RFR(S): 8215483: Off heap memory accesses should be vectorized
vladimir.kozlov at oracle.com
Fri Jan 25 18:29:16 UTC 2019
I missed original RFR e-mail and asked to add explanation for code change in RFE which had only code sample before and
very short statement. Roland added that information now and I am satisfied.
Yes, we should vectorize such cases.
On 1/25/19 1:33 AM, Andrew Haley wrote:
> On 1/24/19 5:50 PM, Vladimir Kozlov wrote:
>> Looks good to me too. But it would be nice to have changes
>> explanation in RFE. Why it helps vectorize off heap memory accesses?
> I don't quite understand what you're asking here, but I guess it's
> about what applications need vertorized MappedByteBuffers. Sometimes
> people use allocateDirect to get a chunk of memory rather than simply
> allocate, and it would be surprising that their programs ran more
> slowly as a result. Apache Lucene, for example, is a large-scale user
> of ByteBuffers, and they need all the performance they can get.
More information about the hotspot-compiler-dev