Akira Saito
Akira Saito
I suspect that there are some cases where the thread fails to restore write protection because a C/C++ exception is thrown between `pthread_jit_write_protect_np()` calls. In such situation, the thread will...
I have been trying to reproduce it on internal Grinder, but not successful so far.
Reproduced on internal Grinder jobs (`job/Grinder/26020/`, `job/Grinder/26022/`).
It seems `AbstractVector.castShape` fails to convert `ByteVector` to `ShortVector`. https://github.com/ibmruntimes/openj9-openjdk-jdk18/blob/0c89bcfb32934a82973a74b49c726382ae98f552/test/jdk/jdk/incubator/vector/VectorReshapeTests.java#L737 ``` JIT inline frame: bp = 0x0000FFFEE40080B8, pc = 0x0000FFFF5AD37524, unwindSP = 0x0000FFFEE4007FA0, cp = 0x0000FFFF94495790, arg0EA = 0x0000000000000000, jitInfo...
It seems either of JIT compiled `Byte128Vector.slice(int)` or `AbstractVector.convert0(char, AbstractSpecies)` has a problem. https://github.com/ibmruntimes/openj9-openjdk-jdk18/blob/0c89bcfb32934a82973a74b49c726382ae98f552/src/jdk.incubator.vector/share/classes/jdk/incubator/vector/AbstractVector.java#L371 ``` int origin = shapeChangeOrigin(vsp, rsp, true, part); //System.out.println("*** origin = "+origin+", part = "+part+", lanewise");...
It seems `Byte128Vector.slice(int)` returns zero vector. In `ByteVector.sliceTemplate(int)`, mask for blending is created but this happens to be all false mask when the test fails. https://github.com/ibmruntimes/openj9-openjdk-jdk18/blob/9bdb9425c548ea2917c5b79ecbb5788b86314c00/src/jdk.incubator.vector/share/classes/jdk/incubator/vector/ByteVector.java#L2173 ``` VectorMask blendMask =...
It seems that Linux kernel OOM killer terminates test java process of `jdk_vector_0`. A certain test spends more than 4GB of memory and multiple tests are running in parallel. Hence,...
Is it possible to use jtreg option to set the concurrency to lower value?
I am seeing the failure on the internal machine with 8 cpu cores and 8 GB memory. The test does not fail on another machine with 8 cpu cores and...
Hmm. the concurrency on aarch64 machine (8GB memory) is 3 while it is 1 on x86 machine (ub20x64rt3) with 4GB memory. This might be the reason why the test only...