openwrt/target/linux/bcm27xx/patches-5.15/950-0467-media-rpivid-Only-create-aux-entries-for-H265-if-nee.patch

From 9b0eddfd7341525e0e3074a7f6202f5ccd431b05 Mon Sep 17 00:00:00 2001
From: John Cox <jc@kynesim.co.uk>
Date: Thu, 6 May 2021 13:48:05 +0100
Subject: [PATCH] media: rpivid: Only create aux entries for H265 if
 needed

Only create aux entries of mv info for frames where that info might
be used by a later frame.  This saves some memory bandwidth and
potentially some memory.

Signed-off-by: John Cox <jc@kynesim.co.uk>
---
 drivers/staging/media/rpivid/rpivid_h265.c | 38 +++++++++++++---------
 1 file changed, 23 insertions(+), 15 deletions(-)

--- a/drivers/staging/media/rpivid/rpivid_h265.c
+++ b/drivers/staging/media/rpivid/rpivid_h265.c
@@ -246,6 +246,8 @@ struct rpivid_dec_state {
 	// Slice vars
 	unsigned int slice_idx;
 	bool slice_temporal_mvp;  /* Slice flag but constant for frame */
+	bool use_aux;
+	bool mk_aux;
 
 	// Temp vars per run - don't actually need to persist
 	u8 *src_buf;
@@ -1657,7 +1659,7 @@ static u32 mk_config2(const struct rpivi
 		c |= BIT(13);
 	if (sps->flags & V4L2_HEVC_SPS_FLAG_STRONG_INTRA_SMOOTHING_ENABLED)
 		c |= BIT(14);
-	if (sps->flags & V4L2_HEVC_SPS_FLAG_SPS_TEMPORAL_MVP_ENABLED)
+	if (s->mk_aux)
 		c |= BIT(15); /* Write motion vectors to external memory */
 	c |= (pps->log2_parallel_merge_level_minus2 + 2) << 16;
 	if (s->slice_temporal_mvp)
@@ -1669,6 +1671,14 @@ static u32 mk_config2(const struct rpivi
 	return c;
 }
 
+static inline bool is_ref_unit_type(const unsigned int nal_unit_type)
+{
+	/* From Table 7-1
+	 * True for 1, 3, 5, 7, 9, 11, 13, 15
+	 */
+	return (nal_unit_type & ~0xe) != 0;
+}
+
 static void rpivid_h265_setup(struct rpivid_ctx *ctx, struct rpivid_run *run)
 {
 	struct rpivid_dev *const dev = ctx->dev;
@@ -1685,7 +1695,6 @@ static void rpivid_h265_setup(struct rpi
 	struct rpivid_dec_env *de = ctx->dec0;
 	unsigned int prev_rs;
 	unsigned int i;
-	int use_aux;
 	int rv;
 	bool slice_temporal_mvp;
 	bool frame_end;
@@ -1828,6 +1837,16 @@ static void rpivid_h265_setup(struct rpi
 		 */
 		s->slice_temporal_mvp = slice_temporal_mvp;
 
+		/*
+		 * Need Aux ents for all (ref) DPB ents if temporal MV could
+		 * be enabled for any pic
+		 */
+		s->use_aux = ((s->sps.flags &
+			       V4L2_HEVC_SPS_FLAG_SPS_TEMPORAL_MVP_ENABLED) != 0);
+		s->mk_aux = s->use_aux &&
+			    (s->sps.sps_max_sub_layers_minus1 >= sh0->nuh_temporal_id_plus1 ||
+			     is_ref_unit_type(sh0->nal_unit_type));
+
 		// Phase 2 reg pre-calc
 		de->rpi_config2 = mk_config2(s);
 		de->rpi_framesize = (s->sps.pic_height_in_luma_samples << 16) |
@@ -1952,15 +1971,6 @@ static void rpivid_h265_setup(struct rpi
 	// Frame end
 	memset(dpb_q_aux, 0,
 	       sizeof(*dpb_q_aux) * V4L2_HEVC_DPB_ENTRIES_NUM_MAX);
-	/*
-	 * Need Aux ents for all (ref) DPB ents if temporal MV could
-	 * be enabled for any pic
-	 * ** At the moment we create aux ents for all pics whether or not
-	 *    they are ref - they should then be discarded by the DPB-aux
-	 *    garbage collection code
-	 */
-	use_aux = ((s->sps.flags &
-		  V4L2_HEVC_SPS_FLAG_SPS_TEMPORAL_MVP_ENABLED) != 0);
 
 	// Locate ref frames
 	// At least in the current implementation this is constant across all
@@ -1994,7 +2004,7 @@ static void rpivid_h265_setup(struct rpi
 			continue;
 		}
 
-		if (use_aux) {
+		if (s->use_aux) {
 			dpb_q_aux[i] = aux_q_ref_idx(ctx, buffer_index);
 			if (!dpb_q_aux[i])
 				v4l2_warn(&dev->v4l2_dev,
@@ -2016,9 +2026,7 @@ static void rpivid_h265_setup(struct rpi
 	// now
 	aux_q_release(ctx, &s->frame_aux);
 
-	if (use_aux) {
-		// New frame so new aux ent
-		// ??? Do we need this if non-ref ??? can we tell
+	if (s->mk_aux) {
 		s->frame_aux = aux_q_new(ctx, run->dst->vb2_buf.index);
 
 		if (!s->frame_aux) {