2018-11-09 07:40:28

by Malathi Gottam

[permalink] [raw]
Subject: [PATCH v2] media: venus: add support for selection rectangles

Handles target type crop by setting the new active rectangle
to hardware. The new rectangle should be within YUV size.

Signed-off-by: Malathi Gottam <[email protected]>
---
drivers/media/platform/qcom/venus/venc.c | 26 ++++++++++++++++++++++----
1 file changed, 22 insertions(+), 4 deletions(-)

diff --git a/drivers/media/platform/qcom/venus/venc.c b/drivers/media/platform/qcom/venus/venc.c
index ce85962..d26c129 100644
--- a/drivers/media/platform/qcom/venus/venc.c
+++ b/drivers/media/platform/qcom/venus/venc.c
@@ -478,16 +478,34 @@ static int venc_g_fmt(struct file *file, void *fh, struct v4l2_format *f)
venc_s_selection(struct file *file, void *fh, struct v4l2_selection *s)
{
struct venus_inst *inst = to_inst(file);
+ int ret;
+ u32 buftype;

if (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT)
return -EINVAL;

switch (s->target) {
case V4L2_SEL_TGT_CROP:
- if (s->r.width != inst->out_width ||
- s->r.height != inst->out_height ||
- s->r.top != 0 || s->r.left != 0)
- return -EINVAL;
+ if (s->r.left != 0) {
+ s->r.width += s->r.left;
+ s->r.left = 0;
+ }
+
+ if (s->r.top != 0) {
+ s->r.height += s->r.top;
+ s->r.top = 0;
+ }
+
+ if (s->r.width > inst->width)
+ s->r.width = inst->width;
+ else
+ inst->width = s->r.width;
+
+ if (s->r.height > inst->height)
+ s->r.height = inst->height;
+ else
+ inst->height = s->r.height;
+
break;
default:
return -EINVAL;
--
1.9.1



2018-11-09 07:57:20

by Tomasz Figa

[permalink] [raw]
Subject: Re: [PATCH v2] media: venus: add support for selection rectangles

Hi Malathi,

On Fri, Nov 9, 2018 at 4:39 PM Malathi Gottam <[email protected]> wrote:
>
> Handles target type crop by setting the new active rectangle
> to hardware. The new rectangle should be within YUV size.
>
> Signed-off-by: Malathi Gottam <[email protected]>
> ---
> drivers/media/platform/qcom/venus/venc.c | 26 ++++++++++++++++++++++----
> 1 file changed, 22 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/media/platform/qcom/venus/venc.c b/drivers/media/platform/qcom/venus/venc.c
> index ce85962..d26c129 100644
> --- a/drivers/media/platform/qcom/venus/venc.c
> +++ b/drivers/media/platform/qcom/venus/venc.c
> @@ -478,16 +478,34 @@ static int venc_g_fmt(struct file *file, void *fh, struct v4l2_format *f)
> venc_s_selection(struct file *file, void *fh, struct v4l2_selection *s)
> {
> struct venus_inst *inst = to_inst(file);
> + int ret;
> + u32 buftype;
>
> if (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT)
> return -EINVAL;
>
> switch (s->target) {
> case V4L2_SEL_TGT_CROP:
> - if (s->r.width != inst->out_width ||
> - s->r.height != inst->out_height ||
> - s->r.top != 0 || s->r.left != 0)
> - return -EINVAL;
> + if (s->r.left != 0) {
> + s->r.width += s->r.left;
> + s->r.left = 0;
> + }
> +
> + if (s->r.top != 0) {
> + s->r.height += s->r.top;
> + s->r.top = 0;
> + }
> +
> + if (s->r.width > inst->width)
> + s->r.width = inst->width;
> + else
> + inst->width = s->r.width;
> +
> + if (s->r.height > inst->height)
> + s->r.height = inst->height;
> + else
> + inst->height = s->r.height;
> +

From semantic point of view, it looks fine, but where is the rectangle
actually set to the hardware?

Best regards,
Tomasz

2018-11-09 09:21:24

by Malathi Gottam

[permalink] [raw]
Subject: Re: [PATCH v2] media: venus: add support for selection rectangles

On 2018-11-09 07:56, Tomasz Figa wrote:
> Hi Malathi,
>
> On Fri, Nov 9, 2018 at 4:39 PM Malathi Gottam <[email protected]>
> wrote:
>>
>> Handles target type crop by setting the new active rectangle
>> to hardware. The new rectangle should be within YUV size.
>>
>> Signed-off-by: Malathi Gottam <[email protected]>
>> ---
>> drivers/media/platform/qcom/venus/venc.c | 26
>> ++++++++++++++++++++++----
>> 1 file changed, 22 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/media/platform/qcom/venus/venc.c
>> b/drivers/media/platform/qcom/venus/venc.c
>> index ce85962..d26c129 100644
>> --- a/drivers/media/platform/qcom/venus/venc.c
>> +++ b/drivers/media/platform/qcom/venus/venc.c
>> @@ -478,16 +478,34 @@ static int venc_g_fmt(struct file *file, void
>> *fh, struct v4l2_format *f)
>> venc_s_selection(struct file *file, void *fh, struct v4l2_selection
>> *s)
>> {
>> struct venus_inst *inst = to_inst(file);
>> + int ret;
>> + u32 buftype;
>>
>> if (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT)
>> return -EINVAL;
>>
>> switch (s->target) {
>> case V4L2_SEL_TGT_CROP:
>> - if (s->r.width != inst->out_width ||
>> - s->r.height != inst->out_height ||
>> - s->r.top != 0 || s->r.left != 0)
>> - return -EINVAL;
>> + if (s->r.left != 0) {
>> + s->r.width += s->r.left;
>> + s->r.left = 0;
>> + }
>> +
>> + if (s->r.top != 0) {
>> + s->r.height += s->r.top;
>> + s->r.top = 0;
>> + }
>> +
>> + if (s->r.width > inst->width)
>> + s->r.width = inst->width;
>> + else
>> + inst->width = s->r.width;
>> +
>> + if (s->r.height > inst->height)
>> + s->r.height = inst->height;
>> + else
>> + inst->height = s->r.height;
>> +
>
> From semantic point of view, it looks fine, but where is the rectangle
> actually set to the hardware?
>
> Best regards,
> Tomasz

As this set selection call occurs before the hfi session initialization,
for now we are holding these values in driver.

As this call is followed by VIDIOC_REQBUFS(), as a part of this
we have venc_init_session

static int venc_init_session(struct venus_inst *inst)
{
int ret;

ret = hfi_session_init(inst, inst->fmt_cap->pixfmt);
if (ret)
return ret;

ret = venus_helper_set_input_resolution(inst, inst->width,
inst->height);
if (ret)
goto deinit;

ret = venus_helper_set_output_resolution(inst, inst->width,
inst->height,
HFI_BUFFER_OUTPUT);
if (ret)
goto deinit;

ret = venus_helper_set_color_format(inst, inst->fmt_out->pixfmt);
if (ret)
goto deinit;

ret = venc_set_properties(inst);


From here we set these values to hardware.

2018-11-09 10:32:43

by Tomasz Figa

[permalink] [raw]
Subject: Re: [PATCH v2] media: venus: add support for selection rectangles

Hi Malathi,

On Fri, Nov 9, 2018 at 6:20 PM <[email protected]> wrote:
>
> On 2018-11-09 07:56, Tomasz Figa wrote:
> > Hi Malathi,
> >
> > On Fri, Nov 9, 2018 at 4:39 PM Malathi Gottam <[email protected]>
> > wrote:
> >>
> >> Handles target type crop by setting the new active rectangle
> >> to hardware. The new rectangle should be within YUV size.
> >>
> >> Signed-off-by: Malathi Gottam <[email protected]>
> >> ---
> >> drivers/media/platform/qcom/venus/venc.c | 26
> >> ++++++++++++++++++++++----
> >> 1 file changed, 22 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/media/platform/qcom/venus/venc.c
> >> b/drivers/media/platform/qcom/venus/venc.c
> >> index ce85962..d26c129 100644
> >> --- a/drivers/media/platform/qcom/venus/venc.c
> >> +++ b/drivers/media/platform/qcom/venus/venc.c
> >> @@ -478,16 +478,34 @@ static int venc_g_fmt(struct file *file, void
> >> *fh, struct v4l2_format *f)
> >> venc_s_selection(struct file *file, void *fh, struct v4l2_selection
> >> *s)
> >> {
> >> struct venus_inst *inst = to_inst(file);
> >> + int ret;
> >> + u32 buftype;
> >>
> >> if (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT)
> >> return -EINVAL;
> >>
> >> switch (s->target) {
> >> case V4L2_SEL_TGT_CROP:
> >> - if (s->r.width != inst->out_width ||
> >> - s->r.height != inst->out_height ||
> >> - s->r.top != 0 || s->r.left != 0)
> >> - return -EINVAL;
> >> + if (s->r.left != 0) {
> >> + s->r.width += s->r.left;
> >> + s->r.left = 0;
> >> + }
> >> +
> >> + if (s->r.top != 0) {
> >> + s->r.height += s->r.top;
> >> + s->r.top = 0;
> >> + }
> >> +
> >> + if (s->r.width > inst->width)
> >> + s->r.width = inst->width;
> >> + else
> >> + inst->width = s->r.width;
> >> +
> >> + if (s->r.height > inst->height)
> >> + s->r.height = inst->height;
> >> + else
> >> + inst->height = s->r.height;
> >> +
> >
> > From semantic point of view, it looks fine, but where is the rectangle
> > actually set to the hardware?
> >
> > Best regards,
> > Tomasz
>
> As this set selection call occurs before the hfi session initialization,
> for now we are holding these values in driver.
>
> As this call is followed by VIDIOC_REQBUFS(), as a part of this
> we have venc_init_session
>
> static int venc_init_session(struct venus_inst *inst)
> {
> int ret;
>
> ret = hfi_session_init(inst, inst->fmt_cap->pixfmt);
> if (ret)
> return ret;
>
> ret = venus_helper_set_input_resolution(inst, inst->width,
> inst->height);
> if (ret)
> goto deinit;
>
> ret = venus_helper_set_output_resolution(inst, inst->width,
> inst->height,
> HFI_BUFFER_OUTPUT);

Something sounds not right here. Shouldn't one of the width/height be
the OUPUT format and the other the selection CROP target rectangle?

> if (ret)
> goto deinit;
>
> ret = venus_helper_set_color_format(inst, inst->fmt_out->pixfmt);
> if (ret)
> goto deinit;
>
> ret = venc_set_properties(inst);
>
>
> From here we set these values to hardware.

Okay, thanks for the explanation. In this case, we must return -EBUSY
if selection is attempted to be set after the session is initialized.

Best regards,
Tomasz

2018-11-12 08:36:32

by Alexandre Courbot

[permalink] [raw]
Subject: Re: [PATCH v2] media: venus: add support for selection rectangles

On Fri, Nov 9, 2018 at 7:31 PM Tomasz Figa <[email protected]> wrote:
>
> Hi Malathi,
>
> On Fri, Nov 9, 2018 at 6:20 PM <[email protected]> wrote:
> >
> > On 2018-11-09 07:56, Tomasz Figa wrote:
> > > Hi Malathi,
> > >
> > > On Fri, Nov 9, 2018 at 4:39 PM Malathi Gottam <[email protected]>
> > > wrote:
> > >>
> > >> Handles target type crop by setting the new active rectangle
> > >> to hardware. The new rectangle should be within YUV size.
> > >>
> > >> Signed-off-by: Malathi Gottam <[email protected]>
> > >> ---
> > >> drivers/media/platform/qcom/venus/venc.c | 26
> > >> ++++++++++++++++++++++----
> > >> 1 file changed, 22 insertions(+), 4 deletions(-)
> > >>
> > >> diff --git a/drivers/media/platform/qcom/venus/venc.c
> > >> b/drivers/media/platform/qcom/venus/venc.c
> > >> index ce85962..d26c129 100644
> > >> --- a/drivers/media/platform/qcom/venus/venc.c
> > >> +++ b/drivers/media/platform/qcom/venus/venc.c
> > >> @@ -478,16 +478,34 @@ static int venc_g_fmt(struct file *file, void
> > >> *fh, struct v4l2_format *f)
> > >> venc_s_selection(struct file *file, void *fh, struct v4l2_selection
> > >> *s)
> > >> {
> > >> struct venus_inst *inst = to_inst(file);
> > >> + int ret;
> > >> + u32 buftype;

Looks like these two variables are not used anymore?

> > >>
> > >> if (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT)
> > >> return -EINVAL;
> > >>
> > >> switch (s->target) {
> > >> case V4L2_SEL_TGT_CROP:
> > >> - if (s->r.width != inst->out_width ||
> > >> - s->r.height != inst->out_height ||
> > >> - s->r.top != 0 || s->r.left != 0)
> > >> - return -EINVAL;
> > >> + if (s->r.left != 0) {
> > >> + s->r.width += s->r.left;
> > >> + s->r.left = 0;
> > >> + }
> > >> +
> > >> + if (s->r.top != 0) {
> > >> + s->r.height += s->r.top;
> > >> + s->r.top = 0;
> > >> + }
> > >> +
> > >> + if (s->r.width > inst->width)
> > >> + s->r.width = inst->width;
> > >> + else
> > >> + inst->width = s->r.width;
> > >> +
> > >> + if (s->r.height > inst->height)
> > >> + s->r.height = inst->height;
> > >> + else
> > >> + inst->height = s->r.height;
> > >> +
> > >
> > > From semantic point of view, it looks fine, but where is the rectangle
> > > actually set to the hardware?
> > >
> > > Best regards,
> > > Tomasz
> >
> > As this set selection call occurs before the hfi session initialization,
> > for now we are holding these values in driver.
> >
> > As this call is followed by VIDIOC_REQBUFS(), as a part of this
> > we have venc_init_session
> >
> > static int venc_init_session(struct venus_inst *inst)
> > {
> > int ret;
> >
> > ret = hfi_session_init(inst, inst->fmt_cap->pixfmt);
> > if (ret)
> > return ret;
> >
> > ret = venus_helper_set_input_resolution(inst, inst->width,
> > inst->height);
> > if (ret)
> > goto deinit;
> >
> > ret = venus_helper_set_output_resolution(inst, inst->width,
> > inst->height,
> > HFI_BUFFER_OUTPUT);
>
> Something sounds not right here. Shouldn't one of the width/height be
> the OUPUT format and the other the selection CROP target rectangle?

Yeah, I don't see where the stride of the input (from HFI perspective)
buffer is set, so unless I missed something that would not capture the
correct part of the buffer.

Also doesn't Venus support non-zero left and top parameters for the
selection rectangle?