All of lore.kernel.org
 help / color / mirror / Atom feed
From: Hans Verkuil <hverkuil@xs4all.nl>
To: Kirill Smelkov <kirr@mns.spb.ru>
Cc: Mauro Carvalho Chehab <mchehab@redhat.com>,
	Mauro Carvalho Chehab <mchehab@infradead.org>,
	linux-media@vger.kernel.org
Subject: Re: [PATCH v6] [media] vivi: Teach it to tune FPS
Date: Fri, 30 Nov 2012 12:10:59 +0100	[thread overview]
Message-ID: <201211301210.59509.hverkuil@xs4all.nl> (raw)
In-Reply-To: <20121130110214.GA10056@tugrik.mns.mnsspb.ru>

On Fri November 30 2012 12:02:14 Kirill Smelkov wrote:
> On Mon, Nov 19, 2012 at 09:52:36AM +0400, Kirill Smelkov wrote:
> > On Sun, Nov 18, 2012 at 07:25:38AM -0200, Mauro Carvalho Chehab wrote:
> > > Em 18-11-2012 07:24, Mauro Carvalho Chehab escreveu:
> > > >Em 17-11-2012 08:45, Kirill Smelkov escreveu:
> > > >>On Fri, Nov 16, 2012 at 01:46:58PM -0200, Mauro Carvalho Chehab wrote:
> > > >>>Em 16-11-2012 11:38, Hans Verkuil escreveu:
> > > >>>>On Wed November 7 2012 12:30:01 Kirill Smelkov wrote:
> > > >>[...]
> > > >>
> > > >>>>>diff --git a/drivers/media/platform/vivi.c b/drivers/media/platform/vivi.c
> > > >>>>>index 37d0af8..5d1b374 100644
> > > >>>>>--- a/drivers/media/platform/vivi.c
> > > >>>>>+++ b/drivers/media/platform/vivi.c
> > > >>>>>@@ -65,8 +65,11 @@ MODULE_PARM_DESC(vid_limit, "capture memory limit in megabytes");
> > > >>>>>  /* Global font descriptor */
> > > >>>>>  static const u8 *font8x16;
> > > >>>>>
> > > >>>>>-/* default to NTSC timeperframe */
> > > >>>>>-static const struct v4l2_fract TPF_DEFAULT = {.numerator = 1001, .denominator = 30000};
> > > >>>>>+/* timeperframe: min/max and default */
> > > >>>>>+static const struct v4l2_fract
> > > >>>>>+    tpf_min     = {.numerator = 1,        .denominator = UINT_MAX},  /* 1/infty */
> > > >>>>>+    tpf_max     = {.numerator = UINT_MAX,    .denominator = 1},         /* infty */
> > > >>>>
> > > >>>>I understand your reasoning here, but I wouldn't go with UINT_MAX here. Something like
> > > >>>>1/1000 tpf (or 1 ms) up to 86400/1 tpf (or once a day). With UINT_MAX I am afraid we
> > > >>>>might hit application errors when they manipulate these values. The shortest time
> > > >>>>between frames is 1 ms anyway.
> > > >>>>
> > > >>>>It's the only comment I have, it looks good otherwise.
> > > >>>
> > > >>>As those will be a arbitrary values, I suggest to declare a macro for it at the
> > > >>>beginning of vivi.c file, with some comment explaining the rationale of the choose,
> > > >>>and what else needs to be changed, if this changes (e. g. less than 1ms would require
> > > >>>changing the image generation logic, to avoid producing frames with equal content).
> > > >>
> > > >>Maybe something like this? (please note, I'm not a good text writer. If
> > > >>this needs adjustment please help me shape the text up)
> > > >>
> > > >>
> > > >>diff --git a/drivers/media/platform/vivi.c b/drivers/media/platform/vivi.c
> > > >>index 5d1b374..45b8a81 100644
> > > >>--- a/drivers/media/platform/vivi.c
> > > >>+++ b/drivers/media/platform/vivi.c
> > > >>@@ -36,6 +36,18 @@
> > > >>
> > > >>  #define VIVI_MODULE_NAME "vivi"
> > > >>
> > > >>+/* Maximum allowed frame rate
> > > >>+ *
> > > >>+ * Vivi will allow setting timeperframe in [1/FPS_MAX - FPS_MAX/1] range.
> > > >>+ *
> > > >>+ * Ideally FPS_MAX should be infinity, i.e. practically UINT_MAX, but that
> > > >>+ * might hit application errors when they manipulate these values.
> > > >>+ *
> > > >>+ * Besides, for tpf < 1ms image-generation logic should be changed, to avoid
> > > >>+ * producing frames with equal content.
> > > >>+ */
> > > >>+#define FPS_MAX 1000
> > > >>+
> > > >>  #define MAX_WIDTH 1920
> > > >>  #define MAX_HEIGHT 1200
> > > >>
> > > >>@@ -67,8 +79,8 @@ static const u8 *font8x16;
> > > >>
> > > >>  /* timeperframe: min/max and default */
> > > >>  static const struct v4l2_fract
> > > >>-    tpf_min     = {.numerator = 1,        .denominator = UINT_MAX},  /* 1/infty */
> > > >>-    tpf_max     = {.numerator = UINT_MAX,    .denominator = 1},         /* infty */
> > > >>+    tpf_min     = {.numerator = 1,        .denominator = FPS_MAX},   /* ~1/infty */
> > > >>+    tpf_max     = {.numerator = FPS_MAX,    .denominator = 1},         /* ~infty */
> > > 
> > > Was too fast answering it... The comments there should also be dropped, as it doesn't
> > > range anymore to infty.
> > 
> > Ok, agree, let's drop those ~infty comments and be done with it.
> > 
> > 
> > ---- 8< ----
> > From: Kirill Smelkov <kirr@mns.spb.ru>
> > Date: Tue, 23 Oct 2012 16:56:59 +0400
> > Subject: [PATCH v6] [media] vivi: Teach it to tune FPS
> > 
> > I was testing my video-over-ethernet subsystem recently, and vivi
> > seemed to be perfect video source for testing when one don't have lots
> > of capture boards and cameras. Only its framerate was hardcoded to
> > NTSC's 30fps, while in my country we usually use PAL (25 fps) and I
> > needed that to precisely simulate bandwidth.
> > 
> > That's why here is this patch with ->enum_frameintervals() and
> > ->{g,s}_parm() implemented as suggested by Hans Verkuil which passes
> > v4l2-compliance and manual testing through v4l2-ctl -P / -p <fps>.
> > 
> > Regarding newly introduced __get_format(u32 pixelformat) I decided not
> > to convert original get_format() to operate on fourcc codes, since >= 3
> > places in driver need to deal with v4l2_format and otherwise it won't be
> > handy.
> > 
> > Thanks.
> > 
> > Signed-off-by: Kirill Smelkov <kirr@mns.spb.ru>
> > ---
> >  drivers/media/platform/vivi.c | 102 ++++++++++++++++++++++++++++++++++++++----
> >  1 file changed, 94 insertions(+), 8 deletions(-)
> > 
> > V6:
> >     - Moved TPF/FPS limit to beginning of vivi.c and added comment there
> >       why that limit is used, to avoid overlows, as suggested by Mauro
> >       Carvalho Chehab.
> > 
> > V5:
> >     - changed  1/infty - infty/1  limits to  1/1000 - 1000/1, to avoid
> >       hitting aplication errors when they try to manipulato those
> >       values, as suggested by Hans Verkuil.
> > 
> > V4:
> >     - corrected fival->stepwise setting and added its check to s_parm();
> >       also cosmetics - all as per Hans Verkuil review.
> > 
> > V3:
> >     - corrected issues with V4L2 spec compliance as pointed by Hans
> >       Verkuil.
> > 
> > V2:
> > 
> >     - reworked FPS setting from module param to via ->{g,s}_parm() as suggested
> >       by Hans Verkuil.
> > 
> > 
> > diff --git a/drivers/media/platform/vivi.c b/drivers/media/platform/vivi.c
> > index 6e6dd25..9ca920e 100644
> > --- a/drivers/media/platform/vivi.c
> > +++ b/drivers/media/platform/vivi.c
> > @@ -36,9 +36,17 @@
> >  
> >  #define VIVI_MODULE_NAME "vivi"
> >  
> > -/* Wake up at about 30 fps */
> > -#define WAKE_NUMERATOR 30
> > -#define WAKE_DENOMINATOR 1001
> > +/* Maximum allowed frame rate
> > + *
> > + * Vivi will allow setting timeperframe in [1/FPS_MAX - FPS_MAX/1] range.
> > + *
> > + * Ideally FPS_MAX should be infinity, i.e. practically UINT_MAX, but that
> > + * might hit application errors when they manipulate these values.
> > + *
> > + * Besides, for tpf < 1ms image-generation logic should be changed, to avoid
> > + * producing frames with equal content.
> > + */
> > +#define FPS_MAX 1000
> >  
> >  #define MAX_WIDTH 1920
> >  #define MAX_HEIGHT 1200
> > @@ -69,6 +77,12 @@ MODULE_PARM_DESC(vid_limit, "capture memory limit in megabytes");
> >  /* Global font descriptor */
> >  static const u8 *font8x16;
> >  
> > +/* timeperframe: min/max and default */
> > +static const struct v4l2_fract
> > +	tpf_min     = {.numerator = 1,		.denominator = FPS_MAX},
> > +	tpf_max     = {.numerator = FPS_MAX,	.denominator = 1},
> > +	tpf_default = {.numerator = 1001,	.denominator = 30000};	/* NTSC */
> > +
> >  #define dprintk(dev, level, fmt, arg...) \
> >  	v4l2_dbg(level, debug, &dev->v4l2_dev, fmt, ## arg)
> >  
> > @@ -150,14 +164,14 @@ static struct vivi_fmt formats[] = {
> >  	},
> >  };
> >  
> > -static struct vivi_fmt *get_format(struct v4l2_format *f)
> > +static struct vivi_fmt *__get_format(u32 pixelformat)
> >  {
> >  	struct vivi_fmt *fmt;
> >  	unsigned int k;
> >  
> >  	for (k = 0; k < ARRAY_SIZE(formats); k++) {
> >  		fmt = &formats[k];
> > -		if (fmt->fourcc == f->fmt.pix.pixelformat)
> > +		if (fmt->fourcc == pixelformat)
> >  			break;
> >  	}
> >  
> > @@ -167,6 +181,11 @@ static struct vivi_fmt *get_format(struct v4l2_format *f)
> >  	return &formats[k];
> >  }
> >  
> > +static struct vivi_fmt *get_format(struct v4l2_format *f)
> > +{
> > +	return __get_format(f->fmt.pix.pixelformat);
> > +}
> > +
> >  /* buffer for one video frame */
> >  struct vivi_buffer {
> >  	/* common v4l buffer stuff -- must be first */
> > @@ -232,6 +251,7 @@ struct vivi_dev {
> >  
> >  	/* video capture */
> >  	struct vivi_fmt            *fmt;
> > +	struct v4l2_fract          timeperframe;
> >  	unsigned int               width, height;
> >  	struct vb2_queue	   vb_vidq;
> >  	unsigned int		   field_count;
> > @@ -691,8 +711,8 @@ static void vivi_thread_tick(struct vivi_dev *dev)
> >  	dprintk(dev, 2, "[%p/%d] done\n", buf, buf->vb.v4l2_buf.index);
> >  }
> >  
> > -#define frames_to_ms(frames)					\
> > -	((frames * WAKE_NUMERATOR * 1000) / WAKE_DENOMINATOR)
> > +#define frames_to_ms(dev, frames)				\
> > +	((frames * dev->timeperframe.numerator * 1000) / dev->timeperframe.denominator)
> >  
> >  static void vivi_sleep(struct vivi_dev *dev)
> >  {
> > @@ -708,7 +728,7 @@ static void vivi_sleep(struct vivi_dev *dev)
> >  		goto stop_task;
> >  
> >  	/* Calculate time to wake up */
> > -	timeout = msecs_to_jiffies(frames_to_ms(1));
> > +	timeout = msecs_to_jiffies(frames_to_ms(dev, 1));
> >  
> >  	vivi_thread_tick(dev);
> >  
> > @@ -1080,6 +1100,68 @@ static int vidioc_s_input(struct file *file, void *priv, unsigned int i)
> >  	return 0;
> >  }
> >  
> > +/* timeperframe is arbitrary and continous */
> > +static int vidioc_enum_frameintervals(struct file *file, void *priv,
> > +					     struct v4l2_frmivalenum *fival)
> > +{
> > +	struct vivi_fmt *fmt;
> > +
> > +	if (fival->index)
> > +		return -EINVAL;
> > +
> > +	fmt = __get_format(fival->pixel_format);
> > +	if (!fmt)
> > +		return -EINVAL;
> > +
> > +	/* regarding width & height - we support any */
> > +
> > +	fival->type = V4L2_FRMIVAL_TYPE_CONTINUOUS;
> > +
> > +	/* fill in stepwise (step=1.0 is requred by V4L2 spec) */
> > +	fival->stepwise.min  = tpf_min;
> > +	fival->stepwise.max  = tpf_max;
> > +	fival->stepwise.step = (struct v4l2_fract) {1, 1};
> > +
> > +	return 0;
> > +}
> > +
> > +static int vidioc_g_parm(struct file *file, void *priv, struct v4l2_streamparm *parm)
> > +{
> > +	struct vivi_dev *dev = video_drvdata(file);
> > +
> > +	if (parm->type != V4L2_BUF_TYPE_VIDEO_CAPTURE)
> > +		return -EINVAL;
> > +
> > +	parm->parm.capture.capability   = V4L2_CAP_TIMEPERFRAME;
> > +	parm->parm.capture.timeperframe = dev->timeperframe;
> > +	parm->parm.capture.readbuffers  = 1;
> > +	return 0;
> > +}
> > +
> > +#define FRACT_CMP(a, OP, b)	\
> > +	( (u64)(a).numerator * (b).denominator  OP  (u64)(b).numerator * (a).denominator )
> > +
> > +static int vidioc_s_parm(struct file *file, void *priv, struct v4l2_streamparm *parm)
> > +{
> > +	struct vivi_dev *dev = video_drvdata(file);
> > +	struct v4l2_fract tpf;
> > +
> > +	if (parm->type != V4L2_BUF_TYPE_VIDEO_CAPTURE)
> > +		return -EINVAL;
> > +
> > +	tpf = parm->parm.capture.timeperframe;
> > +
> > +	/* tpf: {*, 0} resets timing; clip to [min, max]*/
> > +	tpf = tpf.denominator ? tpf : tpf_default;
> > +	tpf = FRACT_CMP(tpf, <, tpf_min) ? tpf_min : tpf;
> > +	tpf = FRACT_CMP(tpf, >, tpf_max) ? tpf_max : tpf;
> > +
> > +	dev->timeperframe = tpf;
> > +	parm->parm.capture.timeperframe = tpf;
> > +	parm->parm.capture.readbuffers  = 1;
> > +	return 0;
> > +}
> > +
> >  /* --- controls ---------------------------------------------- */
> >  
> >  static int vivi_g_volatile_ctrl(struct v4l2_ctrl *ctrl)
> > @@ -1238,6 +1320,9 @@ static const struct v4l2_ioctl_ops vivi_ioctl_ops = {
> >  	.vidioc_enum_input    = vidioc_enum_input,
> >  	.vidioc_g_input       = vidioc_g_input,
> >  	.vidioc_s_input       = vidioc_s_input,
> > +	.vidioc_enum_frameintervals = vidioc_enum_frameintervals,
> > +	.vidioc_g_parm        = vidioc_g_parm,
> > +	.vidioc_s_parm        = vidioc_s_parm,
> >  	.vidioc_streamon      = vb2_ioctl_streamon,
> >  	.vidioc_streamoff     = vb2_ioctl_streamoff,
> >  	.vidioc_log_status    = v4l2_ctrl_log_status,
> > @@ -1296,6 +1381,7 @@ static int __init vivi_create_instance(int inst)
> >  		goto free_dev;
> >  
> >  	dev->fmt = &formats[0];
> > +	dev->timeperframe = tpf_default;
> >  	dev->width = 640;
> >  	dev->height = 480;
> >  	dev->pixelsize = dev->fmt->depth / 8;
> 
> 
> Mauro, Hans, what's the state of this patch? Just a ping, I know you are
> under load.
> 
> Thanks,
> Kirill
> 
> 
> P.S. Hans, if this is ok for you, would you please add your ack?
> 

Looks good to me!

Acked-by: Hans Verkuil <hans.verkuil@cisco.com>

Thanks,

	Hans

  reply	other threads:[~2012-11-30 11:12 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-10-22 13:54 [PATCH 1/2] [media] vivi: Kill BUFFER_TIMEOUT macro Kirill Smelkov
2012-10-22 13:54 ` [PATCH 2/2] [media] vivi: Teach it to tune FPS Kirill Smelkov
2012-10-22 14:16   ` Hans Verkuil
2012-10-22 17:01     ` Kirill Smelkov
2012-10-22 17:29       ` Kirill Smelkov
2012-10-23  6:27         ` Hans Verkuil
2012-10-23  6:40       ` Hans Verkuil
2012-10-23 13:35         ` [PATCH v3] " Kirill Smelkov
2012-11-02 14:09           ` Kirill Smelkov
2012-11-02 14:42           ` Hans Verkuil
2012-11-02 16:44             ` Kirill Smelkov
2012-11-07 11:30             ` [PATCH v4] " Kirill Smelkov
2012-11-12  8:12               ` Kirill Smelkov
2012-11-12  9:46                 ` Hans Verkuil
2012-11-12 10:45                   ` Kirill Smelkov
2012-11-16 13:38               ` Hans Verkuil
2012-11-16 14:48                 ` [PATCH v5] " Kirill Smelkov
2012-11-16 14:51                   ` Hans Verkuil
2012-11-16 15:02                     ` Kirill Smelkov
2012-11-16 15:46                 ` [PATCH v4] " Mauro Carvalho Chehab
2012-11-17 10:45                   ` Kirill Smelkov
2012-11-18  9:24                     ` Mauro Carvalho Chehab
2012-11-18  9:25                       ` Mauro Carvalho Chehab
2012-11-19  5:52                         ` [PATCH v6] " Kirill Smelkov
2012-11-30 11:02                           ` Kirill Smelkov
2012-11-30 11:10                             ` Hans Verkuil [this message]
2012-11-30 12:16                               ` Kirill Smelkov
2012-10-22 14:16 ` [PATCH 1/2] [media] vivi: Kill BUFFER_TIMEOUT macro Hans Verkuil

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=201211301210.59509.hverkuil@xs4all.nl \
    --to=hverkuil@xs4all.nl \
    --cc=kirr@mns.spb.ru \
    --cc=linux-media@vger.kernel.org \
    --cc=mchehab@infradead.org \
    --cc=mchehab@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.