lkml.org 
[lkml]   [2021]   [May]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v2 3/3] PM / EM: Skip inefficient OPPs
From
Date


On 5/25/21 10:46 AM, Vincent Donnefort wrote:
> On Tue, May 25, 2021 at 09:33:01AM +0000, Quentin Perret wrote:
>> On Friday 21 May 2021 at 17:54:24 (+0100), Vincent Donnefort wrote:
>>> @@ -161,6 +162,8 @@ static int em_create_perf_table(struct device *dev, struct em_perf_domain *pd,
>>> table[i].cost = div64_u64(fmax * table[i].power,
>>> table[i].frequency);
>>> if (table[i].cost >= prev_cost) {
>>> + table[i].flags = EM_PERF_STATE_INEFFICIENT;
>>> + pd->flags |= EM_PERF_DOMAIN_INEFFICIENCIES;
>>
>> If we're looking for micro-optimizations, then perhaps you could store
>> the index of the next efficient OPP (which would be 'i' if the current
>> OPP is already efficient), so you can jump to it directly when doing the
>> search.
>
> Wouldn't add any new field compared to this version so yeah might be an
> interesting improvement.
>

That's a few more instructions to parse the 'flags' filed. I'm not sure
if that brings speed improvements vs. if we not parse and have bool
filed with a simple looping. The out-of-order core might even suffer
from this parsing and loop index manipulations...

\
 
 \ /
  Last update: 2021-05-25 13:05    [W:0.106 / U:0.168 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site