| Message ID | 20190628101127.6044-2-john@phrozen.org (mailing list archive) |
|---|---|
| State | RFC |
| Delegated to: | Kalle Valo |
| Headers | show |
| Series | [RFC,1/2] ath11k: drop tx_status_fifo | expand |
John Crispin <john@phrozen.org> writes: > This in a small but measurable performance increase. > > Signed-off-by: John Crispin <john@phrozen.org> > --- > drivers/net/wireless/ath/ath11k/dp_tx.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/wireless/ath/ath11k/dp_tx.c b/drivers/net/wireless/ath/ath11k/dp_tx.c > index b3d5995fc657..8c5db3c88a87 100644 > --- a/drivers/net/wireless/ath/ath11k/dp_tx.c > +++ b/drivers/net/wireless/ath/ath11k/dp_tx.c > @@ -448,7 +448,6 @@ void ath11k_dp_tx_completion_handler(struct ath11k_base *ab, int ring_id) > int hal_ring_id = dp->tx_ring[ring_id].tcl_comp_ring.ring_id; > struct hal_srng *status_ring = &ab->hal.srng_list[hal_ring_id]; > struct sk_buff *msdu; > - struct hal_tx_status ts; > struct dp_tx_ring *tx_ring = &dp->tx_ring[ring_id]; > u32 msdu_id; > u32 *desc; > @@ -459,7 +458,8 @@ void ath11k_dp_tx_completion_handler(struct ath11k_base *ab, int ring_id) > ath11k_hal_srng_access_begin(ab, status_ring); > > while ((desc = ath11k_hal_srng_dst_get_next_entry(ab, status_ring))) { > - memset(&ts, 0, sizeof(ts)); > + struct hal_tx_status ts = { 0 }; Very interesting, so a compiler zeroes memory from stack faster than memset() from "heap" (or whatever kmalloced memory should be called)? Just out of curiosity, any ideas why memset() is so slow? Also some kind of numbers (before and after) would be nice to have in the commit log. It's not uncommon that years after we investigate why a certain change was made.
diff --git a/drivers/net/wireless/ath/ath11k/dp_tx.c b/drivers/net/wireless/ath/ath11k/dp_tx.c index b3d5995fc657..8c5db3c88a87 100644 --- a/drivers/net/wireless/ath/ath11k/dp_tx.c +++ b/drivers/net/wireless/ath/ath11k/dp_tx.c @@ -448,7 +448,6 @@ void ath11k_dp_tx_completion_handler(struct ath11k_base *ab, int ring_id) int hal_ring_id = dp->tx_ring[ring_id].tcl_comp_ring.ring_id; struct hal_srng *status_ring = &ab->hal.srng_list[hal_ring_id]; struct sk_buff *msdu; - struct hal_tx_status ts; struct dp_tx_ring *tx_ring = &dp->tx_ring[ring_id]; u32 msdu_id; u32 *desc; @@ -459,7 +458,8 @@ void ath11k_dp_tx_completion_handler(struct ath11k_base *ab, int ring_id) ath11k_hal_srng_access_begin(ab, status_ring); while ((desc = ath11k_hal_srng_dst_get_next_entry(ab, status_ring))) { - memset(&ts, 0, sizeof(ts)); + struct hal_tx_status ts = { 0 }; + ath11k_hal_tx_status_parse(ab, (struct hal_wbm_release_ring *)desc, &ts);
This in a small but measurable performance increase. Signed-off-by: John Crispin <john@phrozen.org> --- drivers/net/wireless/ath/ath11k/dp_tx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)