Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII;
 format=flowed
Content-Transfer-Encoding: 7bit
Date:   Mon, 19 Feb 2018 16:19:48 +0530
From:   Abhishek Sahu <absahu@codeaurora.org>
To:     Sricharan R <sricharan@codeaurora.org>
Cc:     Andy Gross <andy.gross@linaro.org>,
        Wolfram Sang <wsa@the-dreams.de>,
        David Brown <david.brown@linaro.org>,
        linux-arm-msm@vger.kernel.org, linux-soc@vger.kernel.org,
        linux-i2c@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 07/12] i2c: qup: use the complete transfer length to
 choose DMA mode
In-Reply-To: <4a3d98e2-c340-1a13-eaa4-dde07768e7d8@codeaurora.org>
References: <1517644697-30806-1-git-send-email-absahu@codeaurora.org>
 <1517644697-30806-8-git-send-email-absahu@codeaurora.org>
 <4a3d98e2-c340-1a13-eaa4-dde07768e7d8@codeaurora.org>
Message-ID: <0142c0722cd0d444711918741e3691ab@codeaurora.org>
User-Agent: Roundcube Webmail/1.2.5
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 2018-02-16 10:05, Sricharan R wrote:
> On 2/3/2018 1:28 PM, Abhishek Sahu wrote:
>> Currently each message length in complete transfer is being
>> checked for determining DMA mode and if any of the message length
>> is less than FIFO length then non DMA mode is being used which
>> will increase overhead. DMA can be used for any length and it
>> should be determined with complete transfer length. Now, this
>> patch introduces DMA threshold length and the transfer will be
>> done in DMA mode if the total length is greater than this
>> threshold length.
>> 
>> Signed-off-by: Abhishek Sahu <absahu@codeaurora.org>
>> ---
>>  drivers/i2c/busses/i2c-qup.c | 15 +++++++++------
>>  1 file changed, 9 insertions(+), 6 deletions(-)
>> 
>> diff --git a/drivers/i2c/busses/i2c-qup.c 
>> b/drivers/i2c/busses/i2c-qup.c
>> index 6227a5c..a91fc70 100644
>> --- a/drivers/i2c/busses/i2c-qup.c
>> +++ b/drivers/i2c/busses/i2c-qup.c
>> @@ -192,6 +192,8 @@ struct qup_i2c_dev {
>>  	bool			is_dma;
>>  	/* To check if the current transfer is using DMA */
>>  	bool			use_dma;
>> +	/* The threshold length above which DMA will be used */
>> +	unsigned int		dma_threshold;
>>  	struct			dma_pool *dpool;
>>  	struct			qup_i2c_tag start_tag;
>>  	struct			qup_i2c_bam brx;
>> @@ -1294,7 +1296,8 @@ static int qup_i2c_xfer_v2(struct i2c_adapter 
>> *adap,
>>  			   int num)
>>  {
>>  	struct qup_i2c_dev *qup = i2c_get_adapdata(adap);
>> -	int ret, len, idx = 0;
>> +	int ret, idx = 0;
>> +	unsigned int total_len = 0;
>> 
>>  	qup->bus_err = 0;
>>  	qup->qup_err = 0;
>> @@ -1320,14 +1323,13 @@ static int qup_i2c_xfer_v2(struct i2c_adapter 
>> *adap,
>>  				goto out;
>>  			}
>> 
>> -			len = (msgs[idx].len > qup->out_fifo_sz) ||
>> -			      (msgs[idx].len > qup->in_fifo_sz);
>> -
>> -			if (is_vmalloc_addr(msgs[idx].buf) || !len)
>> +			if (is_vmalloc_addr(msgs[idx].buf))
>>  				break;
>> +
>> +			total_len += msgs[idx].len;
>>  		}
>> 
>> -		if (idx == num)
>> +		if (idx == num && total_len > qup->dma_threshold)
>>  			qup->use_dma = true;
>>  	}
>> 
>> @@ -1587,6 +1589,7 @@ static int qup_i2c_probe(struct platform_device 
>> *pdev)
>> 
>>  	size = QUP_INPUT_FIFO_SIZE(io_mode);
>>  	qup->in_fifo_sz = qup->in_blk_sz * (2 << size);
>> +	qup->dma_threshold = min(qup->out_fifo_sz, qup->in_fifo_sz);
> 
>   The patch is fine. small nit, you can avoid this global 
> dma_threshold. Instead
>   have it locally in qup_i2c_xfer_v2 itself.

  Thanks Sricharan.

  The idea for introducing global dma_threshold was to give option
  for controlling the behavior from one place. If someone wants to
  change the length above which dma will be used, then It can be
  controlled from once place in the probe function itself.

  Also, min(qup->out_fifo_sz, qup->in_fifo_sz) requires comparison
  and adding in qup_i2c_xfer_v2 will make this happen everytime.

  Regards,
  Abhishek