nv_full runtime test failed

[guodonglai]

I followed the virtualplatform instruction guide but when I try one of the sanity test, it failed.

Here is the log file:

./nvdla_runtime --loadable ../../regression/flatbufs/kmd/CONV/CONV_D_L0_0_larg

e_fbuf
creating new runtime context...
Emulator starting
submitting tasks...
[ 62.871762] Enter:dla_read_network_config
[ 62.875498] Exit:dla_read_network_config status=0
[ 62.888309] Enter: dla_initiate_processors
[ 62.903849] Enter: dla_submit_operation
[ 62.905443] Prepare Convolution operation index 0 ROI 0 dep_count 1
[ 62.908877] Enter: dla_prepare_operation
[ 62.911547] processor:Convolution group:0, rdma_group:0 available
[ 62.913815] Enter: dla_read_config
[ 62.915589] Exit: dla_read_config
[ 62.920740] Exit: dla_prepare_operation status=0
[ 62.923508] Enter: dla_program_operation
[ 62.928722] Program Convolution operation index 0 ROI 0 Group[0]
[ 62.941093] no desc get due to index==-1
[ 62.944967] no desc get due to index==-1
[ 62.952742] no desc get due to index==-1
[ 62.966997] no desc get due to index==-1
[ 62.968714] no desc get due to index==-1
[ 62.971225] no desc get due to index==-1
[ 62.972885] Enter: dla_op_programmed
[ 62.979211] Update dependency operation index 1 ROI 0 DEP_COUNT=1
[ 62.982929] enable SDP in dla_update_dependency as depdency are resolved
[ 62.988152] Enter: dla_enable_operation
[ 62.989812] exit dla_enable_operation without actual enable due to processor hasn't been programmed
[ 62.996153] Exit: dla_enable_operation status=0
[ 63.007180] Exit: dla_op_programmed
[ 63.008845] Exit: dla_program_operation status=0
[ 63.011889] Exit: dla_submit_operation
[ 63.013824] Enter: dla_dequeue_operation
[ 63.016482] exit Convolution as there's no further operation
[ 63.022989] Exit: dla_dequeue_operation
[ 63.027468] Enter: dla_submit_operation
[ 63.029075] Prepare SDP operation index 1 ROI 0 dep_count 0
[ 63.039124] Enter: dla_prepare_operation
[ 63.041107] processor:SDP group:0, rdma_group:0 available
[ 63.043659] Enter: dla_read_config
[ 63.045195] Exit: dla_read_config
[ 63.047137] Exit: dla_prepare_operation status=0
[ 63.048941] Enter: dla_program_operation
[ 63.049646] Program SDP operation index 1 ROI 0 Group[0]
[ 63.065550] no desc get due to index==-1
[ 63.072565] no desc get due to index==-1
[ 63.076414] no desc get due to index==-1
[ 63.087045] no desc get due to index==-1
[ 63.088770] no desc get due to index==-1
[ 63.093519] no desc get due to index==-1
[ 63.096387] Enter: dla_op_programmed
[ 63.099191] Exit: dla_op_programmed
[ 63.104234] Exit: dla_program_operation status=0
[ 63.107019] Enter: dla_enable_operation
[ 63.112596] Enable SDP operation index 1 ROI 0
[ 63.118555] Enter: dla_op_enabled
[ 63.120020] Update dependency operation index 0 ROI 0 DEP_COUNT=1
[ 63.137489] enable Convolution in dla_update_dependency as depdency are resolved
[ 63.141712] Enter: dla_enable_operation
[ 63.144328] Enable Convolution operation index 0 ROI 0
[ 63.148278] Enter: dla_op_enabled
[ 63.149705] Exit: dla_op_enabled
[ 63.152031] Exit: dla_enable_operation status=0
[ 63.153873] Exit: dla_op_enabled
[ 63.155912] Exit: dla_enable_operation status=0
[ 63.159603] Exit: dla_submit_operation
[ 63.161200] Enter: dla_dequeue_operation
[ 63.163713] exit SDP as there's no further operation
[ 63.168822] Exit: dla_dequeue_operation
[ 63.171264] Exit: dla_initiate_processors status=0

Info: NV_NVDLA_csc.cpp: 647:NV_NVDLA_csc::SendDataToMacSequencerDirectConvCommon, invalid configuration csc_entries_, actual value is 0xF, it shall be 0x7.

Info: NV_NVDLA_cdma.cpp: 1265:NV_NVDLA_cdma::DirectConvDataResponseSequencerCommon, invalid configuration cdma_entries_, set value is 0xF, it shall be 0x7FF.

Info: NV_NVDLA_cdma.cpp: 1272:NV_NVDLA_cdma::DirectConvDataResponseSequencerCommon, invalid configuration of cdma_line_packed_, cdma_line_packed_ is 1, line_stride == ATOM_CUBE_SIZE * cube_width is 0.

Info: NV_NVDLA_cdma.cpp: 2073:If weight uncompressed, all cdma2sc_wt should have been sent

Any ideas how to fix this?

[guodonglai]

This is the latest version of compiler and I am using nv_full as provided in the hardware/specs.

[prasshantg]

You should be using https://github.com/nvdla/sw/blob/master/regression/flatbufs/kmd/CONV/CONV_D_L0_0_fbuf for nv_full

[guodonglai]

I tried the CONV_D_L0_0_fbuf with opendla_1.ko and five.pgm on the image you provided. The result is still the same.

[guodonglai]

I have also tried to test it on the image I generated using buildroot, but it is still showing the same info:
Info: NV_NVDLA_cdma.cpp: 1265:NV_NVDLA_cdma::DirectConvDataResponseSequencerCommon, invalid configuration cdma_entries_, set value is 0xF, it shall be 0x7FF.
So where exactly does the problem come from? From Cmodel or from the loadable provided in sw folder?

[shazib-summar]

You should be using https://github.com/nvdla/sw/blob/master/regression/flatbufs/kmd/CONV/CONV_D_L0_0_fbuf for nv_full

Which file should I use for nv_small @prasshantg ?

[gitosu67]

@killerzula For nv_small you should be using https://github.com/nvdla/sw/blob/master/regression/flatbufs/kmd/CONV/CONV_D_L0_0_small_fbuf

[ICermanRan]

You should be using https://github.com/nvdla/sw/blob/master/regression/flatbufs/kmd/CONV/CONV_D_L0_0_fbuf for nv_full
I meet the same problem,when I build and compile vp HW_PROJECT without nv_small,such as nv_small_256 or nv_large,the QEMU will directly quit,and faile.
The log file as:
creating new runtime context...
Emulator starting
dlaimg height: 28 x 28 x 1: LS: 896 SS: 0 Size: 25088
submitting tasks...
[ 49.212281] Enter:dla_read_network_config
[ 49.218963] Exit:dla_read_network_config status=0
[ 49.219117] Enter: dla_initiate_processors
[ 49.219451] Enter: dla_submit_operation
[ 49.219660] Prepare Convolution operation index 0 ROI 0 dep_count 1
[ 49.220188] Enter: dla_prepare_operation
[ 49.225101] processor:Convolution group:0, rdma_group:0 available
[ 49.226196] Enter: dla_read_config
[ 49.226997] Exit: dla_read_config
[ 49.227370] Exit: dla_prepare_operation status=0
[ 49.228722] Enter: dla_program_operation
[ 49.229118] Program Convolution operation index 0 ROI 0 Group[0]
[ 49.250708] no desc get due to index==-1
[ 49.250906] no desc get due to index==-1
[ 49.250996] no desc get due to index==-1
[ 49.251109] no desc get due to index==-1
[ 49.251201] no desc get due to index==-1
[ 49.251317] Enter: dla_op_programmed
[ 49.251435] Update dependency operation index 3 ROI 0 DEP_COUNT=3
[ 49.251573] Update dependency operation index 1 ROI 0 DEP_COUNT=1
[ 49.251700] enable SDP in dla_update_dependency as depdency are resolved
[ 49.251863] Enter: dla_enable_operation
[ 49.251967] exit dla_enable_operation without actual enable due to processor hasn't been programmed
[ 49.252199] Exit: dla_enable_operation status=0
[ 49.252320] Exit: dla_op_programmed
[ 49.252408] Exit: dla_program_operation status=0
[ 49.252495] Exit: dla_submit_operation
[ 49.252574] Enter: dla_dequeue_operation
[ 49.252715] Dequeue op from Convolution processor, index=3 ROI=0
[ 49.252887] Enter: dla_submit_operation
[ 49.252991] Prepare Convolution operation index 3 ROI 0 dep_count 2
[ 49.254260] Enter: dla_prepare_operation
[ 49.256650] processor:Convolution group:1, rdma_group:0 available
[ 49.256761] Enter: dla_read_config
[ 49.256861] Exit: dla_read_config
[ 49.256931] Exit: dla_prepare_operation status=0
[ 49.257049] Enter: dla_program_operation
[ 49.258710] Program Convolution operation index 3 ROI 0 Group[1]
[ 49.262756] no desc get due to index==-1
[ 49.262897] no desc get due to index==-1
[ 49.262978] no desc get due to index==-1
[ 49.263073] no desc get due to index==-1
[ 49.263321] no desc get due to index==-1
[ 49.263426] Enter: dla_op_programmed
[ 49.263539] Update dependency operation index 6 ROI 0 DEP_COUNT=3
[ 49.263671] Update dependency operation index 4 ROI 0 DEP_COUNT=2
[ 49.263802] Exit: dla_op_programmed
[ 49.263850] Exit: dla_program_operation status=0
[ 49.263894] Exit: dla_submit_operation
[ 49.263957] Exit: dla_dequeue_operation
[ 49.264004] Enter: dla_submit_operation
[ 49.264043] Prepare SDP operation index 1 ROI 0 dep_count 0
[ 49.264191] Enter: dla_prepare_operation
[ 49.264578] processor:SDP group:0, rdma_group:0 available
[ 49.264705] Enter: dla_read_config
[ 49.264843] Exit: dla_read_config
[ 49.264953] Exit: dla_prepare_operation status=0
[ 49.265061] Enter: dla_program_operation
[ 49.266137] Program SDP operation index 1 ROI 0 Group[0]
[ 49.268374] no desc get due to index==-1
[ 49.268485] no desc get due to index==-1
[ 49.268539] no desc get due to index==-1
[ 49.268661] no desc get due to index==-1
[ 49.268786] Enter: dla_op_programmed
[ 49.268826] Update dependency operation index 4 ROI 0 DEP_COUNT=1
[ 49.268988] enable SDP in dla_update_dependency as depdency are resolved
[ 49.269898] Enter: dla_enable_operation
[ 49.270140] exit dla_enable_operation without actual enable due to processor hasn't been programmed
[ 49.270314] Exit: dla_enable_operation status=0
[ 49.270417] Exit: dla_op_programmed
[ 49.270533] Exit: dla_program_operation status=0
[ 49.270762] Enter: dla_enable_operation
[ 49.270859] Enable SDP operation index 1 ROI 0
[ 49.272428] Enter: dla_op_enabled
[ 49.272524] Update dependency operation index 0 ROI 0 DEP_COUNT=1
[ 49.272668] enable Convolution in dla_update_dependency as depdency are resolved
[ 49.273001] Enter: dla_enable_operation
[ 49.274491] Enable Convolution operation index 0 ROI 0
[ 49.276213] Enter: dla_op_enabled
[ 49.276299] Exit: dla_op_enabled
[ 49.276373] Exit: dla_enable_operation status=0
[ 49.276496] Exit: dla_op_enabled
[ 49.276597] Exit: dla_enable_operation status=0
[ 49.276707] Exit: dla_submit_operation
[ 49.276787] Enter: dla_dequeue_operation
[ 49.276876] Dequeue op from SDP processor, index=4 ROI=0
[ 49.277000] Enter: dla_submit_operation
[ 49.277088] Prepare SDP operation index 4 ROI 0 dep_count 0
[ 49.277948] Enter: dla_prepare_operation

Fatal: NV_NVDLA_csc.cpp: 659:NV_NVDLA_csc::SendDataToMacSequencerDirectConvCommon, invalid configuration csc_entries_, actual value is 0xD, it shall be 0x1B.
In file: ../cmod/csc/NV_NVDLA_csc.cpp:659
In process: nvdla.nvdla_core.nvdla_core.csc.DataLoadSequenceThread @ 49100 ms
Aborted (core dumped)
Any ideas how to fix this? How should I use https://github.com/nvdla/sw/blob/master/regression/flatbufs/kmd/CONV/CONV_D_L0_0_fbuf for nv_full? change kmd?
Waitting for your response! QAQ

[long771]

这是来自QQ邮箱的自动回复邮件。 您好，我是龙欣荣，很高兴收到您的邮件，我将尽快处理。