Re: [PATCH v2 1/3] media: cedrus: Fix decoding for some H264 videos

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On Wed 02 Oct 19, 21:35, Jernej Skrabec wrote:
> It seems that for some H264 videos at least one bitstream parsing
> trigger must be called in order to be decoded correctly. There is no
> explanation why this helps, but it was observed that two sample videos
> with this fix are now decoded correctly and there is no regression with
> others.

I understand there might be some magic going on under the hood here, but I would
be interested in trying to understand what's really going on.

For instance, comparing register dumps of the whole H264 block before/after
calling the hardware parser could help, and comparing that to using the previous
code (without hardware parsing).

I could try and have a look if you have an available sample for testing the
erroneous case!

Another minor thing: do you have some idea of whether the udelay call adds
significant delay in the process?

Cheers and thanks for the patch!

Paul

> Signed-off-by: Jernej Skrabec <jernej.skrabec@xxxxxxxx>
> ---
>  .../staging/media/sunxi/cedrus/cedrus_h264.c  | 30 +++++++++++++++++--
>  .../staging/media/sunxi/cedrus/cedrus_regs.h  |  3 ++
>  2 files changed, 30 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_h264.c b/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> index d6a782703c9b..bd848146eada 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> @@ -6,6 +6,7 @@
>   * Copyright (c) 2018 Bootlin
>   */
>  
> +#include <linux/delay.h>
>  #include <linux/types.h>
>  
>  #include <media/videobuf2-dma-contig.h>
> @@ -289,6 +290,28 @@ static void cedrus_write_pred_weight_table(struct cedrus_ctx *ctx,
>  	}
>  }
>  
> +/*
> + * It turns out that using VE_H264_VLD_OFFSET to skip bits is not reliable. In
> + * rare cases frame is not decoded correctly. However, setting offset to 0 and
> + * skipping appropriate amount of bits with flush bits trigger always works.
> + */
> +static void cedrus_skip_bits(struct cedrus_dev *dev, int num)
> +{
> +	int count = 0;
> +
> +	while (count < num) {
> +		int tmp = min(num - count, 32);
>
> +
> +		cedrus_write(dev, VE_H264_TRIGGER_TYPE,
> +			     VE_H264_TRIGGER_TYPE_FLUSH_BITS |
> +			     VE_H264_TRIGGER_TYPE_N_BITS(tmp));
> +		while (cedrus_read(dev, VE_H264_STATUS) & VE_H264_STATUS_VLD_BUSY)
> +			udelay(1);
> +
> +		count += tmp;
> +	}
> +}
> +
>  static void cedrus_set_params(struct cedrus_ctx *ctx,
>  			      struct cedrus_run *run)
>  {
> @@ -299,12 +322,11 @@ static void cedrus_set_params(struct cedrus_ctx *ctx,
>  	struct vb2_buffer *src_buf = &run->src->vb2_buf;
>  	struct cedrus_dev *dev = ctx->dev;
>  	dma_addr_t src_buf_addr;
> -	u32 offset = slice->header_bit_size;
> -	u32 len = (slice->size * 8) - offset;
> +	u32 len = slice->size * 8;
>  	u32 reg;
>  
>  	cedrus_write(dev, VE_H264_VLD_LEN, len);
> -	cedrus_write(dev, VE_H264_VLD_OFFSET, offset);
> +	cedrus_write(dev, VE_H264_VLD_OFFSET, 0);
>  
>  	src_buf_addr = vb2_dma_contig_plane_dma_addr(src_buf, 0);
>  	cedrus_write(dev, VE_H264_VLD_END,
> @@ -323,6 +345,8 @@ static void cedrus_set_params(struct cedrus_ctx *ctx,
>  	cedrus_write(dev, VE_H264_TRIGGER_TYPE,
>  		     VE_H264_TRIGGER_TYPE_INIT_SWDEC);
>  
> +	cedrus_skip_bits(dev, slice->header_bit_size);
> +
>  	if (((pps->flags & V4L2_H264_PPS_FLAG_WEIGHTED_PRED) &&
>  	     (slice->slice_type == V4L2_H264_SLICE_TYPE_P ||
>  	      slice->slice_type == V4L2_H264_SLICE_TYPE_SP)) ||
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_regs.h b/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> index 3329f9aaf975..b52926a54025 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> @@ -538,13 +538,16 @@
>  					 VE_H264_CTRL_SLICE_DECODE_INT)
>  
>  #define VE_H264_TRIGGER_TYPE		0x224
> +#define VE_H264_TRIGGER_TYPE_N_BITS(x)		(((x) & 0x3f) << 8)
>  #define VE_H264_TRIGGER_TYPE_AVC_SLICE_DECODE	(8 << 0)
>  #define VE_H264_TRIGGER_TYPE_INIT_SWDEC		(7 << 0)
> +#define VE_H264_TRIGGER_TYPE_FLUSH_BITS		(3 << 0)
>  
>  #define VE_H264_STATUS			0x228
>  #define VE_H264_STATUS_VLD_DATA_REQ_INT		VE_H264_CTRL_VLD_DATA_REQ_INT
>  #define VE_H264_STATUS_DECODE_ERR_INT		VE_H264_CTRL_DECODE_ERR_INT
>  #define VE_H264_STATUS_SLICE_DECODE_INT		VE_H264_CTRL_SLICE_DECODE_INT
> +#define VE_H264_STATUS_VLD_BUSY			BIT(8)
>  
>  #define VE_H264_STATUS_INT_MASK			VE_H264_CTRL_INT_MASK
>  
> -- 
> 2.23.0
> 

Attachment: signature.asc
Description: PGP signature


[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux