From ede21fffa0e610e92133b37b701a4bd8b82d5b02 Mon Sep 17 00:00:00 2001
From: jeffqjiangNew <142832361+jeffqjiangNew@users.noreply.github.com>
Date: Tue, 18 Mar 2025 17:19:01 -0400
Subject: [PATCH] HEVC: Added DPB size change handling through decoder
 reconfiguration. (#531)

* * rocDecode/HEVC: Added DPB size change handling through decoder reconfiguration.

* * rocDecode/HEVC: Minor fix.

[ROCm/rocdecode commit: 5988eb16b643bdf34326eb4253529266e629843e]
---
 projects/rocdecode/CHANGELOG.md               |  2 +-
 projects/rocdecode/src/parser/hevc_parser.cpp |  5 ++--
 .../rocdecode/src/parser/roc_video_parser.cpp |  1 +
 .../rocdecode/src/rocdecode/roc_decoder.cpp   |  3 +++
 .../utils/rocvideodecode/roc_video_dec.cpp    | 23 ++++++++++++-------
 .../utils/rocvideodecode/roc_video_dec.h      |  1 +
 6 files changed, 24 insertions(+), 11 deletions(-)

diff --git a/projects/rocdecode/CHANGELOG.md b/projects/rocdecode/CHANGELOG.md
index 07f2e2c343..7d8b64a336 100644
--- a/projects/rocdecode/CHANGELOG.md
+++ b/projects/rocdecode/CHANGELOG.md
@@ -9,7 +9,7 @@ Full documentation for rocDecode is available at [https://rocm.docs.amd.com/proj
 * VP9 IVF container file parsing support in bitstream reader.
 * CTest for VP9 decode on bitstream reader.
 * HEVC stream syntax error handling.
-* HEVC stream bit depth change handling through decoder reconfiguration.
+* HEVC stream bit depth change handling and DPB buffer size change handling through decoder reconfiguration.
 * rocDecode now uses the Cmake CMAKE_PREFIX_PATH directive.
 
 ### Optimized
diff --git a/projects/rocdecode/src/parser/hevc_parser.cpp b/projects/rocdecode/src/parser/hevc_parser.cpp
index 228307ed0a..5a249ed604 100644
--- a/projects/rocdecode/src/parser/hevc_parser.cpp
+++ b/projects/rocdecode/src/parser/hevc_parser.cpp
@@ -1551,7 +1551,7 @@ ParserResult HevcVideoParser::ParseSliceHeader(uint8_t *nalu, size_t size, HevcS
     if (m_active_sps_id_ != pps_ptr->pps_seq_parameter_set_id) {
         m_active_sps_id_ = pps_ptr->pps_seq_parameter_set_id;
         sps_ptr = &sps_list_[m_active_sps_id_];
-        new_seq_activated_ = true;  // Note: clear this flag after the actions are taken.
+        new_seq_activated_ = true;
     }
 
     sps_ptr = &sps_list_[m_active_sps_id_];
@@ -1570,7 +1570,7 @@ ParserResult HevcVideoParser::ParseSliceHeader(uint8_t *nalu, size_t size, HevcS
         pic_width_ = sps_ptr->pic_width_in_luma_samples;
         pic_height_ = sps_ptr->pic_height_in_luma_samples;
         // Take care of the case where a new SPS replaces the old SPS with the same id but with different dimensions
-        new_seq_activated_ = true;  // Note: clear this flag after the actions are taken.
+        new_seq_activated_ = true;
     }
     // Check DPB buffer size change
     uint32_t max_dec_pic_buffering = sps_ptr->sps_max_dec_pic_buffering_minus1[sps_ptr->sps_max_sub_layers_minus1] + 1;
@@ -1578,6 +1578,7 @@ ParserResult HevcVideoParser::ParseSliceHeader(uint8_t *nalu, size_t size, HevcS
         dpb_buffer_.dpb_size = max_dec_pic_buffering;
         dpb_buffer_.dpb_size = dpb_buffer_.dpb_size > HEVC_MAX_DPB_FRAMES ? HEVC_MAX_DPB_FRAMES : dpb_buffer_.dpb_size;
         CheckAndAdjustDecBufPoolSize(dpb_buffer_.dpb_size);
+        new_seq_activated_ = true;
     }
     // Check bit depth change
     if (bit_depth_luma_minus8_ != sps_ptr->bit_depth_luma_minus8 || bit_depth_chroma_minus8_ != sps_ptr->bit_depth_chroma_minus8) {
diff --git a/projects/rocdecode/src/parser/roc_video_parser.cpp b/projects/rocdecode/src/parser/roc_video_parser.cpp
index 695d9e52ee..b3872cb1c1 100644
--- a/projects/rocdecode/src/parser/roc_video_parser.cpp
+++ b/projects/rocdecode/src/parser/roc_video_parser.cpp
@@ -94,6 +94,7 @@ void RocVideoParser::InitDecBufPool() {
 
 void RocVideoParser::CheckAndAdjustDecBufPoolSize(int dpb_size) {
     int min_dec_buf_pool_size = dpb_size + (parser_params_.max_display_delay > DECODE_BUF_POOL_EXTENSION ? parser_params_.max_display_delay : DECODE_BUF_POOL_EXTENSION);
+    // If DPB size decreases, we keep the existing pool and skip reconfiguration.
     if ( dec_buf_pool_size_ < min_dec_buf_pool_size) {
         dec_buf_pool_size_ = min_dec_buf_pool_size;
         decode_buffer_pool_.resize(dec_buf_pool_size_, {0});
diff --git a/projects/rocdecode/src/rocdecode/roc_decoder.cpp b/projects/rocdecode/src/rocdecode/roc_decoder.cpp
index 8e87ff8a80..8e864510f4 100644
--- a/projects/rocdecode/src/rocdecode/roc_decoder.cpp
+++ b/projects/rocdecode/src/rocdecode/roc_decoder.cpp
@@ -93,6 +93,9 @@ rocDecStatus RocDecoder::ReconfigureDecoder(RocdecReconfigureDecoderInfo *reconf
             return rocdec_status;
         }
     }
+    if (hip_interop_.size() != reconfig_params->num_decode_surfaces) {
+        hip_interop_.resize(reconfig_params->num_decode_surfaces);
+    }
     rocdec_status = va_video_decoder_.ReconfigureDecoder(reconfig_params);
     if (rocdec_status != ROCDEC_SUCCESS) {
         ERR("Reconfiguration of the decoder failed.");
diff --git a/projects/rocdecode/utils/rocvideodecode/roc_video_dec.cpp b/projects/rocdecode/utils/rocvideodecode/roc_video_dec.cpp
index e34a98bfa2..323e67291a 100644
--- a/projects/rocdecode/utils/rocvideodecode/roc_video_dec.cpp
+++ b/projects/rocdecode/utils/rocvideodecode/roc_video_dec.cpp
@@ -238,7 +238,6 @@ static void GetSurfaceStrideInternal(rocDecVideoSurfaceFormat surface_format, ui
 *  0: fail, 1: succeeded, > 1: override dpb size of parser (set by RocdecParserParams::max_num_decode_surfaces while creating parser)
 */
 int RocVideoDecoder::HandleVideoSequence(RocdecVideoFormat *p_video_format) {
-
     if (p_video_format == nullptr) {
         ROCDEC_THROW("Rocdec:: Invalid video format in HandleVideoSequence: ", ROCDEC_INVALID_PARAMETER);
         return 0;
@@ -260,8 +259,6 @@ int RocVideoDecoder::HandleVideoSequence(RocdecVideoFormat *p_video_format) {
     ;
     input_video_info_str_ << std::endl;
 
-    int num_decode_surfaces = p_video_format->min_num_decode_surfaces;
-
     RocdecDecodeCaps decode_caps;
     memset(&decode_caps, 0, sizeof(decode_caps));
     decode_caps.codec_type = p_video_format->codec;
@@ -329,6 +326,7 @@ int RocVideoDecoder::HandleVideoSequence(RocdecVideoFormat *p_video_format) {
     disp_rect_.right = p_video_format->display_area.right;
     disp_width_ = p_video_format->display_area.right - p_video_format->display_area.left;
     disp_height_ = p_video_format->display_area.bottom - p_video_format->display_area.top;
+    num_decode_surfaces_ = p_video_format->min_num_decode_surfaces;
 
     // AV1 has max width/height of sequence in sequence header
     if (codec_id_ == rocDecVideoCodec_AV1 && p_video_format->seqhdr_data_length > 0) {
@@ -350,7 +348,7 @@ int RocVideoDecoder::HandleVideoSequence(RocdecVideoFormat *p_video_format) {
     videoDecodeCreateInfo.chroma_format = video_chroma_format_;
     videoDecodeCreateInfo.output_format = video_surface_format_;
     videoDecodeCreateInfo.bit_depth_minus_8 = bitdepth_minus_8_;
-    videoDecodeCreateInfo.num_decode_surfaces = num_decode_surfaces;
+    videoDecodeCreateInfo.num_decode_surfaces = num_decode_surfaces_;
     videoDecodeCreateInfo.width = coded_width_;
     videoDecodeCreateInfo.height = coded_height_;
     videoDecodeCreateInfo.max_width = max_width_;
@@ -419,7 +417,7 @@ int RocVideoDecoder::HandleVideoSequence(RocdecVideoFormat *p_video_format) {
     ROCDEC_API_CALL(rocDecCreateDecoder(&roc_decoder_, &videoDecodeCreateInfo));
     double elapsed_time = StopTimer(start_time);
     AddDecoderSessionOverHead(std::this_thread::get_id(), elapsed_time);
-    return num_decode_surfaces;
+    return num_decode_surfaces_;
 }
 
 /**
@@ -476,13 +474,16 @@ int RocVideoDecoder::ReconfigureDecoder(RocdecVideoFormat *p_video_format) {
         ROCDEC_THROW("Reconfigure Not supported for chroma format change", ROCDEC_NOT_SUPPORTED);
         return 0;
     }
+
     bool is_decode_res_changed = !(p_video_format->coded_width == coded_width_ && p_video_format->coded_height == coded_height_);
     bool is_display_rect_changed = !(p_video_format->display_area.bottom == disp_rect_.bottom &&
                                      p_video_format->display_area.top == disp_rect_.top &&
                                      p_video_format->display_area.left == disp_rect_.left &&
                                      p_video_format->display_area.right == disp_rect_.right);
     bool is_bit_depth_changed = p_video_format->bit_depth_luma_minus8 != bitdepth_minus_8_;
-    if (!is_decode_res_changed && !is_display_rect_changed && !is_bit_depth_changed && !b_force_recofig_flush_) {
+    bool is_dec_surface_num_changed = p_video_format->min_num_decode_surfaces != num_decode_surfaces_;
+
+    if (!is_decode_res_changed && !is_display_rect_changed && !is_bit_depth_changed && !is_dec_surface_num_changed && !b_force_recofig_flush_) {
         return 1;
     }
 
@@ -542,6 +543,7 @@ int RocVideoDecoder::ReconfigureDecoder(RocdecVideoFormat *p_video_format) {
         else if (video_chroma_format_ == rocDecVideoChromaFormat_422)
             video_surface_format_ = bitdepth_minus_8_ ? rocDecVideoSurfaceFormat_YUV422_16Bit : rocDecVideoSurfaceFormat_YUV422;
     }
+    num_decode_surfaces_ = p_video_format->min_num_decode_surfaces;
 
     if (p_video_format->reconfig_options == ROCDEC_RECONFIG_NEW_SURFACES) {
         if (out_mem_type_ == OUT_SURFACE_MEM_DEV_INTERNAL || out_mem_type_ == OUT_SURFACE_MEM_NOT_MAPPED) {
@@ -583,7 +585,7 @@ int RocVideoDecoder::ReconfigureDecoder(RocdecVideoFormat *p_video_format) {
 
     // If the coded_width or coded_height hasn't changed but display resolution has changed, then need to update width and height for
     // correct output with cropping. There is no need to reconfigure the decoder.
-    if (!is_decode_res_changed && is_display_rect_changed && !is_bit_depth_changed) {
+    if (!is_decode_res_changed && is_display_rect_changed && !is_bit_depth_changed && !is_dec_surface_num_changed) {
         return 1;
     }
 
@@ -621,6 +623,9 @@ int RocVideoDecoder::ReconfigureDecoder(RocdecVideoFormat *p_video_format) {
     }
     if (is_bit_depth_changed) {
         input_video_info_str_ << "Input Video Bit Depth Changed:" << std::endl;
+    }
+    if (is_dec_surface_num_changed) {
+        input_video_info_str_ << "Number of decoded surfaces Changed:" << std::endl;
     }
         input_video_info_str_ << "\tCoded size   : [" << p_video_format->coded_width << ", " << p_video_format->coded_height << "]" << std::endl
         << "\tDisplay area : [" << p_video_format->display_area.left << ", " << p_video_format->display_area.top << ", "
@@ -634,7 +639,9 @@ int RocVideoDecoder::ReconfigureDecoder(RocdecVideoFormat *p_video_format) {
     input_video_info_str_ << std::endl;
     std::cout << input_video_info_str_.str();
 
-    is_decoder_reconfigured_ = true;
+    if (is_decode_res_changed || is_bit_depth_changed) {
+        is_decoder_reconfigured_ = true;
+    }
     return 1;
 }
 
diff --git a/projects/rocdecode/utils/rocvideodecode/roc_video_dec.h b/projects/rocdecode/utils/rocvideodecode/roc_video_dec.h
index 3596530ece..b6700d4924 100644
--- a/projects/rocdecode/utils/rocvideodecode/roc_video_dec.h
+++ b/projects/rocdecode/utils/rocvideodecode/roc_video_dec.h
@@ -505,6 +505,7 @@ class RocVideoDecoder {
         uint32_t target_height_ = 0;
         int max_width_ = 0, max_height_ = 0;
         uint32_t chroma_height_ = 0, chroma_width_ = 0;
+        uint32_t num_decode_surfaces_ = 0;
         uint32_t num_chroma_planes_ = 0;
         uint32_t num_components_ = 0;
         uint32_t surface_stride_ = 0;