RFR: 8365844: RISC-V: TestBadFormat.java fails when running without RVV
Emanuel Peter
epeter at openjdk.org
Wed Aug 20 13:05:41 UTC 2025
On Wed, 20 Aug 2025 07:56:19 GMT, Dingli Zhang <dzhang at openjdk.org> wrote:
> Hi,
> Can you help to review this patch? Thanks!
>
> We noticed that testlibrary_tests/ir_framework/tests/TestBadFormat.java fails when running tier4 tests on p550.
> The reason for the error is that the Vector test related to badVectorNodeSize requires RVV on riscv, otherwise the expected passing case will fail and cannot match FailCount.
>
> ### Test (fastdebug)
> - [x] Run testlibrary_tests/ir_framework/tests/TestBadFormat.java on k1/k230/sg2042
test/hotspot/jtreg/testlibrary_tests/ir_framework/tests/TestBadFormat.java line 43:
> 41: * @test
> 42: * @requires vm.debug == true & vm.compiler2.enabled & vm.flagless
> 43: * @requires (os.arch != "riscv64" | (os.arch == "riscv64" & vm.cpu.features ~= ".*rvv.*"))
Generally, it would be prefereable to adjust the IR rules. But I'm not sure if that is preferrable here. So I think that this is the right solution.
@chhagedorn This test may fail on other platforms as well that don't have all the required optimizations, such as vectors and others. Should we accept this solution?
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/26855#discussion_r2288098967
More information about the hotspot-compiler-dev
mailing list