Re: main difference between expandable and singlepage subfile

>> Let me emphasize that I don't know if this is true with IBM POWER

It certainly is. IBM POWER was the first server to gain the ENERGY STAR certification for example. Basically this capability arrived in POWER 5 chips. Prior to that they mostly just added more heatsink and more fan, sorry "Air Movement Device", to prevent melting.

Since then it's incredible how many independent parts of the CPU can be powered up and down in one cycle. I believe the POWER8 chips had (from memory here) 8000 different bits that could be powered down when not needed.

You can also see in WRKSYSACT the average CPU speed the cores are running at as a percent of the maximum for the chip. It is VERY often less than 100.

Now is one of those techniques going to save power vs another?? I rather think it's going to be like daylight savings time and end up making near zero difference.

Additionally I have LONG felt that coding for better readability and maintainability trumps the need for speed or fewer lines of code. There are always exceptions here and there but don't sacrifice these things just because you feel a micro-watt savings might result!


- Larry "DrFranken" Bolhuis

www.Frankeni.com
www.iDevCloud.com - Personal Development IBM i timeshare service.
www.iInTheCloud.com - Commercial IBM i Cloud Hosting.

On 11/25/2019 4:30 AM, Patrik Schindler wrote:

Hello Booth, (?)

Am 24.11.2019 um 22:48 schrieb Booth Martin <booth@xxxxxxxxxxxx>:

With today's speeds and drill-downs the expandable subfile becomes, in my opinion, anachronistic.

On a side note, from a programmer's viewpoint, there's not much difference between a load-all vs. an expandable SFL. As soon as SFLPAG and SFLSIZ are different in value, it's always possible to expand the SFL programmatically to just add records.

So maybe it's anachronistic, but a few lines of code enable the hardware to finish this particular task faster (by stopping before EOF). I agree it makes not much difference in terms of absolute time until the screen arrives at the user's display. But when there are lot of users heavily using the SFL facility, the load of the machine is less.

Maybe somewhat more important but unknown to me if applicable to IBM POWER is that most "consumer grade" CPUs make extensive use of power saving facilities. Stuff the CPU doesn't need gets powered down in fractions of a second, and back up. So it's possible that an idle server with just SSDs consumes about 80W and if all cores are fully loaded with work, it's consuming 200W. If these 200W were drawn for a shorter amount of time because of more efficient programming, less energy is wasted. This makes not much of a difference seen for a single server in one small company. Seen globally with gazillions of machines, it would truly make a difference. Let me emphasize that I don't know if this is true with IBM POWER, too.

I'm somewhat appalled with this "we've enough CPU speed, so why invest in clever programming" attitude…

:wq! PoC

PGP-Key: DDD3 4ABF 6413 38DE - https://www.pocnet.net/poc-key.asc