* [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx
@ 2024-04-10 13:31 Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 02/10] lavc/hevcdec: track local context count separately from WPP thread count Anton Khirnov
` (8 more replies)
0 siblings, 9 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
It is more consistent with our naming conventions.
---
libavcodec/hevcdec.c | 30 +++++++++++++++---------------
libavcodec/hevcdec.h | 2 +-
2 files changed, 16 insertions(+), 16 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index d3b668af00..c70937a756 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -2697,14 +2697,14 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
}
for (i = 1; i < s->threads_number; i++) {
- if (s->HEVClcList[i])
+ if (s->local_ctx[i])
continue;
- s->HEVClcList[i] = av_mallocz(sizeof(HEVCLocalContext));
- if (!s->HEVClcList[i])
+ s->local_ctx[i] = av_mallocz(sizeof(HEVCLocalContext));
+ if (!s->local_ctx[i])
return AVERROR(ENOMEM);
- s->HEVClcList[i]->logctx = s->avctx;
- s->HEVClcList[i]->parent = s;
- s->HEVClcList[i]->common_cabac_state = &s->cabac;
+ s->local_ctx[i]->logctx = s->avctx;
+ s->local_ctx[i]->parent = s;
+ s->local_ctx[i]->common_cabac_state = &s->cabac;
}
offset = (lc->gb.index >> 3);
@@ -2742,8 +2742,8 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
s->data = data;
for (i = 1; i < s->threads_number; i++) {
- s->HEVClcList[i]->first_qp_group = 1;
- s->HEVClcList[i]->qp_y = s->HEVClc->qp_y;
+ s->local_ctx[i]->first_qp_group = 1;
+ s->local_ctx[i]->qp_y = s->HEVClc->qp_y;
}
atomic_store(&s->wpp_err, 0);
@@ -2756,7 +2756,7 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
return AVERROR(ENOMEM);
if (s->ps.pps->entropy_coding_sync_enabled_flag)
- s->avctx->execute2(s->avctx, hls_decode_entry_wpp, s->HEVClcList, ret, s->sh.num_entry_point_offsets + 1);
+ s->avctx->execute2(s->avctx, hls_decode_entry_wpp, s->local_ctx, ret, s->sh.num_entry_point_offsets + 1);
for (i = 0; i <= s->sh.num_entry_point_offsets; i++)
res += ret[i];
@@ -3474,13 +3474,13 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
av_freep(&s->sh.offset);
av_freep(&s->sh.size);
- if (s->HEVClcList) {
+ if (s->local_ctx) {
for (i = 1; i < s->threads_number; i++) {
- av_freep(&s->HEVClcList[i]);
+ av_freep(&s->local_ctx[i]);
}
}
av_freep(&s->HEVClc);
- av_freep(&s->HEVClcList);
+ av_freep(&s->local_ctx);
ff_h2645_packet_uninit(&s->pkt);
@@ -3497,13 +3497,13 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
s->avctx = avctx;
s->HEVClc = av_mallocz(sizeof(HEVCLocalContext));
- s->HEVClcList = av_mallocz(sizeof(HEVCLocalContext*) * s->threads_number);
- if (!s->HEVClc || !s->HEVClcList)
+ s->local_ctx = av_mallocz(sizeof(HEVCLocalContext*) * s->threads_number);
+ if (!s->HEVClc || !s->local_ctx)
return AVERROR(ENOMEM);
s->HEVClc->parent = s;
s->HEVClc->logctx = avctx;
s->HEVClc->common_cabac_state = &s->cabac;
- s->HEVClcList[0] = s->HEVClc;
+ s->local_ctx[0] = s->HEVClc;
s->output_frame = av_frame_alloc();
if (!s->output_frame)
diff --git a/libavcodec/hevcdec.h b/libavcodec/hevcdec.h
index e65a6180ca..9e3e6a8cd7 100644
--- a/libavcodec/hevcdec.h
+++ b/libavcodec/hevcdec.h
@@ -441,7 +441,7 @@ typedef struct HEVCContext {
const AVClass *c; // needed by private avoptions
AVCodecContext *avctx;
- HEVCLocalContext **HEVClcList;
+ HEVCLocalContext **local_ctx;
HEVCLocalContext *HEVClc;
uint8_t threads_type;
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 02/10] lavc/hevcdec: track local context count separately from WPP thread count
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers Anton Khirnov
` (7 subsequent siblings)
8 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
The latter can be lowered while decoding, which would lead to memleaks.
---
libavcodec/hevcdec.c | 7 +++++--
libavcodec/hevcdec.h | 2 ++
2 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index c70937a756..55f72af972 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -2697,11 +2697,13 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
}
for (i = 1; i < s->threads_number; i++) {
- if (s->local_ctx[i])
+ if (i < s->nb_local_ctx)
continue;
s->local_ctx[i] = av_mallocz(sizeof(HEVCLocalContext));
if (!s->local_ctx[i])
return AVERROR(ENOMEM);
+ s->nb_local_ctx++;
+
s->local_ctx[i]->logctx = s->avctx;
s->local_ctx[i]->parent = s;
s->local_ctx[i]->common_cabac_state = &s->cabac;
@@ -3475,7 +3477,7 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
av_freep(&s->sh.size);
if (s->local_ctx) {
- for (i = 1; i < s->threads_number; i++) {
+ for (i = 1; i < s->nb_local_ctx; i++) {
av_freep(&s->local_ctx[i]);
}
}
@@ -3504,6 +3506,7 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
s->HEVClc->logctx = avctx;
s->HEVClc->common_cabac_state = &s->cabac;
s->local_ctx[0] = s->HEVClc;
+ s->nb_local_ctx = 1;
s->output_frame = av_frame_alloc();
if (!s->output_frame)
diff --git a/libavcodec/hevcdec.h b/libavcodec/hevcdec.h
index 9e3e6a8cd7..a881eb9981 100644
--- a/libavcodec/hevcdec.h
+++ b/libavcodec/hevcdec.h
@@ -442,6 +442,8 @@ typedef struct HEVCContext {
AVCodecContext *avctx;
HEVCLocalContext **local_ctx;
+ unsigned nb_local_ctx;
+
HEVCLocalContext *HEVClc;
uint8_t threads_type;
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 02/10] lavc/hevcdec: track local context count separately from WPP thread count Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-17 9:29 ` Andreas Rheinhardt
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 04/10] lavc/hevcdec: drop a useless execute() call with 1 job Anton Khirnov
` (6 subsequent siblings)
8 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
It is more efficient and easier to manage.
---
libavcodec/hevcdec.c | 57 +++++++++++++++++++++++++-------------------
libavcodec/hevcdec.h | 2 +-
2 files changed, 33 insertions(+), 26 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index 55f72af972..47226ef0ab 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -2598,7 +2598,7 @@ static int hls_slice_data(HEVCContext *s)
static int hls_decode_entry_wpp(AVCodecContext *avctxt, void *hevc_lclist,
int job, int self_id)
{
- HEVCLocalContext *lc = ((HEVCLocalContext**)hevc_lclist)[self_id];
+ HEVCLocalContext *lc = &((HEVCLocalContext*)hevc_lclist)[self_id];
const HEVCContext *const s = lc->parent;
int ctb_size = 1 << s->ps.sps->log2_ctb_size;
int more_data = 1;
@@ -2682,7 +2682,7 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
{
const uint8_t *data = nal->data;
int length = nal->size;
- HEVCLocalContext *lc = s->HEVClc;
+ HEVCLocalContext *lc;
int *ret;
int64_t offset;
int64_t startheader, cmpt = 0;
@@ -2696,19 +2696,31 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
return AVERROR_INVALIDDATA;
}
- for (i = 1; i < s->threads_number; i++) {
- if (i < s->nb_local_ctx)
- continue;
- s->local_ctx[i] = av_mallocz(sizeof(HEVCLocalContext));
- if (!s->local_ctx[i])
- return AVERROR(ENOMEM);
- s->nb_local_ctx++;
+ if (s->threads_number > s->nb_local_ctx) {
+ HEVCLocalContext *tmp = av_malloc_array(s->threads_number, sizeof(*s->local_ctx));
- s->local_ctx[i]->logctx = s->avctx;
- s->local_ctx[i]->parent = s;
- s->local_ctx[i]->common_cabac_state = &s->cabac;
+ if (!tmp)
+ return AVERROR(ENOMEM);
+
+ memcpy(tmp, s->local_ctx, sizeof(*s->local_ctx) * s->nb_local_ctx);
+ av_free(s->local_ctx);
+ s->local_ctx = tmp;
+ s->HEVClc = &s->local_ctx[0];
+
+ for (unsigned i = s->nb_local_ctx; i < s->threads_number; i++) {
+ tmp = &s->local_ctx[i];
+
+ memset(tmp, 0, sizeof(*tmp));
+
+ tmp->logctx = s->avctx;
+ tmp->parent = s;
+ tmp->common_cabac_state = &s->cabac;
+ }
+
+ s->nb_local_ctx = s->threads_number;
}
+ lc = &s->local_ctx[0];
offset = (lc->gb.index >> 3);
for (j = 0, cmpt = 0, startheader = offset + s->sh.entry_point_offset[0]; j < nal->skipped_bytes; j++) {
@@ -2744,8 +2756,8 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
s->data = data;
for (i = 1; i < s->threads_number; i++) {
- s->local_ctx[i]->first_qp_group = 1;
- s->local_ctx[i]->qp_y = s->HEVClc->qp_y;
+ s->local_ctx[i].first_qp_group = 1;
+ s->local_ctx[i].qp_y = s->HEVClc->qp_y;
}
atomic_store(&s->wpp_err, 0);
@@ -3476,12 +3488,6 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
av_freep(&s->sh.offset);
av_freep(&s->sh.size);
- if (s->local_ctx) {
- for (i = 1; i < s->nb_local_ctx; i++) {
- av_freep(&s->local_ctx[i]);
- }
- }
- av_freep(&s->HEVClc);
av_freep(&s->local_ctx);
ff_h2645_packet_uninit(&s->pkt);
@@ -3498,15 +3504,16 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
s->avctx = avctx;
- s->HEVClc = av_mallocz(sizeof(HEVCLocalContext));
- s->local_ctx = av_mallocz(sizeof(HEVCLocalContext*) * s->threads_number);
- if (!s->HEVClc || !s->local_ctx)
+ s->local_ctx = av_mallocz(sizeof(*s->local_ctx));
+ if (!s->local_ctx)
return AVERROR(ENOMEM);
+ s->nb_local_ctx = 1;
+
+ s->HEVClc = &s->local_ctx[0];
+
s->HEVClc->parent = s;
s->HEVClc->logctx = avctx;
s->HEVClc->common_cabac_state = &s->cabac;
- s->local_ctx[0] = s->HEVClc;
- s->nb_local_ctx = 1;
s->output_frame = av_frame_alloc();
if (!s->output_frame)
diff --git a/libavcodec/hevcdec.h b/libavcodec/hevcdec.h
index a881eb9981..5a4ed270e8 100644
--- a/libavcodec/hevcdec.h
+++ b/libavcodec/hevcdec.h
@@ -441,7 +441,7 @@ typedef struct HEVCContext {
const AVClass *c; // needed by private avoptions
AVCodecContext *avctx;
- HEVCLocalContext **local_ctx;
+ HEVCLocalContext *local_ctx;
unsigned nb_local_ctx;
HEVCLocalContext *HEVClc;
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 04/10] lavc/hevcdec: drop a useless execute() call with 1 job
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 02/10] lavc/hevcdec: track local context count separately from WPP thread count Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 05/10] lavc/hevc_ps: reduce the size of used_by_curr_pic_lt_sps_flag Anton Khirnov
` (5 subsequent siblings)
8 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
---
libavcodec/hevcdec.c | 12 ++----------
1 file changed, 2 insertions(+), 10 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index 47226ef0ab..6b2c0b75ee 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -2526,9 +2526,8 @@ static void hls_decode_neighbour(HEVCLocalContext *lc, int x_ctb, int y_ctb,
lc->ctb_up_left_flag = ((x_ctb > 0) && (y_ctb > 0) && (ctb_addr_in_slice-1 >= s->ps.sps->ctb_width) && (s->ps.pps->tile_id[ctb_addr_ts] == s->ps.pps->tile_id[s->ps.pps->ctb_addr_rs_to_ts[ctb_addr_rs-1 - s->ps.sps->ctb_width]]));
}
-static int hls_decode_entry(AVCodecContext *avctxt, void *arg)
+static int hls_decode_entry(HEVCContext *s)
{
- HEVCContext *s = avctxt->priv_data;
HEVCLocalContext *const lc = s->HEVClc;
int ctb_size = 1 << s->ps.sps->log2_ctb_size;
int more_data = 1;
@@ -2588,13 +2587,6 @@ static int hls_decode_entry(AVCodecContext *avctxt, void *arg)
return ctb_addr_ts;
}
-static int hls_slice_data(HEVCContext *s)
-{
- int ret = 0;
-
- s->avctx->execute(s->avctx, hls_decode_entry, NULL, &ret , 1, 0);
- return ret;
-}
static int hls_decode_entry_wpp(AVCodecContext *avctxt, void *hevc_lclist,
int job, int self_id)
{
@@ -3133,7 +3125,7 @@ static int decode_nal_unit(HEVCContext *s, const H2645NAL *nal)
if (s->threads_number > 1 && s->sh.num_entry_point_offsets > 0)
ctb_addr_ts = hls_slice_data_wpp(s, nal);
else
- ctb_addr_ts = hls_slice_data(s);
+ ctb_addr_ts = hls_decode_entry(s);
if (ctb_addr_ts >= (s->ps.sps->ctb_width * s->ps.sps->ctb_height)) {
ret = hevc_frame_end(s);
if (ret < 0)
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 05/10] lavc/hevc_ps: reduce the size of used_by_curr_pic_lt_sps_flag
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (2 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 04/10] lavc/hevcdec: drop a useless execute() call with 1 job Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields Anton Khirnov
` (4 subsequent siblings)
8 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
It is currently an array of 32 uint8_t, each storing a single flag. A
single uint32_t is sufficient.
---
libavcodec/hevc_ps.c | 4 +++-
libavcodec/hevc_ps.h | 2 +-
libavcodec/hevcdec.c | 2 +-
libavcodec/vulkan_hevc.c | 3 +--
4 files changed, 6 insertions(+), 5 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index 25f087ed75..8d5fc0d0ca 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -1094,9 +1094,11 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
sps->num_long_term_ref_pics_sps);
return AVERROR_INVALIDDATA;
}
+
+ sps->used_by_curr_pic_lt = 0;
for (i = 0; i < sps->num_long_term_ref_pics_sps; i++) {
sps->lt_ref_pic_poc_lsb_sps[i] = get_bits(gb, sps->log2_max_poc_lsb);
- sps->used_by_curr_pic_lt_sps_flag[i] = get_bits1(gb);
+ sps->used_by_curr_pic_lt |= get_bits1(gb) * (1 << i);
}
}
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index 61a0fe2219..b2f3a8dbd1 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -230,7 +230,7 @@ typedef struct HEVCSPS {
uint8_t long_term_ref_pics_present_flag;
uint16_t lt_ref_pic_poc_lsb_sps[HEVC_MAX_LONG_TERM_REF_PICS];
- uint8_t used_by_curr_pic_lt_sps_flag[HEVC_MAX_LONG_TERM_REF_PICS];
+ uint32_t used_by_curr_pic_lt;
uint8_t num_long_term_ref_pics_sps;
struct {
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index 6b2c0b75ee..4a7046bdec 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -295,7 +295,7 @@ static int decode_lt_rps(HEVCContext *s, LongTermRPS *rps, GetBitContext *gb)
lt_idx_sps = get_bits(gb, av_ceil_log2(sps->num_long_term_ref_pics_sps));
rps->poc[i] = sps->lt_ref_pic_poc_lsb_sps[lt_idx_sps];
- rps->used[i] = sps->used_by_curr_pic_lt_sps_flag[lt_idx_sps];
+ rps->used[i] = !!(sps->used_by_curr_pic_lt & (1 << lt_idx_sps));
} else {
rps->poc[i] = get_bits(gb, sps->log2_max_poc_lsb);
rps->used[i] = get_bits1(gb);
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index 2705a965b9..9b40f5ad58 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -386,11 +386,10 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
}
*ltr = (StdVideoH265LongTermRefPicsSps) {
- .used_by_curr_pic_lt_sps_flag = 0x0,
+ .used_by_curr_pic_lt_sps_flag = sps->used_by_curr_pic_lt,
};
for (int i = 0; i < sps->num_long_term_ref_pics_sps; i++) {
- ltr->used_by_curr_pic_lt_sps_flag |= sps->used_by_curr_pic_lt_sps_flag[i] << i;
ltr->lt_ref_pic_poc_lsb_sps[i] = sps->lt_ref_pic_poc_lsb_sps[i];
}
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (3 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 05/10] lavc/hevc_ps: reduce the size of used_by_curr_pic_lt_sps_flag Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-11 11:55 ` Andreas Rheinhardt
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 07/10] lavc/hevc_ps: fix variable signedness in ff_hevc_decode_short_term_rps() Anton Khirnov
` (3 subsequent siblings)
8 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
Reduces sizeof(HEVCSPS) by 96 bytes.
Also improve flag names: drop redundant suffixes and prefixes, and
consistently use disabled/enabled.
---
libavcodec/dxva2_hevc.c | 24 ++++-----
libavcodec/hevc_cabac.c | 36 ++++++-------
libavcodec/hevc_filter.c | 8 +--
libavcodec/hevc_parser.c | 2 +-
libavcodec/hevc_ps.c | 95 +++++++++++++++++-----------------
libavcodec/hevc_ps.h | 62 +++++++++++-----------
libavcodec/hevcdec.c | 10 ++--
libavcodec/hevcpred_template.c | 4 +-
libavcodec/mips/hevcpred_msa.c | 6 +--
libavcodec/nvdec_hevc.c | 42 +++++++--------
libavcodec/qsvenc_hevc.c | 2 +-
libavcodec/vaapi_hevc.c | 42 +++++++--------
libavcodec/vdpau_hevc.c | 36 ++++++-------
libavcodec/vulkan_hevc.c | 56 ++++++++++----------
14 files changed, 212 insertions(+), 213 deletions(-)
diff --git a/libavcodec/dxva2_hevc.c b/libavcodec/dxva2_hevc.c
index 31d74a7164..b500d7917a 100644
--- a/libavcodec/dxva2_hevc.c
+++ b/libavcodec/dxva2_hevc.c
@@ -72,7 +72,7 @@ void ff_dxva2_hevc_fill_picture_parameters(const AVCodecContext *avctx, AVDXVACo
pp->PicHeightInMinCbsY = sps->min_cb_height;
pp->wFormatAndSequenceInfoFlags = (sps->chroma_format_idc << 0) |
- (sps->separate_colour_plane_flag << 2) |
+ (sps->separate_colour_plane << 2) |
((sps->bit_depth - 8) << 3) |
((sps->bit_depth - 8) << 6) |
((sps->log2_max_poc_lsb - 4) << 9) |
@@ -99,18 +99,18 @@ void ff_dxva2_hevc_fill_picture_parameters(const AVCodecContext *avctx, AVDXVACo
pp->wNumBitsForShortTermRPSInSlice = h->sh.short_term_ref_pic_set_size;
}
- pp->dwCodingParamToolFlags = (sps->scaling_list_enable_flag << 0) |
- (sps->amp_enabled_flag << 1) |
+ pp->dwCodingParamToolFlags = (sps->scaling_list_enabled << 0) |
+ (sps->amp_enabled << 1) |
(sps->sao_enabled << 2) |
- (sps->pcm_enabled_flag << 3) |
- ((sps->pcm_enabled_flag ? (sps->pcm.bit_depth - 1) : 0) << 4) |
- ((sps->pcm_enabled_flag ? (sps->pcm.bit_depth_chroma - 1) : 0) << 8) |
- ((sps->pcm_enabled_flag ? (sps->pcm.log2_min_pcm_cb_size - 3) : 0) << 12) |
- ((sps->pcm_enabled_flag ? (sps->pcm.log2_max_pcm_cb_size - sps->pcm.log2_min_pcm_cb_size) : 0) << 14) |
- (sps->pcm.loop_filter_disable_flag << 16) |
- (sps->long_term_ref_pics_present_flag << 17) |
- (sps->sps_temporal_mvp_enabled_flag << 18) |
- (sps->sps_strong_intra_smoothing_enable_flag << 19) |
+ (sps->pcm_enabled << 3) |
+ ((sps->pcm_enabled ? (sps->pcm.bit_depth - 1) : 0) << 4) |
+ ((sps->pcm_enabled ? (sps->pcm.bit_depth_chroma - 1) : 0) << 8) |
+ ((sps->pcm_enabled ? (sps->pcm.log2_min_pcm_cb_size - 3) : 0) << 12) |
+ ((sps->pcm_enabled ? (sps->pcm.log2_max_pcm_cb_size - sps->pcm.log2_min_pcm_cb_size) : 0) << 14) |
+ (sps->pcm_loop_filter_disabled << 16) |
+ (sps->long_term_ref_pics_present << 17) |
+ (sps->temporal_mvp_enabled << 18) |
+ (sps->strong_intra_smoothing_enabled << 19) |
(pps->dependent_slice_segments_enabled_flag << 20) |
(pps->output_flag_present_flag << 21) |
(pps->num_extra_slice_header_bits << 22) |
diff --git a/libavcodec/hevc_cabac.c b/libavcodec/hevc_cabac.c
index 63ffb3d37c..2e639a7e41 100644
--- a/libavcodec/hevc_cabac.c
+++ b/libavcodec/hevc_cabac.c
@@ -408,7 +408,7 @@ void ff_hevc_save_states(HEVCLocalContext *lc, int ctb_addr_ts)
(s->ps.sps->ctb_width == 2 &&
ctb_addr_ts % s->ps.sps->ctb_width == 0))) {
memcpy(lc->common_cabac_state->state, lc->cabac_state, HEVC_CONTEXTS);
- if (s->ps.sps->persistent_rice_adaptation_enabled_flag) {
+ if (s->ps.sps->persistent_rice_adaptation_enabled) {
memcpy(lc->common_cabac_state->stat_coeff, lc->stat_coeff, HEVC_STAT_COEFFS);
}
}
@@ -417,7 +417,7 @@ void ff_hevc_save_states(HEVCLocalContext *lc, int ctb_addr_ts)
static void load_states(HEVCLocalContext *lc, const HEVCContext *s)
{
memcpy(lc->cabac_state, lc->common_cabac_state->state, HEVC_CONTEXTS);
- if (s->ps.sps->persistent_rice_adaptation_enabled_flag) {
+ if (s->ps.sps->persistent_rice_adaptation_enabled) {
memcpy(lc->stat_coeff, lc->common_cabac_state->stat_coeff, HEVC_STAT_COEFFS);
}
}
@@ -683,7 +683,7 @@ int ff_hevc_part_mode_decode(HEVCLocalContext *lc, int log2_cb_size)
return PART_NxN; // 000
}
- if (!lc->parent->ps.sps->amp_enabled_flag) {
+ if (!lc->parent->ps.sps->amp_enabled) {
if (GET_CABAC(PART_MODE_OFFSET + 1)) // 01
return PART_2NxN;
return PART_Nx2N;
@@ -1091,7 +1091,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
scale_m = 16; // default when no custom scaling lists.
dc_scale = 16;
- if (s->ps.sps->scaling_list_enable_flag && !(transform_skip_flag && log2_trafo_size > 2)) {
+ if (s->ps.sps->scaling_list_enabled && !(transform_skip_flag && log2_trafo_size > 2)) {
const ScalingList *sl = s->ps.pps->scaling_list_data_present_flag ?
&s->ps.pps->scaling_list : &s->ps.sps->scaling_list;
int matrix_id = lc->cu.pred_mode != MODE_INTRA;
@@ -1109,7 +1109,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
dc_scale = 0;
}
- if (lc->cu.pred_mode == MODE_INTER && s->ps.sps->explicit_rdpcm_enabled_flag &&
+ if (lc->cu.pred_mode == MODE_INTER && s->ps.sps->explicit_rdpcm_enabled &&
(transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) {
explicit_rdpcm_flag = explicit_rdpcm_flag_decode(lc, c_idx);
if (explicit_rdpcm_flag) {
@@ -1240,7 +1240,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
};
const uint8_t *ctx_idx_map_p;
int scf_offset = 0;
- if (s->ps.sps->transform_skip_context_enabled_flag &&
+ if (s->ps.sps->transform_skip_context_enabled &&
(transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) {
ctx_idx_map_p = &ctx_idx_map[4 * 16];
if (c_idx == 0) {
@@ -1281,7 +1281,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
}
}
if (implicit_non_zero_coeff == 0) {
- if (s->ps.sps->transform_skip_context_enabled_flag &&
+ if (s->ps.sps->transform_skip_context_enabled &&
(transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) {
if (c_idx == 0) {
scf_offset = 42;
@@ -1326,7 +1326,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
// initialize first elem of coeff_bas_level_greater1_flag
int ctx_set = (i > 0 && c_idx == 0) ? 2 : 0;
- if (s->ps.sps->persistent_rice_adaptation_enabled_flag) {
+ if (s->ps.sps->persistent_rice_adaptation_enabled) {
if (!transform_skip_flag && !lc->cu.cu_transquant_bypass_flag)
sb_type = 2 * (c_idx == 0 ? 1 : 0);
else
@@ -1355,7 +1355,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
if (lc->cu.cu_transquant_bypass_flag ||
(lc->cu.pred_mode == MODE_INTRA &&
- s->ps.sps->implicit_rdpcm_enabled_flag && transform_skip_flag &&
+ s->ps.sps->implicit_rdpcm_enabled && transform_skip_flag &&
(pred_mode_intra == 10 || pred_mode_intra == 26 )) ||
explicit_rdpcm_flag)
sign_hidden = 0;
@@ -1381,8 +1381,8 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
trans_coeff_level += last_coeff_abs_level_remaining;
if (trans_coeff_level > (3 << c_rice_param))
- c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled_flag ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4);
- if (s->ps.sps->persistent_rice_adaptation_enabled_flag && !rice_init) {
+ c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4);
+ if (s->ps.sps->persistent_rice_adaptation_enabled && !rice_init) {
int c_rice_p_init = lc->stat_coeff[sb_type] / 4;
if (last_coeff_abs_level_remaining >= (3 << c_rice_p_init))
lc->stat_coeff[sb_type]++;
@@ -1397,8 +1397,8 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
trans_coeff_level = 1 + last_coeff_abs_level_remaining;
if (trans_coeff_level > (3 << c_rice_param))
- c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled_flag ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4);
- if (s->ps.sps->persistent_rice_adaptation_enabled_flag && !rice_init) {
+ c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4);
+ if (s->ps.sps->persistent_rice_adaptation_enabled && !rice_init) {
int c_rice_p_init = lc->stat_coeff[sb_type] / 4;
if (last_coeff_abs_level_remaining >= (3 << c_rice_p_init))
lc->stat_coeff[sb_type]++;
@@ -1417,7 +1417,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
trans_coeff_level = -trans_coeff_level;
coeff_sign_flag <<= 1;
if(!lc->cu.cu_transquant_bypass_flag) {
- if (s->ps.sps->scaling_list_enable_flag && !(transform_skip_flag && log2_trafo_size > 2)) {
+ if (s->ps.sps->scaling_list_enabled && !(transform_skip_flag && log2_trafo_size > 2)) {
if(y_c || x_c || log2_trafo_size < 4) {
switch(log2_trafo_size) {
case 3: pos = (y_c << 3) + x_c; break;
@@ -1445,15 +1445,15 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
}
if (lc->cu.cu_transquant_bypass_flag) {
- if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag &&
+ if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled &&
(pred_mode_intra == 10 || pred_mode_intra == 26))) {
- int mode = s->ps.sps->implicit_rdpcm_enabled_flag ? (pred_mode_intra == 26) : explicit_rdpcm_dir_flag;
+ int mode = s->ps.sps->implicit_rdpcm_enabled ? (pred_mode_intra == 26) : explicit_rdpcm_dir_flag;
s->hevcdsp.transform_rdpcm(coeffs, log2_trafo_size, mode);
}
} else {
if (transform_skip_flag) {
- int rot = s->ps.sps->transform_skip_rotation_enabled_flag &&
+ int rot = s->ps.sps->transform_skip_rotation_enabled &&
log2_trafo_size == 2 &&
lc->cu.pred_mode == MODE_INTRA;
if (rot) {
@@ -1463,7 +1463,7 @@ void ff_hevc_hls_residual_coding(HEVCLocalContext *lc, int x0, int y0,
s->hevcdsp.dequant(coeffs, log2_trafo_size);
- if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag &&
+ if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled &&
lc->cu.pred_mode == MODE_INTRA &&
(pred_mode_intra == 10 || pred_mode_intra == 26))) {
int mode = explicit_rdpcm_flag ? explicit_rdpcm_dir_flag : (pred_mode_intra == 26);
diff --git a/libavcodec/hevc_filter.c b/libavcodec/hevc_filter.c
index 0c45310ea6..70d3ca588a 100644
--- a/libavcodec/hevc_filter.c
+++ b/libavcodec/hevc_filter.c
@@ -224,7 +224,7 @@ static void restore_tqb_pixels(const HEVCContext *s,
int x0, int y0, int width, int height, int c_idx)
{
if ( s->ps.pps->transquant_bypass_enable_flag ||
- (s->ps.sps->pcm.loop_filter_disable_flag && s->ps.sps->pcm_enabled_flag)) {
+ (s->ps.sps->pcm_loop_filter_disabled && s->ps.sps->pcm_enabled)) {
int x, y;
int min_pu_size = 1 << s->ps.sps->log2_min_pu_size;
int hshift = s->ps.sps->hshift[c_idx];
@@ -330,7 +330,7 @@ static void sao_filter_CTB(HEVCLocalContext *lc, const HEVCContext *s, int x, in
copy_CTB_to_hv(s, src, stride_src, x0, y0, width, height, c_idx,
x_ctb, y_ctb);
if (s->ps.pps->transquant_bypass_enable_flag ||
- (s->ps.sps->pcm.loop_filter_disable_flag && s->ps.sps->pcm_enabled_flag)) {
+ (s->ps.sps->pcm_loop_filter_disabled && s->ps.sps->pcm_enabled)) {
dst = lc->edge_emu_buffer;
stride_dst = 2*MAX_PB_SIZE;
copy_CTB(dst, src, width << s->ps.sps->pixel_shift, height, stride_dst, stride_src);
@@ -500,8 +500,8 @@ static void deblocking_filter_CTB(const HEVCContext *s, int x0, int y0)
int cur_beta_offset = s->deblock[ctb].beta_offset;
int left_tc_offset, left_beta_offset;
int tc_offset, beta_offset;
- int pcmf = (s->ps.sps->pcm_enabled_flag &&
- s->ps.sps->pcm.loop_filter_disable_flag) ||
+ int pcmf = (s->ps.sps->pcm_enabled &&
+ s->ps.sps->pcm_loop_filter_disabled) ||
s->ps.pps->transquant_bypass_enable_flag;
if (x0) {
diff --git a/libavcodec/hevc_parser.c b/libavcodec/hevc_parser.c
index f174d419e4..73fc5d6372 100644
--- a/libavcodec/hevc_parser.c
+++ b/libavcodec/hevc_parser.c
@@ -150,7 +150,7 @@ static int hevc_parse_slice_header(AVCodecParserContext *s, H2645NAL *nal,
if (ps->pps->output_flag_present_flag)
skip_bits1(gb); // pic_output_flag
- if (ps->sps->separate_colour_plane_flag)
+ if (ps->sps->separate_colour_plane)
skip_bits(gb, 2); // colour_plane_id
if (!IS_IDR_NAL(nal)) {
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index 8d5fc0d0ca..1af691414e 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -895,7 +895,7 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
return AVERROR_INVALIDDATA;
}
- sps->temporal_id_nesting_flag = get_bits(gb, 1);
+ sps->temporal_id_nesting = get_bits(gb, 1);
if ((ret = parse_ptl(gb, avctx, &sps->ptl, sps->max_sub_layers)) < 0)
return ret;
@@ -913,9 +913,9 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
}
if (sps->chroma_format_idc == 3)
- sps->separate_colour_plane_flag = get_bits1(gb);
+ sps->separate_colour_plane = get_bits1(gb);
- if (sps->separate_colour_plane_flag)
+ if (sps->separate_colour_plane)
sps->chroma_format_idc = 0;
sps->width = get_ue_golomb_long(gb);
@@ -924,8 +924,8 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
sps->height, 0, avctx)) < 0)
return ret;
- sps->conformance_window_flag = get_bits1(gb);
- if (sps->conformance_window_flag) {
+ sps->conformance_window = get_bits1(gb);
+ if (sps->conformance_window) {
int vert_mult = hevc_sub_height_c[sps->chroma_format_idc];
int horiz_mult = hevc_sub_width_c[sps->chroma_format_idc];
sps->pic_conf_win.left_offset = get_ue_golomb_long(gb) * horiz_mult;
@@ -982,8 +982,8 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
return AVERROR_INVALIDDATA;
}
- sps->sublayer_ordering_info_flag = get_bits1(gb);
- start = sps->sublayer_ordering_info_flag ? 0 : sps->max_sub_layers - 1;
+ sps->sublayer_ordering_info = get_bits1(gb);
+ start = sps->sublayer_ordering_info ? 0 : sps->max_sub_layers - 1;
for (i = start; i < sps->max_sub_layers; i++) {
sps->temporal_layer[i].max_dec_pic_buffering = get_ue_golomb_long(gb) + 1;
sps->temporal_layer[i].num_reorder_pics = get_ue_golomb_long(gb);
@@ -1004,7 +1004,7 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
}
}
- if (!sps->sublayer_ordering_info_flag) {
+ if (!sps->sublayer_ordering_info) {
for (i = 0; i < start; i++) {
sps->temporal_layer[i].max_dec_pic_buffering = sps->temporal_layer[start].max_dec_pic_buffering;
sps->temporal_layer[i].num_reorder_pics = sps->temporal_layer[start].num_reorder_pics;
@@ -1043,8 +1043,8 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
sps->max_transform_hierarchy_depth_inter = get_ue_golomb_long(gb);
sps->max_transform_hierarchy_depth_intra = get_ue_golomb_long(gb);
- sps->scaling_list_enable_flag = get_bits1(gb);
- if (sps->scaling_list_enable_flag) {
+ sps->scaling_list_enabled = get_bits1(gb);
+ if (sps->scaling_list_enabled) {
set_default_scaling_list_data(&sps->scaling_list);
if (get_bits1(gb)) {
@@ -1054,11 +1054,11 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
}
}
- sps->amp_enabled_flag = get_bits1(gb);
- sps->sao_enabled = get_bits1(gb);
+ sps->amp_enabled = get_bits1(gb);
+ sps->sao_enabled = get_bits1(gb);
- sps->pcm_enabled_flag = get_bits1(gb);
- if (sps->pcm_enabled_flag) {
+ sps->pcm_enabled = get_bits1(gb);
+ if (sps->pcm_enabled) {
sps->pcm.bit_depth = get_bits(gb, 4) + 1;
sps->pcm.bit_depth_chroma = get_bits(gb, 4) + 1;
sps->pcm.log2_min_pcm_cb_size = get_ue_golomb_long(gb) + 3;
@@ -1071,7 +1071,7 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
return AVERROR_INVALIDDATA;
}
- sps->pcm.loop_filter_disable_flag = get_bits1(gb);
+ sps->pcm_loop_filter_disabled = get_bits1(gb);
}
sps->nb_st_rps = get_ue_golomb_long(gb);
@@ -1086,8 +1086,8 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
return ret;
}
- sps->long_term_ref_pics_present_flag = get_bits1(gb);
- if (sps->long_term_ref_pics_present_flag) {
+ sps->long_term_ref_pics_present = get_bits1(gb);
+ if (sps->long_term_ref_pics_present) {
sps->num_long_term_ref_pics_sps = get_ue_golomb_long(gb);
if (sps->num_long_term_ref_pics_sps > HEVC_MAX_LONG_TERM_REF_PICS) {
av_log(avctx, AV_LOG_ERROR, "Too many long term ref pics: %d.\n",
@@ -1102,54 +1102,53 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
}
}
- sps->sps_temporal_mvp_enabled_flag = get_bits1(gb);
- sps->sps_strong_intra_smoothing_enable_flag = get_bits1(gb);
+ sps->temporal_mvp_enabled = get_bits1(gb);
+ sps->strong_intra_smoothing_enabled = get_bits1(gb);
sps->vui.common.sar = (AVRational){0, 1};
sps->vui_present = get_bits1(gb);
if (sps->vui_present)
decode_vui(gb, avctx, apply_defdispwin, sps);
- sps->sps_extension_present_flag = get_bits1(gb);
- if (sps->sps_extension_present_flag) {
- sps->sps_range_extension_flag = get_bits1(gb);
- sps->sps_multilayer_extension_flag = get_bits1(gb);
- sps->sps_3d_extension_flag = get_bits1(gb);
- sps->sps_scc_extension_flag = get_bits1(gb);
+ sps->extension_present = get_bits1(gb);
+ if (sps->extension_present) {
+ sps->range_extension = get_bits1(gb);
+ sps->multilayer_extension = get_bits1(gb);
+ sps->sps_3d_extension = get_bits1(gb);
+ sps->scc_extension = get_bits1(gb);
skip_bits(gb, 4); // sps_extension_4bits
- if (sps->sps_range_extension_flag) {
- sps->transform_skip_rotation_enabled_flag = get_bits1(gb);
- sps->transform_skip_context_enabled_flag = get_bits1(gb);
- sps->implicit_rdpcm_enabled_flag = get_bits1(gb);
+ if (sps->range_extension) {
+ sps->transform_skip_rotation_enabled = get_bits1(gb);
+ sps->transform_skip_context_enabled = get_bits1(gb);
+ sps->implicit_rdpcm_enabled = get_bits1(gb);
+ sps->explicit_rdpcm_enabled = get_bits1(gb);
- sps->explicit_rdpcm_enabled_flag = get_bits1(gb);
-
- sps->extended_precision_processing_flag = get_bits1(gb);
- if (sps->extended_precision_processing_flag)
+ sps->extended_precision_processing = get_bits1(gb);
+ if (sps->extended_precision_processing)
av_log(avctx, AV_LOG_WARNING,
"extended_precision_processing_flag not yet implemented\n");
- sps->intra_smoothing_disabled_flag = get_bits1(gb);
- sps->high_precision_offsets_enabled_flag = get_bits1(gb);
- if (sps->high_precision_offsets_enabled_flag)
+ sps->intra_smoothing_disabled = get_bits1(gb);
+ sps->high_precision_offsets_enabled = get_bits1(gb);
+ if (sps->high_precision_offsets_enabled)
av_log(avctx, AV_LOG_WARNING,
"high_precision_offsets_enabled_flag not yet implemented\n");
- sps->persistent_rice_adaptation_enabled_flag = get_bits1(gb);
+ sps->persistent_rice_adaptation_enabled = get_bits1(gb);
- sps->cabac_bypass_alignment_enabled_flag = get_bits1(gb);
- if (sps->cabac_bypass_alignment_enabled_flag)
+ sps->cabac_bypass_alignment_enabled = get_bits1(gb);
+ if (sps->cabac_bypass_alignment_enabled)
av_log(avctx, AV_LOG_WARNING,
"cabac_bypass_alignment_enabled_flag not yet implemented\n");
}
- if (sps->sps_multilayer_extension_flag) {
+ if (sps->multilayer_extension) {
skip_bits1(gb); // inter_view_mv_vert_constraint_flag
av_log(avctx, AV_LOG_WARNING,
"sps_multilayer_extension_flag not yet implemented\n");
}
- if (sps->sps_3d_extension_flag) {
+ if (sps->sps_3d_extension) {
for (i = 0; i <= 1; i++) {
skip_bits1(gb); // iv_di_mc_enabled_flag
skip_bits1(gb); // iv_mv_scal_enabled_flag
@@ -1173,15 +1172,15 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
"sps_3d_extension_flag not yet implemented\n");
}
- if (sps->sps_scc_extension_flag) {
- sps->sps_curr_pic_ref_enabled_flag = get_bits1(gb);
- sps->palette_mode_enabled_flag = get_bits1(gb);
- if (sps->palette_mode_enabled_flag) {
+ if (sps->scc_extension) {
+ sps->curr_pic_ref_enabled = get_bits1(gb);
+ sps->palette_mode_enabled = get_bits1(gb);
+ if (sps->palette_mode_enabled) {
sps->palette_max_size = get_ue_golomb(gb);
sps->delta_palette_max_predictor_size = get_ue_golomb(gb);
- sps->sps_palette_predictor_initializers_present_flag = get_bits1(gb);
+ sps->palette_predictor_initializers_present = get_bits1(gb);
- if (sps->sps_palette_predictor_initializers_present_flag) {
+ if (sps->palette_predictor_initializers_present) {
sps->sps_num_palette_predictor_initializers = get_ue_golomb(gb) + 1;
if (sps->sps_num_palette_predictor_initializers > HEVC_MAX_PALETTE_PREDICTOR_SIZE) {
av_log(avctx, AV_LOG_ERROR,
@@ -1198,7 +1197,7 @@ int ff_hevc_parse_sps(HEVCSPS *sps, GetBitContext *gb, unsigned int *sps_id,
}
}
sps->motion_vector_resolution_control_idc = get_bits(gb, 2);
- sps->intra_boundary_filtering_disabled_flag = get_bits1(gb);
+ sps->intra_boundary_filtering_disabled = get_bits1(gb);
}
}
if (apply_defdispwin) {
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index b2f3a8dbd1..9801b4347f 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -188,11 +188,9 @@ typedef struct ScalingList {
typedef struct HEVCSPS {
unsigned vps_id;
int chroma_format_idc;
- uint8_t separate_colour_plane_flag;
HEVCWindow output_window;
- uint8_t conformance_window_flag;
HEVCWindow pic_conf_win;
HEVCHdrParams hdr;
@@ -203,32 +201,23 @@ typedef struct HEVCSPS {
enum AVPixelFormat pix_fmt;
unsigned int log2_max_poc_lsb;
- int pcm_enabled_flag;
- uint8_t sublayer_ordering_info_flag;
int max_sub_layers;
struct {
int max_dec_pic_buffering;
int num_reorder_pics;
int max_latency_increase;
} temporal_layer[HEVC_MAX_SUB_LAYERS];
- uint8_t temporal_id_nesting_flag;
int vui_present;
VUI vui;
PTL ptl;
- uint8_t sps_extension_present_flag;
- uint8_t scaling_list_enable_flag;
ScalingList scaling_list;
unsigned int nb_st_rps;
ShortTermRPS st_rps[HEVC_MAX_SHORT_TERM_REF_PIC_SETS];
- uint8_t amp_enabled_flag;
- uint8_t sao_enabled;
-
- uint8_t long_term_ref_pics_present_flag;
uint16_t lt_ref_pic_poc_lsb_sps[HEVC_MAX_LONG_TERM_REF_PICS];
uint32_t used_by_curr_pic_lt;
uint8_t num_long_term_ref_pics_sps;
@@ -238,10 +227,7 @@ typedef struct HEVCSPS {
uint8_t bit_depth_chroma;
unsigned int log2_min_pcm_cb_size;
unsigned int log2_max_pcm_cb_size;
- uint8_t loop_filter_disable_flag;
} pcm;
- uint8_t sps_temporal_mvp_enabled_flag;
- uint8_t sps_strong_intra_smoothing_enable_flag;
unsigned int log2_min_cb_size;
unsigned int log2_diff_max_min_coding_block_size;
@@ -254,30 +240,44 @@ typedef struct HEVCSPS {
int max_transform_hierarchy_depth_inter;
int max_transform_hierarchy_depth_intra;
- int sps_range_extension_flag;
- int transform_skip_rotation_enabled_flag;
- int transform_skip_context_enabled_flag;
- int implicit_rdpcm_enabled_flag;
- int explicit_rdpcm_enabled_flag;
- int extended_precision_processing_flag;
- int intra_smoothing_disabled_flag;
- int high_precision_offsets_enabled_flag;
- int persistent_rice_adaptation_enabled_flag;
- int cabac_bypass_alignment_enabled_flag;
+ unsigned separate_colour_plane:1;
+ unsigned conformance_window:1;
+ unsigned pcm_enabled:1;
+ unsigned pcm_loop_filter_disabled:1;
+ unsigned sublayer_ordering_info:1;
+ unsigned temporal_id_nesting:1;
+ unsigned extension_present:1;
+ unsigned scaling_list_enabled:1;
+ unsigned amp_enabled:1;
+ unsigned sao_enabled:1;
+ unsigned long_term_ref_pics_present:1;
+ unsigned temporal_mvp_enabled:1;
+ unsigned strong_intra_smoothing_enabled:1;
+ unsigned range_extension:1;
+ unsigned transform_skip_rotation_enabled:1;
+ unsigned transform_skip_context_enabled:1;
+ unsigned implicit_rdpcm_enabled:1;
+ unsigned explicit_rdpcm_enabled:1;
+ unsigned extended_precision_processing:1;
+ unsigned intra_smoothing_disabled:1;
+ unsigned high_precision_offsets_enabled:1;
+ unsigned persistent_rice_adaptation_enabled:1;
+ unsigned cabac_bypass_alignment_enabled:1;
- int sps_multilayer_extension_flag;
- int sps_3d_extension_flag;
+ unsigned multilayer_extension:1;
+ unsigned sps_3d_extension:1;
+
+ unsigned scc_extension:1;
+ unsigned curr_pic_ref_enabled:1;
+ unsigned palette_mode_enabled:1;
+ unsigned palette_predictor_initializers_present:1;
+ unsigned intra_boundary_filtering_disabled:1;
- int sps_scc_extension_flag;
- int sps_curr_pic_ref_enabled_flag;
- int palette_mode_enabled_flag;
int palette_max_size;
int delta_palette_max_predictor_size;
- int sps_palette_predictor_initializers_present_flag;
int sps_num_palette_predictor_initializers;
int sps_palette_predictor_initializer[3][HEVC_MAX_PALETTE_PREDICTOR_SIZE];
int motion_vector_resolution_control_idc;
- int intra_boundary_filtering_disabled_flag;
///< coded frame dimension in various units
int width;
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index 4a7046bdec..eb03c45e8a 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -272,7 +272,7 @@ static int decode_lt_rps(HEVCContext *s, LongTermRPS *rps, GetBitContext *gb)
int i;
rps->nb_refs = 0;
- if (!sps->long_term_ref_pics_present_flag)
+ if (!sps->long_term_ref_pics_present)
return 0;
if (sps->num_long_term_ref_pics_sps > 0)
@@ -705,7 +705,7 @@ static int hls_slice_header(HEVCContext *s)
if (s->ps.pps->output_flag_present_flag)
sh->pic_output_flag = get_bits1(gb);
- if (s->ps.sps->separate_colour_plane_flag)
+ if (s->ps.sps->separate_colour_plane)
sh->colour_plane_id = get_bits(gb, 2);
if (!IS_IDR(s)) {
@@ -753,7 +753,7 @@ static int hls_slice_header(HEVCContext *s)
}
sh->long_term_ref_pic_set_size = pos - get_bits_left(gb);
- if (s->ps.sps->sps_temporal_mvp_enabled_flag)
+ if (s->ps.sps->temporal_mvp_enabled)
sh->slice_temporal_mvp_enabled_flag = get_bits1(gb);
else
sh->slice_temporal_mvp_enabled_flag = 0;
@@ -2294,7 +2294,7 @@ static int hls_coding_unit(HEVCLocalContext *lc, const HEVCContext *s, int x0, i
}
if (lc->cu.pred_mode == MODE_INTRA) {
- if (lc->cu.part_mode == PART_2Nx2N && s->ps.sps->pcm_enabled_flag &&
+ if (lc->cu.part_mode == PART_2Nx2N && s->ps.sps->pcm_enabled &&
log2_cb_size >= s->ps.sps->pcm.log2_min_pcm_cb_size &&
log2_cb_size <= s->ps.sps->pcm.log2_max_pcm_cb_size) {
pcm_flag = ff_hevc_pcm_flag_decode(lc);
@@ -2302,7 +2302,7 @@ static int hls_coding_unit(HEVCLocalContext *lc, const HEVCContext *s, int x0, i
if (pcm_flag) {
intra_prediction_unit_default_value(lc, x0, y0, log2_cb_size);
ret = hls_pcm_sample(lc, x0, y0, log2_cb_size);
- if (s->ps.sps->pcm.loop_filter_disable_flag)
+ if (s->ps.sps->pcm_loop_filter_disabled)
set_deblocking_bypass(s, x0, y0, log2_cb_size);
if (ret < 0)
diff --git a/libavcodec/hevcpred_template.c b/libavcodec/hevcpred_template.c
index 46bd806523..9082a816e0 100644
--- a/libavcodec/hevcpred_template.c
+++ b/libavcodec/hevcpred_template.c
@@ -285,14 +285,14 @@ do { \
top[-1] = left[-1];
// Filtering process
- if (!s->ps.sps->intra_smoothing_disabled_flag && (c_idx == 0 || s->ps.sps->chroma_format_idc == 3)) {
+ if (!s->ps.sps->intra_smoothing_disabled && (c_idx == 0 || s->ps.sps->chroma_format_idc == 3)) {
if (mode != INTRA_DC && size != 4){
int intra_hor_ver_dist_thresh[] = { 7, 1, 0 };
int min_dist_vert_hor = FFMIN(FFABS((int)(mode - 26U)),
FFABS((int)(mode - 10U)));
if (min_dist_vert_hor > intra_hor_ver_dist_thresh[log2_size - 3]) {
int threshold = 1 << (BIT_DEPTH - 5);
- if (s->ps.sps->sps_strong_intra_smoothing_enable_flag && c_idx == 0 &&
+ if (s->ps.sps->strong_intra_smoothing_enabled && c_idx == 0 &&
log2_size == 5 &&
FFABS(top[-1] + top[63] - 2 * top[31]) < threshold &&
FFABS(left[-1] + left[63] - 2 * left[31]) < threshold) {
diff --git a/libavcodec/mips/hevcpred_msa.c b/libavcodec/mips/hevcpred_msa.c
index b056ee986b..d9137519ad 100644
--- a/libavcodec/mips/hevcpred_msa.c
+++ b/libavcodec/mips/hevcpred_msa.c
@@ -2369,7 +2369,7 @@ void ff_intra_pred_8_16x16_msa(HEVCLocalContext *lc, int x0, int y0, int c_idx)
top[-1] = left[-1];
- if (!s->ps.sps->intra_smoothing_disabled_flag
+ if (!s->ps.sps->intra_smoothing_disabled
&& (c_idx == 0 || s->ps.sps->chroma_format_idc == 3)) {
if (mode != INTRA_DC && 16 != 4) {
int intra_hor_ver_dist_thresh[] = { 7, 1, 0 };
@@ -2886,7 +2886,7 @@ void ff_intra_pred_8_32x32_msa(HEVCLocalContext *lc, int x0, int y0, int c_idx)
top[-1] = left[-1];
- if (!s->ps.sps->intra_smoothing_disabled_flag
+ if (!s->ps.sps->intra_smoothing_disabled
&& (c_idx == 0 || s->ps.sps->chroma_format_idc == 3)) {
if (mode != INTRA_DC && 32 != 4) {
int intra_hor_ver_dist_thresh[] = { 7, 1, 0 };
@@ -2901,7 +2901,7 @@ void ff_intra_pred_8_32x32_msa(HEVCLocalContext *lc, int x0, int y0, int c_idx)
0 ? ((int) (mode - 26U)) : (-((int) (mode - 26U))))));
if (min_dist_vert_hor > intra_hor_ver_dist_thresh[5 - 3]) {
int threshold = 1 << (8 - 5);
- if (s->ps.sps->sps_strong_intra_smoothing_enable_flag
+ if (s->ps.sps->strong_intra_smoothing_enabled
&& c_idx == 0
&& ((top[-1] + top[63] - 2 * top[31]) >=
0 ? (top[-1] + top[63] -
diff --git a/libavcodec/nvdec_hevc.c b/libavcodec/nvdec_hevc.c
index f26ced71ef..3d704666d0 100644
--- a/libavcodec/nvdec_hevc.c
+++ b/libavcodec/nvdec_hevc.c
@@ -105,44 +105,44 @@ static int nvdec_hevc_start_frame(AVCodecContext *avctx,
.log2_diff_max_min_luma_coding_block_size = sps->log2_diff_max_min_coding_block_size,
.log2_min_transform_block_size_minus2 = sps->log2_min_tb_size - 2,
.log2_diff_max_min_transform_block_size = sps->log2_max_trafo_size - sps->log2_min_tb_size,
- .pcm_enabled_flag = sps->pcm_enabled_flag,
- .log2_min_pcm_luma_coding_block_size_minus3 = sps->pcm_enabled_flag ? sps->pcm.log2_min_pcm_cb_size - 3 : 0,
+ .pcm_enabled_flag = sps->pcm_enabled,
+ .log2_min_pcm_luma_coding_block_size_minus3 = sps->pcm_enabled ? sps->pcm.log2_min_pcm_cb_size - 3 : 0,
.log2_diff_max_min_pcm_luma_coding_block_size = sps->pcm.log2_max_pcm_cb_size - sps->pcm.log2_min_pcm_cb_size,
- .pcm_sample_bit_depth_luma_minus1 = sps->pcm_enabled_flag ? sps->pcm.bit_depth - 1 : 0,
- .pcm_sample_bit_depth_chroma_minus1 = sps->pcm_enabled_flag ? sps->pcm.bit_depth_chroma - 1 : 0,
+ .pcm_sample_bit_depth_luma_minus1 = sps->pcm_enabled ? sps->pcm.bit_depth - 1 : 0,
+ .pcm_sample_bit_depth_chroma_minus1 = sps->pcm_enabled ? sps->pcm.bit_depth_chroma - 1 : 0,
#if NVDECAPI_CHECK_VERSION(8, 1)
.log2_max_transform_skip_block_size_minus2 = pps->log2_max_transform_skip_block_size - 2,
.log2_sao_offset_scale_luma = pps->log2_sao_offset_scale_luma,
.log2_sao_offset_scale_chroma = pps->log2_sao_offset_scale_chroma,
- .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled_flag,
+ .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled,
#endif
- .pcm_loop_filter_disabled_flag = sps->pcm.loop_filter_disable_flag,
- .strong_intra_smoothing_enabled_flag = sps->sps_strong_intra_smoothing_enable_flag,
+ .pcm_loop_filter_disabled_flag = sps->pcm_loop_filter_disabled,
+ .strong_intra_smoothing_enabled_flag = sps->strong_intra_smoothing_enabled,
.max_transform_hierarchy_depth_intra = sps->max_transform_hierarchy_depth_intra,
.max_transform_hierarchy_depth_inter = sps->max_transform_hierarchy_depth_inter,
- .amp_enabled_flag = sps->amp_enabled_flag,
- .separate_colour_plane_flag = sps->separate_colour_plane_flag,
+ .amp_enabled_flag = sps->amp_enabled,
+ .separate_colour_plane_flag = sps->separate_colour_plane,
.log2_max_pic_order_cnt_lsb_minus4 = sps->log2_max_poc_lsb - 4,
.num_short_term_ref_pic_sets = sps->nb_st_rps,
- .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present_flag,
+ .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present,
.num_long_term_ref_pics_sps = sps->num_long_term_ref_pics_sps,
- .sps_temporal_mvp_enabled_flag = sps->sps_temporal_mvp_enabled_flag,
+ .sps_temporal_mvp_enabled_flag = sps->temporal_mvp_enabled,
.sample_adaptive_offset_enabled_flag = sps->sao_enabled,
- .scaling_list_enable_flag = sps->scaling_list_enable_flag,
+ .scaling_list_enable_flag = sps->scaling_list_enabled,
.IrapPicFlag = IS_IRAP(s),
.IdrPicFlag = IS_IDR(s),
.bit_depth_luma_minus8 = sps->bit_depth - 8,
.bit_depth_chroma_minus8 = sps->bit_depth - 8,
#if NVDECAPI_CHECK_VERSION(9, 0)
- .sps_range_extension_flag = sps->sps_range_extension_flag,
- .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled_flag,
- .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled_flag,
- .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled_flag,
- .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled_flag,
- .extended_precision_processing_flag = sps->extended_precision_processing_flag,
- .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled_flag,
- .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled_flag,
- .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled_flag,
+ .sps_range_extension_flag = sps->range_extension,
+ .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled,
+ .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled,
+ .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled,
+ .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled,
+ .extended_precision_processing_flag = sps->extended_precision_processing,
+ .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled,
+ .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled,
+ .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled,
#endif
.dependent_slice_segments_enabled_flag = pps->dependent_slice_segments_enabled_flag,
diff --git a/libavcodec/qsvenc_hevc.c b/libavcodec/qsvenc_hevc.c
index 2ccbe0464b..40cff6335a 100644
--- a/libavcodec/qsvenc_hevc.c
+++ b/libavcodec/qsvenc_hevc.c
@@ -108,7 +108,7 @@ static int generate_fake_vps(QSVEncContext *q, AVCodecContext *avctx)
/* generate the VPS */
vps.vps_max_layers = 1;
vps.vps_max_sub_layers = sps.max_sub_layers;
- vps.vps_temporal_id_nesting_flag = sps.temporal_id_nesting_flag;
+ vps.vps_temporal_id_nesting_flag = sps.temporal_id_nesting;
memcpy(&vps.ptl, &sps.ptl, sizeof(vps.ptl));
vps.vps_sub_layer_ordering_info_present_flag = 1;
for (i = 0; i < HEVC_MAX_SUB_LAYERS; i++) {
diff --git a/libavcodec/vaapi_hevc.c b/libavcodec/vaapi_hevc.c
index 3bdd2dd1b8..5af297c2af 100644
--- a/libavcodec/vaapi_hevc.c
+++ b/libavcodec/vaapi_hevc.c
@@ -171,12 +171,12 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
.pic_fields.bits = {
.chroma_format_idc = sps->chroma_format_idc,
.tiles_enabled_flag = pps->tiles_enabled_flag,
- .separate_colour_plane_flag = sps->separate_colour_plane_flag,
- .pcm_enabled_flag = sps->pcm_enabled_flag,
- .scaling_list_enabled_flag = sps->scaling_list_enable_flag,
+ .separate_colour_plane_flag = sps->separate_colour_plane,
+ .pcm_enabled_flag = sps->pcm_enabled,
+ .scaling_list_enabled_flag = sps->scaling_list_enabled,
.transform_skip_enabled_flag = pps->transform_skip_enabled_flag,
- .amp_enabled_flag = sps->amp_enabled_flag,
- .strong_intra_smoothing_enabled_flag = sps->sps_strong_intra_smoothing_enable_flag,
+ .amp_enabled_flag = sps->amp_enabled,
+ .strong_intra_smoothing_enabled_flag = sps->strong_intra_smoothing_enabled,
.sign_data_hiding_enabled_flag = pps->sign_data_hiding_flag,
.constrained_intra_pred_flag = pps->constrained_intra_pred_flag,
.cu_qp_delta_enabled_flag = pps->cu_qp_delta_enabled_flag,
@@ -186,12 +186,12 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
.entropy_coding_sync_enabled_flag = pps->entropy_coding_sync_enabled_flag,
.pps_loop_filter_across_slices_enabled_flag = pps->seq_loop_filter_across_slices_enabled_flag,
.loop_filter_across_tiles_enabled_flag = pps->loop_filter_across_tiles_enabled_flag,
- .pcm_loop_filter_disabled_flag = sps->pcm.loop_filter_disable_flag,
+ .pcm_loop_filter_disabled_flag = sps->pcm_loop_filter_disabled,
},
.slice_parsing_fields.bits = {
.lists_modification_present_flag = pps->lists_modification_present_flag,
- .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present_flag,
- .sps_temporal_mvp_enabled_flag = sps->sps_temporal_mvp_enabled_flag,
+ .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present,
+ .sps_temporal_mvp_enabled_flag = sps->temporal_mvp_enabled,
.cabac_init_present_flag = pps->cabac_init_present_flag,
.output_flag_present_flag = pps->output_flag_present_flag,
.dependent_slice_segments_enabled_flag = pps->dependent_slice_segments_enabled_flag,
@@ -231,15 +231,15 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
avctx->profile == AV_PROFILE_HEVC_SCC) {
pic->pic_param.rext = (VAPictureParameterBufferHEVCRext) {
.range_extension_pic_fields.bits = {
- .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled_flag,
- .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled_flag,
- .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled_flag,
- .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled_flag,
- .extended_precision_processing_flag = sps->extended_precision_processing_flag,
- .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled_flag,
- .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled_flag,
- .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled_flag,
- .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled_flag,
+ .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled,
+ .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled,
+ .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled,
+ .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled,
+ .extended_precision_processing_flag = sps->extended_precision_processing,
+ .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled,
+ .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled,
+ .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled,
+ .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled,
.cross_component_prediction_enabled_flag = pps->cross_component_prediction_enabled_flag,
.chroma_qp_offset_list_enabled_flag = pps->chroma_qp_offset_list_enabled_flag,
},
@@ -258,7 +258,7 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
pre_palette_size = pps->pps_palette_predictor_initializers_present_flag ?
pps->pps_num_palette_predictor_initializers :
- (sps->sps_palette_predictor_initializers_present_flag ?
+ (sps->palette_predictor_initializers_present ?
sps->sps_num_palette_predictor_initializers :
0);
@@ -266,9 +266,9 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
pic->pic_param.scc = (VAPictureParameterBufferHEVCScc) {
.screen_content_pic_fields.bits = {
.pps_curr_pic_ref_enabled_flag = pps->pps_curr_pic_ref_enabled_flag,
- .palette_mode_enabled_flag = sps->palette_mode_enabled_flag,
+ .palette_mode_enabled_flag = sps->palette_mode_enabled,
.motion_vector_resolution_control_idc = sps->motion_vector_resolution_control_idc,
- .intra_boundary_filtering_disabled_flag = sps->intra_boundary_filtering_disabled_flag,
+ .intra_boundary_filtering_disabled_flag = sps->intra_boundary_filtering_disabled,
.residual_adaptive_colour_transform_enabled_flag
= pps->residual_adaptive_colour_transform_enabled_flag,
.pps_slice_act_qp_offsets_present_flag = pps->pps_slice_act_qp_offsets_present_flag,
@@ -305,7 +305,7 @@ static int vaapi_hevc_start_frame(AVCodecContext *avctx,
if (pps->scaling_list_data_present_flag)
scaling_list = &pps->scaling_list;
- else if (sps->scaling_list_enable_flag)
+ else if (sps->scaling_list_enabled)
scaling_list = &sps->scaling_list;
if (scaling_list) {
diff --git a/libavcodec/vdpau_hevc.c b/libavcodec/vdpau_hevc.c
index 4cd7ce5621..b029f9c42e 100644
--- a/libavcodec/vdpau_hevc.c
+++ b/libavcodec/vdpau_hevc.c
@@ -53,7 +53,7 @@ static int vdpau_hevc_start_frame(AVCodecContext *avctx,
/* SPS */
info->chroma_format_idc = sps->chroma_format_idc;
- info->separate_colour_plane_flag = sps->separate_colour_plane_flag;
+ info->separate_colour_plane_flag = sps->separate_colour_plane;
info->pic_width_in_luma_samples = sps->width;
info->pic_height_in_luma_samples = sps->height;
info->bit_depth_luma_minus8 = sps->bit_depth - 8;
@@ -68,7 +68,7 @@ static int vdpau_hevc_start_frame(AVCodecContext *avctx,
info->log2_diff_max_min_transform_block_size = sps->log2_max_trafo_size - sps->log2_min_tb_size;
info->max_transform_hierarchy_depth_inter = sps->max_transform_hierarchy_depth_inter;
info->max_transform_hierarchy_depth_intra = sps->max_transform_hierarchy_depth_intra;
- info->scaling_list_enabled_flag = sps->scaling_list_enable_flag;
+ info->scaling_list_enabled_flag = sps->scaling_list_enabled;
/* Scaling lists, in diagonal order, to be used for this frame. */
for (size_t i = 0; i < 6; i++) {
for (size_t j = 0; j < 16; j++) {
@@ -100,9 +100,9 @@ static int vdpau_hevc_start_frame(AVCodecContext *avctx,
info->ScalingListDCCoeff32x32[i] = sl->sl_dc[1][i * 3];
}
}
- info->amp_enabled_flag = sps->amp_enabled_flag;
+ info->amp_enabled_flag = sps->amp_enabled;
info->sample_adaptive_offset_enabled_flag = sps->sao_enabled;
- info->pcm_enabled_flag = sps->pcm_enabled_flag;
+ info->pcm_enabled_flag = sps->pcm_enabled;
if (info->pcm_enabled_flag) {
/* Only needs to be set if pcm_enabled_flag is set. Ignored otherwise. */
info->pcm_sample_bit_depth_luma_minus1 = sps->pcm.bit_depth - 1;
@@ -113,17 +113,17 @@ static int vdpau_hevc_start_frame(AVCodecContext *avctx,
/* Only needs to be set if pcm_enabled_flag is set. Ignored otherwise. */
info->log2_diff_max_min_pcm_luma_coding_block_size = sps->pcm.log2_max_pcm_cb_size - sps->pcm.log2_min_pcm_cb_size;
/* Only needs to be set if pcm_enabled_flag is set. Ignored otherwise. */
- info->pcm_loop_filter_disabled_flag = sps->pcm.loop_filter_disable_flag;
+ info->pcm_loop_filter_disabled_flag = sps->pcm_loop_filter_disabled;
}
/* Per spec, when zero, assume short_term_ref_pic_set_sps_flag
is also zero. */
info->num_short_term_ref_pic_sets = sps->nb_st_rps;
- info->long_term_ref_pics_present_flag = sps->long_term_ref_pics_present_flag;
+ info->long_term_ref_pics_present_flag = sps->long_term_ref_pics_present;
/* Only needed if long_term_ref_pics_present_flag is set. Ignored
otherwise. */
info->num_long_term_ref_pics_sps = sps->num_long_term_ref_pics_sps;
- info->sps_temporal_mvp_enabled_flag = sps->sps_temporal_mvp_enabled_flag;
- info->strong_intra_smoothing_enabled_flag = sps->sps_strong_intra_smoothing_enable_flag;
+ info->sps_temporal_mvp_enabled_flag = sps->temporal_mvp_enabled;
+ info->strong_intra_smoothing_enabled_flag = sps->strong_intra_smoothing_enabled;
/* Copy the HEVC Picture Parameter Set bitstream fields. */
info->dependent_slice_segments_enabled_flag = pps->dependent_slice_segments_enabled_flag;
@@ -360,17 +360,17 @@ static int vdpau_hevc_start_frame(AVCodecContext *avctx,
}
#ifdef VDP_YCBCR_FORMAT_Y_U_V_444
- if (sps->sps_range_extension_flag) {
+ if (sps->range_extension) {
info2->sps_range_extension_flag = 1;
- info2->transformSkipRotationEnableFlag = sps->transform_skip_rotation_enabled_flag;
- info2->transformSkipContextEnableFlag = sps->transform_skip_context_enabled_flag;
- info2->implicitRdpcmEnableFlag = sps->implicit_rdpcm_enabled_flag;
- info2->explicitRdpcmEnableFlag = sps->explicit_rdpcm_enabled_flag;
- info2->extendedPrecisionProcessingFlag = sps->extended_precision_processing_flag;
- info2->intraSmoothingDisabledFlag = sps->intra_smoothing_disabled_flag;
- info2->highPrecisionOffsetsEnableFlag = sps->high_precision_offsets_enabled_flag;
- info2->persistentRiceAdaptationEnableFlag = sps->persistent_rice_adaptation_enabled_flag;
- info2->cabacBypassAlignmentEnableFlag = sps->cabac_bypass_alignment_enabled_flag;
+ info2->transformSkipRotationEnableFlag = sps->transform_skip_rotation_enabled;
+ info2->transformSkipContextEnableFlag = sps->transform_skip_context_enabled;
+ info2->implicitRdpcmEnableFlag = sps->implicit_rdpcm_enabled;
+ info2->explicitRdpcmEnableFlag = sps->explicit_rdpcm_enabled;
+ info2->extendedPrecisionProcessingFlag = sps->extended_precision_processing;
+ info2->intraSmoothingDisabledFlag = sps->intra_smoothing_disabled;
+ info2->highPrecisionOffsetsEnableFlag = sps->high_precision_offsets_enabled;
+ info2->persistentRiceAdaptationEnableFlag = sps->persistent_rice_adaptation_enabled;
+ info2->cabacBypassAlignmentEnableFlag = sps->cabac_bypass_alignment_enabled;
} else {
info2->sps_range_extension_flag = 0;
}
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index 9b40f5ad58..df86049d22 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -395,36 +395,36 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
*vksps = (StdVideoH265SequenceParameterSet) {
.flags = (StdVideoH265SpsFlags) {
- .sps_temporal_id_nesting_flag = sps->temporal_id_nesting_flag,
- .separate_colour_plane_flag = sps->separate_colour_plane_flag,
- .conformance_window_flag = sps->conformance_window_flag,
- .sps_sub_layer_ordering_info_present_flag = sps->sublayer_ordering_info_flag,
- .scaling_list_enabled_flag = sps->scaling_list_enable_flag,
- .sps_scaling_list_data_present_flag = sps->scaling_list_enable_flag,
- .amp_enabled_flag = sps->amp_enabled_flag,
+ .sps_temporal_id_nesting_flag = sps->temporal_id_nesting,
+ .separate_colour_plane_flag = sps->separate_colour_plane,
+ .conformance_window_flag = sps->conformance_window,
+ .sps_sub_layer_ordering_info_present_flag = sps->sublayer_ordering_info,
+ .scaling_list_enabled_flag = sps->scaling_list_enabled,
+ .sps_scaling_list_data_present_flag = sps->scaling_list_enabled,
+ .amp_enabled_flag = sps->amp_enabled,
.sample_adaptive_offset_enabled_flag = sps->sao_enabled,
- .pcm_enabled_flag = sps->pcm_enabled_flag,
- .pcm_loop_filter_disabled_flag = sps->pcm.loop_filter_disable_flag,
- .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present_flag,
- .sps_temporal_mvp_enabled_flag = sps->sps_temporal_mvp_enabled_flag,
- .strong_intra_smoothing_enabled_flag = sps->sps_strong_intra_smoothing_enable_flag,
+ .pcm_enabled_flag = sps->pcm_enabled,
+ .pcm_loop_filter_disabled_flag = sps->pcm_loop_filter_disabled,
+ .long_term_ref_pics_present_flag = sps->long_term_ref_pics_present,
+ .sps_temporal_mvp_enabled_flag = sps->temporal_mvp_enabled,
+ .strong_intra_smoothing_enabled_flag = sps->strong_intra_smoothing_enabled,
.vui_parameters_present_flag = sps->vui_present,
- .sps_extension_present_flag = sps->sps_extension_present_flag,
- .sps_range_extension_flag = sps->sps_range_extension_flag,
- .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled_flag,
- .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled_flag,
- .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled_flag,
- .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled_flag,
- .extended_precision_processing_flag = sps->extended_precision_processing_flag,
- .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled_flag,
- .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled_flag,
- .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled_flag,
- .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled_flag,
- .sps_scc_extension_flag = sps->sps_scc_extension_flag,
- .sps_curr_pic_ref_enabled_flag = sps->sps_curr_pic_ref_enabled_flag,
- .palette_mode_enabled_flag = sps->palette_mode_enabled_flag,
- .sps_palette_predictor_initializers_present_flag = sps->sps_palette_predictor_initializers_present_flag,
- .intra_boundary_filtering_disabled_flag = sps->intra_boundary_filtering_disabled_flag,
+ .sps_extension_present_flag = sps->extension_present,
+ .sps_range_extension_flag = sps->range_extension,
+ .transform_skip_rotation_enabled_flag = sps->transform_skip_rotation_enabled,
+ .transform_skip_context_enabled_flag = sps->transform_skip_context_enabled,
+ .implicit_rdpcm_enabled_flag = sps->implicit_rdpcm_enabled,
+ .explicit_rdpcm_enabled_flag = sps->explicit_rdpcm_enabled,
+ .extended_precision_processing_flag = sps->extended_precision_processing,
+ .intra_smoothing_disabled_flag = sps->intra_smoothing_disabled,
+ .high_precision_offsets_enabled_flag = sps->high_precision_offsets_enabled,
+ .persistent_rice_adaptation_enabled_flag = sps->persistent_rice_adaptation_enabled,
+ .cabac_bypass_alignment_enabled_flag = sps->cabac_bypass_alignment_enabled,
+ .sps_scc_extension_flag = sps->scc_extension,
+ .sps_curr_pic_ref_enabled_flag = sps->curr_pic_ref_enabled,
+ .palette_mode_enabled_flag = sps->palette_mode_enabled,
+ .sps_palette_predictor_initializers_present_flag = sps->palette_predictor_initializers_present,
+ .intra_boundary_filtering_disabled_flag = sps->intra_boundary_filtering_disabled,
},
.chroma_format_idc = sps->chroma_format_idc,
.pic_width_in_luma_samples = sps->width,
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 07/10] lavc/hevc_ps: fix variable signedness in ff_hevc_decode_short_term_rps()
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (4 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 08/10] lavc/hevc_ps: do not store delta_poc_s[01] in ShortTermRPS Anton Khirnov
` (2 subsequent siblings)
8 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
It is actually supposed to go negative in the loop over num_negative
pics, but underflow does not break anything as the result is then
assigned to a signed int.
---
libavcodec/hevc_ps.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index 1af691414e..d90f172c46 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -197,7 +197,8 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
}
} else {
- unsigned int prev, nb_positive_pics;
+ unsigned int nb_positive_pics;
+
rps->num_negative_pics = get_ue_golomb_long(gb);
nb_positive_pics = get_ue_golomb_long(gb);
@@ -209,7 +210,8 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
rps->num_delta_pocs = rps->num_negative_pics + nb_positive_pics;
if (rps->num_delta_pocs) {
- prev = 0;
+ int prev = 0;
+
for (i = 0; i < rps->num_negative_pics; i++) {
delta_poc = rps->delta_poc_s0[i] = get_ue_golomb_long(gb) + 1;
if (delta_poc < 1 || delta_poc > 32768) {
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 08/10] lavc/hevc_ps: do not store delta_poc_s[01] in ShortTermRPS
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (5 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 07/10] lavc/hevc_ps: fix variable signedness in ff_hevc_decode_short_term_rps() Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS Anton Khirnov
8 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
They are only used in vulkan_hevc and are not actually needed, as they
can be computed from delta_poc.
Reduces sizeof(HEVCSPS) by 16kB.
---
libavcodec/hevc_ps.c | 4 ++--
libavcodec/hevc_ps.h | 2 --
libavcodec/vulkan_hevc.c | 7 +++++--
3 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index d90f172c46..a6b0021bc3 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -213,7 +213,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
int prev = 0;
for (i = 0; i < rps->num_negative_pics; i++) {
- delta_poc = rps->delta_poc_s0[i] = get_ue_golomb_long(gb) + 1;
+ delta_poc = get_ue_golomb_long(gb) + 1;
if (delta_poc < 1 || delta_poc > 32768) {
av_log(avctx, AV_LOG_ERROR,
"Invalid value of delta_poc: %d\n",
@@ -226,7 +226,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
prev = 0;
for (i = 0; i < nb_positive_pics; i++) {
- delta_poc = rps->delta_poc_s1[i] = get_ue_golomb_long(gb) + 1;
+ delta_poc = get_ue_golomb_long(gb) + 1;
if (delta_poc < 1 || delta_poc > 32768) {
av_log(avctx, AV_LOG_ERROR,
"Invalid value of delta_poc: %d\n",
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index 9801b4347f..6ef29a8ea7 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -78,8 +78,6 @@ typedef struct ShortTermRPS {
unsigned int num_negative_pics;
int num_delta_pocs;
int rps_idx_num_delta_pocs;
- int32_t delta_poc_s0[32];
- int32_t delta_poc_s1[32];
int32_t delta_poc[32];
uint8_t used[32];
} ShortTermRPS;
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index df86049d22..5d7c6b1b64 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -351,6 +351,8 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
pal->PredictorPaletteEntries[i][j] = sps->sps_palette_predictor_initializer[i][j];
for (int i = 0; i < sps->nb_st_rps; i++) {
+ const ShortTermRPS *st_rps = &sps->st_rps[i];
+
str[i] = (StdVideoH265ShortTermRefPicSet) {
.flags = (StdVideoH265ShortTermRefPicSetFlags) {
.inter_ref_pic_set_prediction_flag = sps->st_rps[i].rps_predict,
@@ -375,12 +377,13 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
for (int j = 0; j < str[i].num_negative_pics; j++) {
- str[i].delta_poc_s0_minus1[j] = sps->st_rps[i].delta_poc_s0[j] - 1;
+ str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
}
for (int j = 0; j < str[i].num_positive_pics; j++) {
- str[i].delta_poc_s1_minus1[j] = sps->st_rps[i].delta_poc_s1[j] - 1;
+ str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[st_rps->num_negative_pics + j] -
+ (j ? st_rps->delta_poc[st_rps->num_negative_pics + j - 1] : 0) - 1;
str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[str[i].num_negative_pics + j] << j;
}
}
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (6 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 08/10] lavc/hevc_ps: do not store delta_poc_s[01] in ShortTermRPS Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-10 13:42 ` James Almer
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS Anton Khirnov
8 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
It is currently an array of 32 uint8_t, each storing a single flag. A
single uint32_t is sufficient.
Reduces sizeof(HEVCSPS) by 1792 bytes.
---
libavcodec/hevc_ps.c | 33 +++++++++++++++++++--------------
libavcodec/hevc_ps.h | 2 +-
libavcodec/hevc_refs.c | 6 +++---
libavcodec/vulkan_hevc.c | 8 ++++----
4 files changed, 27 insertions(+), 22 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index a6b0021bc3..76fe507e7b 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -107,6 +107,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
int k = 0;
int i;
+ rps->used = 0;
rps->rps_predict = 0;
if (rps != sps->st_rps && sps->nb_st_rps)
@@ -114,6 +115,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
if (rps->rps_predict) {
const ShortTermRPS *rps_ridx;
+ uint8_t used[32] = { 0 };
int delta_rps;
if (is_slice_header) {
@@ -139,13 +141,13 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
delta_rps = (1 - (rps->delta_rps_sign << 1)) * rps->abs_delta_rps;
for (i = 0; i <= rps_ridx->num_delta_pocs; i++) {
- int used = rps->used[k] = get_bits1(gb);
+ used[k] = get_bits1(gb);
rps->use_delta_flag = 0;
- if (!used)
+ if (!used[k])
rps->use_delta_flag = get_bits1(gb);
- if (used || rps->use_delta_flag) {
+ if (used[k] || rps->use_delta_flag) {
if (i < rps_ridx->num_delta_pocs)
delta_poc = delta_rps + rps_ridx->delta_poc[i];
else
@@ -157,7 +159,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
}
- if (k >= FF_ARRAY_ELEMS(rps->used)) {
+ if (k >= FF_ARRAY_ELEMS(used)) {
av_log(avctx, AV_LOG_ERROR,
"Invalid num_delta_pocs: %d\n", k);
return AVERROR_INVALIDDATA;
@@ -167,35 +169,38 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
rps->num_negative_pics = k0;
// sort in increasing order (smallest first)
if (rps->num_delta_pocs != 0) {
- int used, tmp;
+ int u, tmp;
for (i = 1; i < rps->num_delta_pocs; i++) {
delta_poc = rps->delta_poc[i];
- used = rps->used[i];
+ u = used[i];
for (k = i - 1; k >= 0; k--) {
tmp = rps->delta_poc[k];
if (delta_poc < tmp) {
rps->delta_poc[k + 1] = tmp;
- rps->used[k + 1] = rps->used[k];
+ used[k + 1] = used[k];
rps->delta_poc[k] = delta_poc;
- rps->used[k] = used;
+ used[k] = u;
}
}
}
}
if ((rps->num_negative_pics >> 1) != 0) {
- int used;
+ int u;
k = rps->num_negative_pics - 1;
// flip the negative values to largest first
for (i = 0; i < rps->num_negative_pics >> 1; i++) {
delta_poc = rps->delta_poc[i];
- used = rps->used[i];
+ u = used[i];
rps->delta_poc[i] = rps->delta_poc[k];
- rps->used[i] = rps->used[k];
+ used[i] = used[k];
rps->delta_poc[k] = delta_poc;
- rps->used[k] = used;
+ used[k] = u;
k--;
}
}
+
+ for (unsigned i = 0; i < FF_ARRAY_ELEMS(used); i++)
+ rps->used |= used[i] * (1 << i);
} else {
unsigned int nb_positive_pics;
@@ -222,7 +227,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
prev -= delta_poc;
rps->delta_poc[i] = prev;
- rps->used[i] = get_bits1(gb);
+ rps->used |= get_bits1(gb) * (1 << i);
}
prev = 0;
for (i = 0; i < nb_positive_pics; i++) {
@@ -235,7 +240,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
prev += delta_poc;
rps->delta_poc[rps->num_negative_pics + i] = prev;
- rps->used[rps->num_negative_pics + i] = get_bits1(gb);
+ rps->used |= get_bits1(gb) * (1 << (rps->num_negative_pics + i));
}
}
}
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index 6ef29a8ea7..92b85115f7 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -79,7 +79,7 @@ typedef struct ShortTermRPS {
int num_delta_pocs;
int rps_idx_num_delta_pocs;
int32_t delta_poc[32];
- uint8_t used[32];
+ uint32_t used;
} ShortTermRPS;
typedef struct HEVCWindow {
diff --git a/libavcodec/hevc_refs.c b/libavcodec/hevc_refs.c
index aed649933d..19f3fa81da 100644
--- a/libavcodec/hevc_refs.c
+++ b/libavcodec/hevc_refs.c
@@ -501,7 +501,7 @@ int ff_hevc_frame_rps(HEVCContext *s)
int poc = s->poc + short_rps->delta_poc[i];
int list;
- if (!short_rps->used[i])
+ if (!(short_rps->used & (1 << i)))
list = ST_FOLL;
else if (i < short_rps->num_negative_pics)
list = ST_CURR_BEF;
@@ -540,9 +540,9 @@ int ff_hevc_frame_nb_refs(const HEVCContext *s)
if (rps) {
for (i = 0; i < rps->num_negative_pics; i++)
- ret += !!rps->used[i];
+ ret += !!(rps->used & (1 << i));
for (; i < rps->num_delta_pocs; i++)
- ret += !!rps->used[i];
+ ret += !!(rps->used & (1 << i));
}
if (long_rps) {
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index 5d7c6b1b64..c2b65fc201 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -374,17 +374,17 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
/* NOTE: This is the predicted, and *reordered* version.
* Probably incorrect, but the spec doesn't say which version to use. */
for (int j = 0; j < sps->st_rps[i].num_delta_pocs; j++)
- str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
+ str[i].used_by_curr_pic_flag |= st_rps->used;
for (int j = 0; j < str[i].num_negative_pics; j++) {
- str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
- str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
+ str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
+ str[i].used_by_curr_pic_s0_flag |= st_rps->used & ((1 << str[i].num_negative_pics) - 1);
}
for (int j = 0; j < str[i].num_positive_pics; j++) {
str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[st_rps->num_negative_pics + j] -
(j ? st_rps->delta_poc[st_rps->num_negative_pics + j - 1] : 0) - 1;
- str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[str[i].num_negative_pics + j] << j;
+ str[i].used_by_curr_pic_s0_flag |= st_rps->used >> str[i].num_negative_pics;
}
}
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
` (7 preceding siblings ...)
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used Anton Khirnov
@ 2024-04-10 13:31 ` Anton Khirnov
2024-04-11 12:35 ` James Almer
8 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-04-10 13:31 UTC (permalink / raw)
To: ffmpeg-devel
Do not use larger fields than needed, use size-1 bitfields for flags.
Reduces sizeof(HEVCSPS) by 1280 bytes.
---
libavcodec/hevc_ps.c | 6 +++---
libavcodec/hevc_ps.h | 19 +++++++++++--------
libavcodec/vulkan_hevc.c | 2 +-
3 files changed, 15 insertions(+), 12 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index 76fe507e7b..7b486ce0af 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -143,11 +143,11 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
for (i = 0; i <= rps_ridx->num_delta_pocs; i++) {
used[k] = get_bits1(gb);
- rps->use_delta_flag = 0;
+ rps->use_delta = 0;
if (!used[k])
- rps->use_delta_flag = get_bits1(gb);
+ rps->use_delta = get_bits1(gb);
- if (used[k] || rps->use_delta_flag) {
+ if (used[k] || rps->use_delta) {
if (i < rps_ridx->num_delta_pocs)
delta_poc = delta_rps + rps_ridx->delta_poc[i];
else
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index 92b85115f7..a8d07aa1b2 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -70,16 +70,19 @@ typedef struct HEVCHdrParams {
} HEVCHdrParams;
typedef struct ShortTermRPS {
- uint8_t rps_predict;
- unsigned int delta_idx;
- uint8_t use_delta_flag;
- uint8_t delta_rps_sign;
- unsigned int abs_delta_rps;
- unsigned int num_negative_pics;
- int num_delta_pocs;
- int rps_idx_num_delta_pocs;
int32_t delta_poc[32];
uint32_t used;
+
+ uint8_t delta_idx;
+ uint8_t num_negative_pics;
+ uint8_t num_delta_pocs;
+ uint8_t rps_idx_num_delta_pocs;
+
+ uint16_t abs_delta_rps;
+ unsigned delta_rps_sign:1;
+
+ unsigned rps_predict:1;
+ unsigned use_delta:1;
} ShortTermRPS;
typedef struct HEVCWindow {
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index c2b65fc201..b109475194 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -359,7 +359,7 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
.delta_rps_sign = sps->st_rps[i].delta_rps_sign,
},
.delta_idx_minus1 = sps->st_rps[i].delta_idx - 1,
- .use_delta_flag = sps->st_rps[i].use_delta_flag,
+ .use_delta_flag = sps->st_rps[i].use_delta,
.abs_delta_rps_minus1 = sps->st_rps[i].abs_delta_rps - 1,
.used_by_curr_pic_flag = 0x0,
.used_by_curr_pic_s0_flag = 0x0,
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used Anton Khirnov
@ 2024-04-10 13:42 ` James Almer
2024-05-24 9:11 ` Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: James Almer @ 2024-04-10 13:42 UTC (permalink / raw)
To: ffmpeg-devel
On 4/10/2024 10:31 AM, Anton Khirnov wrote:
> It is currently an array of 32 uint8_t, each storing a single flag. A
> single uint32_t is sufficient.
>
> Reduces sizeof(HEVCSPS) by 1792 bytes.
> ---
> libavcodec/hevc_ps.c | 33 +++++++++++++++++++--------------
> libavcodec/hevc_ps.h | 2 +-
> libavcodec/hevc_refs.c | 6 +++---
> libavcodec/vulkan_hevc.c | 8 ++++----
> 4 files changed, 27 insertions(+), 22 deletions(-)
>
> diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
> index a6b0021bc3..76fe507e7b 100644
> --- a/libavcodec/hevc_ps.c
> +++ b/libavcodec/hevc_ps.c
> @@ -107,6 +107,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> int k = 0;
> int i;
>
> + rps->used = 0;
> rps->rps_predict = 0;
>
> if (rps != sps->st_rps && sps->nb_st_rps)
> @@ -114,6 +115,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
>
> if (rps->rps_predict) {
> const ShortTermRPS *rps_ridx;
> + uint8_t used[32] = { 0 };
> int delta_rps;
>
> if (is_slice_header) {
> @@ -139,13 +141,13 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> }
> delta_rps = (1 - (rps->delta_rps_sign << 1)) * rps->abs_delta_rps;
> for (i = 0; i <= rps_ridx->num_delta_pocs; i++) {
> - int used = rps->used[k] = get_bits1(gb);
> + used[k] = get_bits1(gb);
>
> rps->use_delta_flag = 0;
> - if (!used)
> + if (!used[k])
> rps->use_delta_flag = get_bits1(gb);
>
> - if (used || rps->use_delta_flag) {
> + if (used[k] || rps->use_delta_flag) {
> if (i < rps_ridx->num_delta_pocs)
> delta_poc = delta_rps + rps_ridx->delta_poc[i];
> else
> @@ -157,7 +159,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> }
> }
>
> - if (k >= FF_ARRAY_ELEMS(rps->used)) {
> + if (k >= FF_ARRAY_ELEMS(used)) {
> av_log(avctx, AV_LOG_ERROR,
> "Invalid num_delta_pocs: %d\n", k);
> return AVERROR_INVALIDDATA;
> @@ -167,35 +169,38 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> rps->num_negative_pics = k0;
> // sort in increasing order (smallest first)
> if (rps->num_delta_pocs != 0) {
> - int used, tmp;
> + int u, tmp;
> for (i = 1; i < rps->num_delta_pocs; i++) {
> delta_poc = rps->delta_poc[i];
> - used = rps->used[i];
> + u = used[i];
> for (k = i - 1; k >= 0; k--) {
> tmp = rps->delta_poc[k];
> if (delta_poc < tmp) {
> rps->delta_poc[k + 1] = tmp;
> - rps->used[k + 1] = rps->used[k];
> + used[k + 1] = used[k];
> rps->delta_poc[k] = delta_poc;
> - rps->used[k] = used;
> + used[k] = u;
> }
> }
> }
> }
> if ((rps->num_negative_pics >> 1) != 0) {
> - int used;
> + int u;
> k = rps->num_negative_pics - 1;
> // flip the negative values to largest first
> for (i = 0; i < rps->num_negative_pics >> 1; i++) {
> delta_poc = rps->delta_poc[i];
> - used = rps->used[i];
> + u = used[i];
> rps->delta_poc[i] = rps->delta_poc[k];
> - rps->used[i] = rps->used[k];
> + used[i] = used[k];
> rps->delta_poc[k] = delta_poc;
> - rps->used[k] = used;
> + used[k] = u;
> k--;
> }
> }
> +
> + for (unsigned i = 0; i < FF_ARRAY_ELEMS(used); i++)
> + rps->used |= used[i] * (1 << i);
> } else {
> unsigned int nb_positive_pics;
>
> @@ -222,7 +227,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> }
> prev -= delta_poc;
> rps->delta_poc[i] = prev;
> - rps->used[i] = get_bits1(gb);
> + rps->used |= get_bits1(gb) * (1 << i);
> }
> prev = 0;
> for (i = 0; i < nb_positive_pics; i++) {
> @@ -235,7 +240,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> }
> prev += delta_poc;
> rps->delta_poc[rps->num_negative_pics + i] = prev;
> - rps->used[rps->num_negative_pics + i] = get_bits1(gb);
> + rps->used |= get_bits1(gb) * (1 << (rps->num_negative_pics + i));
> }
> }
> }
> diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
> index 6ef29a8ea7..92b85115f7 100644
> --- a/libavcodec/hevc_ps.h
> +++ b/libavcodec/hevc_ps.h
> @@ -79,7 +79,7 @@ typedef struct ShortTermRPS {
> int num_delta_pocs;
> int rps_idx_num_delta_pocs;
> int32_t delta_poc[32];
> - uint8_t used[32];
> + uint32_t used;
> } ShortTermRPS;
>
> typedef struct HEVCWindow {
> diff --git a/libavcodec/hevc_refs.c b/libavcodec/hevc_refs.c
> index aed649933d..19f3fa81da 100644
> --- a/libavcodec/hevc_refs.c
> +++ b/libavcodec/hevc_refs.c
> @@ -501,7 +501,7 @@ int ff_hevc_frame_rps(HEVCContext *s)
> int poc = s->poc + short_rps->delta_poc[i];
> int list;
>
> - if (!short_rps->used[i])
> + if (!(short_rps->used & (1 << i)))
> list = ST_FOLL;
> else if (i < short_rps->num_negative_pics)
> list = ST_CURR_BEF;
> @@ -540,9 +540,9 @@ int ff_hevc_frame_nb_refs(const HEVCContext *s)
>
> if (rps) {
> for (i = 0; i < rps->num_negative_pics; i++)
> - ret += !!rps->used[i];
> + ret += !!(rps->used & (1 << i));
> for (; i < rps->num_delta_pocs; i++)
> - ret += !!rps->used[i];
> + ret += !!(rps->used & (1 << i));
> }
>
> if (long_rps) {
> diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
> index 5d7c6b1b64..c2b65fc201 100644
> --- a/libavcodec/vulkan_hevc.c
> +++ b/libavcodec/vulkan_hevc.c
> @@ -374,17 +374,17 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
> /* NOTE: This is the predicted, and *reordered* version.
> * Probably incorrect, but the spec doesn't say which version to use. */
> for (int j = 0; j < sps->st_rps[i].num_delta_pocs; j++)
> - str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
> + str[i].used_by_curr_pic_flag |= st_rps->used;
>
> for (int j = 0; j < str[i].num_negative_pics; j++) {
> - str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
> - str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
> + str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
> + str[i].used_by_curr_pic_s0_flag |= st_rps->used & ((1 << str[i].num_negative_pics) - 1);
av_mod_uintp2(st_rps->used, str[i].num_negative_pics).
> }
>
> for (int j = 0; j < str[i].num_positive_pics; j++) {
> str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[st_rps->num_negative_pics + j] -
> (j ? st_rps->delta_poc[st_rps->num_negative_pics + j - 1] : 0) - 1;
> - str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[str[i].num_negative_pics + j] << j;
> + str[i].used_by_curr_pic_s0_flag |= st_rps->used >> str[i].num_negative_pics;
> }
> }
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields Anton Khirnov
@ 2024-04-11 11:55 ` Andreas Rheinhardt
2024-05-24 9:07 ` Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: Andreas Rheinhardt @ 2024-04-11 11:55 UTC (permalink / raw)
To: ffmpeg-devel
Anton Khirnov:
> Reduces sizeof(HEVCSPS) by 96 bytes.
>
> Also improve flag names: drop redundant suffixes and prefixes, and
> consistently use disabled/enabled.
> ---
> libavcodec/dxva2_hevc.c | 24 ++++-----
> libavcodec/hevc_cabac.c | 36 ++++++-------
> libavcodec/hevc_filter.c | 8 +--
> libavcodec/hevc_parser.c | 2 +-
> libavcodec/hevc_ps.c | 95 +++++++++++++++++-----------------
> libavcodec/hevc_ps.h | 62 +++++++++++-----------
> libavcodec/hevcdec.c | 10 ++--
> libavcodec/hevcpred_template.c | 4 +-
> libavcodec/mips/hevcpred_msa.c | 6 +--
> libavcodec/nvdec_hevc.c | 42 +++++++--------
> libavcodec/qsvenc_hevc.c | 2 +-
> libavcodec/vaapi_hevc.c | 42 +++++++--------
> libavcodec/vdpau_hevc.c | 36 ++++++-------
> libavcodec/vulkan_hevc.c | 56 ++++++++++----------
> 14 files changed, 212 insertions(+), 213 deletions(-)
>
Making this a bitfield is not worth it. It will necessitate masking
operations on every access to these fields which may increase the size
of .text by more than 96B; it may even be that this more than offsets
the savings of 96B from using a bitfield.
- Andreas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS Anton Khirnov
@ 2024-04-11 12:35 ` James Almer
2024-05-24 9:19 ` Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: James Almer @ 2024-04-11 12:35 UTC (permalink / raw)
To: ffmpeg-devel
On 4/10/2024 10:31 AM, Anton Khirnov wrote:
> Do not use larger fields than needed, use size-1 bitfields for flags.
>
> Reduces sizeof(HEVCSPS) by 1280 bytes.
> ---
> libavcodec/hevc_ps.c | 6 +++---
> libavcodec/hevc_ps.h | 19 +++++++++++--------
> libavcodec/vulkan_hevc.c | 2 +-
> 3 files changed, 15 insertions(+), 12 deletions(-)
>
> diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
> index 76fe507e7b..7b486ce0af 100644
> --- a/libavcodec/hevc_ps.c
> +++ b/libavcodec/hevc_ps.c
> @@ -143,11 +143,11 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
> for (i = 0; i <= rps_ridx->num_delta_pocs; i++) {
> used[k] = get_bits1(gb);
>
> - rps->use_delta_flag = 0;
> + rps->use_delta = 0;
> if (!used[k])
> - rps->use_delta_flag = get_bits1(gb);
> + rps->use_delta = get_bits1(gb);
>
> - if (used[k] || rps->use_delta_flag) {
> + if (used[k] || rps->use_delta) {
> if (i < rps_ridx->num_delta_pocs)
> delta_poc = delta_rps + rps_ridx->delta_poc[i];
> else
> diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
> index 92b85115f7..a8d07aa1b2 100644
> --- a/libavcodec/hevc_ps.h
> +++ b/libavcodec/hevc_ps.h
> @@ -70,16 +70,19 @@ typedef struct HEVCHdrParams {
> } HEVCHdrParams;
>
> typedef struct ShortTermRPS {
> - uint8_t rps_predict;
> - unsigned int delta_idx;
> - uint8_t use_delta_flag;
> - uint8_t delta_rps_sign;
> - unsigned int abs_delta_rps;
> - unsigned int num_negative_pics;
> - int num_delta_pocs;
> - int rps_idx_num_delta_pocs;
> int32_t delta_poc[32];
> uint32_t used;
> +
> + uint8_t delta_idx;
> + uint8_t num_negative_pics;
> + uint8_t num_delta_pocs;
> + uint8_t rps_idx_num_delta_pocs;
> +
> + uint16_t abs_delta_rps;
> + unsigned delta_rps_sign:1;
Wont the compiler add two bytes of padding if you put this here?
> +
> + unsigned rps_predict:1;
> + unsigned use_delta:1;
> } ShortTermRPS;
>
> typedef struct HEVCWindow {
> diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
> index c2b65fc201..b109475194 100644
> --- a/libavcodec/vulkan_hevc.c
> +++ b/libavcodec/vulkan_hevc.c
> @@ -359,7 +359,7 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
> .delta_rps_sign = sps->st_rps[i].delta_rps_sign,
> },
> .delta_idx_minus1 = sps->st_rps[i].delta_idx - 1,
> - .use_delta_flag = sps->st_rps[i].use_delta_flag,
> + .use_delta_flag = sps->st_rps[i].use_delta,
> .abs_delta_rps_minus1 = sps->st_rps[i].abs_delta_rps - 1,
> .used_by_curr_pic_flag = 0x0,
> .used_by_curr_pic_s0_flag = 0x0,
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers Anton Khirnov
@ 2024-04-17 9:29 ` Andreas Rheinhardt
2024-05-24 9:03 ` Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: Andreas Rheinhardt @ 2024-04-17 9:29 UTC (permalink / raw)
To: ffmpeg-devel
Anton Khirnov:
> It is more efficient and easier to manage.
> ---
Allocating structures used by slice contexts jointly has the potential
downside of false sharing if the structures are not sufficiently
aligned/padded.
- Andreas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers
2024-04-17 9:29 ` Andreas Rheinhardt
@ 2024-05-24 9:03 ` Anton Khirnov
2024-05-27 13:10 ` Andreas Rheinhardt
0 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-05-24 9:03 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Quoting Andreas Rheinhardt (2024-04-17 11:29:18)
> Anton Khirnov:
> > It is more efficient and easier to manage.
> > ---
>
> Allocating structures used by slice contexts jointly has the potential
> downside of false sharing if the structures are not sufficiently
> aligned/padded.
What do you suggest? Align first member to cacheline size?
--
Anton Khirnov
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields
2024-04-11 11:55 ` Andreas Rheinhardt
@ 2024-05-24 9:07 ` Anton Khirnov
0 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-05-24 9:07 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Quoting Andreas Rheinhardt (2024-04-11 13:55:05)
> Anton Khirnov:
> > Reduces sizeof(HEVCSPS) by 96 bytes.
> >
> > Also improve flag names: drop redundant suffixes and prefixes, and
> > consistently use disabled/enabled.
> > ---
> > libavcodec/dxva2_hevc.c | 24 ++++-----
> > libavcodec/hevc_cabac.c | 36 ++++++-------
> > libavcodec/hevc_filter.c | 8 +--
> > libavcodec/hevc_parser.c | 2 +-
> > libavcodec/hevc_ps.c | 95 +++++++++++++++++-----------------
> > libavcodec/hevc_ps.h | 62 +++++++++++-----------
> > libavcodec/hevcdec.c | 10 ++--
> > libavcodec/hevcpred_template.c | 4 +-
> > libavcodec/mips/hevcpred_msa.c | 6 +--
> > libavcodec/nvdec_hevc.c | 42 +++++++--------
> > libavcodec/qsvenc_hevc.c | 2 +-
> > libavcodec/vaapi_hevc.c | 42 +++++++--------
> > libavcodec/vdpau_hevc.c | 36 ++++++-------
> > libavcodec/vulkan_hevc.c | 56 ++++++++++----------
> > 14 files changed, 212 insertions(+), 213 deletions(-)
> >
>
> Making this a bitfield is not worth it. It will necessitate masking
> operations on every access to these fields which may increase the size
> of .text by more than 96B; it may even be that this more than offsets
> the savings of 96B from using a bitfield.
Okay, then changed to uint8_t locally.
--
Anton Khirnov
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used
2024-04-10 13:42 ` James Almer
@ 2024-05-24 9:11 ` Anton Khirnov
2024-05-24 11:54 ` James Almer
0 siblings, 1 reply; 22+ messages in thread
From: Anton Khirnov @ 2024-05-24 9:11 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Quoting James Almer (2024-04-10 15:42:51)
>
>
> On 4/10/2024 10:31 AM, Anton Khirnov wrote:
> > diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
> > index 5d7c6b1b64..c2b65fc201 100644
> > --- a/libavcodec/vulkan_hevc.c
> > +++ b/libavcodec/vulkan_hevc.c
> > @@ -374,17 +374,17 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
> > /* NOTE: This is the predicted, and *reordered* version.
> > * Probably incorrect, but the spec doesn't say which version to use. */
> > for (int j = 0; j < sps->st_rps[i].num_delta_pocs; j++)
> > - str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
> > + str[i].used_by_curr_pic_flag |= st_rps->used;
> >
> > for (int j = 0; j < str[i].num_negative_pics; j++) {
> > - str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
> > - str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
> > + str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
> > + str[i].used_by_curr_pic_s0_flag |= st_rps->used & ((1 << str[i].num_negative_pics) - 1);
>
> av_mod_uintp2(st_rps->used, str[i].num_negative_pics).
unrelated change
--
Anton Khirnov
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS
2024-04-11 12:35 ` James Almer
@ 2024-05-24 9:19 ` Anton Khirnov
0 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-05-24 9:19 UTC (permalink / raw)
To: FFmpeg development discussions and patches
Quoting James Almer (2024-04-11 14:35:37)
> > diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
> > index 92b85115f7..a8d07aa1b2 100644
> > --- a/libavcodec/hevc_ps.h
> > +++ b/libavcodec/hevc_ps.h
> > @@ -70,16 +70,19 @@ typedef struct HEVCHdrParams {
> > } HEVCHdrParams;
> >
> > typedef struct ShortTermRPS {
> > - uint8_t rps_predict;
> > - unsigned int delta_idx;
> > - uint8_t use_delta_flag;
> > - uint8_t delta_rps_sign;
> > - unsigned int abs_delta_rps;
> > - unsigned int num_negative_pics;
> > - int num_delta_pocs;
> > - int rps_idx_num_delta_pocs;
> > int32_t delta_poc[32];
> > uint32_t used;
> > +
> > + uint8_t delta_idx;
> > + uint8_t num_negative_pics;
> > + uint8_t num_delta_pocs;
> > + uint8_t rps_idx_num_delta_pocs;
> > +
> > + uint16_t abs_delta_rps;
> > + unsigned delta_rps_sign:1;
>
> Wont the compiler add two bytes of padding if you put this here?
No. Pahole says
struct ShortTermRPS {
int32_t delta_poc[32]; /* 0 128 */
/* --- cacheline 2 boundary (128 bytes) --- */
uint32_t used; /* 128 4 */
uint8_t delta_idx; /* 132 1 */
uint8_t num_negative_pics; /* 133 1 */
uint8_t num_delta_pocs; /* 134 1 */
uint8_t rps_idx_num_delta_pocs; /* 135 1 */
uint16_t abs_delta_rps; /* 136 2 */
/* Bitfield combined with previous fields */
unsigned int delta_rps_sign:1; /* 136:16 4 */
unsigned int rps_predict:1; /* 136:17 4 */
unsigned int use_delta:1; /* 136:18 4 */
/* size: 140, cachelines: 3, members: 10 */
/* bit_padding: 13 bits */
/* last cacheline: 12 bytes */
};
--
Anton Khirnov
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used
2024-05-24 9:11 ` Anton Khirnov
@ 2024-05-24 11:54 ` James Almer
2024-05-29 8:05 ` [FFmpeg-devel] [PATCH v2 09-10] " Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: James Almer @ 2024-05-24 11:54 UTC (permalink / raw)
To: ffmpeg-devel
On 5/24/2024 6:11 AM, Anton Khirnov wrote:
> Quoting James Almer (2024-04-10 15:42:51)
>>
>>
>> On 4/10/2024 10:31 AM, Anton Khirnov wrote:
>>> diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
>>> index 5d7c6b1b64..c2b65fc201 100644
>>> --- a/libavcodec/vulkan_hevc.c
>>> +++ b/libavcodec/vulkan_hevc.c
>>> @@ -374,17 +374,17 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
>>> /* NOTE: This is the predicted, and *reordered* version.
>>> * Probably incorrect, but the spec doesn't say which version to use. */
>>> for (int j = 0; j < sps->st_rps[i].num_delta_pocs; j++)
>>> - str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
>>> + str[i].used_by_curr_pic_flag |= st_rps->used;
>>>
>>> for (int j = 0; j < str[i].num_negative_pics; j++) {
>>> - str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
>>> - str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
>>> + str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
>>> + str[i].used_by_curr_pic_s0_flag |= st_rps->used & ((1 << str[i].num_negative_pics) - 1);
>>
>> av_mod_uintp2(st_rps->used, str[i].num_negative_pics).
>
> unrelated change
It's a nit, but not unrelated. You're adding a "i & ((1 << j) - 1)"
line, which can be simplified (and optimized) by using av_mod_uintp2()
instead.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers
2024-05-24 9:03 ` Anton Khirnov
@ 2024-05-27 13:10 ` Andreas Rheinhardt
2024-05-28 13:54 ` [FFmpeg-devel] [PATCH v2 " Anton Khirnov
0 siblings, 1 reply; 22+ messages in thread
From: Andreas Rheinhardt @ 2024-05-27 13:10 UTC (permalink / raw)
To: ffmpeg-devel
Anton Khirnov:
> Quoting Andreas Rheinhardt (2024-04-17 11:29:18)
>> Anton Khirnov:
>>> It is more efficient and easier to manage.
>>> ---
>>
>> Allocating structures used by slice contexts jointly has the potential
>> downside of false sharing if the structures are not sufficiently
>> aligned/padded.
>
> What do you suggest? Align first member to cacheline size?
>
This is problematic, because av_malloc is not necessarily aligned to the
chacheline size; for the same reason it is not possible to simply
DECLARE_ALIGNED_64 for it (see e.g.
7945d30e91b96d2f4f5b612048169087d214d41e). Given that the structure we
are talking about is already pretty big, the easiest way is to add
explicit padding at the end. Use __GCC_DESTRUCTIVE_SIZE if that is
defined or 64 if not (or maybe always use 128B?
https://stackoverflow.com/questions/72126606/should-the-cache-padding-size-of-x86-64-be-128-bytes).
- Andreas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH v2 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers
2024-05-27 13:10 ` Andreas Rheinhardt
@ 2024-05-28 13:54 ` Anton Khirnov
0 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-05-28 13:54 UTC (permalink / raw)
To: ffmpeg-devel
It is more efficient and easier to manage.
---
libavcodec/hevcdec.c | 57 +++++++++++++++++++++++++-------------------
libavcodec/hevcdec.h | 6 ++++-
2 files changed, 37 insertions(+), 26 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index e84f45e3f8..88a481c043 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -2598,7 +2598,7 @@ static int hls_slice_data(HEVCContext *s)
static int hls_decode_entry_wpp(AVCodecContext *avctxt, void *hevc_lclist,
int job, int self_id)
{
- HEVCLocalContext *lc = ((HEVCLocalContext**)hevc_lclist)[self_id];
+ HEVCLocalContext *lc = &((HEVCLocalContext*)hevc_lclist)[self_id];
const HEVCContext *const s = lc->parent;
int ctb_size = 1 << s->ps.sps->log2_ctb_size;
int more_data = 1;
@@ -2682,7 +2682,7 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
{
const uint8_t *data = nal->data;
int length = nal->size;
- HEVCLocalContext *lc = s->HEVClc;
+ HEVCLocalContext *lc;
int *ret;
int64_t offset;
int64_t startheader, cmpt = 0;
@@ -2696,19 +2696,31 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
return AVERROR_INVALIDDATA;
}
- for (i = 1; i < s->threads_number; i++) {
- if (i < s->nb_local_ctx)
- continue;
- s->local_ctx[i] = av_mallocz(sizeof(HEVCLocalContext));
- if (!s->local_ctx[i])
- return AVERROR(ENOMEM);
- s->nb_local_ctx++;
+ if (s->threads_number > s->nb_local_ctx) {
+ HEVCLocalContext *tmp = av_malloc_array(s->threads_number, sizeof(*s->local_ctx));
- s->local_ctx[i]->logctx = s->avctx;
- s->local_ctx[i]->parent = s;
- s->local_ctx[i]->common_cabac_state = &s->cabac;
+ if (!tmp)
+ return AVERROR(ENOMEM);
+
+ memcpy(tmp, s->local_ctx, sizeof(*s->local_ctx) * s->nb_local_ctx);
+ av_free(s->local_ctx);
+ s->local_ctx = tmp;
+ s->HEVClc = &s->local_ctx[0];
+
+ for (unsigned i = s->nb_local_ctx; i < s->threads_number; i++) {
+ tmp = &s->local_ctx[i];
+
+ memset(tmp, 0, sizeof(*tmp));
+
+ tmp->logctx = s->avctx;
+ tmp->parent = s;
+ tmp->common_cabac_state = &s->cabac;
+ }
+
+ s->nb_local_ctx = s->threads_number;
}
+ lc = &s->local_ctx[0];
offset = (lc->gb.index >> 3);
for (j = 0, cmpt = 0, startheader = offset + s->sh.entry_point_offset[0]; j < nal->skipped_bytes; j++) {
@@ -2744,8 +2756,8 @@ static int hls_slice_data_wpp(HEVCContext *s, const H2645NAL *nal)
s->data = data;
for (i = 1; i < s->threads_number; i++) {
- s->local_ctx[i]->first_qp_group = 1;
- s->local_ctx[i]->qp_y = s->HEVClc->qp_y;
+ s->local_ctx[i].first_qp_group = 1;
+ s->local_ctx[i].qp_y = s->HEVClc->qp_y;
}
atomic_store(&s->wpp_err, 0);
@@ -3474,12 +3486,6 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
av_freep(&s->sh.offset);
av_freep(&s->sh.size);
- if (s->local_ctx) {
- for (i = 1; i < s->nb_local_ctx; i++) {
- av_freep(&s->local_ctx[i]);
- }
- }
- av_freep(&s->HEVClc);
av_freep(&s->local_ctx);
ff_h2645_packet_uninit(&s->pkt);
@@ -3496,15 +3502,16 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
s->avctx = avctx;
- s->HEVClc = av_mallocz(sizeof(HEVCLocalContext));
- s->local_ctx = av_mallocz(sizeof(HEVCLocalContext*) * s->threads_number);
- if (!s->HEVClc || !s->local_ctx)
+ s->local_ctx = av_mallocz(sizeof(*s->local_ctx));
+ if (!s->local_ctx)
return AVERROR(ENOMEM);
+ s->nb_local_ctx = 1;
+
+ s->HEVClc = &s->local_ctx[0];
+
s->HEVClc->parent = s;
s->HEVClc->logctx = avctx;
s->HEVClc->common_cabac_state = &s->cabac;
- s->local_ctx[0] = s->HEVClc;
- s->nb_local_ctx = 1;
s->output_frame = av_frame_alloc();
if (!s->output_frame)
diff --git a/libavcodec/hevcdec.h b/libavcodec/hevcdec.h
index ca68fb54a7..5aa3d40450 100644
--- a/libavcodec/hevcdec.h
+++ b/libavcodec/hevcdec.h
@@ -439,13 +439,17 @@ typedef struct HEVCLocalContext {
/* properties of the boundary of the current CTB for the purposes
* of the deblocking filter */
int boundary_flags;
+
+ // an array of these structs is used for per-thread state - pad its size
+ // to avoid false sharing
+ char padding[128];
} HEVCLocalContext;
typedef struct HEVCContext {
const AVClass *c; // needed by private avoptions
AVCodecContext *avctx;
- HEVCLocalContext **local_ctx;
+ HEVCLocalContext *local_ctx;
unsigned nb_local_ctx;
HEVCLocalContext *HEVClc;
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
* [FFmpeg-devel] [PATCH v2 09-10] lavc/hevc_ps: reduce the size of ShortTermRPS.used
2024-05-24 11:54 ` James Almer
@ 2024-05-29 8:05 ` Anton Khirnov
0 siblings, 0 replies; 22+ messages in thread
From: Anton Khirnov @ 2024-05-29 8:05 UTC (permalink / raw)
To: ffmpeg-devel
It is currently an array of 32 uint8_t, each storing a single flag. A
single uint32_t is sufficient.
Reduces sizeof(HEVCSPS) by 1792 bytes.
---
libavcodec/hevc_ps.c | 33 +++++++++++++++++++--------------
libavcodec/hevc_ps.h | 2 +-
libavcodec/hevc_refs.c | 6 +++---
libavcodec/vulkan_hevc.c | 13 +++++--------
4 files changed, 28 insertions(+), 26 deletions(-)
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index a6b0021bc3..76fe507e7b 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -107,6 +107,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
int k = 0;
int i;
+ rps->used = 0;
rps->rps_predict = 0;
if (rps != sps->st_rps && sps->nb_st_rps)
@@ -114,6 +115,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
if (rps->rps_predict) {
const ShortTermRPS *rps_ridx;
+ uint8_t used[32] = { 0 };
int delta_rps;
if (is_slice_header) {
@@ -139,13 +141,13 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
delta_rps = (1 - (rps->delta_rps_sign << 1)) * rps->abs_delta_rps;
for (i = 0; i <= rps_ridx->num_delta_pocs; i++) {
- int used = rps->used[k] = get_bits1(gb);
+ used[k] = get_bits1(gb);
rps->use_delta_flag = 0;
- if (!used)
+ if (!used[k])
rps->use_delta_flag = get_bits1(gb);
- if (used || rps->use_delta_flag) {
+ if (used[k] || rps->use_delta_flag) {
if (i < rps_ridx->num_delta_pocs)
delta_poc = delta_rps + rps_ridx->delta_poc[i];
else
@@ -157,7 +159,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
}
- if (k >= FF_ARRAY_ELEMS(rps->used)) {
+ if (k >= FF_ARRAY_ELEMS(used)) {
av_log(avctx, AV_LOG_ERROR,
"Invalid num_delta_pocs: %d\n", k);
return AVERROR_INVALIDDATA;
@@ -167,35 +169,38 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
rps->num_negative_pics = k0;
// sort in increasing order (smallest first)
if (rps->num_delta_pocs != 0) {
- int used, tmp;
+ int u, tmp;
for (i = 1; i < rps->num_delta_pocs; i++) {
delta_poc = rps->delta_poc[i];
- used = rps->used[i];
+ u = used[i];
for (k = i - 1; k >= 0; k--) {
tmp = rps->delta_poc[k];
if (delta_poc < tmp) {
rps->delta_poc[k + 1] = tmp;
- rps->used[k + 1] = rps->used[k];
+ used[k + 1] = used[k];
rps->delta_poc[k] = delta_poc;
- rps->used[k] = used;
+ used[k] = u;
}
}
}
}
if ((rps->num_negative_pics >> 1) != 0) {
- int used;
+ int u;
k = rps->num_negative_pics - 1;
// flip the negative values to largest first
for (i = 0; i < rps->num_negative_pics >> 1; i++) {
delta_poc = rps->delta_poc[i];
- used = rps->used[i];
+ u = used[i];
rps->delta_poc[i] = rps->delta_poc[k];
- rps->used[i] = rps->used[k];
+ used[i] = used[k];
rps->delta_poc[k] = delta_poc;
- rps->used[k] = used;
+ used[k] = u;
k--;
}
}
+
+ for (unsigned i = 0; i < FF_ARRAY_ELEMS(used); i++)
+ rps->used |= used[i] * (1 << i);
} else {
unsigned int nb_positive_pics;
@@ -222,7 +227,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
prev -= delta_poc;
rps->delta_poc[i] = prev;
- rps->used[i] = get_bits1(gb);
+ rps->used |= get_bits1(gb) * (1 << i);
}
prev = 0;
for (i = 0; i < nb_positive_pics; i++) {
@@ -235,7 +240,7 @@ int ff_hevc_decode_short_term_rps(GetBitContext *gb, AVCodecContext *avctx,
}
prev += delta_poc;
rps->delta_poc[rps->num_negative_pics + i] = prev;
- rps->used[rps->num_negative_pics + i] = get_bits1(gb);
+ rps->used |= get_bits1(gb) * (1 << (rps->num_negative_pics + i));
}
}
}
diff --git a/libavcodec/hevc_ps.h b/libavcodec/hevc_ps.h
index 1d3bdca4c6..ed6372c747 100644
--- a/libavcodec/hevc_ps.h
+++ b/libavcodec/hevc_ps.h
@@ -79,7 +79,7 @@ typedef struct ShortTermRPS {
int num_delta_pocs;
int rps_idx_num_delta_pocs;
int32_t delta_poc[32];
- uint8_t used[32];
+ uint32_t used;
} ShortTermRPS;
typedef struct HEVCWindow {
diff --git a/libavcodec/hevc_refs.c b/libavcodec/hevc_refs.c
index 8da9ec982a..d6dc2f9e0a 100644
--- a/libavcodec/hevc_refs.c
+++ b/libavcodec/hevc_refs.c
@@ -497,7 +497,7 @@ int ff_hevc_frame_rps(HEVCContext *s)
int poc = s->poc + short_rps->delta_poc[i];
int list;
- if (!short_rps->used[i])
+ if (!(short_rps->used & (1 << i)))
list = ST_FOLL;
else if (i < short_rps->num_negative_pics)
list = ST_CURR_BEF;
@@ -536,9 +536,9 @@ int ff_hevc_frame_nb_refs(const HEVCContext *s)
if (rps) {
for (i = 0; i < rps->num_negative_pics; i++)
- ret += !!rps->used[i];
+ ret += !!(rps->used & (1 << i));
for (; i < rps->num_delta_pocs; i++)
- ret += !!rps->used[i];
+ ret += !!(rps->used & (1 << i));
}
if (long_rps) {
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index 21cf49c0ec..a35f3d992d 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -373,19 +373,16 @@ static void set_sps(const HEVCSPS *sps, int sps_idx,
/* NOTE: This is the predicted, and *reordered* version.
* Probably incorrect, but the spec doesn't say which version to use. */
- for (int j = 0; j < sps->st_rps[i].num_delta_pocs; j++)
- str[i].used_by_curr_pic_flag |= sps->st_rps[i].used[j] << j;
+ str[i].used_by_curr_pic_flag = st_rps->used;
+ str[i].used_by_curr_pic_s0_flag = av_mod_uintp2(st_rps->used, str[i].num_negative_pics);
+ str[i].used_by_curr_pic_s1_flag = st_rps->used >> str[i].num_negative_pics;
- for (int j = 0; j < str[i].num_negative_pics; j++) {
+ for (int j = 0; j < str[i].num_negative_pics; j++)
str[i].delta_poc_s0_minus1[j] = st_rps->delta_poc[j] - (j ? st_rps->delta_poc[j - 1] : 0) - 1;
- str[i].used_by_curr_pic_s0_flag |= sps->st_rps[i].used[j] << j;
- }
- for (int j = 0; j < str[i].num_positive_pics; j++) {
+ for (int j = 0; j < str[i].num_positive_pics; j++)
str[i].delta_poc_s1_minus1[j] = st_rps->delta_poc[st_rps->num_negative_pics + j] -
(j ? st_rps->delta_poc[st_rps->num_negative_pics + j - 1] : 0) - 1;
- str[i].used_by_curr_pic_s1_flag |= sps->st_rps[i].used[str[i].num_negative_pics + j] << j;
- }
}
*ltr = (StdVideoH265LongTermRefPicsSps) {
--
2.43.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
^ permalink raw reply [flat|nested] 22+ messages in thread
end of thread, other threads:[~2024-05-29 8:06 UTC | newest]
Thread overview: 22+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-04-10 13:31 [FFmpeg-devel] [PATCH 01/10] lavc/hevcdec: rename HEVCContext.HEVClcList to local_ctx Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 02/10] lavc/hevcdec: track local context count separately from WPP thread count Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 03/10] lavc/hevcdec: allocate local_ctx as array of structs rather than pointers Anton Khirnov
2024-04-17 9:29 ` Andreas Rheinhardt
2024-05-24 9:03 ` Anton Khirnov
2024-05-27 13:10 ` Andreas Rheinhardt
2024-05-28 13:54 ` [FFmpeg-devel] [PATCH v2 " Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 04/10] lavc/hevcdec: drop a useless execute() call with 1 job Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 05/10] lavc/hevc_ps: reduce the size of used_by_curr_pic_lt_sps_flag Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 06/10] lavc/hevc_ps/HEVCSPS: change flags into size-1 bitfields Anton Khirnov
2024-04-11 11:55 ` Andreas Rheinhardt
2024-05-24 9:07 ` Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 07/10] lavc/hevc_ps: fix variable signedness in ff_hevc_decode_short_term_rps() Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 08/10] lavc/hevc_ps: do not store delta_poc_s[01] in ShortTermRPS Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 09/10] lavc/hevc_ps: reduce the size of ShortTermRPS.used Anton Khirnov
2024-04-10 13:42 ` James Almer
2024-05-24 9:11 ` Anton Khirnov
2024-05-24 11:54 ` James Almer
2024-05-29 8:05 ` [FFmpeg-devel] [PATCH v2 09-10] " Anton Khirnov
2024-04-10 13:31 ` [FFmpeg-devel] [PATCH 10/10] lavc/hevc_ps: compactify ShortTermRPS Anton Khirnov
2024-04-11 12:35 ` James Almer
2024-05-24 9:19 ` Anton Khirnov
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git