Does TVM support RK3288?

[wjtan99]

Hi, I saw you do the benchmark test on RK3399. RK3288 has a Mali-T764 GPU. We tried it but the opencl has some problem recognizing the GPU. Do you have idea if TVM supports RK3288?

[merrymercy]

rk3288 is similar to rk3399, so tvm supports it. There maybe some problems in your environment configuration, like opencl driver, tvm build option.

Could you post your error message?

[wjtan99]

Here is the error message

python -m tvm.exec.rpc_server --host 0.0.0.0 --port=9090
INFO:root:RPCServer: bind to 0.0.0.0:9090
INFO:root:RPCServer: connection from ('192.168.7.48', 33508)
INFO:root:Connection from ('192.168.7.48', 33508)
[01:12:39] src/runtime/rpc/rpc_server_env.cc:23: Upload /tmp/tmpSITYtx/net.tar... nbytes=421284
[01:12:54] src/runtime/opencl/opencl_device_api.cc:197: Initialize OpenCL platform 'ARM Platform'
ERROR: The DDK is not compatible with any of the Mali GPUs on the system.
The DDK was built for 0x750 r0p0 status range [1..1], but none of the GPUs matched:
/dev/mali0 detected as 0x750 r1p0 status 0
[01:12:54] src/runtime/opencl/opencl_device_api.cc:202: No OpenCL device any device matched given the options: gpu mode
[01:12:54] src/runtime/opencl/opencl_device_api.cc:203: Now try OpenCL cpu mode
ERROR: The DDK is not compatible with any of the Mali GPUs on the system.
The DDK was built for 0x750 r0p0 status range [1..1], but none of the GPUs matched:
/dev/mali0 detected as 0x750 r1p0 status 0
[01:12:54] src/runtime/opencl/opencl_device_api.cc:206: No OpenCL device any device matched given the options: cpu mode
[01:12:54] src/runtime/rpc/rpc_session.cc:751: Shutdown...
INFO:root:Finish serving ('192.168.7.48', 33508)****

[wjtan99]

“make USE_OPENCL=1 LLVM_CONFIG=llvm-config-4.0 -j4”

[merrymercy]

It seems the problem is your opencl driver, not the tvm runtime.

[liurs1990]

When I run the test on RK3288, I get the error message:

Traceback (most recent call last):
File "mali_imagenet_bench.py", line 109, in
run_case(args.model, args.dtype)
File "mali_imagenet_bench.py", line 55, in run_case
module = runtime.create(graph, rlib, ctx)
File "/work/python-tvm/local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-x86_64.egg/tvm/contrib/graph_runtime.py", line 42, in create
return GraphModule(fcreate(graph_json_str, hmod, device_type, device_id), ctx)
File "/work/python-tvm/local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-x86_64.egg/tvm/_ffi/_ctypes/function.py", line 183, in call
ctypes.byref(ret_val), ctypes.byref(ret_tcode)))
File "/work/python-tvm/local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-x86_64.egg/tvm/_ffi/base.py", line 66, in check_call
raise TVMError(py_str(_LIB.TVMGetLastError()))
tvm._ffi.base.TVMError: Except caught from RPC call: [02:02:46] src/runtime/graph/graph_runtime.cc:485: Check failed: storage_id >= 0 (-1 vs. 0) Do not support runtime shape op

Stack trace returned 10 entries:
[bt] (0) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(dmlc::StackTraceabi:cxx11+0x37) [0xb5c9fa4c]
[bt] (1) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(dmlc::LogMessageFatal::~LogMessageFatal()+0x21) [0xb5ca011a]
[bt] (2) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::GraphRuntime::SetupStorage()+0x50f) [0xb5f601c0]
[bt] (3) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::GraphRuntimeCreate(std::__cxx11::basic_string<char, std::char_traits, std::allocator >, tvm::runtime::Module, int, int)+0x1c1) [0xb5f60c92]
[bt] (4) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(+0x4321bc) [0xb5f611bc]
[bt] (5) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandlePackedCall()+0x3d3) [0xb5f4ad5c]
[bt] (6) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::RPCSession::EventHandler::SwitchToState(tvm::runtime::RPCSession::EventHandler::State)+0x1a7) [0xb5f4e7f8]
[bt] (7) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandleRecvPackedSeqArg()+0x20d) [0xb5f4efce]
[bt] (8) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandleNextEvent(tvm::runtime::TVMRetValue*, bool, tvm::runtime::PackedFunc const*)+0x1d7) [0xb5f4f474]
[bt] (9) /home/firefly/.local/lib/python2.7/site-packages/tvm-0.2.0-py2.7-linux-armv7l.egg/tvm/libtvm.so(tvm::runtime::RPCSession::HandleUntilReturnEvent(tvm::runtime::TVMRetValue*, bool, tvm::runtime::PackedFunc const*)+0xf7) [0xb5f4bf18]

@merrymercy Could you give some help ?

[merrymercy]

cc @tqchen

[tqchen]

Please use the latest nnvm along with latest tvm