argrento opened a new pull request #10372:
URL: https://github.com/apache/tvm/pull/10372


   Currently in some cases debug messages are shown in a unformatted way as 
shown below.
   
   ```
   DEBUG:autotvm:No: 151        GFLOPS: 0.00/0.00       result: 
MeasureResult(costs=('Traceback (most recent call last):\n  [bt] (8) 
/path/tvm/build/libtvm.so(TVMFuncCall+0x63) [0x7f2995ef4fa3]\n  [bt] (7) 
/path/tvm/build/libtvm.so(+0x1619457) [0x7f2995f65457]\n  [bt] (6) 
/path/tvm/build/libtvm.so(tvm::runtime::RPCWrappedFunc::operator()(tvm::runtime::TVMArgs,
 tvm::runtime::TVMRetValue*) const+0x365) [0x7f2995f6bd35]\n  [bt] (5) 
/path/tvm/build/libtvm.so(tvm::runtime::RPCClientSession::CallFunc(void*, 
TVMValue const*, int const*, int, std::function<void (tvm::runtime::TVMArgs)> 
const&)+0x57) [0x7f2995f5dcf7]\n  [bt] (4) 
/path/tvm/build/libtvm.so(tvm::runtime::RPCEndpoint::CallFunc(void*, TVMValue 
const*, int const*, int, std::function<void (tvm::runtime::TVMArgs)>)+0x515) 
[0x7f2995f51f95]\n  [bt] (3) 
/path/tvm/build/libtvm.so(tvm::runtime::RPCEndpoint::HandleUntilReturnEvent(bool,
 std::function<void (tvm::runtime::TVMArgs)>)+0x2d5) [0x7f2995f50f95]\n  [bt] 
(2) /path/tvm/build/libtvm.so(tvm::runti
 me::RPCEndpoint::EventHandler::HandleNextEvent(bool, bool, std::function<void 
(tvm::runtime::TVMArgs)>)+0x459) [0x7f2995f5dac9]\n  [bt] (1) 
/path/tvm/build/libtvm.so(tvm::runtime::RPCEndpoint::EventHandler::HandleReturn(tvm::runtime::RPCCode,
 std::function<void (tvm::runtime::TVMArgs)>)+0x13c) [0x7f2995f5c8dc]\n  [bt] 
(0) /path/tvm/build/libtvm.so(+0x1602e78) [0x7f2995f4ee78]\n  File 
"/path/tvm/src/runtime/opencl/opencl_module.cc", line 240\n  File 
"/path/tvm/src/runtime/rpc/rpc_endpoint.cc", line 378\nRPCError: Error caught 
from RPC call:\n[18:13:48] /path/tvm/src/runtime/library_module.cc:78: 
\n---------------------------------------------------------------\nAn internal 
invariant was violated during the execution of TVM.\nPlease read TVM\'s error 
reporting guidelines.\nMore details can be found here: 
https://discuss.tvm.ai/t/error-reporting/7793.\n---------------------------------------------------------------\n\n
  Check failed: ret == 0 (-1 vs. 0) : TVMError: OpenCL build error f
 or device=0x6d783918c8\nBC-src-code:9:110: error: expected \')\'\n__kernel 
void default_function_kernel1(__write_only image2d_t kernel_texture_weight, 
__global half* restrict kernel) {\n                                             
                                                                
^\nBC-src-code:9:39: note: to match this \'(\'\n__kernel void 
default_function_kernel1(__write_only image2d_t kernel_texture_weight, __global 
half* restrict kernel) {\n                                      
^\nBC-src-code:9:110: error: parameter name omitted\n__kernel void 
default_function_kernel1(__write_only image2d_t kernel_texture_weight, __global 
half* restrict kernel) {\n                                                      
                                                       ^\nBC-src-code:10:158: 
error: expected expression\n  (void)write_imageh(kernel_texture_weight, 
(int2)((((int)get_local_id(0)) % 27), ((((int)get_group_id(0)) * 2) + 
(((int)get_local_id(0)) / 27))), vload4(0, kernel
  + ((((int)get_group_id(0)) * 216) + (((int)get_local_id(0)) * 4))));\n        
                                                                                
                                                                     ^\n3 
diagnostic(s) generated.',), error_no=7, all_cost=15, 
timestamp=1645715629.9118824)   [('tile_fc', [-1, 1, 8]), ('tile_y', [-1, 7, 
8]), ('tile_x', [-1, 4, 2]), ('tile_rcc', [-1, 1]), ('tile_ry', [-1, 3]), 
('tile_rx', [-1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 
0)],None,14882
   ```
   This PR makes the following changes:
   
   1. Modify debug output to make it more readable
   2. Replace magic number with a variable `error_ct_threshold`
   3. Add function to set error counter threshold externally for debug purposes
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to