Re: SM501: Implement acceleration features

[Ben Dooks <ben@simtec.co.uk>]

Vincent Sanders wrote:
> On Wed, Nov 11, 2009 at 12:58:25AM +0100, Krzysztof Helt wrote:
>> On Tue, 10 Nov 2009 17:18:10 +0000
>> Ben Dooks <ben@simtec.co.uk> wrote:
>>
>>> From: Vincent Sanders <vince@simtec.co.uk>
>>>
>>> This patch provides the acceleration entry points for the SM501
>>> framebuffer driver.
>>>
>>> This patch provides the sync, copyarea and fillrect entry points,
>>> using the SM501's 2D acceleration engine to perform the operations
>>> in-chip rather than across the bus.
>>>
>>> Signed-off-by: Vincent Sanders <vince@simtec.co.uk>
>>> Signed-off-by: Simtec Linux Team <linux@simtec.co.uk>
>>> Signed-off-by: Ben Dooks <ben@simtec.co.uk>
>>>
>>> ---
>>>  drivers/video/sm501fb.c    |  238 ++++++++++++++++++++++++++++++++++++++++++---
>>>  include/linux/sm501-regs.h |    2 
>>>  2 files changed, 226 insertions(+), 14 deletions(-)
>>>
>>> Index: b/drivers/video/sm501fb.c
>>> ===================================================================
>>> --- a/drivers/video/sm501fb.c	2009-11-03 11:19:44.000000000 +0000
>>> +++ b/drivers/video/sm501fb.c	2009-11-03 11:19:46.000000000 +0000
> 
> I have snipped all but small amount for context
> 
>>> +	/* wait for the 2d engine to be ready */
>>> +	while ((count > 0) &&
>>> +	       (readl(fbi->regs + SM501_SYSTEM_CONTROL) &
>>> +		SM501_SYSCTRL_2D_ENGINE_STATUS) != 0)
>>> +		count--;
>>> +
>> You may add cpu_relax() inside this loop.
>>
> 
> ok, would need to test this thoroughly as the SM501 has some...odd
> behaviours (see later on).
> 
>>> +
>>> +	if (sm501fb_sync(info))
>>> +		return;
>>> +
>> Please check if you need to wait for the blit engine before writting
>> to any register.  Usually, the values in the bit engine registers
>> are shadowed after the engine is started (ie. the blitting operation
>> is started) and the next set of values can be written into the regs.
> 
> indeed, if it were sane I would agree, it isnt in all circumstances, see later

It might be worth caching the data that relies on bpp after set_par
call, and thus avoiding another swtich on bpp in the acceleration
routines.

Otherwise not much else to say.

-- 
Ben Dooks, Software Engineer, Simtec Electronics

http://www.simtec.co.uk/