Author Topic: Linux Distros Optimized For C3k's Xscale Mmx? (Read 4712 times)

Neverwinter · « **on:** December 28, 2004, 04:09:45 am »

Are any of the Linux distros (esp. the original Linux Rom) optimized to take advantage of the XScale's Wireless MMX or Speed Step technology?

amrein · « **Reply #1 on:** December 28, 2004, 07:39:43 am »

None.

And using those Intell MMX specific instructions will broke ARM v5 compatibilty with other device.

The initial Sharp ROM uses them only in kernel and DSL to speed the framebuffer processing I guess.

Neverwinter · « **Reply #2 on:** December 28, 2004, 07:45:28 am »

lol Even though I am a programmer, I didn't get what you just said. If you are saying that none do, then if they did, wouldn't that possibly increase the battery life by plenty? If not even the original ROM does, then I am surprised...

ThC · « **Reply #3 on:** December 28, 2004, 07:50:27 am »

Quote

None.

And using those Intell MMX specific instructions will broke ARM v5 compatibilty with other device.

The initial Sharp ROM uses them only in kernel and DSL to speed the framebuffer processing I guess.

btw it's obvious that a lot of programs will take advantage of it so it may be interresting to have a special feed for c3000 devices in the future (because the mmx is not the only part of the c3000 which is new ... we'll need optimised programs for the graphic chip as it's not a w100 no more ...)

shanks · « **Reply #4 on:** December 28, 2004, 09:53:28 am »

hi

maybe it wouldn't be "too difficult" to port netbsd evbarm who has been designed to work on
xscale http://www.netbsd.org/Ports/evbarm/
I'm waiting for my c3k and would be interested to port it, if i'm not alone
it would rocks to have netbsd on zaurus

excuse my bad english

Neverwinter · « **Reply #5 on:** December 28, 2004, 11:02:42 am »

Quote

Quote
None.

And using those Intell MMX specific instructions will broke ARM v5 compatibilty with other device.

The initial Sharp ROM uses them only in kernel and DSL to speed the framebuffer processing I guess.
btw it's obvious that a lot of programs will take advantage of it so it may be interresting to have a special feed for c3000 devices in the future (because the mmx is not the only part of the c3000 which is new ... we'll need optimised programs for the graphic chip as it's not a w100 no more ...)

I thought it was a bit confusing as to wether the C3000 actually has its own GPU or not. I thought that it didn't and for all graphics/video related functions it uses the XScale CPU?

amrein · « **Reply #6 on:** December 28, 2004, 03:42:46 pm »

Don't forget the fact that they are still using gcc2.95.3. There is no optimisation for xcale nor MMX support in this old compiler. Binary compatibility reasons...

Neverwinter · « **Reply #7 on:** December 28, 2004, 05:33:19 pm »

So if I were to assume that the optimizations does not exist, then if the optimizations were made, how much longer would the battery last at the CPU's lowest speed? Isn't the CPU able to run at 100 MHz? That is about 1/4 of 416 MHz... and if so, then you would multiply the battery life times 4? 7 hours x 4 = 28 hours?

sriley · « **Reply #8 on:** December 28, 2004, 06:38:29 pm »

Quote

That is about 1/4 of 416 MHz... and if so, then you would multiply the battery life times 4? 7 hours x 4 = 28 hours?

If the CPU was the only thing using battery power, then yes. But the screen and drive are still going to be using power at the normal rate, right?

wormwood · « **Reply #9 on:** December 28, 2004, 11:44:12 pm »

Quote

shanks Posted on Dec 28 2004, 02:53 PM
hi

maybe it wouldn't be "too difficult" to port netbsd evbarm who has been designed to work on
xscale http://www.netbsd.org/Ports/evbarm/
I'm waiting for my c3k and would be interested to port it, if i'm not alone
it would rocks to have netbsd on zaurus

excuse my bad english

BSD on the Zaurus would definately be very nice . Keep us informed shanks!

Neverwinter · « **Reply #10 on:** December 29, 2004, 12:14:43 am »

Quote

Quote
That is about 1/4 of 416 MHz... and if so, then you would multiply the battery life times 4? 7 hours x 4 = 28 hours?
If the CPU was the only thing using battery power, then yes. But the screen and drive are still going to be using power at the normal rate, right?

...I don't know. Does it?

sriley · « **Reply #11 on:** December 29, 2004, 11:21:43 pm »

Quote

Quote

Quote
That is about 1/4 of 416 MHz... and if so, then you would multiply the battery life times 4? 7 hours x 4 = 28 hours?
If the CPU was the only thing using battery power, then yes. But the screen and drive are still going to be using power at the normal rate, right?
...I don't know. Does it?

Underclocking the CPU to 100mhz certainly extends the battery life on my 860 running Cacko 1.21b, but not by a multiple of four. Since my 3000 adds a microdrive to the equation I'd imagine the effect of underclocking would be even less pronounced.

Unless the screen was 1/4 of its normal brightness and the drive spun at 1/4 its normal RPM, I'd say they were still consuming battery power at their normal rates. I think those two would probably have a greater impact on battery life than the CPU, but I'm just guessing and have no data to back that up other than personal observation.

Neverwinter · « **Reply #12 on:** December 29, 2004, 11:43:00 pm »

Quote

Quote

Quote

Quote
That is about 1/4 of 416 MHz... and if so, then you would multiply the battery life times 4?Â 7 hours x 4 = 28 hours?
If the CPU was the only thing using battery power, then yes. But the screen and drive are still going to be using power at the normal rate, right?
...I don't know. Does it?
Underclocking the CPU to 100mhz certainly extends the battery life on my 860 running Cacko 1.21b, but not by a multiple of four. Since my 3000 adds a microdrive to the equation I'd imagine the effect of underclocking would be even less pronounced.

Unless the screen was 1/4 of its normal brightness and the drive spun at 1/4 its normal RPM, I'd say they were still consuming battery power at their normal rates. I think those two would probably have a greater impact on battery life than the CPU, but I'm just guessing and have no data to back that up other than personal observation.

Underclocking you c860? Could you give an example about the battery life differences?

sriley · « **Reply #13 on:** December 29, 2004, 11:55:27 pm »

Typo, sorry. It's a 760 not an 860.

When I underclock it from 400mhz to 100mhz using the utility included in Cacko 1.21b, my battery life roughly doubles. I say roughly because I only underclock to read. When I'm underclocked I don't ever use a network card or play games or anything else that would draw extra power, so it's a guess at best.

Oddly (and off topic, wrong forum, etc.), I've noticed that underclocking my 760 affects my screen calibration, throwing it slightly off to one side. When I reclock to 400mhz it goes back to normal. I think that's weird.

xatax · « **Reply #14 on:** December 30, 2004, 10:54:09 am »

Quote

[...]
Oddly (and off topic, wrong forum, etc.), I've noticed that underclocking my 760 affects my screen calibration, throwing it slightly off to one side. When I reclock to 400mhz it goes back to normal. I think that's weird.

Same here, I have a C750. It is indeed weird, I must tap about 3mm to the right everything on the screen...

News:

Author Topic: Linux Distros Optimized For C3k's Xscale Mmx? (Read 4712 times)