compiler/loopopts/TestUnrollLimitPreciseType.java#test1 fails on 32-bit platforms with Unknown VM option 'MaxVectorSize=64'

[smlambert]

Seen in July triage, compiler/loopopts/TestUnrollLimitPreciseType.java#test1 fails for x86-32_windows and arm_linux.

#5441 (comment)

Do not find any upstream issue but presume since this option is not recognized by either platform that the test should be excluded on the 2 platforms.

[smlambert]

Ideally this testcase exclusion would be reported upstream (in which case our problem lists can be changed to refer to the upstream issue and this one be closed) and/or added to upstream problem lists, in which case we can remove the exclusion from our problemlists, since we use both ours and upstream problem lists when executing test material.

[jiekang]

I will write a JBS report for this.

[jiekang]

Though first I will see what happened in 11+ compared to 8...

[jiekang]

This test is in hotspot_jre which only runs for jdk 8 and 9:

		<versions>
			<version>8</version>
			<version>9</version>
		</versions>

aqa-tests/openjdk/playlist.xml

Line 499 in 6c93937

<version>8</version>

[jiekang]

Also noting for history, this test was introduced via backports of https://bugs.openjdk.org/browse/JDK-8303466

[sophia-guo]

Though first I will see what happened in 11+ compared to 8...

July release the test was disabled for jdk11+.

Grinder with hotspot_custom shows the test will be skipped with jdk11+ on win32 , passed on other platform
https://ci.adoptium.net/view/Test_grinder/job/Grinder/10727/ - jdk11 win32
https://ci.adoptium.net/view/Test_grinder/job/Grinder/10728/ - jdk17 win32
https://ci.adoptium.net/view/Test_grinder/job/Grinder/10729/ -jdkk11 win64

The test will not be skipped and failed with Unrecognized VM option 'MaxVectorSize=64' with jdk8 win32. Passed with jdk8 linux-x64
https://ci.adoptium.net/view/Test_grinder/job/Grinder/10735/console
https://ci.adoptium.net/view/Test_grinder/job/Grinder/10736/

If 32 bits cannot work with MaxVectorSize=64 it should be skipped on 32bits. Looks like for jdk11+ @requires vm.compiler2.enabled skips win32 platform while jdk8 doesn't support property [vm.compiler2.enabled](https://github.com/adoptium/jdk8u/blob/master/test/jtreg-ext/requires/VMProps.java)

There will be two solution for JBS reports. One is update jdk8 VMProps.java or add @requires vm.bits != "32" to test https://github.com/adoptium/jdk8u/blob/master/hotspot/test/compiler/loopopts/TestUnrollLimitPreciseType.java#L29

[jiekang]

Excellent, thank you for the analysis @sophia-guo I will take this info and open a JBS to try to have an appropriate upstream fix.

[jiekang]

Upstream JBS: https://bugs.openjdk.org/browse/JDK-8338125

[sophia-guo]

Closed in favour of https://bugs.openjdk.org/browse/JDK-8338125 and #5494