jump to navigation

Is this what people get paid for…? April 9, 2006

Posted by Florian in GPE, Maemo, Source, World.

I just read Lorn Potter’s article here and wondered what he gets paid for. I’m not surprised that he didn’t publish my comment to his blog. In fact it is possible to make all GTK based applications on a SL-5000d terribly slow using a kernel with a fixed size ramdisk, but that’s the missing detail in his text. If you have 32MB of physical memory and use a more common setup, GTK based applications are much faster than in this comparison. Suddenly the startup time for gpe-todo decreases from 13 seconds to 1 second.

However… independent from the technical issue i really wonder why things like this are necessary. Are these guys afraid of GTK that much?


1. lpotter - April 9, 2006

I used the same kernel for all these tests. Straight off the shelf openzaurus 3.5.4, I was trying to be as unbiased and fair as possible. I had nothing to do with the configuration or compilation of the system.
I used the same exact machine also – Zaurus 5000d. I used this machine just because it is slow, and lacking memory. Perfect embedded environment for testing application startup speed.

I did this research of my own accord because I got tired of hearing that such-and-such environment is so much faster than this other one. that X is faster than framebuffer.
No one had ever done such a test. I would love for you to do a similar test with whatever device you choose.

I have nothing to do whether your comments get published or not.

2. zack - April 10, 2006

Instead of accusing lpotter, you should do an unbiased test and publish the results. Let the readers decide.

3. Jussi Kukkonen - April 10, 2006

So he didn’t test what you would have wanted him to test. I don’t think that’s a good reason to get condescending… I mean, he might have other reasons for benchmarking than “fearing GTK” (like the ones he stated in the post)?

By the way, why do you think benchmarks aren’t necessary in this context? I’m not picking a fight or anything, just interested. In other contexts I’ve found benchmarks with competitors quite useful, as they may pinpoint areas of “easiest optimizations” — not to mention that the “general public” might be interested…

4. fl0rian - April 10, 2006

For some strange reason my SL-5000d is much faster… with a stock OZ image and the recommended kernel 32-0 kernel for this device. No idea how the 0.1s precise measurement is done, but even on the old SL-5000d e.g. gpe-todo takes less than 2 seconds to start.
But many thanks for your comments – that’s _exactly_ what i expected.
Ah and Lorn – if you ran into trouble switching between applications you’ll maybe like my new gpe-windowlist applet. ;-)

5. lpotter - April 10, 2006

I will redo this test with another machine – a c700, which is less memory constrained, and has a faster xscale cpu.

The 0.1 is me pushing the button twice, because it is instantaneous. Of course, these tests were done by hand and not some script.

The 32-0 kernel is a gross hack in which you need a separate memory card for doing anything.

Good to hear about the windowlist applet. :)

6. fl0rian - May 9, 2006

Looks like it is a little bit more complicated to make GPE slow on a c700, eh? ;-)

7. cris - August 8, 2006

I’m one of those thinking that GPE was ‘terribly slow’ as it was pretty not usable on palm devices (TX/32Mb): the only way to run it at an acceptable speed was using a swap fs.

We have recently started using initramfs for the boot process and … it increased dramatically GPE’s speed; a lot of memory gets freed up and the swap file is not even used at all.

PIM apps seem to need about 2 seconds to start (3-4 if the rootfs is on an ext2 image).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: