Opened 4 years ago

Closed 4 years ago

#664 closed defect (fixed)

Mutiple slices/tiles is not working

Reported by: Weijia Owned by:
Priority: major Milestone: VTM-7.0
Component: VTM Version: VTM-7.0rc1
Keywords: Cc: ksuehring, XiangLi, fbossen, jvet@…

Description

When using the cfg files under the folder of nonCTC-SliceConfigExamples, the encoder and decoder crash. The tests follows CE2 test condition where IBC=1, PLT=1, BDPCM=2 and HashME=1 are additionally added. The problem affects the CE2 tests.

Attachments (7)

encoder_randomaccess_vtm_RasterScanSlice.cfg (12.6 KB) - added by Weijia 4 years ago.
encoder_randomaccess_vtm_RectangularSlice.cfg (12.7 KB) - added by Weijia 4 years ago.
encoder_randomaccess_vtm_SingleTilePerSlice.cfg (12.7 KB) - added by Weijia 4 years ago.
CE2.cfg (36 bytes) - added by Weijia 4 years ago.
encoder_randomaccess_vtm_RectangularSlice_r1.cfg (13.4 KB) - added by Weijia 4 years ago.
log.txt (2.8 KB) - added by Weijia 4 years ago.
encoder_randomaccess_vtm_RectangularSlice.2.cfg (13.5 KB) - added by XiangLi 4 years ago.
cfg file to duplicate the reported crash on top of VTM-7rc1 for the first I frame

Download all attachments as: .zip

Change history (14)

Changed 4 years ago by Weijia

comment:1 Changed 4 years ago by Weijia

Submitted a merge !1118 request to solve the problem. AI seems OK now, but RA and LDB still have a crash.

Version 1, edited 4 years ago by Weijia (previous) (next) (diff)

comment:2 follow-ups: Changed 4 years ago by ksuehring

I cannot reproduce the problem. Can you specify the exact parameters that you are using? Is this for 4:4:4? BDPCM=2 does not work for 4:2:0, so I changed this to 1. The rectangular slice config didn't work for encoder_randomaccess_vtm_SingleTilePerSlice.cfg, but that seems to be a different problem.

comment:3 in reply to: ↑ 2 Changed 4 years ago by Weijia

Yes. It's for 4:4:4.
Replying to ksuehring:

I cannot reproduce the problem. Can you specify the exact parameters that you are using? Is this for 4:4:4? BDPCM=2 does not work for 4:2:0, so I changed this to 1. The rectangular slice config didn't work for encoder_randomaccess_vtm_SingleTilePerSlice.cfg, but that seems to be a different problem.

Changed 4 years ago by Weijia

comment:4 in reply to: ↑ 2 Changed 4 years ago by Weijia

I used the attached cfg file encoder_randomaccess_vtm_RectangularSlice_r1.cfg and put the results to the attached log.txt. The decoder has crash. You could use that cfg file to reproduce the problem.
Replying to ksuehring:

I cannot reproduce the problem. Can you specify the exact parameters that you are using? Is this for 4:4:4? BDPCM=2 does not work for 4:2:0, so I changed this to 1. The rectangular slice config didn't work for encoder_randomaccess_vtm_SingleTilePerSlice.cfg, but that seems to be a different problem.

comment:5 Changed 4 years ago by XiangLi

  • Version set to VTM-7.0rc1

The issue may be duplicated with VTM-7rc1 with following settings and attached cfg

EncoderApp -c encoder_randomaccess_vtm_RectangularSlice.cfg -i sc_web_browsing_1280x720_30_8bit_300_444_r1.yuv -wdt 1280 -hgt 720 --InputBitDepth=8 -fr 30

The issue is that in multiple tiles, IBC buffer is reset at decoder and the access to left tiles is invalid. However, at encoder no such check is performed when searching BV. It looks https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/merge_requests/1118 fixed this issue for I frames. But it was reported there was still crash for RA with the fix in MR1118. To further duplicate such crash and check.

Changed 4 years ago by XiangLi

cfg file to duplicate the reported crash on top of VTM-7rc1 for the first I frame

comment:6 Changed 4 years ago by XiangLi

The key issue of crashing in inter slices looks due to tile index is not properly set in function xCheckRDCostHashInter() which is used a lot in SCC inter frame encoding. In addition, title index and related information are not properly initialized. Also, tile boundary is not properly considered in deblocking. A new MR is created https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/merge_requests/1128.

comment:7 Changed 4 years ago by ksuehring

  • Milestone set to VTM-7.0
  • Resolution set to fixed
  • Status changed from new to closed

The MR was merged

Note: See TracTickets for help on using tickets.