Re: [PATCH 04/11] [media] s5p-mfc: Support MFCv10.10 buffer requirements

From: Smitha T Murthy
Date: Mon Feb 06 2017 - 03:57:17 EST

Next message: Smitha T Murthy: "Re: [PATCH 06/11] [media] videodev2.h: Add v4l2 definition for HEVC"
Previous message: Smitha T Murthy: "Re: [PATCH 05/11] [media] s5p-mfc: Add support for HEVC decoder"
In reply to: Andrzej Hajda: "Re: [PATCH 04/11] [media] s5p-mfc: Support MFCv10.10 buffer requirements"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Thu, 2017-02-02 at 09:30 +0100, Andrzej Hajda wrote:
> Hi Smitha,
>
> Ups, I have missed this patch, I hope it wont influence the review :)
>
>
> On 18.01.2017 11:02, Smitha T Murthy wrote:
> > Aligning the luma_dpb_size, chroma_dpb_size, mv_size and me_buffer_size
> > for MFCv10.10.
> >
> > Signed-off-by: Smitha T Murthy <smitha.t@xxxxxxxxxxx>
> > ---
> > drivers/media/platform/s5p-mfc/regs-mfc-v10.h | 13 +++
> > drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c | 97 ++++++++++++++++++-----
> > drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.h | 2 +
> > 3 files changed, 91 insertions(+), 21 deletions(-)
> >
> > diff --git a/drivers/media/platform/s5p-mfc/regs-mfc-v10.h b/drivers/media/platform/s5p-mfc/regs-mfc-v10.h
> > index bd671a5..153ee68 100644
> > --- a/drivers/media/platform/s5p-mfc/regs-mfc-v10.h
> > +++ b/drivers/media/platform/s5p-mfc/regs-mfc-v10.h
> > @@ -32,5 +32,18 @@
> > #define MFC_VERSION_V10 0xA0
> > #define MFC_NUM_PORTS_V10 1
> >
> > +/* Encoder buffer size for MFC v10.0 */
> > +#define ENC_V100_H264_ME_SIZE(x, y) \
> > + (((x + 3) * (y + 3) * 8) \
> > + + ((((x * y) + 63) / 64) * 32) \
> > + + (((y * 64) + 1280) * (x + 7) / 8))
> > +#define ENC_V100_MPEG4_ME_SIZE(x, y) \
> > + (((x + 3) * (y + 3) * 8) \
> > + + ((((x * y) + 127) / 128) * 16) \
> > + + (((y * 64) + 1280) * (x + 7) / 8))
> > +#define ENC_V100_VP8_ME_SIZE(x, y) \
> > + (((x + 3) * (y + 3) * 8) \
> > + + (((y * 64) + 1280) * (x + 7) / 8))
> > +
>
> Crazy, cryptic math here, I guess you can make it more readable by using
> DIV_ROUND_UP macro and abstracting out common parts, for example:
>
> #define ENC_V100_BASE_SIZE(x, y) \
> (((x + 3) * (y + 3) * 8) \
> + ((y * 64) + 1280) * DIV_ROUND_UP(x, 8))
>
> #define ENC_V100_H264_ME_SIZE(x, y) \
> (ENC_V100_BASE_SIZE(x, y)
> + DIV_ROUND_UP(x * y, 64) * 32)
>
> #define ENC_V100_MPEG4_ME_SIZE(x, y) \
> (ENC_V100_BASE_SIZE(x, y)
> + DIV_ROUND_UP(x * y, 128) * 16)
>
> #define ENC_V100_VP8_ME_SIZE(x, y) \
> ENC_V100_BASE_SIZE(x, y)
>
>
I put the equation as I had found in the User Manual, so that it will
help in quick reference check in future. But I will change it as per
your suggestion.

> > #endif /*_REGS_MFC_V10_H*/
> >
> > diff --git a/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c b/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c
> > index faceee6..369210a 100644
> > --- a/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c
> > +++ b/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c
> > @@ -64,6 +64,7 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > {
> > struct s5p_mfc_dev *dev = ctx->dev;
> > unsigned int mb_width, mb_height;
> > + unsigned int lcu_width = 0, lcu_height = 0;
> > int ret;
> >
> > mb_width = MB_WIDTH(ctx->img_width);
> > @@ -74,7 +75,9 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > ctx->luma_size, ctx->chroma_size, ctx->mv_size);
> > mfc_debug(2, "Totals bufs: %d\n", ctx->total_dpb_count);
> > } else if (ctx->type == MFCINST_ENCODER) {
> > - if (IS_MFCV8_PLUS(dev))
> > + if (IS_MFCV10(dev)) {
> > + ctx->tmv_buffer_size = 0;
> > + } else if (IS_MFCV8_PLUS(dev))
> > ctx->tmv_buffer_size = S5P_FIMV_NUM_TMV_BUFFERS_V6 *
> > ALIGN(S5P_FIMV_TMV_BUFFER_SIZE_V8(mb_width, mb_height),
> > S5P_FIMV_TMV_BUFFER_ALIGN_V6);
> > @@ -82,13 +85,36 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > ctx->tmv_buffer_size = S5P_FIMV_NUM_TMV_BUFFERS_V6 *
> > ALIGN(S5P_FIMV_TMV_BUFFER_SIZE_V6(mb_width, mb_height),
> > S5P_FIMV_TMV_BUFFER_ALIGN_V6);
> > -
> > - ctx->luma_dpb_size = ALIGN((mb_width * mb_height) *
> > - S5P_FIMV_LUMA_MB_TO_PIXEL_V6,
> > - S5P_FIMV_LUMA_DPB_BUFFER_ALIGN_V6);
> > - ctx->chroma_dpb_size = ALIGN((mb_width * mb_height) *
> > - S5P_FIMV_CHROMA_MB_TO_PIXEL_V6,
> > - S5P_FIMV_CHROMA_DPB_BUFFER_ALIGN_V6);
> > + if (IS_MFCV10(dev)) {
> > + lcu_width = enc_lcu_width(ctx->img_width);
> > + lcu_height = enc_lcu_height(ctx->img_height);
> > + if (ctx->codec_mode != S5P_FIMV_CODEC_HEVC_ENC) {
> > + ctx->luma_dpb_size =
> > + ALIGN((((mb_width * 16) + 63) / 64)
> > + * 64 * (((mb_height * 16) + 31)
> > + / 32) * 32 + 64, 64);
> > + ctx->chroma_dpb_size =
> > + ALIGN((((mb_width * 16) + 63) / 64)
> > + * 64 * (mb_height * 8)
> > + + 64, 64);
> > + } else {
> > + ctx->luma_dpb_size =
> > + ALIGN((((lcu_width * 32) + 63) / 64)
> > + * 64 * (((lcu_height * 32) + 31)
> > + / 32) * 32 + 64, 64);
> > + ctx->chroma_dpb_size =
> > + ALIGN((((lcu_width * 32) + 63) / 64)
> > + * 64 * (lcu_height * 16)
> > + + 64, 64);
>
> Again crazy math, please try to avoid repetitive calculations, replace
> all constructs of form (x + n - 1) / n, with DIV_ROUND_UP(x, n), verify
> if ALIGN is necessary.

I will add the DIV_ROUND_UP. Since ALIGN was there for other versions
too, I followed the same.
>
> > + }
> > + } else {
> > + ctx->luma_dpb_size = ALIGN((mb_width * mb_height) *
> > + S5P_FIMV_LUMA_MB_TO_PIXEL_V6,
> > + S5P_FIMV_LUMA_DPB_BUFFER_ALIGN_V6);
> > + ctx->chroma_dpb_size = ALIGN((mb_width * mb_height) *
> > + S5P_FIMV_CHROMA_MB_TO_PIXEL_V6,
> > + S5P_FIMV_CHROMA_DPB_BUFFER_ALIGN_V6);
> > + }
> > if (IS_MFCV8_PLUS(dev))
> > ctx->me_buffer_size = ALIGN(S5P_FIMV_ME_BUFFER_SIZE_V8(
> > ctx->img_width, ctx->img_height,
> > @@ -197,6 +223,8 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > case S5P_MFC_CODEC_H264_ENC:
> > if (IS_MFCV10(dev)) {
> > mfc_debug(2, "Use min scratch buffer size\n");
> > + ctx->me_buffer_size =
> > + ALIGN(ENC_V100_H264_ME_SIZE(mb_width, mb_height), 16);
>
> This and similar entries below should be in previous patch, again check
> if ALIGN is needed.

I will take these changes in the previous patch in the next version.

> > } else if (IS_MFCV8_PLUS(dev))
> > ctx->scratch_buf_size =
> > S5P_FIMV_SCRATCH_BUF_SIZE_H264_ENC_V8(
> > @@ -219,6 +247,9 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > case S5P_MFC_CODEC_H263_ENC:
> > if (IS_MFCV10(dev)) {
> > mfc_debug(2, "Use min scratch buffer size\n");
> > + ctx->me_buffer_size =
> > + ALIGN(ENC_V100_MPEG4_ME_SIZE(mb_width,
> > + mb_height), 16);
> > } else
> > ctx->scratch_buf_size =
> > S5P_FIMV_SCRATCH_BUF_SIZE_MPEG4_ENC_V6(
> > @@ -235,7 +266,10 @@ static int s5p_mfc_alloc_codec_buffers_v6(struct s5p_mfc_ctx *ctx)
> > case S5P_MFC_CODEC_VP8_ENC:
> > if (IS_MFCV10(dev)) {
> > mfc_debug(2, "Use min scratch buffer size\n");
> > - } else if (IS_MFCV8_PLUS(dev))
> > + ctx->me_buffer_size =
> > + ALIGN(ENC_V100_VP8_ME_SIZE(mb_width, mb_height),
> > + 16);
> > + } else if (IS_MFCV8_PLUS(dev))
> > ctx->scratch_buf_size =
> > S5P_FIMV_SCRATCH_BUF_SIZE_VP8_ENC_V8(
> > mb_width,
> > @@ -395,13 +429,15 @@ static void s5p_mfc_dec_calc_dpb_size_v6(struct s5p_mfc_ctx *ctx)
> >
> > if (ctx->codec_mode == S5P_MFC_CODEC_H264_DEC ||
> > ctx->codec_mode == S5P_MFC_CODEC_H264_MVC_DEC) {
> > - if (IS_MFCV10(dev))
> > + if (IS_MFCV10(dev)) {
> > ctx->mv_size = S5P_MFC_DEC_MV_SIZE_V10(ctx->img_width,
> > ctx->img_height);
> > - else
> > + ctx->mv_size = ALIGN(ctx->mv_size, 32);
> > + } else {
> > ctx->mv_size = S5P_MFC_DEC_MV_SIZE_V6(ctx->img_width,
> > ctx->img_height);
> > - ctx->mv_size = ALIGN(ctx->mv_size, 16);
> > + ctx->mv_size = ALIGN(ctx->mv_size, 16);
> > + }
> > } else {
> > ctx->mv_size = 0;
> > }
> > @@ -607,15 +643,34 @@ static int s5p_mfc_set_enc_ref_buffer_v6(struct s5p_mfc_ctx *ctx)
> >
> > mfc_debug(2, "Buf1: %p (%d)\n", (void *)buf_addr1, buf_size1);
> >
> > - for (i = 0; i < ctx->pb_count; i++) {
> > - writel(buf_addr1, mfc_regs->e_luma_dpb + (4 * i));
> > - buf_addr1 += ctx->luma_dpb_size;
> > - writel(buf_addr1, mfc_regs->e_chroma_dpb + (4 * i));
> > - buf_addr1 += ctx->chroma_dpb_size;
> > - writel(buf_addr1, mfc_regs->e_me_buffer + (4 * i));
> > - buf_addr1 += ctx->me_buffer_size;
> > - buf_size1 -= (ctx->luma_dpb_size + ctx->chroma_dpb_size +
> > - ctx->me_buffer_size);
> > + if (IS_MFCV10(dev)) {
> > + /* start address of per buffer is aligned */
> > + for (i = 0; i < ctx->pb_count; i++) {
> > + writel(buf_addr1, mfc_regs->e_luma_dpb + (4 * i));
> > + buf_addr1 += ctx->luma_dpb_size;
> > + buf_size1 -= ctx->luma_dpb_size;
> > + }
> > + for (i = 0; i < ctx->pb_count; i++) {
> > + writel(buf_addr1, mfc_regs->e_chroma_dpb + (4 * i));
> > + buf_addr1 += ctx->chroma_dpb_size;
> > + buf_size1 -= ctx->chroma_dpb_size;
> > + }
> > + for (i = 0; i < ctx->pb_count; i++) {
> > + writel(buf_addr1, mfc_regs->e_me_buffer + (4 * i));
> > + buf_addr1 += ctx->me_buffer_size;
> > + buf_size1 -= ctx->me_buffer_size;
> > + }
> > + } else {
> > + for (i = 0; i < ctx->pb_count; i++) {
> > + writel(buf_addr1, mfc_regs->e_luma_dpb + (4 * i));
> > + buf_addr1 += ctx->luma_dpb_size;
> > + writel(buf_addr1, mfc_regs->e_chroma_dpb + (4 * i));
> > + buf_addr1 += ctx->chroma_dpb_size;
> > + writel(buf_addr1, mfc_regs->e_me_buffer + (4 * i));
> > + buf_addr1 += ctx->me_buffer_size;
> > + buf_size1 -= (ctx->luma_dpb_size + ctx->chroma_dpb_size
> > + + ctx->me_buffer_size);
> > + }
> > }
> >
> > writel(buf_addr1, mfc_regs->e_scratch_buffer_addr);
> > diff --git a/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.h b/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.h
> > index fcc2368..2b5a9f4 100644
> > --- a/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.h
> > +++ b/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.h
> > @@ -26,6 +26,8 @@
> > (((MB_HEIGHT(y)+1)/2)*2) * 64 + 128)
> > #define S5P_MFC_DEC_MV_SIZE_V10(x, y) (MB_WIDTH(x) * \
> > (((MB_HEIGHT(y)+1)/2)*2) * 64 + 512)
> > +#define enc_lcu_width(x_size) ((x_size + 31) / 32)
> > +#define enc_lcu_height(y_size) ((y_size + 31) / 32)
>
> DIV_ROUND_UP
>
>
> Regards
> Andrzej
>
Yes I will correct it.

Thank you for the review.
Regards,
Smitha
> >
> > /* Definition */
> > #define ENC_MULTI_SLICE_MB_MAX ((1 << 30) - 1)
>
>
>

Next message: Smitha T Murthy: "Re: [PATCH 06/11] [media] videodev2.h: Add v4l2 definition for HEVC"
Previous message: Smitha T Murthy: "Re: [PATCH 05/11] [media] s5p-mfc: Add support for HEVC decoder"
In reply to: Andrzej Hajda: "Re: [PATCH 04/11] [media] s5p-mfc: Support MFCv10.10 buffer requirements"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]