Good news, bad news

Charles Oliver Nutter headius at headius.com
Fri May 20 13:21:33 PDT 2011


Meh, false alarm. This is just the default empty Object[] assigned to
all objects. I still can't explain the perf degradation.

Here's the full assembly for fib_ruby: https://gist.github.com/983721

It seems like stuff is inlining, so I have no idea why performance is
so terrible.

- Charlie

On Fri, May 20, 2011 at 2:59 PM, Charles Oliver Nutter
<headius at headius.com> wrote:
> Another update...I managed to turn off the ricochet frames and have it
> still work (???). The with-ricochet mode is faster, but still 2-3x
> slower than no indy at all (using JRuby's IC), and 4-5x slower than
> invokedynamic was recently...
>
> HOWEVER...an assembly dump I just did contains some very odd code that
> might indicate a problem in my code. Hold the presses on digging into
> it on your end until I can explain this. (this = apparently standing
> up an instance variable table for Fixnum objects...*definitely* should
> not be happening).
>
> 0x02868f2a: mov [eax+0x1C], ebx  ;*invokespecial <init>
>                                        ; -
> org.jruby.RubyBasicObject::<init>@1 (line 218)
>                                        ; -
> org.jruby.RubyObject::<init>@2 (line 116)
>                                        ; -
> org.jruby.RubyNumeric::<init>@2 (line 114)
>                                        ; -
> org.jruby.RubyInteger::<init>@2 (line 95)
>                                        ; -
> org.jruby.RubyFixnum::<init>@5 (line 112)
>                                        ; -
> org.jruby.RubyFixnum::newFixnum at 25 (line 173)
>                                        ; -
> org.jruby.RubyFixnum::op_minus at 34 (line 395)
>                                        ; -
> org.jruby.runtime.callsite.MinusCallSite::call at 24 (line 16)
>  0x02868f2d: mov       ebx, [esp+0x14]
>  0x02868f31: mov       ecx, [ebx+0xBC]
>  0x02868f37: mov       edx, eax
>  0x02868f39: shr       edx, 9
>  0x02868f3c: mov       [edx+0x394000], 0x00  ;*putfield varTable
>                                        ; -
> org.jruby.RubyBasicObject::<init>@8 (line 84)
>                                        ; -
> org.jruby.RubyObject::<init>@2 (line 116)
>                                        ; -
> org.jruby.RubyNumeric::<init>@2 (line 114)
>                                        ; -
> org.jruby.RubyInteger::<init>@2 (line 95)
>                                        ; -
> org.jruby.RubyFixnum::<init>@5 (line 112)
>                                        ; -
> org.jruby.RubyFixnum::newFixnum at 25 (line 173)
>                                        ; -
> org.jruby.RubyFixnum::op_minus at 34 (line 395)
>                                        ; -
> org.jruby.runtime.callsite.MinusCallSite::call at 24 (line 16)
>  0x02868f43: mov       edx, a 'java/lang/Class' = 'org/jruby/RubyBasicObject'
>                                        ;   {oop(a 'java/lang/Class' =
> 'org/jruby/RubyBasicObject')}
>  0x02868f48: mov       ebx, [edx+0x68]  ;*getstatic NULL_OBJECT_ARRAY
>                                        ; -
> org.jruby.RubyBasicObject::<init>@5 (line 84)
>                                        ; -
> org.jruby.RubyObject::<init>@2 (line 116)
>                                        ; -
> org.jruby.RubyNumeric::<init>@2 (line 114)
>                                        ; -
> org.jruby.RubyInteger::<init>@2 (line 95)
>                                        ; -
> org.jruby.RubyFixnum::<init>@5 (line 112)
>                                        ; -
> org.jruby.RubyFixnum::newFixnum at 25 (line 173)
>                                        ; -
> org.jruby.RubyFixnum::op_minus at 34 (line 395)
>                                        ; -
> org.jruby.runtime.callsite.MinusCallSite::call at 24 (line 16)
>  0x02868f4b: mov       [eax+0x10], ebx
>  0x02868f4e: lock add  [esp], 0x00     ;*putfield varTable
>                                        ; -
> org.jruby.RubyBasicObject::<init>@8 (line 84)
>                                        ; -
> org.jruby.RubyObject::<init>@2 (line 116)
>                                        ; -
> org.jruby.RubyNumeric::<init>@2 (line 114)
>                                        ; -
> org.jruby.RubyInteger::<init>@2 (line 95)
>                                        ; -
> org.jruby.RubyFixnum::<init>@5 (line 112)
>                                        ; -
> org.jruby.RubyFixnum::newFixnum at 25 (line 173)
>                                        ; -
> org.jruby.RubyFixnum::op_minus at 34 (line 395)
>                                        ; -
> org.jruby.runtime.callsite.MinusCallSite::call at 24 (line 16)
>
> On Fri, May 20, 2011 at 8:10 AM, Charles Oliver Nutter
> <headius at headius.com> wrote:
>> On Fri, May 20, 2011 at 1:07 AM, Ola Bini <ola.bini at gmail.com> wrote:
>>>> Try -XX:+UnlockDiag* -XX:-UseRicochetFrames for differential testing.
>>>>
>>>> If you can find a microbenchmark which shows differential slowdown, we can debug it.
>>>
>>> When trying the above all my tests fail with this:
>>>  java.lang.RuntimeException: java.lang.InternalError: NYI
>>
>> Same here.
>>
>> Reproducing it is pretty easy...
>>
>> * Clone JRuby: git clone https://github.com/jruby/jruby.git
>> * Build JRuby: cd jruby ; ant
>> * Run benchmarks: bin/jruby bench/bench_fib_recursive.rb 10 35
>>
>> There are many benchmarks under the bench dir, almost all of which
>> show this degradation. You can disable invokedynamic use in JRuby by
>> passing -Xcompile.invokedynamic=false, as seen here:
>>
>>
>> ~/projects/jruby ➔ jruby -Xcompile.invokedynamic=false
>> bench/bench_fib_recursive.rb 5 35
>> WARNING: Both jruby-complete.jar and jruby.jar are present in the
>> 'lib' directory. Will use jruby.jar
>> 9227465
>>  1.698000   0.000000   1.698000 (  1.645000)
>> 9227465
>>  1.352000   0.000000   1.352000 (  1.352000)
>> 9227465
>>  1.364000   0.000000   1.364000 (  1.364000)
>> 9227465
>>  1.344000   0.000000   1.344000 (  1.344000)
>> 9227465
>>  1.336000   0.000000   1.336000 (  1.336000)
>>
>> ~/projects/jruby ➔ jruby -Xcompile.invokedynamic=true
>> bench/bench_fib_recursive.rb 5 35
>> WARNING: Both jruby-complete.jar and jruby.jar are present in the
>> 'lib' directory. Will use jruby.jar
>> 9227465
>>  3.182000   0.000000   3.182000 (  3.120000)
>> 9227465
>>  3.019000   0.000000   3.019000 (  3.019000)
>> 9227465
>>  3.195000   0.000000   3.195000 (  3.195000)
>> 9227465
>>  3.163000   0.000000   3.163000 (  3.163000)
>> 9227465
>>  3.656000   0.000000   3.656000 (  3.656000)
>>
>> I'm sure something's just broken...this benchmark was easily almost 2x
>> faster than the non-invokedynamic version just a few days ago, and
>> it's now 2-3x slower than non-indy logic.
>>
>> I'm going to be online all day, standing by to help in any way I can.
>> I did do a quick LogCompilation last night and the main thing I
>> noticed was that while it seemed like handles are inlining, the
>> eventual DMH does not inline. That would explain some of the perf
>> degradation, I think...but hopefully not all of it.
>>
>> Whatever you need me to run or whatever information...let me know.
>>
>> ALSO...I got the same error as Ola when running JRuby with a
>> standalone jar file:
>>
>> ~/projects/jruby ➔ java -jar lib/jruby-complete.jar bench/bench_fib_recursive.rb
>> MethodHandle.java:-1:in `invokeExact': java.lang.NoClassDefFoundError:
>> org/jruby/runtime/ThreadContext
>>        from MethodHandle.java:-1:in `invokeExact'
>>        from bench/bench_fib_recursive.rb:7:in `method__0$RUBY$fib_ruby'
>>        from bench/bench_fib_recursive.rb:7:in `method__0$RUBY$fib_ruby'
>>        from bench/bench_fib_recursive.rb:7:in `method__0$RUBY$fib_ruby'
>>        from bench/bench_fib_recursive.rb:7:in `method__0$RUBY$fib_ruby'
>>
>> As in Ola's case, ThreadContext is a core JRuby class. I do *not* get
>> this error when JRuby is loaded from the bootstrap classloader, so I
>> imagine there's some mix-up with generated bytecode and the
>> classloaders that that bytecode gets loaded into.
>>
>> - Charlie
>>
>


More information about the mlvm-dev mailing list