AVC encode summary shows incorrectly formatted avgQP values. (misaligned decimal separation)

[DimkaTsv]

Issue:
Quite simple this time.
For AVC specifically encode summary after completion shows misaligned numbers for it's avgQP measurements. (shifted dot)
HEVC works fine, and AV1 encode does not show avgQP values in summary.

Issue is also specific to VBR encodes. CQP reports avgQP as expected (as CQP, of course).

Command:
%~dp0\VCEEncC64.exe -i %1 -o "%~n1_test%~x1" --avsw --codec avc --vbr 15000 --max-bitrate 45000 --bframes 2 --audio-copy
Observed behaviour:

encoded 500 frames, 517.60 fps, 15780.06 kbps, 18.81 MB
encode time 0:00:00, CPULoad: 0.4%
frame type IDR   1
frame type I     1,  avgQP  24576.00,  total size   0.22 MB
frame type P   167,  avgQP  31641.29,  total size  16.41 MB
frame type B   332,  avgQP  35065.83,  total size   2.18 MB

avgQP values are improperly rounded.
24.58 is presented as 24576.00,
31.64 is presented as 31641.29
35.07 is presented as 35065.83

Expected behaviour:
avgQP should be presented properly.

[rigaya]

As I seem to have no problem here using RX 7900 XT with the same command line.

avgQP is average of eport returned each frame by AMF, regardless of rate control mode (CQP or VBR...), so bit weird why this issue happens.

I'll keep it as-is, as this should not have any impact on encoded result itself.

encoded 4550 frames, 396.93 fps, 15007.00 kbps, 271.60 MB
encode time 0:00:11, CPU: 7.4%, GPU: 4.0%, VE: 99.2%
frame type IDR   16
frame type I     16,  avgQP  11.81,  total size    2.97 MB
frame type P   1517,  avgQP  12.12,  total size  149.31 MB
frame type B   3017,  avgQP  15.03,  total size  119.33 MB

[DimkaTsv]

I'll keep it as-is, as this should not have any impact on encoded result itself.

Ok, it's fine. I understand that it is just statistics presentation problem and low priority issue.

It may be something specific to RDNA4 then.
Weird... Maybe some AMF/driver issue? I guess i will report this to AMF devs.
I wonder how it must be reported, though. I will try figuring something out.

Is there way to check what value AMF returns for each frame? Maybe it just started internally to return 5+ digit values?
But then... It should be an integer anyways, not a float. So this makes digit count to have even less sense.

Then i will keep this open for the time being, and write AMF report with reference.
It's not like issue is critical, so it can wait.