Hi Hans,
On Tuesday 17 December 2013 01:30 PM, Hans Verkuil wrote:
On 12/12/2013 09:35 AM, Archit Taneja wrote:
The VPE and VIP IPs in DRA7x contain common scaler and color conversion hardware
blocks. We create libraries for these components such that the vpe driver and
the vip driver(in future) can use these library funcs to configure the blocks.
There are some points for which I would like comments:
- For VPE, setting the format and colorspace for the source and destination
queues is enough to determine how these blocks need to be configured and
whether they need to be bypassed or not. So it didn't make sense to represent
them as media controller entities. For VIP(driver not upstream yet), it's
possible that there are multiple data paths which may or may not include these
blocks. However, the current use cases don't require such flexibility. There
may be a need to re-consider a media controller like setup once we work on the
VIP driver. Is it a good idea in terms of user-space compatibilty if we use
media controller framework in the future.
As long as you don't need the mc, then there is no need to implement it.
The thing is that we want to use these blocks for a future capture
hardware called VIP. A VIP port can capture multiple streams from
different sensors in time slices, and each of those streams might be
sharing the same hardware, but probably in different configurations. I'm
not completely aware of media controller details, and whether it can
help us here.
I was just wondering if it would be a problem if we later realise that
mc might be useful for some use cases. Would implementing it then break
previous user space applications which don't call mc ioctls?
- These blocks will also require some custom control commands later on. For
example, we may want to tell the scaler later on to perform bi-linear
scaling, or perform peaking at a particular frequency.
- The current series keeps the default scaler coefficients in a header file.
These coefficients add a lot of lines of code in the kernel. Does it make more
sense for the user application to pass the co-efficients to the kernel using
an ioctl? Is there any driver which currenlty does this?
I think it is good to keep it in the driver. Otherwise apps would be forced to
set up the table. It's about 11 kilobyte in memory, which isn't that bad.
Okay.
The series is based on the branch:
git://linuxtv.org/media_tree.git master
Archit Taneja (8):
v4l: ti-vpe: create a scaler block library
v4l: ti-vpe: support loading of scaler coefficients
v4l: ti-vpe: make vpe driver load scaler coefficients
v4l: ti-vpe: enable basic scaler support
v4l: ti-vpe: create a color space converter block library
v4l: ti-vpe: Add helper to perform color conversion
v4l: ti-vpe: enable CSC support for VPE
v4l: ti-vpe: Add a type specifier to describe vpdma data format type
drivers/media/platform/ti-vpe/Makefile | 2 +-
drivers/media/platform/ti-vpe/csc.c | 196 +++++
drivers/media/platform/ti-vpe/csc.h | 68 ++
drivers/media/platform/ti-vpe/sc.c | 311 +++++++
drivers/media/platform/ti-vpe/sc.h | 208 +++++
drivers/media/platform/ti-vpe/sc_coeff.h | 1342 ++++++++++++++++++++++++++++++
drivers/media/platform/ti-vpe/vpdma.c | 36 +-
drivers/media/platform/ti-vpe/vpdma.h | 7 +
drivers/media/platform/ti-vpe/vpe.c | 251 ++++--
drivers/media/platform/ti-vpe/vpe_regs.h | 187 -----
10 files changed, 2335 insertions(+), 273 deletions(-)
create mode 100644 drivers/media/platform/ti-vpe/csc.c
create mode 100644 drivers/media/platform/ti-vpe/csc.h
create mode 100644 drivers/media/platform/ti-vpe/sc.c
create mode 100644 drivers/media/platform/ti-vpe/sc.h
create mode 100644 drivers/media/platform/ti-vpe/sc_coeff.h
For this patch series:
Acked-by: Hans Verkuil <hans.verkuil@xxxxxxxxx>
Thanks for the review.
Archit
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
