2015-06-16 08:40:22

by Zefir Kurtisi

[permalink] [raw]
Subject: [PATCH] ath: DFS - limit number of potential PRI sequences

In the PRI detector, after the current radar pulse
has been checked agains existing PRI sequences, it
is considered as part of a new potential sequence.

Previously, the condition to accept a new sequence
was to have at least the same number of pulses as
the longest matching sequence. This was wrong,
since it led to duplicates of PRI sequences.

This patch changes the acceptance criteria for new
potential sequences from 'at least' to 'more than'
the longest existing.

Detection performance remains unaffected, while
the number of PRI sequences accounted at runtime
(and with it CPU load) is reduced by up to 50%.

Signed-off-by: Zefir Kurtisi <[email protected]>
---
drivers/net/wireless/ath/dfs_pri_detector.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/wireless/ath/dfs_pri_detector.c b/drivers/net/wireless/ath/dfs_pri_detector.c
index 1b5ad19..cc5c592 100644
--- a/drivers/net/wireless/ath/dfs_pri_detector.c
+++ b/drivers/net/wireless/ath/dfs_pri_detector.c
@@ -273,7 +273,7 @@ static bool pseq_handler_create_sequences(struct pri_detector *pde,
tmp_false_count++;
}
}
- if (ps.count < min_count)
+ if (ps.count <= min_count)
/* did not reach minimum count, drop sequence */
continue;

--
1.9.1



2015-06-30 12:34:43

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] ath: DFS - limit number of potential PRI sequences

Zefir Kurtisi <[email protected]> writes:

> In the PRI detector, after the current radar pulse
> has been checked agains existing PRI sequences, it
> is considered as part of a new potential sequence.
>
> Previously, the condition to accept a new sequence
> was to have at least the same number of pulses as
> the longest matching sequence. This was wrong,
> since it led to duplicates of PRI sequences.
>
> This patch changes the acceptance criteria for new
> potential sequences from 'at least' to 'more than'
> the longest existing.
>
> Detection performance remains unaffected, while
> the number of PRI sequences accounted at runtime
> (and with it CPU load) is reduced by up to 50%.
>
> Signed-off-by: Zefir Kurtisi <[email protected]>

Thanks, applied to ath.git.

--
Kalle Valo