I have been unable to reproduce the issue. The encode completed for me without error (see below). Are you able provide a log of the encode?

$ ~/project/build/jctvc-a124-amd64/jctvc-a124-enc -c ~/project/jctvc-a124/cfg/cfp/cfg/RaceHorsesC.cfg -ip 32 -f 96 -g 8 -r 4 -rb0 2 -rb1 2 -s 128 -h 5 -q 32 -t 8 -ltd 0 -utd 1 -1 FEN -i /project/hvc/anchors/RaceHorses_832x480_30.yuv

TMuC SW candidate: Encoder Version [0.1][Linux][GCC 4.5.1][64 bit]

...
POC   34 ( B-SLICE, QP 35 )      21632 bits [Y 31.7250 dB    U 36.9214 dB    V 38.1342 dB]  [ET    55 ] [L0 32 30 ] [L1 36 40 ]
POC   38 ( B-SLICE, QP 35 )      18632 bits [Y 31.9794 dB    U 37.3608 dB    V 38.3262 dB]  [ET    34 ] [L0 36 34 ] [L1 40 36 ]
POC   33 ( B-SLICE, QP 36 )       9256 bits [Y 31.4215 dB    U 36.8232 dB    V 38.0069 dB]  [ET    27 ] [L0 32 30 ] [L1 34 36 ]
...

My log is as below

TMuC SW candidate: Encoder Version [0.1][Windows][VS 1600][64 bit]

...
POC 34 ( B-SLICE, QP 35 ) 21840 bits [Y 31.7093 dB U 36.9497 dB V 38.1170 dB] [ET 75 ] [L0 32 30 ] [L1 36 40 ]
POC 38 ( B-SLICE, QP 35 ) 18992 bits [Y 31.9933 dB U 37.5095 dB V 38.4196 dB] [ET 53 ] [L0 36 34 ] [L1 40 36 ]

and the software ends, no more information.

BTW, the encoding results are not exactly the same.

BTW, other QP, such as 22, 27, 37, and other 832x480 sequence also fails.

Please could you provide the full encoder log. Does the encoder crash with an windows exception, if so, what exception?

Please could you check that the md5sum of the source video is:

0a351df99f22d837bc528bd4901c6968 *RaceHorses_832x480_30.yuv

Replying to anonymous:

Encoding 832x480 sequence, using the command line below, after encoding POC 38, the software will fail.

-c ./cfg/RaceHorsesC.cfg -ip 32 -f 96 -g 8 -r 4 -rb0 2 -rb1 2 -s 128 -h 5 -q 32 -t 8 -ltd 0 -utd 1 -1 FEN

I guess that '-t 8' (max allowable transform size) can be the problem source. Original A124 supports this feature (large block size, small transform size) when LOT (logical transform) is turned on, however, during the stripping stage from A124 to TMuC code-base, LOT was removed and now '-t' option is not working for intra mode. But I'm not fully sure since the crash occurs after encoding several slices.

Could you please check it without '-t 8' option?

By the way, if you really want to test the small transform size while keeping large block size, quadtree TU tool by HHI may be useful for that purpose. The following is a part of the configuration file from the recently integrated software.

QuadtreeTUFlag : 1
# Use quadtree-based TU coding

QuadtreeTULog2MaxSize : 6
# Log2 of maximum transform size for quadtree-based TU coding (2...6)

QuadtreeTULog2MinSize : 2
# Log2 of minimum transform size for quadtree-based TU coding (2...6)

I guess that QuadtreeTULog2MaxSize should be equal to 3 to limit the maximum transform size to 8. For the integrated software, I belive that David will apply the latest changes to the source repository soon.

Thanks for all your help. I have tried other max transform sizes, all are OK.
The md5sum of my source video is correct. The encoder crash without any useful windows exception. And I have uploaded my full log file.

I have not been able to reproduce this bug. Please can you retry with a newer version of the encoder. Several memory leak issues have since been fixed that might have been causing this problem.