[PATCH] TX4927 processor can support different speeds

[PATCH] TX4927 processor can support different speeds

[Manish Lachwani]

[-- Attachment #1: Type: text/plain, Size: 224 bytes --]

Hi Ralf,

Based on the feedback from Toshiba, the TX4927 processor can support different 
speeds. Attached patch takes care of that. If you find this approach reasonable, 
can you please check it in

Thanks
Manish Lachwani


[-- Attachment #2: common_mips_tx4927_cpuspeed.patch --]
[-- Type: text/plain, Size: 1639 bytes --]

Source: MontaVista Software, Inc. | http://www.mvista.com | Manish Lachwani <mlachwani@mvista.com>
MR: 9936
Type: Enhancement
Disposition: Submitted to linux-mips.org
Keywords:
Signed-off-by: Manish Lachwani  <mlachwani@mvista.com>
Description:
	Based on suggestion from Toshiba, TX4927 can operate at different
	speeds

Index: linux-2.6.10/arch/mips/Kconfig
===================================================================
--- linux-2.6.10.orig/arch/mips/Kconfig
+++ linux-2.6.10/arch/mips/Kconfig
@@ -901,6 +901,14 @@
 	  This Toshiba board is based on the TX4927 processor. Say Y here to
 	  support this machine type
 
+config TOSHIBA_TX4927_CPU_SPEED
+	int "CPU speed of the TX4927 processor (MHz)"
+	depends on TOSHIBA_RBTX4927
+	default 200
+	help
+	  This sets the speed for the TX4927 processor. The default speed
+	  is 200 MHz. 
+
 config TOSHIBA_FPCIB0
 	bool "FPCIB0 Backplane Support"
 	depends on TOSHIBA_RBTX4927
Index: linux-2.6.10/arch/mips/tx4927/toshiba_rbtx4927/toshiba_rbtx4927_setup.c
===================================================================
--- linux-2.6.10.orig/arch/mips/tx4927/toshiba_rbtx4927/toshiba_rbtx4927_setup.c
+++ linux-2.6.10/arch/mips/tx4927/toshiba_rbtx4927/toshiba_rbtx4927_setup.c
@@ -1008,7 +1008,11 @@
 				       mips_hpt_frequency,
 				       mips_hpt_frequency / 1000000);
 #else
-	mips_hpt_frequency = 100000000;
+	/*
+	 * Default TX4927 processor speed is 200 MHz. However, it 
+	 * can be configured by the user
+	 */
+	mips_hpt_frequency = (CONFIG_TOSHIBA_TX4927_CPU_SPEED * 1000000) / 2;
 #endif
 
 	TOSHIBA_RBTX4927_SETUP_DPRINTK(TOSHIBA_RBTX4927_SETUP_TIME_INIT, "+\n");

Re: [PATCH] TX4927 processor can support different speeds

[Thiemo Seufer]

Manish Lachwani wrote:
> Hi Ralf,
> 
> Based on the feedback from Toshiba, the TX4927 processor can support different 
> speeds. Attached patch takes care of that. If you find this approach reasonable, 
> can you please check it in

Shoudn't this better be tunable via /proc?


Thiemo

Re: [PATCH] TX4927 processor can support different speeds

[Ralf Baechle]

On Sun, Jan 23, 2005 at 08:41:40PM +0100, Thiemo Seufer wrote:

> > Based on the feedback from Toshiba, the TX4927 processor can support different 
> > speeds. Attached patch takes care of that. If you find this approach reasonable, 
> > can you please check it in
> 
> Shoudn't this better be tunable via /proc?

By the time that interface becomes available interrupt timers would
already have been missprogrammed.  Try to calibrate the CPU timer against
some external timer such as the RTC instead.  It's already being done
on the Indy.

  Ralf

Re: [PATCH] TX4927 processor can support different speeds

[Manish Lachwani]

Ralf Baechle wrote:

>On Sun, Jan 23, 2005 at 08:41:40PM +0100, Thiemo Seufer wrote:
>
>  
>
>>>Based on the feedback from Toshiba, the TX4927 processor can support different 
>>>speeds. Attached patch takes care of that. If you find this approach reasonable, 
>>>can you please check it in
>>>      
>>>
>>Shoudn't this better be tunable via /proc?
>>    
>>
>
>By the time that interface becomes available interrupt timers would
>already have been missprogrammed.  Try to calibrate the CPU timer against
>some external timer such as the RTC instead.  It's already being done
>on the Indy.
>
>  Ralf
>
>  
>
Hi Ralf,

Why is this approach (in the patch) bad?

Thanks
Manish Lachwani

Re: [PATCH] TX4927 processor can support different speeds

[Ralf Baechle]

On Sun, Jan 23, 2005 at 12:39:42PM -0800, Manish Lachwani wrote:

> Why is this approach (in the patch) bad?

It's fragile because clock frequencies are changing faster in today's
world of electronics than the weather in April.

  Ralf

Re: [PATCH] TX4927 processor can support different speeds

[Manish Lachwani]

Ralf Baechle wrote:

>On Sun, Jan 23, 2005 at 12:39:42PM -0800, Manish Lachwani wrote:
>
>  
>
>>Why is this approach (in the patch) bad?
>>    
>>
>
>It's fragile because clock frequencies are changing faster in today's
>world of electronics than the weather in April.
>
>  Ralf
>
>  
>
So? Can you be a little more clear?

Thanks
Manish Lachwani

Re: [PATCH] TX4927 processor can support different speeds

[Maciej W. Rozycki]

On Mon, 24 Jan 2005, Manish Lachwani wrote:

> > > Why is this approach (in the patch) bad?
[...]
> > It's fragile because clock frequencies are changing faster in today's
> > world of electronics than the weather in April.
[...]
> So? Can you be a little more clear?

 Oh well, how can you assure a given binary will be booted on a CPU driven 
by the right frequency?  Is the clock source inside the chip containing 
the CPU at least?

  Maciej

Re: [PATCH] TX4927 processor can support different speeds

[Manish Lachwani]

Maciej W. Rozycki wrote:

>On Mon, 24 Jan 2005, Manish Lachwani wrote:
>
>  
>
>>>>Why is this approach (in the patch) bad?
>>>>        
>>>>
>[...]
>  
>
>>>It's fragile because clock frequencies are changing faster in today's
>>>world of electronics than the weather in April.
>>>      
>>>
>[...]
>  
>
>>So? Can you be a little more clear?
>>    
>>
>
> Oh well, how can you assure a given binary will be booted on a CPU driven 
>by the right frequency?  Is the clock source inside the chip containing 
>the CPU at least?
>
>  Maciej
>
>  
>
Ok, that make things clear. Thanks Maciej,

Manish Lachwani

RE: [PATCH] TX4927 processor can support different speeds

[Brad Larson]

> On Mon, 24 Jan 2005, Manish Lachwani wrote:
> 
> > > > Why is this approach (in the patch) bad?
> [...]
> > > It's fragile because clock frequencies are changing 
> faster in today's
> > > world of electronics than the weather in April.
> [...]
> > So? Can you be a little more clear?
> 
>  Oh well, how can you assure a given binary will be booted on 
> a CPU driven 
> by the right frequency?  Is the clock source inside the chip 
> containing 
> the CPU at least?
> 
>   Maciej

I've seen more boards with slightly, sometimes dramatically, modified cpu clock source for performance and production reliability reasons than not having an external RTC.  External RTC present and one kernel can accurately enough report the cpu frequency for Toshiba "stock" values and others as well.

--Brad

RE: [PATCH] TX4927 processor can support different speeds

[Maciej W. Rozycki]

On Mon, 24 Jan 2005, Brad Larson wrote:

> I've seen more boards with slightly, sometimes dramatically, modified 
> cpu clock source for performance and production reliability reasons than 
> not having an external RTC.  External RTC present and one kernel can 
> accurately enough report the cpu frequency for Toshiba "stock" values 
> and others as well.

 And you don't need a full-blown RTC even -- any reasonably accessible 
reference clock with a known frequency will do.

  Maciej