[Intel-gfx] [PATCH 2/2]drm/i915: Use?best PLL timing values for G4X chipsets

[Fu Michael]

Zhenyu Wang wrote:
> On 2009.03.10 21:00:25 +0800, Ma Ling wrote:
>   
>> +
>> +	memset (best_clock, 0, sizeof (*best_clock));
>> +	max_n = limit->n.max;
>> +	/* based on hardware requriment prefer smaller n to precision */
>> +	for (clock.n = limit->n.min; clock.n <= max_n; clock.n++) {
>> +		/* based on hardware requirment prefere larger m1,m2, p1 */
>> +		for (clock.m1 = limit->m1.max;
>> +		     clock.m1 >= limit->m1.min; clock.m1--) {
>> +			for (clock.m2 = limit->m2.max;
>> +			     clock.m2 >= limit->m2.min; clock.m2--) {
>> +				for (clock.p1 = limit->p1.max;
>> +				     clock.p1 >= limit->p1.min; clock.p1--) {
>> +					int this_err;
>> +
>> +					intel_clock (refclk, &clock);
>> +					if (!intel_PLL_is_valid(crtc, &clock))
>> +						continue;
>> +					this_err = abs(clock.dot - target) ;
>> +					if (this_err < err_most) {
>> +						*best_clock = clock;
>> +						err_most = this_err;
>> +						/*chose smaller n*/
>> +						max_n = clock.n;
>> +						found = true;
>>     
>
> If we should prefer larger m1,m2,p1, why not exit the cycle here and return? 
> Here we still try to look for the smallest error for target but might not be preferred
> largest m1,m2,p1?
>   
precision override m1,m2,p1 ( if a smaller m1,m2,p1 can get even better 
precision, we go with smaller combination )
N override precision ( if a bigger N can get even better precision in 
the math, we still go with smaller N )

> And is it possible if no clock can be found that meets the error precision of target?
> Should we guard for that case?
>
> And how about your test result on g4x with KMS?
>
>