Thread: what change in postgres 9.5 improvements for multi-CPU machines

On Fri, Mar 25, 2016 at 5:11 AM, 657985552@qq.com <657985552@qq.com> wrote:
> hello everyone:
>       I was bothered by the postgres spinlock for a long time . How to
> understand this sentence "postgres 9.5 performance improvements for
> multi-CPU machines"
> at present my database is 9.3.5 。  Can  it solve the  spinlock problem, if
> upgrade it to 9.5 ?

What problem?  Spinlock contention is a symptom, not a problem in and
of itself.  Spinlocks are good target for micro-optimization though,
and we're always looking for strategies to make things work better
if/when we get into high contention scenarios.

merlin

On 2016-03-25 18:11:21 +0800, 657985552@qq.com wrote:
> hello everyone:
>       I was bothered by the postgres spinlock for a long time . How to understand this sentence "postgres 9.5
performanceimprovements for multi-CPU machines" 
> at present my database is 9.3.5 。  Can  it solve the  spinlock problem, if upgrade it to 9.5 ?

It's quite possible that the upgrade helps. But without additional data
it's hard to say. The change is that postgres internal reader/writer
lock now, in many cases, avoid the use of a spinlock, relying on atomic
operations instead.

Regards,

Andres