[PATCH] input: egalax_ts: optmize driver for more fluency touch feelling

[PATCH] input: egalax_ts: optmize driver for more fluency touch feelling

[Zhang Jiejing]

egalax_ts use threaded_irq to report touch event, it will wake up the
irq thread and stop the thread irq after process one touch event, so
even user touch the screen very fast, the fastest event report speed
is depends on the scheduler, so the event reported is a uniform speed.

This patch will let thread irq check the gpio's status, if the gpio
is down, means still have more events, it will not exit the irq handler
and continue process another new event, it can improve user experience.

Signed-off-by: Zhang Jiejing <jiejing.zhang@freescale.com>
---
 drivers/input/touchscreen/egalax_ts.c |   63 +++++++++++++++++---------------
 1 files changed, 33 insertions(+), 30 deletions(-)

diff --git a/drivers/input/touchscreen/egalax_ts.c b/drivers/input/touchscreen/egalax_ts.c
index f694535..21bd2ec 100644
--- a/drivers/input/touchscreen/egalax_ts.c
+++ b/drivers/input/touchscreen/egalax_ts.c
@@ -74,45 +74,48 @@ static irqreturn_t egalax_ts_interrupt(int irq, void *dev_id)
 	u8 state;
 
 	do {
-		ret = i2c_master_recv(client, buf, MAX_I2C_DATA_LEN);
-	} while (ret == -EAGAIN && tries++ < EGALAX_MAX_TRIES);
+		do {
+			ret = i2c_master_recv(client, buf, MAX_I2C_DATA_LEN);
+		} while (ret == -EAGAIN && tries++ < EGALAX_MAX_TRIES);
 
-	if (ret < 0)
-		return IRQ_HANDLED;
+		if (ret < 0)
+			return IRQ_HANDLED;
 
-	if (buf[0] != REPORT_MODE_MTTOUCH) {
-		/* ignore mouse events and vendor events */
-		return IRQ_HANDLED;
-	}
+		if (buf[0] != REPORT_MODE_MTTOUCH) {
+			/* ignore mouse events and vendor events */
+			return IRQ_HANDLED;
+		}
 
-	state = buf[1];
-	x = (buf[3] << 8) | buf[2];
-	y = (buf[5] << 8) | buf[4];
-	z = (buf[7] << 8) | buf[6];
+		state = buf[1];
+		x = (buf[3] << 8) | buf[2];
+		y = (buf[5] << 8) | buf[4];
+		z = (buf[7] << 8) | buf[6];
 
-	valid = state & EVENT_VALID_MASK;
-	id = (state & EVENT_ID_MASK) >> EVENT_ID_OFFSET;
-	down = state & EVENT_DOWN_UP;
+		valid = state & EVENT_VALID_MASK;
+		id = (state & EVENT_ID_MASK) >> EVENT_ID_OFFSET;
+		down = state & EVENT_DOWN_UP;
 
-	if (!valid || id > MAX_SUPPORT_POINTS) {
-		dev_dbg(&client->dev, "point invalid\n");
-		return IRQ_HANDLED;
-	}
+		if (!valid || id > MAX_SUPPORT_POINTS) {
+			dev_dbg(&client->dev, "point invalid\n");
+			return IRQ_HANDLED;
+		}
 
-	input_mt_slot(input_dev, id);
-	input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, down);
+		input_mt_slot(input_dev, id);
+		input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, down);
 
-	dev_dbg(&client->dev, "%s id:%d x:%d y:%d z:%d",
-		(down ? "down" : "up"), id, x, y, z);
+		dev_dbg(&client->dev, "%s id:%d x:%d y:%d z:%d",
+			(down ? "down" : "up"), id, x, y, z);
 
-	if (down) {
-		input_report_abs(input_dev, ABS_MT_POSITION_X, x);
-		input_report_abs(input_dev, ABS_MT_POSITION_Y, y);
-		input_report_abs(input_dev, ABS_MT_PRESSURE, z);
-	}
+		if (down) {
+			input_report_abs(input_dev, ABS_MT_POSITION_X, x);
+			input_report_abs(input_dev, ABS_MT_POSITION_Y, y);
+			input_report_abs(input_dev, ABS_MT_PRESSURE, z);
+		}
+
+		input_mt_report_pointer_emulation(input_dev, true);
+		input_sync(input_dev);
+	} while (gpio_get_value(irq_to_gpio(client->irq)) == 0);
 
-	input_mt_report_pointer_emulation(input_dev, true);
-	input_sync(input_dev);
 	return IRQ_HANDLED;
 }
 
-- 
1.7.1

Re: [PATCH] input: egalax_ts: optmize driver for more fluency touch feelling

[Dmitry Torokhov]

Hi Zhang,

On Wed, Dec 07, 2011 at 05:10:02PM +0800, Zhang Jiejing wrote:
> egalax_ts use threaded_irq to report touch event, it will wake up the
> irq thread and stop the thread irq after process one touch event, so
> even user touch the screen very fast, the fastest event report speed
> is depends on the scheduler, so the event reported is a uniform speed.

I am not sure what you mean by "the event reported is a uniform speed"
but since I2C operations generally sleep you are at the mercy of
scheduler anyways. But since IRQ thread is a real-time one it should get
scheduled immediately.

> 
> This patch will let thread irq check the gpio's status, if the gpio
> is down, means still have more events, it will not exit the irq handler
> and continue process another new event, it can improve user experience.

Does it really or it is your opinion that it would?

Also, your change means that the driver can not be stopped/unloaded as
long as user had finger on the touchpad. This would need to be corrected
(but first you need to prove that the change is actually needed).

Thanks.

-- 
Dmitry

Re: [PATCH] input: egalax_ts: optmize driver for more fluency touch feelling

[Jiejing.Zhang]

Hi Dimitry,

2011/12/12 Dmitry Torokhov <dmitry.torokhov@gmail.com>
>
> Hi Zhang,
>
> On Wed, Dec 07, 2011 at 05:10:02PM +0800, Zhang Jiejing wrote:
> > egalax_ts use threaded_irq to report touch event, it will wake up the
> > irq thread and stop the thread irq after process one touch event, so
> > even user touch the screen very fast, the fastest event report speed
> > is depends on the scheduler, so the event reported is a uniform speed.
>
> I am not sure what you mean by "the event reported is a uniform speed"
> but since I2C operations generally sleep you are at the mercy of
> scheduler anyways. But since IRQ thread is a real-time one it should get
> scheduled immediately.

I'm using 2.6.38 kernel and Android 4.0 and with a Freescale ARM
A9(1.0GHz) Soc doing this test, but in my test, if I sweep on the
touch screen with very fast speed, my UI can't move as fast as my
finger,
I first suspect our GPU not powerful to drive UI, but after test with
USB mouse doing same sweep operation, the UI was moving quite fast.

So I made this change, after made this change, I can't feed any lay
between finger and the UI animation.
I think the threaded_irq not quite real time.

I checked kernel/irq/handle.c file,
I found in handle_IRQ_event() functin, it will use
wake_up_processs(action->thread);  and the thread was created by
kthread_create() in __setup_irq() function.

I think in this implementation, it will be not so realtime, it depends
on schedule time.
>
>
> >
> > This patch will let thread irq check the gpio's status, if the gpio
> > is down, means still have more events, it will not exit the irq handler
> > and continue process another new event, it can improve user experience.
>
> Does it really or it is your opinion that it would?
>
As I mentioned in above, I got this result with a real machine and
environment  test.
>
> Also, your change means that the driver can not be stopped/unloaded as
> long as user had finger on the touchpad. This would need to be corrected
> (but first you need to prove that the change is actually needed).
>


I need test in the driver unload case and suspend case.

Thanks for you advance.

BR,
Jiejing

>
> Thanks.
>
> --
> Dmitry
> --
> To unsubscribe from this list: send the line "unsubscribe linux-input" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html