On 2019-09-27 10:35, Toke Høiland-Jørgensen wrote:
> Felix Fietkau <nbd@xxxxxxxx> writes:
>> + switch (status->encoding) {
>> + case RX_ENC_LEGACY:
>> + if (WARN_ON_ONCE(status->band > NL80211_BAND_5GHZ))
>> + return 0;
>> +
>> + sband = dev->hw->wiphy->bands[status->band];
>> + if (!sband || status->rate_idx > sband->n_bitrates)
>> + return 0;
>> +
>> + rate = &sband->bitrates[status->rate_idx];
>> +
>> + return mt76_calc_legacy_rate_duration(rate, sp, len);
>> + case RX_ENC_VHT:
>> + streams = status->nss;
>> + idx = status->rate_idx;
>> + group = VHT_GROUP_IDX(streams, sgi, bw);
>> + break;
>> + case RX_ENC_HT:
>> + streams = ((status->rate_idx >> 3) & 3) + 1;
>> + idx = status->rate_idx & 7;
>> + group = HT_GROUP_IDX(streams, sgi, bw);
>> + break;
>> + default:
>> + WARN_ON_ONCE(1);
>> + return 0;
>> + }
>> +
>> + if (WARN_ON_ONCE(streams > 4))
>> + return 0;
>> +
>> + duration = airtime_mcs_groups[group].duration[idx];
>> + duration <<= airtime_mcs_groups[group].shift;
>> + duration *= len;
>> + duration /= AVG_PKT_SIZE;
>> + duration /= 1024;
>
> On an earlier patch of mine you expressed concern over divisions in the
> fast path. Does this mean this is no longer a concern? Or is the
> compiler doing fancy things with the constant division here? :)
I made the constants power-of-2, so that the compiler can turn them into
bit shifts. I verified the disassembly: no divisions left in the code.
- Felix
