RFR(XXS) : 8205490 : update ProblemList-graal

[Igor Ignatyev]

here is the latest webrev w/ more tests problem listed -- http://cr.openjdk.java.net/~iignatyev//8205490/webrev.02/index.html

-- Igor

> On Jun 21, 2018, at 4:29 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
> 
> On 6/21/18 2:57 PM, Igor Ignatyev wrote:
>>> On Jun 21, 2018, at 12:49 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com <mailto:vladimir.kozlov at oracle.com>> wrote:
>>> 
>>> I would suggest to add all tests to Graal's problem list which use -XX:CompileCommand=compileonly,compiler.whitebox.SimpleTestCaseHelper::*
>>> 
>>> We can't run test which use CompileOnly and expect that Graal will compile specified methods in time. We can add Graal's and JVMCI's methods to CompileOnly list. Or we should make general exception for Graal's and JVMCI's methods.
>> adding these methods to CompileOnly list in each test doesn't sound like a right thing to do, and making general exception for Graal's and JVMCI's methods doesn't either as it will make it impossible/harder to exclude graal/jvmci methods from compilation should it be necessary. adding a special diagnostic-only flag to keep graal/jvmci methods compilable regardless compileOnly might be a good enough compromise.
> 
> Good idea. I agree with diagnostic flag.
> 
>>> 
>>> 
>>> I found next tests are missing in list:
>>> 
>>> compiler/tiered/LevelTransitionTest.java
>>> compiler/whitebox/DeoptimizeAllTest.java
>>> compiler/whitebox/DeoptimizeMethodTest.java
>>> compiler/whitebox/ForceNMethodSweepTest.java:
>>> compiler/whitebox/GetNMethodTest.java
>>> compiler/whitebox/IsMethodCompilableTest.java
>>> compiler/whitebox/LockCompilationTest.java
>>> compiler/whitebox/SetDontInlineMethodTest.java
>>> compiler/whitebox/SetForceInlineMethodTest.java
>> AFAIK, we haven't seen these tests (except DeoptimizeAllTest.java which I have already added to 2nd version of webrev) failing w/ Graal as JIT, but yes they seem to be affected by this problem as well, will add them to the problem list.
> 
> Thanks,
> Vladimir
> 
>>> 
>>> May be there are more other tests.
>>> 
>>> Thanks,
>>> Vladimir
>>> 
>>> On 6/21/18 11:51 AM, Igor Ignatyev wrote:
>>>> http://cr.openjdk.java.net/~iignatyev//8205490/webrev.00/index.html
>>>>> 14 lines changed: 11 ins; 0 del; 3 mod;
>>>> Hi all,
>>>> compiler/intrinsics/mathexact/sanity tests are known to fail when run w/ Graal as JIT, this patch adds them into graal specific problem list file.
>>>> webrev: http://cr.openjdk.java.net/~iignatyev//8205490/webrev.00/index.html
>>>> JBS: https://bugs.openjdk.java.net/browse/JDK-8205490
>>>> Thanks,
>>>> -- Igor

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20180621/19f73efa/attachment.html>