[aarch64-port-dev ] Benchmarks for RewriteBytecodes and RewriteFrequentPairs optimizations

Yumin Qi yumin.qi at oracle.com
Wed May 13 23:07:42 UTC 2015


specjbb2000 and specjbb2005:
	-UseSharedSpace 	+UseSharedSpace 	
Benchmark 	Samples 	Mean 	Stdev 	Samples 	Mean 	Stdev 	Diff %
jetstream 	5 	165.81 	1.04 	20 	163.83 	1.87 	-1.194
scimark 	5 	343.32 	0.99 	20 	336.85 	1.14 	-1.885
specjbb2000 	5 	212125.65 	1923.19 	20 	206717.28 	2233.34 	-2.550
specjbb2005 	5 	97381.88 	650.17 	20 	94713.17 	739.8 	-2.740
specjvm98 	5 	378.15 	3.43 	20 	372.23 	2.84 	-1.566
volano25 	5 	128839 	1764.68 	20 	132124 	2142.89 	2.550



On 5/13/2015 4:05 AM, Alexeev, Alexander wrote:
> Hi, Yumin
>
> You've opened "Enable RewriteBytecodes when VM runs with CDS" [1] 
> record and implemented it for x86. Right? Currently I am going to 
> measure performance benefits from the similar change on aarch64 and 
> need benchmarks that actually can show that.
>
> In JIRA you've mentioned "As a result, some benchmarks showed 2% 
> degradation when running under C2 and with CDS enabled. " Could you 
> tell what are those benchmarks?
>
> Thanks,
> Alexander
>
> [1] https://bugs.openjdk.java.net/browse/JDK-8074345



More information about the aarch64-port-dev mailing list