VTM can result in different kbps under different platform

[johnnyyang]

The bitrate mismatch between two different platform (windows and linux) is discovered in the following sequences and QPs on RA configuration:

	Sequence	QP
Class A1	FoodMarket4	22
	Campfire	22
	Campfire	32
Class B	RitualDance	32
	Cactus	27
Class F	SlideShow	27

The difference are negligible so that it find no issue in summary table. However, it might still cause mismatch while companies are doing crosscheck in the following CE and JVET meeting. After checking with several companies (e.g. HiSilicon and ITRI), the issues were confirmed but there are no solutions so far.

[johnnyyang]

encode log of windows platform

[johnnyyang]

encode log of Linux platform

[johnnyyang]

Summary report of different platform

[johnnyyang]

Attached Summary report(VTM-3.0_R_Foxconn_T_HHI.xlsm​)
I had mark the difference in orange in RA page.

[ksuehring]

Issue #129 reported also a mismatch between Windows and Linux. Can you check with current master, if the fix for #129 also fixes the issues that you are seeing?

[johnnyyang]

As Xiang mentioned, issue #129 related to CPR is not enabled in default in VTM.
If our computing power is not tense,
We would like to volunteer to help check the results on top of Windows platform when this issue is fixed.
Please let us know if any.

[XiangLi]

Please check whether merge request !135 (​https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/merge_requests/135) fixes the issue.

[johnnyyang]

RA CTC result for VTM3.0 ver.10a973e

[johnnyyang]

This issue can be solved at VTM3 git version (10a973e)

Please find attached(VTM3.0_10a973e_RA.xlsm)
Reference: RA result based on Linux platform (Thanks Tencent)
Tested: RA result based on Windows platform (Thanks Foxconn)

[XiangLi]

Fixed in ​https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/merge_requests/135