Post a Comment
Isn't it a bit ironic that the hall-of-shame list contains applications such as gnome-power-manager and S06cpuspeed - who's purpose, as I understand, includes saving power?
Oh, and the ibm-acpi module - required by IBM laptops to check battery, control LCD brightness, etc. - is there, too.
I really don't understand how people can think that polling is ever the right solution. It's one of the first things you learn when studying low-power circuitry: asynchronous circuits generally save power. The same thing applies to software. Asynchronous, event-based programming keeps the machine quiescent when nothing is happening.
I think this is one of the few penalties of the distributed method of software development. There is no core set of practices and ideas on how to solve problems that are produced by interactions up and down the entire stack. And the loose-coupling of components often means that the event one desires to wait for is not exposed by the interface for the underlying component.
> Why so much polling?
Because noone thought about the power usage when programming the application?
I bet that some parts of the code of a Linux distribution was written before laptops existed..
Sure, you're right that highly distributed sw development like in free software leads to poor coherency, but the one nice thing about free software is that if you need an asynchronous notification not present in the original sw then you can add it, with proprietary sw you're stuck having to find a workaround.
I think it is more likely that at the time some of this code has been written there hadn't been any asynchronous method of doing the desired task.
For example take file change monitoring. If our operating system does not provide you a notification mechanism on file changes, you're only option left is to do polling.
Or RSS feeds: news aggregators have to poll their URIs since they can't just broadcast that they have new content available.
With the direct cooperation of a great manufacturer, I think Linux will be able to close quickly this "power" gap (usually about one hour of battery life) because it will give a tool to manage the settings of ram hungry software.
I do not know how far Intel will go also with BIOS settings and ACPI...but I do not complain, this move of course goes in the right way.
Everybody wins...(well, nearly everybody I mean ;-)
This is an interesting development. You don't see many hardware giants pushing projects exclusively at Linux laptop users (both of us!).
Am I right in thinking the Intel platform has stolen a lot of ground in the high-performance server/mainframe area from Sun and other competitors in recent years? I wonder if this is the main reason we've seen them do so much lately for *nix compatibility with their hardware. Saving a few minutes of battery life on a laptop would I imagine translate to a pretty massive utility-bill saving for the owner of a Linux supercomputer with thousands of CPUs...
Whatever the reasons, I couldn't be more pleased with Intel's courting of the OSS community. My next computer will almost definitely have an Intel graphics card, and I'll be keeping an eye on their wireless offerings as well.



