[
https://issues.apache.org/jira/browse/TS-3178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14202104#comment-14202104
]
Alan M. Carroll commented on TS-3178:
-------------------------------------
I think we may want to change the original proxy allocator configuration value
so the pair is something like
{code}
proxy.config.allocator.thread_free_list.high_watermark
proxy.config.allocator.thread_free_list.low_watermark
{code}
We could do this for 5.X and put in some transitional code to read the old tag
for size/high_watermark for compatibility.
Use pre-increment and decrement. That is, {{--(l.allocated);}} etc.
Rather than enclosing the {{return}} in the {{ifdef}}, why on enclose the rest
in a {{#if !defined}}?
{code}
#if !defined(TS_USE_RECLAIMABLE_FREELIST)
if (unlikely(count == 1)) {
a.free(head);
} else if (count > 0) {
a.free_bulk(head, tail, count);
}
ink_assert(l.allocated >= thread_freelist_low_watermark);
#endif
}
{code}
You should put a comment on the count argument to the low level free functions
that it is there only for accounting, to keep the memory counters correct.
> ProxyAllocator improvements
> ---------------------------
>
> Key: TS-3178
> URL: https://issues.apache.org/jira/browse/TS-3178
> Project: Traffic Server
> Issue Type: Improvement
> Components: Core
> Reporter: Brian Geffon
> Assignee: Cynthia Gu
> Attachments: patch.diff
>
>
> Currently when a ProxyAllocator (Thread Local) has more than a
> configurable number of elements it will return them one-by-one to a
> ClassAllocator (Global Freelist). Returning every item in this fashion is
> inefficient as we'll likely need more items in the future. Therefore we
> should maintain a low watermark (a minimum number) of items that should be in
> a ProxyAllocator at any one time. When the number of elements reaches high
> watermark, the free up is triggers to keep it below the low
> watermark. Additionally, the free should be a block free instead of
> one-by-one as we can reduce several hundred compare-and-swap operations to a
> single CAS
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
