[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[pieterg]

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20100708/12fcc7a6/attachment.ksh>

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[Eric Miao]

On Fri, Jul 9, 2010 at 1:04 AM, pieterg <pieterg@gmx.com> wrote:
>
> Signed-off-by: pieter <p.grimmerink@inepro.com>
> ---
> ?arch/arm/mach-pxa/cpufreq-pxa3xx.c | ? 13 ++++++++++++-
> ?1 files changed, 12 insertions(+), 1 deletions(-)
>
> diff --git a/arch/arm/mach-pxa/cpufreq-pxa3xx.c
> b/arch/arm/mach-pxa/cpufreq-pxa3xx.c
> index 27fa329..8cc10c3 100644
> --- a/arch/arm/mach-pxa/cpufreq-pxa3xx.c
> +++ b/arch/arm/mach-pxa/cpufreq-pxa3xx.c
> @@ -76,6 +76,14 @@ static struct pxa3xx_freq_info pxa300_freqs[] = {
> ? ? ? ?OP(624, 24, 2, 208, 260, 208, 312, 3, 1375, 1400), /* 624MHz */
> ?};
>
> +static struct pxa3xx_freq_info pxa310_freqs[] = {
> + ? ? ? /* ?CPU XL XN ?HSS DMEM SMEM SRAM DFI VCC_CORE VCC_SRAM */
> + ? ? ? OP(104, ?8, 1, 104, 260, ?78, 104, 3, 1000, 1100), /* 104MHz */
> + ? ? ? OP(208, 16, 1, 104, 260, 104, 156, 2, 1000, 1100), /* 208MHz */
> + ? ? ? OP(416, 16, 2, 156, 260, 104, 208, 2, 1100, 1200), /* 416MHz */
> + ? ? ? OP(624, 24, 2, 312, 260, 208, 312, 3, 1375, 1400), /* 624MHz */

What's wrong with 208MHz HSS on PXA310?

> +};
> +
> ?static struct pxa3xx_freq_info pxa320_freqs[] = {
> ? ? ? ?/* ?CPU XL XN ?HSS DMEM SMEM SRAM DFI VCC_CORE VCC_SRAM */
> ? ? ? ?OP(104, ?8, 1, 104, 260, ?78, 104, 3, 1000, 1100), /* 104MHz */
> @@ -214,9 +222,12 @@ static __init int pxa3xx_cpufreq_init(struct
> cpufreq_policy *policy)
> ? ? ? ?policy->cpuinfo.transition_latency = 1000; /* FIXME: 1 ms, assumed */
> ? ? ? ?policy->cur = policy->min = policy->max = get_clk_frequency_khz(0);
>
> - ? ? ? if (cpu_is_pxa300() || cpu_is_pxa310())
> + ? ? ? if (cpu_is_pxa300())
> ? ? ? ? ? ? ? ?ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa300_freqs));
>
> + ? ? ? if (cpu_is_pxa310())
> + ? ? ? ? ? ? ? ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa310_freqs));
> +
> ? ? ? ?if (cpu_is_pxa320())
> ? ? ? ? ? ? ? ?ret = setup_freqs_table(policy, ARRAY_AND_SIZE(pxa320_freqs));
>
> --
> 1.6.5.rc1.44.ga1675
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[pieterg]

On Friday 09 July 2010 10:02:16 Eric Miao wrote:
> What's wrong with 208MHz HSS on PXA310?

I haven't done any benchmarking, but Marvell found it worth to introduce 
312MHz for the pxa310. (I assume it might have something to do with the 
video accelerator, which is unique to the pxa310).

And to me it makes sense to use the highest supported speeds in the top 
cpufreq entry.

Rgds, Pieter

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[Igor Grinberg]

On 07/09/10 11:31, pieterg wrote:
> On Friday 09 July 2010 10:02:16 Eric Miao wrote:
>   
>> What's wrong with 208MHz HSS on PXA310?
>>     
> I haven't done any benchmarking, but Marvell found it worth to introduce 
> 312MHz for the pxa310. (I assume it might have something to do with the 
> video accelerator, which is unique to the pxa310).
>
> And to me it makes sense to use the highest supported speeds in the top 
> cpufreq entry.
>
>   

According to Marvell's PXA3xx Specification Update document
(page 68 - SCr#3), the above is correct starting from silicon revision A2.
What about earlier revisions? Shouldn't we support all of them?

> Rgds, Pieter
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
>   

-- 
Regards,
Igor.

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[pieterg]

On Sunday 11 July 2010 09:14:22 Igor Grinberg wrote:
> On 07/09/10 11:31, pieterg wrote:
> > On Friday 09 July 2010 10:02:16 Eric Miao wrote:
> >> What's wrong with 208MHz HSS on PXA310?
> >
> > I haven't done any benchmarking, but Marvell found it worth to
> > introduce 312MHz for the pxa310. (I assume it might have something to
> > do with the video accelerator, which is unique to the pxa310).
> >
> > And to me it makes sense to use the highest supported speeds in the top
> > cpufreq entry.
>
> According to Marvell's PXA3xx Specification Update document
> (page 68 - SCr#3), the above is correct starting from silicon revision
> A2. What about earlier revisions? Shouldn't we support all of them?

OK, I missed that. In that case we should leave this one out.

Rgds, Pieter

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[Igor Grinberg]

On 07/12/10 11:37, pieterg wrote:
> On Sunday 11 July 2010 09:14:22 Igor Grinberg wrote:
>   
>> On 07/09/10 11:31, pieterg wrote:
>>     
>>> On Friday 09 July 2010 10:02:16 Eric Miao wrote:
>>>       
>>>> What's wrong with 208MHz HSS on PXA310?
>>>>         
>>> I haven't done any benchmarking, but Marvell found it worth to
>>> introduce 312MHz for the pxa310. (I assume it might have something to
>>> do with the video accelerator, which is unique to the pxa310).
>>>
>>> And to me it makes sense to use the highest supported speeds in the top
>>> cpufreq entry.
>>>       
>> According to Marvell's PXA3xx Specification Update document
>> (page 68 - SCr#3), the above is correct starting from silicon revision
>> A2. What about earlier revisions? Shouldn't we support all of them?
>>     
> OK, I missed that. In that case we should leave this one out.
>   

If you really want this to go, you can check the cpu stepping (cpuid
register)
and fixup the table accordingly. Something like:

if (cpu_is_pxa310())
        ret = setup_freqs_table(policy, (read_cpuid_id() & 0xf) ?
                                                       
ARRAY_AND_SIZE(pxa310_freqs) :
                                                       
ARRAY_AND_SIZE(pxa300_freqs));

but I don't know if Eric will like it ;)

> Rgds, Pieter
>
>   

-- 
Regards,
Igor.

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[Eric Miao]

On Mon, Jul 12, 2010 at 5:13 PM, Igor Grinberg <grinberg@compulab.co.il> wrote:
>
>
> On 07/12/10 11:37, pieterg wrote:
>> On Sunday 11 July 2010 09:14:22 Igor Grinberg wrote:
>>
>>> On 07/09/10 11:31, pieterg wrote:
>>>
>>>> On Friday 09 July 2010 10:02:16 Eric Miao wrote:
>>>>
>>>>> What's wrong with 208MHz HSS on PXA310?
>>>>>
>>>> I haven't done any benchmarking, but Marvell found it worth to
>>>> introduce 312MHz for the pxa310. (I assume it might have something to
>>>> do with the video accelerator, which is unique to the pxa310).
>>>>
>>>> And to me it makes sense to use the highest supported speeds in the top
>>>> cpufreq entry.
>>>>
>>> According to Marvell's PXA3xx Specification Update document
>>> (page 68 - SCr#3), the above is correct starting from silicon revision
>>> A2. What about earlier revisions? Shouldn't we support all of them?
>>>
>> OK, I missed that. In that case we should leave this one out.
>>
>
> If you really want this to go, you can check the cpu stepping (cpuid
> register)
> and fixup the table accordingly. Something like:
>
> if (cpu_is_pxa310())
> ? ? ? ?ret = setup_freqs_table(policy, (read_cpuid_id() & 0xf) ?
>
> ARRAY_AND_SIZE(pxa310_freqs) :
>
> ARRAY_AND_SIZE(pxa300_freqs));
>
> but I don't know if Eric will like it ;)
>

Nah, I was thinking of introducing a cpu_is_pxa310_a2() or something,
but that's really hard to make generic, since for this case, we'd
have to use cpu_is_pxa310_a2_or_above(), which is #!$!%!@#$

So to put it simple, I think a command line parameter would make
more sense in this case. e.g. pxa310_use_312mhz.

>> Rgds, Pieter
>>
>>
>
> --
> Regards,
> Igor.
>
>

[PATCH 1/5] cpufreq-pxa3xx: use HSIO 312MHz on the pxa310

[Igor Grinberg]

On 07/12/10 12:29, Eric Miao wrote:
> On Mon, Jul 12, 2010 at 5:13 PM, Igor Grinberg <grinberg@compulab.co.il> wrote:
>   
>>
>> On 07/12/10 11:37, pieterg wrote:
>>     
>>> On Sunday 11 July 2010 09:14:22 Igor Grinberg wrote:
>>>
>>>       
>>>> On 07/09/10 11:31, pieterg wrote:
>>>>
>>>>         
>>>>> On Friday 09 July 2010 10:02:16 Eric Miao wrote:
>>>>>
>>>>>           
>>>>>> What's wrong with 208MHz HSS on PXA310?
>>>>>>
>>>>>>             
>>>>> I haven't done any benchmarking, but Marvell found it worth to
>>>>> introduce 312MHz for the pxa310. (I assume it might have something to
>>>>> do with the video accelerator, which is unique to the pxa310).
>>>>>
>>>>> And to me it makes sense to use the highest supported speeds in the top
>>>>> cpufreq entry.
>>>>>
>>>>>           
>>>> According to Marvell's PXA3xx Specification Update document
>>>> (page 68 - SCr#3), the above is correct starting from silicon revision
>>>> A2. What about earlier revisions? Shouldn't we support all of them?
>>>>
>>>>         
>>> OK, I missed that. In that case we should leave this one out.
>>>
>>>       
>> If you really want this to go, you can check the cpu stepping (cpuid
>> register)
>> and fixup the table accordingly. Something like:
>>
>> if (cpu_is_pxa310())
>>        ret = setup_freqs_table(policy, (read_cpuid_id() & 0xf) ?
>>
>> ARRAY_AND_SIZE(pxa310_freqs) :
>>
>> ARRAY_AND_SIZE(pxa300_freqs));
>>
>> but I don't know if Eric will like it ;)
>>
>>     
> Nah, I was thinking of introducing a cpu_is_pxa310_a2() or something,
> but that's really hard to make generic, since for this case, we'd
> have to use cpu_is_pxa310_a2_or_above(), which is #!$!%!@#$
>
> So to put it simple, I think a command line parameter would make
> more sense in this case. e.g. pxa310_use_312mhz.
>   

Indeed, command line parameter pxa310.hss=312/208 sounds good

>   
>>> Rgds, Pieter
>>>
>>>
>>>       
>> --
>> Regards,
>> Igor.
>>
>>
>>     
>   

-- 
Regards,
Igor.