Use swap() instead of implementing it since it makes code cleaner.
Signed-off-by: Salah Triki <salah.triki@gmail.com>
Link: https://lore.kernel.org/r/20210713140521.GA1873885@pc
[p.zabel@pengutronix.de: add "image-convert:" prefix to commit description]
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Use a bit-mask of EOF irqs to determine when all required idmac
channel EOFs have been received for a tile conversion, and only do
tile completion processing after all EOFs have been received. Otherwise
it was found that a conversion would stall after the completion of a
tile and the start of the next tile, because the input/read idmac
channel had not completed and entered idle state, thus locking up the
channel when attempting to re-start it for the next tile.
Fixes: 0537db801b ("gpu: ipu-v3: image-convert: reconfigure IC per tile")
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Combine the rotate_irq() and norotate_irq() handlers into a single
eof_irq() handler.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The first pixel of the next tile is only sampled by the hardware if the
fractional input position corresponding to the last written output pixel
is not an integer position.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Burst aligned input and output width can be calculated once per column,
instead of repeatedly for each tile in the column. The same goes for
input and output height per row. Also don't round up the same values
repeatedly.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
If we managed to create tiles sized 0x0 because of a bug in the seam
calculation, return with an error message instead of letting the driver
run into a division by zero later. Also check for tile sizes that are
larger than supported by the hardware.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
This patch effectively reverts commit 912bbf7e9c ("gpu: ipu-v3:
image-convert: Fix image downsize coefficients") and replaces it with a
different solution based on the preceding patches.
The previous fix tried to solve the problem of intermediate tile size
between IC downsizing and main processing sections not being limited to
1024 pixels by downsizing the input image to a smaller intermediate size
in the downsizing box filter. This causes unnecessary blurring,
especially for scaling factors close to 1.
Now that the seam position calculation makes sure that the 1024 pixel
intermediate tile size limit is not exceeded, calculate the number of
tiles from the maximum of intermediate size and output size and avoid
unnecessary downsizing.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Limit the input seam position to an interval that guarantees the tile
size does not exceed 1024 pixels after the IC downsizing section and
that space is left for the next tile.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
This fixes a failure to determine any seam if the output size is
exactly 1024 multiplied by the number of tiles in a given direction.
In that case an empty interval out_start == out_end is being passed
to find_best_seam, which looks for a seam out_start <= x < out_end.
Also reduce the interval for all but the left column / top row, to
avoid returning position 0 as best fit.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
For the write channels with 4:2:0 subsampled YUV formats, avoid chroma
overdraw by only writing chroma for even lines (skip odd chroma rows).
This reduces necessary write memory bandwidth by at least 25% (more
with rotation enabled).
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Only providing the input and output RGB/YUV space to the IC task init
functions is not sufficient. To fully characterize a colorspace
conversion, the Y'CbCr encoding standard, and quantization also
need to be specified.
Define a 'struct ipu_ic_colorspace' that includes all the above.
This allows to actually enforce the fact that the IC:
- can only encode to/from YUV and RGB full range. A follow-up patch will
remove this restriction.
- can only encode using BT.601 standard. A follow-up patch will add
Rec.709 encoding support.
The determination of the CSC coefficients based on the input/output
'struct ipu_ic_colorspace' are moved to a new exported function
ipu_ic_calc_csc(), and 'struct ic_csc_params' is exported as
'struct ipu_ic_csc_params'. ipu_ic_calc_csc() fills a 'struct ipu_ic_csc'
with the input/output 'struct ipu_ic_colorspace' and the calculated
'struct ic_csc_params' from those input/output colorspaces.
The functions ipu_ic_task_init(_rsc)() now take a filled 'struct
ipu_ic_csc'.
The existing CSC coefficient tables and ipu_ic_calc_csc() are moved
to a new module ipu-ic-csc.c. This is in preparation for adding more
coefficient tables for limited range quantization and more encoding
standards.
The existing ycbcr2rgb and inverse rgb2ycbcr tables defined the BT.601
Y'CbCr encoding coefficients. The rgb2ycbcr table specifically described
the BT.601 encoding from full range RGB to full range YUV. Table
comments have been added in ipu-ic-csc.c to make this more clear.
The ycbcr2rgb inverse table described encoding YUV limited range to RGB
full range. To be consistent with the rgb2ycbcr table, this table is
converted to YUV full range to RGB full range, and the comments are
expanded in ipu-ic-csc.c.
The ic_csc_rgb2rgb table was just an identity matrix, so it is renamed
'identity' in ipu-ic-csc.c.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
[p.zabel@pengutronix.de: removed a superfluous blank line]
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The output of the IC downsizer unit in both dimensions must be <= 1024
before being passed to the IC resizer unit. This was causing corrupted
images when:
input_dim > 1024, and
input_dim / 2 < output_dim < input_dim
Some broken examples were 1920x1080 -> 1024x768 and 1920x1080 ->
1280x1080.
Fixes: 70b9b6b3bc ("gpu: ipu-v3: image-convert: calculate per-tile
resize coefficients")
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The input bytesperline calculation for packed pixel formats was
incorrect. The min/max clamping values must be multiplied by the
packed bits-per-pixel. This was causing corrupted converted images
when the input format was RGB4 (probably also other input packed
formats).
Fixes: d966e23d61 ("gpu: ipu-v3: image-convert: fix bytesperline
adjustment")
Reported-by: Harsha Manjula Mallikarjun <Harsha.ManjulaMallikarjun@in.bosch.com>
Suggested-by: Harsha Manjula Mallikarjun <Harsha.ManjulaMallikarjun@in.bosch.com>
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The output width and height alignment values were being used in the
input bytesperline calculation. Fix by separating local vars w_align
and h_align into w_align_in, h_align_in, w_align_out, and h_align_out.
Fixes: d966e23d61 ("gpu: ipu-v3: image-convert: fix bytesperline
adjustment")
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Based on 3 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version [author] [kishon] [vijay] [abraham]
[i] [kishon]@[ti] [com] this program is distributed in the hope that
it will be useful but without any warranty without even the implied
warranty of merchantability or fitness for a particular purpose see
the gnu general public license for more details
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version [author] [graeme] [gregory]
[gg]@[slimlogic] [co] [uk] [author] [kishon] [vijay] [abraham] [i]
[kishon]@[ti] [com] [based] [on] [twl6030]_[usb] [c] [author] [hema]
[hk] [hemahk]@[ti] [com] this program is distributed in the hope
that it will be useful but without any warranty without even the
implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 1105 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Richard Fontana <rfontana@redhat.com>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070033.202006027@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If width or height are in the [2049, 3072] range, allow to
use just three tiles in this dimension, instead of four.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Double-buffering only works if tile sizes are the same and the resizing
coefficient does not change between tiles, even for non-planar formats.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Visualize the scaling and rotation pipeline with some ASCII art
diagrams. Remove the FIXME comment about missing seam prevention.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
For planar formats, bytesperline does not depend on BPP. It must always
be larger than width and aligned to tile width alignment restrictions.
The input bytesperline to ipu_image_convert_adjust() may be
uninitialized, so don't rely on input bytesperline as the
minimum value for clamp_align(). Use 2 << w_align as the minimum
instead.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
[slongerbeam@gmail.com: clamp input bytesperline]
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
For the planar but U/V-packed formats NV12 and NV16, 8 pixel width
alignment is good enough to fulfill the 8 byte stride requirement.
If we allow the input 8-pixel DMA bursts to overshoot the end of the
line, the only input alignment restrictions are dictated by the pixel
format and 8-byte aligned line start address.
Since different tile sizes are allowed, the output tile with / height
alignment doesn't need to be multiplied by number of columns / rows.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
[slongerbeam@gmail.com: Bring in the fixes to format width and
height alignment restrictions from imx-media-mem2mem.c.]
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Since tile dimensions now vary between tiles, add debug output for each
tile's position and dimensions.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Select seam positions that minimize distortions during seam hiding while
satifying input and output IDMAC, rotator, and image format constraints.
This code looks for aligned output seam positions that minimize the
difference between the fractional corresponding ideal input positions
and the input positions rounded to alignment requirements.
Since now tiles can be sized differently, alignment restrictions of the
complete image can be relaxed in the next step.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Move tile_width_align and tile_height_align up so they
can be used by the tile edge position calculation code.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
This will allow to calculate seam positions after initializing the
ipu_image base structure but before calculating tile dimensions.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Store tile top/left position in pixels in the tile structure.
This will allow overlapping tiles with different sizes later.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
For differently sized tiles or if the resizing coefficients change,
we have to stop, reconfigure, and restart the IC between tiles.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Slightly modifying resize coefficients per-tile allows to completely
hide the seams between tiles and to sample the correct input pixels at
the bottom and right edges of the image.
Tiling requires a bilinear interpolator reset at each tile start, which
causes the image to be slightly shifted if the starting pixel should not
have been sampled from an integer pixel position in the source image
according to the full image resizing ratio. To work around this
hardware limitation, calculate per-tile resizing coefficients that make
sure that the correct input pixels are sampled at the tile end.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Let convert_start start from a given tile index, allocate intermediate
tile with maximum tile size.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Steve Longerbeam <slongerbeam@gmail.com>
Catch calculated tile offsets that are not 8-byte aligned as required by the
IDMAC engine and return error in calc_tile_offsets().
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The need_abort flag is not really needed anymore in
__ipu_image_convert_abort(), remove it.
No functional changes.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Allow reentrancy into ipu_image_convert_abort(), by moving re-init
of ctx->aborted completion under the spin lock, and only if there is
an active run, and complete all waiters do_bh(). Note:
ipu_image_convert_unprepare() is still _not_ reentrant, and can't
be made reentrant.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Only wait for the ctx->aborted completion if there is an active run
in progress, otherwise the wait will just timeout after 10 seconds.
If there is no active run in progress, the done queue just needs to
be emptied.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Prevent possible race by parallel threads between ipu_image_convert_run()
and ipu_image_convert_unprepare(). This involves setting ctx->aborting
to true unconditionally so that no new job runs can be queued during
unprepare, and holding the ctx->aborting flag until the context is freed.
Note that the "normal" ipu_image_convert_abort() case (e.g. not during
context unprepare) should clear the ctx->aborting flag after aborting
any active run and clearing the context's pending queue. This is because
it should be possible to continue to use the conversion context and queue
more runs after an abort.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Allow the caller of ipu_cpmem_set_image() to override the latters
calculation of the chroma plane offsets, by adding override U/V
plane offsets to 'struct ipu_image'.
Signed-off-by: Steve Longerbeam <slongerbeam@gmail.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Using non-zero AXI IDs for anything other than the display channels
collides with the PRG AXI snooping, so only do this if there is no
PRG present.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Use ERR_CAST inlined function instead of ERR_PTR(PTR_ERR(...)).
Generated by: scripts/coccinelle/api/err_cast.cocci
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
This patch implements image conversion support using the IC tasks, with
tiling to support scaling to and from images up to 4096x4096. Image
rotation is also supported. Image conversion requests are added to
a run queue under the IC tasks.
The internal API is subsystem agnostic (no V4L2 dependency except
for the use of V4L2 fourcc pixel formats).
Callers prepare for image conversion by calling
ipu_image_convert_prepare(), which initializes the parameters of
the conversion. The caller passes in the ipu and IC task to use for
the conversion, the input and output image formats, a rotation mode,
and a completion callback and completion context pointer:
struct ipu_image_converter_ctx *
ipu_image_convert_prepare(struct ipu_soc *ipu, enum ipu_ic_task ic_task,
struct ipu_image *in, struct ipu_image *out,
enum ipu_rotate_mode rot_mode,
ipu_image_converter_cb_t complete,
void *complete_context);
A new conversion context is created that is added to an IC task
context queue. The caller is given the new conversion context,
which can then be passed to the further APIs:
int ipu_image_convert_queue(struct ipu_image_converter_run *run);
This queues the given image conversion request run to a run queue,
and starts the conversion immediately if the run queue is empty. Only
the physaddr's of the input and output image buffers are needed,
since the conversion context was created previously with
ipu_image_convert_prepare(). When the conversion completes, the run
pointer is returned to the completion callback.
void ipu_image_convert_abort(struct ipu_image_converter_ctx *ctx);
This will abort any active or pending conversions for this context.
Any currently active or pending runs belonging to this context are
returned via the completion callback with an error status.
void ipu_image_convert_unprepare(struct ipu_image_converter_ctx *ctx);
Unprepares the conversion context. Any active or pending runs will
be aborted by calling ipu_image_convert_abort().
Signed-off-by: Steve Longerbeam <steve_longerbeam@mentor.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
