[
https://issues.apache.org/jira/browse/CASSANDRA-8762?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robert Stupp resolved CASSANDRA-8762.
-------------------------------------
Resolution: Won't Fix
Sorry - was wrong. Forgot about the intrinsics...
{noformat}
[java] Benchmark Mode Samples
Score Error Units
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getByte thrpt 5
80,858 ± 6,366 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getDouble thrpt 5
69,058 ± 3,982 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getFloat thrpt 5
68,358 ± 5,091 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getInt thrpt 5
80,450 ± 4,335 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getLong thrpt 5
80,956 ± 1,780 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_getShort thrpt 5
81,100 ± 3,992 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putByte thrpt 5
68,207 ± 6,179 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putDouble thrpt 5
69,082 ± 7,061 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putFloat thrpt 5
69,650 ± 4,308 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putInt thrpt 5
58,813 ± 5,546 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putLong thrpt 5
62,122 ± 3,496 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.native_putShort thrpt 5
69,906 ± 1,968 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getByte thrpt 5
492,669 ± 54,257 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getDouble thrpt 5
501,847 ± 22,772 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getFloat thrpt 5
490,597 ± 39,931 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getInt thrpt 5
502,179 ± 66,369 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getLong thrpt 5
498,300 ± 31,606 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_getShort thrpt 5
489,958 ± 44,623 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putByte thrpt 5
479,133 ± 180,038 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putDouble thrpt 5
494,137 ± 185,270 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putFloat thrpt 5
507,778 ± 209,803 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putInt thrpt 5
502,569 ± 216,881 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putLong thrpt 5
501,504 ± 204,266 ops/us
[java] o.a.c.t.m.MemoryUtilMicroBench.unsafe_putShort thrpt 5
497,612 ± 184,581 ops/us
{noformat}
> Check if JNA Native gives good performance gain over Unsafe
> -----------------------------------------------------------
>
> Key: CASSANDRA-8762
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8762
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Robert Stupp
> Assignee: Robert Stupp
> Priority: Minor
>
> As noted in CASSANDRA-8714, Unsafe has some invocation overhead compared to
> JNI which is used by JNA.
> This ticket is to check whether it's worth to rewrite the invocations from
> Unsafe to Native - i.e. NativeCell et al
> /cc [~benedict]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
