline Counter.record(R) method ! jcstress-core/src/main/java/org/openjdk/jcstress/infra/processors/JCStressTestProcessor.java ! jcstress-core/src/main/java/org/openjdk/jcstress/util/Counter.java ! jcstress-core/src/test/java/org/openjdk/jcstress/util/CounterTest.java Changeset: e01c5f83 Author: Aleksey Shipil?v Committer: GitHub Date: 2022-09-08 16:15:52 +0000 URL: https://git.openjdk.org/jcstress/commit/e01c5f8341cec302eb615f37a9a4aaf4bdf92667 7903291: jcstress: Optimize jcstress trap handling ! jcstress-core/src/main/java/org/openjdk/jcstress/infra/processors/JCStressTestProcessor.java Changeset: d1187759 Author: Aleksey Shipil?v Committer: GitHub Date: 2022-09-08 16:16:21 +0000 URL: https://git.openjdk.org/jcstress/commit/d118775943666d46ca48a50f21b4e07b9ec1f7ed 7903292: jcstress: Allocation profiling is not available on older JDKs ! jcstress-core/src/main/java/org/openjdk/jcstress/vm/AllocProfileSupport.java From shade at openjdk.org Thu Sep 8 14:22:04 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 14:22:04 GMT Subject: Withdrawn: CODETOOLS-7903291: jcstress: Optimize jcstress trap handling In-Reply-To: <1U5ORFLSIy6gtRQgJR1TGv3DF1EwgDj-61yALaR6D6U=.d47a12ac-1867-409c-9b00-dfcd61ca0eda@github.com> References: <1U5ORFLSIy6gtRQgJR1TGv3DF1EwgDj-61yALaR6D6U=.d47a12ac-1867-409c-9b00-dfcd61ca0eda@github.com> Message-ID: On Thu, 8 Sep 2022 10:03:46 GMT, Aleksey Shipilev wrote: > Currently, jcstress does `jcstress_sink(...)` method to let early NP-checks for field accesses persist. But this is not required: we can just store the value in the local variable, and the NP-checking effects would persist. This saves cycles calling the method in interpeter mode. This pull request has been closed without being integrated. ------------- PR: https://git.openjdk.org/jcstress/pull/121 From shade at openjdk.org Thu Sep 8 14:22:05 2022 From: shade at openjdk.org (Aleksey Shipilev) Date: Thu, 8 Sep 2022 14:22:05 GMT Subject: Withdrawn: CODETOOLS-7903290: jcstress: Inline Counter.record(R) method In-Reply-To: References: Message-ID: On Thu, 8 Sep 2022 09:27:32 GMT, Aleksey Shipilev wrote: > Currently, jcstress uses a "convenience" method on a hot path, which delegates to the real one. This relies on JIT compilers to inline it through. We don't have to rely on this, and can call the full method directly. This pull request has been closed without being integrated. ------------- PR: https://git.openjdk.org/jcstress/pull/120