Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Michael Zaidman <michael.zaidman@gmail.com>
To:     jikos@kernel.org
Cc:     linux-kernel@vger.kernel.org, linux-input@vger.kernel.org,
        Michael Zaidman <michael.zaidman@gmail.com>,
        Guillaume Champagne <champagne.guillaume.c@gmail.com>
Subject: [PATCH v2 2/7] HID: ft260: improve i2c write performance
Date:   Wed, 28 Sep 2022 17:48:49 +0300
Message-Id: <20220928144854.5580-3-michael.zaidman@gmail.com>
In-Reply-To: <20220928144854.5580-1-michael.zaidman@gmail.com>
References: <20220928144854.5580-1-michael.zaidman@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

The patch improves i2c writing performance by about 30 percent by revising
the sleep time in the ft260_hid_output_report_check_status() in the
following ways:

1. Reduce the sleep time and start to poll earlier:

  Before:
    $ sudo ./i2cperf -f 2 -o 2 -s 32 -r 0-0xff 13 0x51 -S

      Fill block with increment via i2ctransfer by chunks
      -------------------------------------------------------------------
      data rate(bps)  efficiency(%)  data size(B)  total IOs   IO size(B)
      -------------------------------------------------------------------
      40510           80             256           8           32

  After:
    $ sudo ./i2cperf -f 2 -o 2 -s 32 -r 0-0xff 13 0x51 -S

      Fill block with increment via i2ctransfer by chunks
      -------------------------------------------------------------------
      data rate(bps)  efficiency(%)  data size(B)  total IOs   IO size(B)
      -------------------------------------------------------------------
      52584           80             256           8           32

2. Do not sleep when the calculated sleep time is below 2 ms:

  Before:
    $ sudo ./i2cperf -f 2 -o 2 -s 16 -r 0-0xff 13 0x51 -S

      Fill block with increment via i2ctransfer by chunks
      -------------------------------------------------------------------
      data rate(bps)  efficiency(%)  data size(B)  total IOs   IO size(B)
      -------------------------------------------------------------------
      26707           73             256           16          16

  After:
    $ sudo ./i2cperf -f 2 -o 2 -s 16 -r 0-0xff 13 0x51 -S

      Fill block with increment via i2ctransfer by chunks
      -------------------------------------------------------------------
      data rate(bps)  efficiency(%)  data size(B)  total IOs   IO size(B)
      -------------------------------------------------------------------
      37034           73             256           16          16

Link to the i2cperf - https://github.com/MichaelZaidman/i2cperf

Signed-off-by: Michael Zaidman <michael.zaidman@gmail.com>
Tested-by: Guillaume Champagne <champagne.guillaume.c@gmail.com>
---
 drivers/hid/hid-ft260.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/hid/hid-ft260.c b/drivers/hid/hid-ft260.c
index a35201d68b15..44106cadd746 100644
--- a/drivers/hid/hid-ft260.c
+++ b/drivers/hid/hid-ft260.c
@@ -345,7 +345,7 @@ static int ft260_hid_output_report(struct hid_device *hdev, u8 *data,
 static int ft260_hid_output_report_check_status(struct ft260_device *dev,
 						u8 *data, int len)
 {
-	int ret, usec, try = 3;
+	int ret, usec, try = 100;
 	struct hid_device *hdev = dev->hdev;
 
 	ret = ft260_hid_output_report(hdev, data, len);
@@ -356,10 +356,14 @@ static int ft260_hid_output_report_check_status(struct ft260_device *dev,
 		return ret;
 	}
 
-	/* transfer time = 1 / clock(KHz) * 10 bits * bytes */
-	usec = 10000 / dev->clock * len;
-	usleep_range(usec, usec + 100);
-	ft260_dbg("wait %d usec, len %d\n", usec, len);
+	/* transfer time = 1 / clock(KHz) * 9 bits * bytes */
+	usec = len * 9000 / dev->clock;
+	if (usec > 2000) {
+		usec -= 1500;
+		usleep_range(usec, usec + 100);
+		ft260_dbg("wait %d usec, len %d\n", usec, len);
+	}
+
 	do {
 		ret = ft260_xfer_status(dev);
 		if (ret != -EAGAIN)
-- 
2.34.1