lkml.org 
[lkml]   [2022]   [Aug]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
SubjectRe: [RESEND PATCH 2/2] net: macb: Optimize reading HW timestamp
Date
On 16.08.2022 14:55, Harini Katakam wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>
> From: Harini Katakam <harini.katakam@xilinx.com>
>
> The seconds input from BD (6 bits) just needs to be ORed with the
> upper bits from timer in this function. Avoid +/- operations every
> single time. Check for seconds rollover at BIT 5 and subtract the
> overhead only in that case.
>
> Signed-off-by: Harini Katakam <harini.katakam@xilinx.com>
> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
> Acked-by: Richard Cochran <richardcochran@gmail.com>
> ---
> drivers/net/ethernet/cadence/macb_ptp.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/cadence/macb_ptp.c b/drivers/net/ethernet/cadence/macb_ptp.c
> index e6cb20aaa76a..674002661366 100644
> --- a/drivers/net/ethernet/cadence/macb_ptp.c
> +++ b/drivers/net/ethernet/cadence/macb_ptp.c
> @@ -247,6 +247,7 @@ static int gem_hw_timestamp(struct macb *bp, u32 dma_desc_ts_1,
> u32 dma_desc_ts_2, struct timespec64 *ts)
> {
> struct timespec64 tsu;
> + bool sec_rollover = false;
>
> ts->tv_sec = (GEM_BFEXT(DMA_SECH, dma_desc_ts_2) << GEM_DMA_SECL_SIZE) |
> GEM_BFEXT(DMA_SECL, dma_desc_ts_1);
> @@ -264,9 +265,12 @@ static int gem_hw_timestamp(struct macb *bp, u32 dma_desc_ts_1,
> */
> if ((ts->tv_sec & (GEM_DMA_SEC_TOP >> 1)) &&
> !(tsu.tv_sec & (GEM_DMA_SEC_TOP >> 1)))
> - ts->tv_sec -= GEM_DMA_SEC_TOP;
> + sec_rollover = true;
> +
> + ts->tv_sec |= ((~GEM_DMA_SEC_MASK) & tsu.tv_sec);

It looks to me that this instruction could be moved before

> if ((ts->tv_sec & (GEM_DMA_SEC_TOP >> 1)) &&
> !(tsu.tv_sec & (GEM_DMA_SEC_TOP >> 1)))

and get rid of sec_rolover extra variable thus, in the end, the diff could
look like this:

diff --git a/drivers/net/ethernet/cadence/macb_ptp.c
b/drivers/net/ethernet/cadence/macb_ptp.c
index e6cb20aaa76a..f9db4501b995 100644
--- a/drivers/net/ethernet/cadence/macb_ptp.c
+++ b/drivers/net/ethernet/cadence/macb_ptp.c
@@ -258,6 +258,8 @@ static int gem_hw_timestamp(struct macb *bp, u32
dma_desc_ts_1,
*/
gem_tsu_get_time(&bp->ptp_clock_info, &tsu, NULL);

+ ts->tv_sec |= ((~GEM_DMA_SEC_MASK) & tsu.tv_sec);
+
/* If the top bit is set in the timestamp,
* but not in 1588 timer, it has rolled over,
* so subtract max size
@@ -266,8 +268,6 @@ static int gem_hw_timestamp(struct macb *bp, u32
dma_desc_ts_1,
!(tsu.tv_sec & (GEM_DMA_SEC_TOP >> 1)))
ts->tv_sec -= GEM_DMA_SEC_TOP;

- ts->tv_sec += ((~GEM_DMA_SEC_MASK) & tsu.tv_sec);
-
return 0;
}

>
> - ts->tv_sec += ((~GEM_DMA_SEC_MASK) & tsu.tv_sec);
> + if (sec_rollover)
> + ts->tv_sec -= GEM_DMA_SEC_TOP;
>
> return 0;
> }
> --
> 2.17.1
>

\
 
 \ /
  Last update: 2022-08-19 10:22    [W:0.064 / U:0.108 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site