On Tue, 29 Oct 2024 14:51:33 GMT, Per Minborg <pminb...@openjdk.org> wrote:
> This PR proposes to improve `MemorySegment::ofBuffer` making it more amenable
> to inlining and generally improving performance.
>
> Testing successfully on tier1-3
I've added a very similar benchmark to the PR which I've run with a Mac M1:
$ make test TEST="micro:java.lang.foreign.SegmentOfBuffer" MICRO="OPTIONS=-prof
gc"
Benchmark Mode Cnt Score
Error Units
SegmentOfBuffer.ofBuffer avgt 30 1.225 ?
0.025 ns/op
SegmentOfBuffer.ofBuffer:gc.alloc.rate avgt 30 0.013 ?
0.001 MB/sec
SegmentOfBuffer.ofBuffer:gc.alloc.rate.norm avgt 30 ? 10??
B/op
SegmentOfBuffer.ofBuffer:gc.count avgt 30 ? 0
counts
SegmentOfBuffer.ofBufferInlineFalse avgt 30 1.221 ?
0.019 ns/op
SegmentOfBuffer.ofBufferInlineFalse:gc.alloc.rate avgt 30 0.013 ?
0.001 MB/sec
SegmentOfBuffer.ofBufferInlineFalse:gc.alloc.rate.norm avgt 30 ? 10??
B/op
SegmentOfBuffer.ofBufferInlineFalse:gc.count avgt 30 ? 0
counts
SegmentOfBuffer.ofBufferInlineTrue avgt 30 1.212 ?
0.010 ns/op
SegmentOfBuffer.ofBufferInlineTrue:gc.alloc.rate avgt 30 0.013 ?
0.001 MB/sec
SegmentOfBuffer.ofBufferInlineTrue:gc.alloc.rate.norm avgt 30 ? 10??
B/op
SegmentOfBuffer.ofBufferInlineTrue:gc.count avgt 30 ? 0
counts
-------------
PR Comment: https://git.openjdk.org/jdk/pull/21764#issuecomment-2449342833
