Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-28 Thread Michael Schnell
An additional idea: perhaps the system can run in one of of these new non volatile RAMs (ferro and competing technologies). This should provide really instant on. Of course after power on some hardware devices will need some initialization. -Michael

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-28 Thread Allon Stern
On Feb 25, 2008, at 3:20 AM, Michael Schnell wrote: An additional idea: perhaps the system can run in one of of these new non volatile RAMs (ferro and competing technologies). Core. - allon ___ uClinux-dev mailing list uClinux-dev@uclinux.org

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-18 Thread Michael Schnell
Did you take a look at my fastram patch? It allows to move cpu intensive kernel functions into the internal ram, so that they are executed much faster than from flash or external ram (typically 4x -10x faster!!). Sounds GREAT ! Does this need XIP ? I maybe would like to use this with NIOS

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-18 Thread Erwin Authried
Am Montag, den 18.02.2008, 11:55 +0100 schrieb Michael Schnell: Did you take a look at my fastram patch? It allows to move cpu intensive kernel functions into the internal ram, so that they are executed much faster than from flash or external ram (typically 4x -10x faster!!). Sounds

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-18 Thread Erwin Authried
Am Montag, den 18.02.2008, 13:07 +0100 schrieb Michael Schnell: no, you can use it with XIP or non-XIP kernels. You just have to modify the kernel's ld script so that there is a new fastram section. Sounds really good ! The kernel startup code must copy __fastram_image to

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-18 Thread Michael Schnell
If your cpu uses a cache, don't expect too much speedup. With ISRs the cache does not help, but fast ram should reduce the latency greatly But it's worth to try for sure, and I'd be courious to hear about your results! Right now I not even have any hardware to play with. So it will be

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-15 Thread Erwin Authried
Am Donnerstag, den 14.02.2008, 07:55 -0700 schrieb Robert Warner: Eriwn, I was afraid that you are thinking about boot times in the msec range. I have got ~1sec boot time with a 40MHz Atmel AT91 board just with using XIP for kernel + applications. In addition, I had to take out the HW

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-15 Thread Robert Warner
Erwin, I was afraid that you are thinking about boot times in the msec range. I have got ~1sec boot time with a 40MHz Atmel AT91 board just with using XIP for kernel + applications. In addition, I had to take out the HW reset for PCMCIA that was done with a ~0.5sec delay. The delay loop

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Erwin Authried
Hi Bob, what is the boot time that you require? With a few optimization tricks like making the kernel XIP from flash and using a patch from celinux to speed up the delay loop calibration you can easily get boot times of a few 100 msecs, depending on your cpu horsepower. Regards, Erwin Am

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Robert Warner
Erwin, Now this is what i'd like to here. Less then 1 sec for startup time would be absolutely perfect. This device is a hand held radio. The people using these radios are used to typical analog radios, which at power on are ready to go immediately. A less then one second boot and

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Erwin Authried
Bob, I was afraid that you are thinking about boot times in the msec range. I have got ~1sec boot time with a 40MHz Atmel AT91 board just with using XIP for kernel + applications. In addition, I had to take out the HW reset for PCMCIA that was done with a ~0.5sec delay. The delay loop calibration

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Robert Warner
Eriwn, I was afraid that you are thinking about boot times in the msec range. I have got ~1sec boot time with a 40MHz Atmel AT91 board just with using XIP for kernel + applications. In addition, I had to take out the HW reset for PCMCIA that was done with a ~0.5sec delay. The delay loop

Re: SPAM-LOW: Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Robert Warner
Jamie, Now this is what i'd like to here. Less then 1 sec for startup time would be absolutely perfect. This device is a hand held radio. The people using these radios are used to typical analog radios, which at power on are ready to go immediately. A less then one second boot and

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Greg Ungerer
Hi Bob, Jamie Lokier wrote: Robert Warner wrote: Now this is what i'd like to here. Less then 1 sec for startup time would be absolutely perfect. This device is a hand held radio. The people using these radios are used to typical analog radios, which at power on are ready to go

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-14 Thread Robert Warner
Greg, Greg Ungerer wrote: Jamie Lokier wrote: Robert Warner wrote: As a reference point, our early ColdFire based snapgear routers booted in around 1s (from power on to shell prompt/login). That wasn't tweaked in any special way either. (Those units where reasonably fast CPU 66MHz with

[uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-13 Thread Robert Warner
Hi I'm looking into the prospect of preloading all of the necessary tables for applications and kernel such that when the device is turned on it doesn't boot it just starts executing as if it had already gone through all the booting and application(s) initialization(s). Has this aspect of

Re: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-13 Thread Greg Ungerer
Hi Bob, Robert Warner wrote: I'm looking into the prospect of preloading all of the necessary tables for applications and kernel such that when the device is turned on it doesn't boot it just starts executing as if it had already gone through all the booting and application(s)

RE: [uClinux-dev] 'Instant-On' ... and uCLinux

2008-02-13 Thread Gavin Lambert
Quoth Robert Warner: I'm looking into the prospect of preloading all of the necessary tables for applications and kernel such that when the device is turned on it doesn't boot it just starts executing as if it had already gone through all the booting and application(s) initialization(s). Has