More performance explorations

John Rose john.r.rose at oracle.com
Fri Jun 3 23:47:31 PDT 2011 

On Jun 3, 2011, at 7:07 AM, Ola Bini wrote:

> Is there anything I can do to help out with finding this problem?

I can't reproduce the VM crash yet.  Like Christian, I got through problems 1 and 2, on bsd (both 32-bit and 64-bit).

Problem 3 generates huge output, so it's hard to see what's going on.  It is running on my machine but seems to take forever.  FWIW, it got through this line at least:
  - (small numbers)                39.867609000

It would help to have a smaller test case, preferably one that doesn't generate any output before the crash.

-- John

On May 24, 2011, at 8:58 PM, Ola Bini wrote:

> Hi,
> 
> There are at least three problems that are still there. They might be
> connected, or not.
> (I will tell you how to reproduce these at the end)
> 
> I just built a new JVM:
> openjdk version "1.7.0-internal"
> OpenJDK Runtime Environment (build
> 1.7.0-internal-olabini_2011_05_25_08_06-b00)
> OpenJDK Server VM (build 21.0-b09, mixed mode)
> 
> I get that weird missing class error when running my test suite:
> java.lang.NoClassDefFoundError: seph/lang/SephObject
> java.lang.RuntimeException: java.lang.NoClassDefFoundError:
> seph/lang/SephObject
> 	at seph.lang.Runtime.evaluateStream(Runtime.java:132)
> 	at seph.lang.Runtime.evaluateString(Runtime.java:146)
> 	at
> seph.lang.code.BasicSanityTest.recursive_odd_and_even_that_should_blow_the_stack(BasicSanityTest.java:214)
> Caused by: java.lang.NoClassDefFoundError: seph/lang/SephObject
> 	at java.lang.invoke.MethodHandle.invokeExact(MethodHandle.java)
> 	at java.lang.invoke.MethodHandle.invokeExact(MethodHandle.java)
> 	at seph$gen$abstraction$41$even?.argument_0_0(<eval>:7)
> 	at seph.lang.compiler.Bootstrap.intrinsic_if(Bootstrap.java:654)
> 	at seph$gen$abstraction$41$even?.even?(<eval>:7)
> 	at seph$gen$abstraction$39$toplevel.toplevel(<eval>:25)
> 	at seph.lang.Runtime.evaluateStream(Runtime.java:125)
> 
> This is obviously a complete blocker.
> 
> Turning off the ricochet frames hits an NYI error when running the test
> suite.
> 
> I'm still seeing a crash in ricochet frames:
> #
> # A fatal error has been detected by the Java Runtime Environment:
> #
> #  SIGBUS (0xa) at pc=0x0104884f, pid=85043, tid=2953850880
> #
> # JRE version: 7.0
> # Java VM: OpenJDK Client VM (21.0-b09 mixed mode bsd-x86 )
> # Problematic frame:
> # V  [libjvm.dylib+0x4884f]  MethodHandles::ricochet_frame_oops_do(frame
> const&, OopClosure*, RegisterMap const*)+0x12f
> #
> # Failed to write core dump. Core dumps have been disabled. To enable
> core dumping, try "ulimit -c unlimited" before starting Java again
> #
> # An error report file with more information is saved as:
> # /Users/olabini/workspace/seph/hs_err_pid85043.log
> #
> # If you would like to submit a bug report, please visit:
> #   http://bugreport.sun.com/bugreport/crash.jsp
> #
> 
> I've attached the error file.
> 
> The third error (in frame.cpp) only happens on master of seph, but there
> are some other problems with master that means you get lots of weird output.
> 
> In order to reproduce problem 1 and 2:
> git clone git://github.com/seph-lang/seph.git
> git checkout 9075c0f4ffe1adac0657057aee2193f16ad12a43
> (build.xml: remove lines with <jvmarg value="-Xint"/>)
> 
> problem 1:
> ant
>   This should show you one entry of the missing class problem
> 
> problem 2:
> ant jar-notest
> bin/seph bench/bench_arithmetic.sp
>   This should show you problem 2.
>   If you for reference want to see the benchmark run correctly, do
> bin/seph -J-Xint bench/bench_arithmetic.sp
> 
> problem 3:
> git checkout master
> git checkout 99c8f2609d468835390e39b68c73f21cc78e5ab5
> ant clean jar-notest
> bin/seph bench/bench_arithmetic.sp
>   This should show you problem 3. You will also see loads of other
> exceptions, since that point generates slightly bad bytecode. That
> shouldn't make the JVM crash though, I assume - and I've seen the
> frame.cpp should not reach here without seeing those exceptions.
> 
> All of these require that JAVA_HOME points to the Java 7 you want to use
> 
> Cheers
> 
>> 
>> tom
>> 
>> On May 23, 2011, at 7:33 PM, Ola Bini wrote:
>> 
>>> Hi,
>>> 
>>> I'm happy to see that the performance degradation is getting
>>> addressed. I would like to point out that there is still a serious
>>> crash in the machinery too... Have you seen any reason why that
>>> happens?
>>> 
>>> Cheers
>>> 
>>> 
>>> On 2011-05-24 06.11, John Rose wrote:
>>>> On May 23, 2011, at 3:44 PM, Tom Rodriguez wrote:
>>>> 
>>>>>> I'd *love* for intermediate static Java snippits like this
>>>>>> to inline straight through, even if invokeExact would be
>>>>>> megamorphic under normal circumstances...but I don't think
>>>>>> that's the case right now, right?
>>>>> 
>>>>> I haven't been following 292 that closely but it used to be a
>>>>> piece of code that did precisely that.
>>>>> 
>>>>> +        private Object invoke_L0() throws Throwable { + if
>>>>> ((boolean) test.invokeExact()) +                return 
>>>>> target.invokeExact(); +            return
>>>>> fallback.invokeExact(); + }
>>>>> 
>>>>> It looks like it was reworked at some point to use 
>>>>> selectAlternative but the optimizer was never updated to deal
>>>>> with it properly.  It's not particularly hard, we just need to
>>>>> generate code like we would for a bimorphic call site.  The
>>>>> current code expects to either get a constant or something else
>>>>> and doesn't inline if it's something else.  In this case we
>>>>> have a Phi of two constants which we just need to split.
>>>> 
>>>> Yes, it would be a quasi-bimorphic call site keyed from a phi of
>>>> two constants, instead of a profile of two receiver classes.
>>>> 
>>>>> I'm still unclear why you couldn't write your own variant of 
>>>>> guardWithTest and have it work but my knowledge of what's
>>>>> really allowed is somewhat limited.
>>>> 
>>>> Those snippets will inline (I think), but at some point Charlie
>>>> will want to fetch a bit of constant stuff out of an instance
>>>> variable. That will look non-constant to the optimizer, even if
>>>> the instance variable is final and the enclosing object is a
>>>> constant reference. We made sure this happens for
>>>> java.lang.invoke classes, but we haven't extended it yet to user
>>>> code, in part because it would have its own bug tail to work
>>>> through.
>>>> 
>>>> -- John _______________________________________________ mlvm-dev 
>>>> mailing list mlvm-dev at openjdk.java.net 
>>>> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
>>>> 
>>> 
>>> 
>>> -- Ola Bini (http://olabini.com) Ioke - JRuby - ThoughtWorks
>>> 
>>> "Yields falsehood when quined" yields falsehood when quined.
>>> 
>>> _______________________________________________ mlvm-dev mailing
>>> list mlvm-dev at openjdk.java.net 
>>> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
>> 
>> _______________________________________________ mlvm-dev mailing
>> list mlvm-dev at openjdk.java.net 
>> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
>> 
> 
> 
> -- 
> Ola Bini (http://olabini.com)
>  Ioke - JRuby - ThoughtWorks
> 
> "Yields falsehood when quined" yields falsehood when quined.
> 
> <hs_err_pid85367.log>_______________________________________________
> mlvm-dev mailing list
> mlvm-dev at openjdk.java.net
> http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev

More information about the mlvm-dev mailing list