Performance comparison: Dacapo benchmarks: tiered vs standard

[Albert Noll]

Hi all,
I compared to performance of TieredCompilation vs. the standard 
compilation using the Dacapo benchmarks.
The results are attached to this mail.

machine:
sc14a501

Executive summary:
The benchmarks show a significantly reduced startup time with 
TieredCompilation enabled.
Steady state performance is equal or better with TieredCcompilation 
compared to standard compilation.
One exception: jython benchmark. For jython, TieredCompilation takes 
until run 20 to come close
to the performance of standard compilation. The reason for this behavior 
is to be determined.

Chart explanation:
x-axis: number of runs of the benchmark executed in the same JVM instance.
y-axis: benchmark execution time in milliseconds.

no_tiered (standard compilation) and tiered (TieredCompilation enabled) 
are executed in
seperate JVM instances.

Best,
Albert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dacapo_tiered_vs_standard.pdf
Type: application/pdf
Size: 31944 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20130503/59785538/dacapo_tiered_vs_standard-0001.pdf