300 nanoseconds

[Anton Gavriliuk]

Hi all

I have read nice article - https://pmem.io/2019/12/19/performance.html

But want clarify this

"Depending on who you ask, you might get a different answer. The broadest definition that we commonly use is that it’s memory that is non-volatile and has access latency low enough that it is reasonable to stall the CPU while waiting for reads or writes to complete."

May be NOT to stall the CPU ??, with Persistent Memory unstalled execution should be higher, making much more performance :-)

Anton

[Andy Rudoff]

Hi Anton,

In the SNIA technical work group, when we were trying to decide on a definition for "persistent memory", we decided if reading from it took so long that it made sense to context switch and do other work while waiting for the read to complete, then it wasn't persistent memory (it was just storage).  For example, reading from NAND takes long enough that it makes sense to switch away and do other work while waiting for the result.  But if the read is so fast that there isn't really time for a context switch, then you just "stall" and wait for the data to show up.  When you choose to stall the load instruction itself, waiting for the result, then that's the persistent memory programming model.

I believe that's the logic that Piotr was summarizing in the blog you quoted.

Hope that helps,

-andy

[Anton Gavriliuk]

Hi Andy

Here is the slide from my pmem presentation based on 24/7 I'm spending on my optane based box.

Anton

сб, 11 янв. 2020 г. в 19:12, Andy Rudoff <an...@rudoff.com>:

--
You received this message because you are subscribed to the Google Groups "pmem" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pmem+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pmem/95069dbd-43ec-4d7f-b255-7770f2b58e2d%40googlegroups.com.