[RFC] ff-memless instant starting

[RFC] ff-memless instant starting

[Jari Vanhala]

[-- Attachment #1: Type: text/plain, Size: 723 bytes --]

Hi,

Looks like starting effect with low HZ machine takes considerably time
due need to wait for system timer. Simplest way to make things faster is
to do calculation for effect always instantly, like in stopping, but
this makes starting multiple effects more costly.

Other solution would be making it a module parameter, so behavior could
be controlled per machine. Optimize possible cost or speed. I attached
small patch which will do that.

If starting priority of individual effect is needed to controlled, you
probably need to add flag for that to effect itself... I don't know
which way FF-api is developing, so this could be either good or bad
thing.

Any opinions or suggestions? Other than rising HZ. =)

	++Jam


[-- Attachment #2: 0001-Input-Add-instant-option-to-ff-memless.patch --]
[-- Type: text/x-patch, Size: 1345 bytes --]

>From d3b6fa1eaac42f878f7c9b89b22ddfa85d900f38 Mon Sep 17 00:00:00 2001
From: Jari Vanhala <ext-jari.vanhala@nokia.com>
Date: Tue, 22 Dec 2009 16:49:29 +0200
Subject: [PATCH] Input: Add instant option to ff-memless

Allows Force Feedback effect to be started immediately
and not to wait for timer. Useful in low HZ systems.

Signed-off-by: Jari Vanhala <ext-jari.vanhala@nokia.com>
---
 drivers/input/ff-memless.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)

diff --git a/drivers/input/ff-memless.c b/drivers/input/ff-memless.c
index b483b29..dfd5830 100644
--- a/drivers/input/ff-memless.c
+++ b/drivers/input/ff-memless.c
@@ -67,6 +67,10 @@ struct ml_device {
 			   struct ff_effect *effect);
 };
 
+static int instant;
+module_param(instant, int, 0);
+MODULE_PARM_DESC(instant, "Start new effect instantly");
+
 static const struct ff_envelope *get_envelope(const struct ff_effect *effect)
 {
 	static const struct ff_envelope empty_envelope;
@@ -411,7 +415,10 @@ static int ml_ff_playback(struct input_dev *dev, int effect_id, int value)
 				 msecs_to_jiffies(state->effect->replay.length);
 		state->adj_at = state->play_at;
 
-		ml_schedule_timer(ml);
+		if (instant && !state->effect->replay.delay)
+			ml_play_effects(ml);
+		else
+			ml_schedule_timer(ml);
 
 	} else {
 		debug("initiated stop");
-- 
1.6.3.3

Re: [RFC] ff-memless instant starting

[Dmitry Torokhov]

On Wed, Dec 23, 2009 at 03:31:14PM +0200, Jari Vanhala wrote:
> Hi,
> 
> Looks like starting effect with low HZ machine takes considerably time
> due need to wait for system timer. Simplest way to make things faster is
> to do calculation for effect always instantly, like in stopping, but
> this makes starting multiple effects more costly.
> 
> Other solution would be making it a module parameter, so behavior could
> be controlled per machine. Optimize possible cost or speed. I attached
> small patch which will do that.
> 
> If starting priority of individual effect is needed to controlled, you
> probably need to add flag for that to effect itself... I don't know
> which way FF-api is developing, so this could be either good or bad
> thing.
> 
> Any opinions or suggestions? Other than rising HZ. =)
> 

I think we should just start playing the effect immediately, without any
additional module parameters. I'll queue such patch unless I hear
otherwise from Anssi.

-- 
Dmitry

Re: [RFC] ff-memless instant starting

[Anssi Hannula]

On 23.12.2009 21:38, Dmitry Torokhov wrote:
> On Wed, Dec 23, 2009 at 03:31:14PM +0200, Jari Vanhala wrote:
>> Hi,
>>
>> Looks like starting effect with low HZ machine takes considerably time
>> due need to wait for system timer. Simplest way to make things faster is
>> to do calculation for effect always instantly, like in stopping, but
>> this makes starting multiple effects more costly.
>>
>> Other solution would be making it a module parameter, so behavior could
>> be controlled per machine. Optimize possible cost or speed. I attached
>> small patch which will do that.
>>
>> If starting priority of individual effect is needed to controlled, you
>> probably need to add flag for that to effect itself... I don't know
>> which way FF-api is developing, so this could be either good or bad
>> thing.
>>
>> Any opinions or suggestions? Other than rising HZ. =)
>>
> 
> I think we should just start playing the effect immediately, without any
> additional module parameters. I'll queue such patch unless I hear
> otherwise from Anssi.

Indeed that is ok, no need for a module parameter.

-- 
Anssi Hannula