On 08. 08. 23, 21:42, Mahmoud Matook wrote: >> On 08/08, Jiri Slaby wrote: >> >>>> On 07. 08. 23, 22:10, Mahmoud Maatuq wrote: >>>> coccinielle reported the following: >>>> ./drivers/net/wireless/ath/ath5k/phy.c:1573:25-26: WARNING opportunity for swap() >>>> >>>> >>> OK, once again: >>> https://lore.kernel.org/all/0c3acbd4-6ab2-5cc5-6293-54e30093cce2@xxxxxxxxxx/ >> >> >> I had a look at the commit history, to see why the original developer >> didn't use sort() function, and name the array variable sort but found >> nothing. >> I have some doubts that he might did that intentionally, so not to call >> sort() function for such small array > It happens once in 10 s and in a work. No worries about that. if it's the case so make sense to use sort(). but should I send the modifications on the same thread? as the subject needs to be changed. ex: repalce double loop with sort function replace bubble sort with heap sort. >> and avoid the cost of context switching. > What context switching? I used the wrong expression, I meant function call overhead e.g store/load registers. >>>> diff --git a/drivers/net/wireless/ath/ath5k/phy.c b/drivers/net/wireless/ath/ath5k/phy.c >>>> index 5797ef9c73d7..f87eb684f223 100644 >>>> --- a/drivers/net/wireless/ath/ath5k/phy.c >>>> +++ b/drivers/net/wireless/ath/ath5k/phy.c >>>> @@ -1562,16 +1562,13 @@ static s16 >>>> ath5k_hw_get_median_noise_floor(struct ath5k_hw *ah) >>>> { >>>> s16 sort[ATH5K_NF_CAL_HIST_MAX]; >>>> - s16 tmp; >>>> int i, j; >>>> memcpy(sort, ah->ah_nfcal_hist.nfval, sizeof(sort)); >>>> for (i = 0; i < ATH5K_NF_CAL_HIST_MAX - 1; i++) { >>>> for (j = 1; j < ATH5K_NF_CAL_HIST_MAX - i; j++) { >>>> if (sort[j] > sort[j - 1]) { >>>> - tmp = sort[j]; >>>> - sort[j] = sort[j - 1]; >>>> - sort[j - 1] = tmp; >>>> + swap(sort[j], sort[j - 1]); >>>> } >>>> } >>>> } >>> >>> -- >>> js >>> suse labs > > -- > js > suse labs
