linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
To: Jernej Skrabec <jernej.skrabec@siol.net>
Cc: mripard@kernel.org, mchehab@kernel.org, hverkuil-cisco@xs4all.nl,
	gregkh@linuxfoundation.org, wens@csie.org,
	linux-media@vger.kernel.org, devel@driverdev.osuosl.org,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org, linux-sunxi@googlegroups.com
Subject: Re: [PATCH 2/3] media: cedrus: Fix H264 4k support
Date: Mon, 4 Nov 2019 11:13:19 +0100	[thread overview]
Message-ID: <20191104101319.GE502900@aptenodytes> (raw)
In-Reply-To: <20191026074959.1073512-3-jernej.skrabec@siol.net>

[-- Attachment #1: Type: text/plain, Size: 8081 bytes --]

Hi,

On Sat 26 Oct 19, 09:49, Jernej Skrabec wrote:
> H264 decoder needs additional or bigger buffers in order to decode 4k
> videos.

Thanks for the fixup, we hadn't looked into those bits at all during initial
bringup of H.264!

See a few minor comments below.

> Signed-off-by: Jernej Skrabec <jernej.skrabec@siol.net>
> ---
>  drivers/staging/media/sunxi/cedrus/cedrus.h   |  7 ++
>  .../staging/media/sunxi/cedrus/cedrus_h264.c  | 83 +++++++++++++++++--
>  .../staging/media/sunxi/cedrus/cedrus_regs.h  | 11 +++
>  3 files changed, 93 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus.h b/drivers/staging/media/sunxi/cedrus/cedrus.h
> index c45fb9a7ad07..96765555ab8a 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus.h
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus.h
> @@ -116,8 +116,15 @@ struct cedrus_ctx {
>  			ssize_t		mv_col_buf_size;
>  			void		*pic_info_buf;
>  			dma_addr_t	pic_info_buf_dma;
> +			ssize_t		pic_info_buf_size;
>  			void		*neighbor_info_buf;
>  			dma_addr_t	neighbor_info_buf_dma;
> +			void		*deblk_buf;
> +			dma_addr_t	deblk_buf_dma;
> +			ssize_t		deblk_buf_size;
> +			void		*intra_pred_buf;
> +			dma_addr_t	intra_pred_buf_dma;
> +			ssize_t		intra_pred_buf_size;
>  		} h264;
>  		struct {
>  			void		*mv_col_buf;
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_h264.c b/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> index d2c854ecdf15..19962f4213d4 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_h264.c
> @@ -39,7 +39,6 @@ struct cedrus_h264_sram_ref_pic {
>  #define CEDRUS_H264_FRAME_NUM		18
>  
>  #define CEDRUS_NEIGHBOR_INFO_BUF_SIZE	(16 * SZ_1K)
> -#define CEDRUS_PIC_INFO_BUF_SIZE	(128 * SZ_1K)

Could we keep a define with the minimum size that you are using later
(increased to 130 * SZ_1K)?

>  static void cedrus_h264_write_sram(struct cedrus_dev *dev,
>  				   enum cedrus_h264_sram_off off,
> @@ -342,6 +341,20 @@ static void cedrus_set_params(struct cedrus_ctx *ctx,
>  		     VE_H264_VLD_ADDR_FIRST | VE_H264_VLD_ADDR_VALID |
>  		     VE_H264_VLD_ADDR_LAST);
>  
> +	if (ctx->src_fmt.width > 2048) {
> +		cedrus_write(dev, VE_BUF_CTRL,
> +			     VE_BUF_CTRL_INTRAPRED_MIXED_RAM |
> +			     VE_BUF_CTRL_DBLK_MIXED_RAM);
> +		cedrus_write(dev, VE_DBLK_DRAM_BUF_ADDR,
> +			     ctx->codec.h264.deblk_buf_dma);
> +		cedrus_write(dev, VE_INTRAPRED_DRAM_BUF_ADDR,
> +			     ctx->codec.h264.intra_pred_buf_dma);
> +	} else {
> +		cedrus_write(dev, VE_BUF_CTRL,
> +			     VE_BUF_CTRL_INTRAPRED_INT_SRAM |
> +			     VE_BUF_CTRL_DBLK_INT_SRAM);
> +	}
> +
>  	/*
>  	 * FIXME: Since the bitstream parsing is done in software, and
>  	 * in userspace, this shouldn't be needed anymore. But it
> @@ -502,18 +515,28 @@ static void cedrus_h264_setup(struct cedrus_ctx *ctx,
>  static int cedrus_h264_start(struct cedrus_ctx *ctx)
>  {
>  	struct cedrus_dev *dev = ctx->dev;
> +	unsigned int pic_info_size;
>  	unsigned int field_size;
>  	unsigned int mv_col_size;
>  	int ret;
>  

Maybe add a comment here this is a half-magic sub-optimal formula?

> +	if (ctx->src_fmt.width > 2048)
> +		pic_info_size = CEDRUS_H264_FRAME_NUM * 0x4000;
> +	else
> +		pic_info_size = CEDRUS_H264_FRAME_NUM * 0x1000;
> +
>  	/*
> -	 * FIXME: It seems that the H6 cedarX code is using a formula
> -	 * here based on the size of the frame, while all the older
> -	 * code is using a fixed size, so that might need to be
> -	 * changed at some point.
> +	 * FIXME: If V4L2_H264_SPS_FLAG_FRAME_MBS_ONLY is set,
> +	 * there is no need to multiply by 2.
>  	 */
> +	pic_info_size += ctx->src_fmt.height * 2 * 64;
> +
> +	if (pic_info_size < 130 * SZ_1K)
> +		pic_info_size = 130 * SZ_1K;

This is where I think we could have a "minimum pic info size" define.

> +
> +	ctx->codec.h264.pic_info_buf_size = pic_info_size;
>  	ctx->codec.h264.pic_info_buf =
> -		dma_alloc_coherent(dev->dev, CEDRUS_PIC_INFO_BUF_SIZE,
> +		dma_alloc_coherent(dev->dev, ctx->codec.h264.pic_info_buf_size,
>  				   &ctx->codec.h264.pic_info_buf_dma,
>  				   GFP_KERNEL);
>  	if (!ctx->codec.h264.pic_info_buf)
> @@ -566,15 +589,51 @@ static int cedrus_h264_start(struct cedrus_ctx *ctx)
>  		goto err_neighbor_buf;
>  	}
>  
> +	if (ctx->src_fmt.width > 2048) {

Feel free to add a comment here to explain where the 12 below comes from if you
have some idea, or that it's a somewhat magical value that generally works.

> +		ctx->codec.h264.deblk_buf_size =
> +			ALIGN(ctx->src_fmt.width, 32) * 12;
> +		ctx->codec.h264.deblk_buf =
> +			dma_alloc_coherent(dev->dev,
> +					   ctx->codec.h264.deblk_buf_size,
> +					   &ctx->codec.h264.deblk_buf_dma,
> +					   GFP_KERNEL);
> +		if (!ctx->codec.h264.deblk_buf) {
> +			ret = -ENOMEM;
> +			goto err_mv_col_buf;
> +		}
> +

Same here, a comment would be welcome about the 5 value below.

Cheers,

Paul

> +		ctx->codec.h264.intra_pred_buf_size =
> +			ALIGN(ctx->src_fmt.width, 64) * 5;
> +		ctx->codec.h264.intra_pred_buf =
> +			dma_alloc_coherent(dev->dev,
> +					   ctx->codec.h264.intra_pred_buf_size,
> +					   &ctx->codec.h264.intra_pred_buf_dma,
> +					   GFP_KERNEL);
> +		if (!ctx->codec.h264.intra_pred_buf) {
> +			ret = -ENOMEM;
> +			goto err_deblk_buf;
> +		}
> +	}
> +
>  	return 0;
>  
> +err_deblk_buf:
> +	dma_free_coherent(dev->dev, ctx->codec.h264.deblk_buf_size,
> +			  ctx->codec.h264.deblk_buf,
> +			  ctx->codec.h264.deblk_buf_dma);
> +
> +err_mv_col_buf:
> +	dma_free_coherent(dev->dev, ctx->codec.h264.mv_col_buf_size,
> +			  ctx->codec.h264.mv_col_buf,
> +			  ctx->codec.h264.mv_col_buf_dma);
> +
>  err_neighbor_buf:
>  	dma_free_coherent(dev->dev, CEDRUS_NEIGHBOR_INFO_BUF_SIZE,
>  			  ctx->codec.h264.neighbor_info_buf,
>  			  ctx->codec.h264.neighbor_info_buf_dma);
>  
>  err_pic_buf:
> -	dma_free_coherent(dev->dev, CEDRUS_PIC_INFO_BUF_SIZE,
> +	dma_free_coherent(dev->dev, ctx->codec.h264.pic_info_buf_size,
>  			  ctx->codec.h264.pic_info_buf,
>  			  ctx->codec.h264.pic_info_buf_dma);
>  	return ret;
> @@ -590,9 +649,17 @@ static void cedrus_h264_stop(struct cedrus_ctx *ctx)
>  	dma_free_coherent(dev->dev, CEDRUS_NEIGHBOR_INFO_BUF_SIZE,
>  			  ctx->codec.h264.neighbor_info_buf,
>  			  ctx->codec.h264.neighbor_info_buf_dma);
> -	dma_free_coherent(dev->dev, CEDRUS_PIC_INFO_BUF_SIZE,
> +	dma_free_coherent(dev->dev, ctx->codec.h264.pic_info_buf_size,
>  			  ctx->codec.h264.pic_info_buf,
>  			  ctx->codec.h264.pic_info_buf_dma);
> +	if (ctx->codec.h264.deblk_buf_size)
> +		dma_free_coherent(dev->dev, ctx->codec.h264.deblk_buf_size,
> +				  ctx->codec.h264.deblk_buf,
> +				  ctx->codec.h264.deblk_buf_dma);
> +	if (ctx->codec.h264.intra_pred_buf_size)
> +		dma_free_coherent(dev->dev, ctx->codec.h264.intra_pred_buf_size,
> +				  ctx->codec.h264.intra_pred_buf,
> +				  ctx->codec.h264.intra_pred_buf_dma);
>  }
>  
>  static void cedrus_h264_trigger(struct cedrus_ctx *ctx)
> diff --git a/drivers/staging/media/sunxi/cedrus/cedrus_regs.h b/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> index ace3d49fcd82..7beb03d3bb39 100644
> --- a/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> +++ b/drivers/staging/media/sunxi/cedrus/cedrus_regs.h
> @@ -46,6 +46,17 @@
>  #define VE_MODE_DEC_H264			(0x01 << 0)
>  #define VE_MODE_DEC_MPEG			(0x00 << 0)
>  
> +#define VE_BUF_CTRL				0x50
> +
> +#define VE_BUF_CTRL_INTRAPRED_EXT_RAM		(0x02 << 2)
> +#define VE_BUF_CTRL_INTRAPRED_MIXED_RAM		(0x01 << 2)
> +#define VE_BUF_CTRL_INTRAPRED_INT_SRAM		(0x00 << 2)
> +#define VE_BUF_CTRL_DBLK_EXT_RAM		(0x02 << 0)
> +#define VE_BUF_CTRL_DBLK_MIXED_RAM		(0x01 << 0)
> +#define VE_BUF_CTRL_DBLK_INT_SRAM		(0x00 << 0)
> +
> +#define VE_DBLK_DRAM_BUF_ADDR			0x54
> +#define VE_INTRAPRED_DRAM_BUF_ADDR		0x58
>  #define VE_PRIMARY_CHROMA_BUF_LEN		0xc4
>  #define VE_PRIMARY_FB_LINE_STRIDE		0xc8
>  
> -- 
> 2.23.0
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

  reply	other threads:[~2019-11-04 10:13 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-26  7:49 [PATCH 0/3] media: cedrus: Add support for 4k videos Jernej Skrabec
2019-10-26  7:49 ` [PATCH 1/3] media: cedrus: Properly signal size in mode register Jernej Skrabec
2019-11-04 10:02   ` Paul Kocialkowski
2019-11-04 16:33     ` Jernej Škrabec
2019-11-05  8:10       ` Paul Kocialkowski
2019-11-06 17:41         ` Jernej Škrabec
     [not found]         ` <jwv1ruj7on7.fsf-monnier+gmane.comp.hardware.netbook.arm.sunxi@gnu.org>
2019-11-09 12:56           ` [linux-sunxi] " Paul Kocialkowski
2019-10-26  7:49 ` [PATCH 2/3] media: cedrus: Fix H264 4k support Jernej Skrabec
2019-11-04 10:13   ` Paul Kocialkowski [this message]
2019-11-04 16:53     ` Jernej Škrabec
2019-11-05  8:07       ` Paul Kocialkowski
2019-10-26  7:49 ` [PATCH 3/3] media: cedrus: Increase maximum supported size Jernej Skrabec
2019-11-04 10:14   ` Paul Kocialkowski

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=20191104101319.GE502900@aptenodytes \
    --to=paul.kocialkowski@bootlin.com \
    --cc=devel@driverdev.osuosl.org \
    --cc=gregkh@linuxfoundation.org \
    --cc=hverkuil-cisco@xs4all.nl \
    --cc=jernej.skrabec@siol.net \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=linux-sunxi@googlegroups.com \
    --cc=mchehab@kernel.org \
    --cc=mripard@kernel.org \
    --cc=wens@csie.org \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).