[jdk21u-dev] RFR: 8320682: [AArch64] C1 compilation fails with "Field too big for insn" [v2]
Paul Hohensee
phh at openjdk.org
Thu Sep 26 16:58:31 UTC 2024
On Thu, 19 Sep 2024 20:57:19 GMT, Boris Ulasevich <bulasevich at openjdk.org> wrote:
>> Hi all,
>>
>> This is a backport of [JDK-8320682](https://github.com/openjdk/jdk/pull/16951) change to fix C1 internal error caused by adr limit being broken by too large NMethodSizeLimit option value.
>>
>> The backported patch includes a change to the TestC1Globals.java that does not yet exist in the jdk21u codebase. In the mainline the test has four updates:
>> - [1] https://github.com/openjdk/jdk/commit/7df73a23
>> - [2] https://github.com/openjdk/jdk/commit/c36ec2ca
>> - [3] https://github.com/openjdk/jdk/commit/69014cd5 (this change)
>> - [4] https://github.com/openjdk/jdk/commit/27d5f5c2
>>
>> With this change I simply copy version [3] of TestC1Globals.java corresponding to this change in jdk-23. Why not backport [1]+[2]+[3] to have a clean incremental patch?
>> - I do not think [1] fits the backporting rules (not a crash fix, and not very important)
>> - change [3] is stronger than change [2] (once NMethodSizeLimit is limited, no need to gracefully handle large values)
>>
>> Testing: jtreg tier1-3.
>> Without the patch `$ fastdebug/bin/java -XX:NMethodSizeLimit=4M` causes the `guarantee(chk == -1 || chk == 0) failed: Field too big for insn`. With the change the 4M value becomes invalid.
>
> Boris Ulasevich has updated the pull request with a new target base due to a merge or a rebase. The pull request now contains one commit:
>
> Backport 69014cd55b59a0a63f4918fad575a6887640573e
Marked as reviewed by phh (Reviewer).
-------------
PR Review: https://git.openjdk.org/jdk21u-dev/pull/956#pullrequestreview-2331876457
More information about the jdk-updates-dev
mailing list