Nsight System fail to start profiling on win10

[xiaowei.liu.april]

RuntimeError (120) {
TargetSideError (1100) {
ErrorText: Throw location unknown (consider using BOOST_THROW_EXCEPTION)
Dynamic exception type: class boost::exception_detail::clone_impl<class boost::exception_detail::current_exception_std_exception_wrapper >
std::exception::what: bad conversion
class std::range_error
ServiceName: AnalysisService
MethodName: Init
InternalError (105) {
OriginalExceptionClass: class boost::exception_detail::clone_impl<class boost::exception_detail::current_exception_std_exception_wrapper >
}
}
}

Nsight Systems Version: 2020.3.2.6-87e152c Windows-x64.
cuda 10.2
Nvidia Studio Driver 461.40

What does the error above mean? How can I fix it?

[Dark_Photon]

Two things you might try:

1. Update to the latest Nsight Systems and retry (2021.1 currently)
2. Run Nsight Systems as Administrator and retry.

I suggest #2 as I hit a problem a few months back on Win10 where on older version of Nsight Systems would crash after profiling collection (with boost APIs on the stack) unless run as an elevated user. That’s been fixed in more recent versions, but you’re running an older version. See this link for details:

Nsight Systems 2020.5 Crashes When Run as Normal User

Whenever I run the Nsight Systems GUI app as an ordinary user, it typically crashes after a collecting profiling data on an application. This crash occurs after the profiling run, immediately after loading and processing all the samples, but just before displaying the Nsight Graphics GUI showing the collected profiling data). The stack trace is shown below. However, I noticed Friday that if I run the Nsight Systems GUI app as an elevated user, then it does not crash. The Nsight Systems GUI i…

[xiaowei.liu.april]

I tried to run it as an administrator, doesn’t work.
I updated to NVIDIA Nsight Systems, 2020.5.1.85-5ee086b Windows-x64, ran it as an administrator, still doesn’t work.
The latest version I see here is 2020.5 NVIDIA Nsight Systems | NVIDIA Developer.

Here are the error details:
Full error information:
StartAnalysisError (4105) {
RuntimeError (120) {
OriginalExceptionClass: class boost::exception_detail::clone_impl
OriginalFile: B:\Src\Default\QuadD\Host\Analysis\Clients\AnalysisHelper\AnalysisStatus.cpp
OriginalLine: 80
OriginalFunction: class Nvidia::QuadD::Analysis::Data::AnalysisStatusInfo __cdecl QuadDAnalysis::AnalysisHelper::AnalysisStatus::MakeFromErrorString(enum Nvidia::QuadD::Analysis::Data::AnalysisStatus,enum Nvidia::QuadD::Analysis::Data::AnalysisErrorType::Type,const class std::basic_string<char,struct std::char_traits,class std::allocator > &,const class boost::intrusive_ptr &)
ErrorText: Failed to start analysis.
}
}

Profiling options:
DeviceId: “Local”
EventTypes {
Items: CpuCycles
Items: Cuda
}
RateHz: 1000
HowToStart: Delayed
HowToStop: Manual
DeviceType: Windows
DeviceDisplayName: “XIAOWEI-LIU”
WindowsPerfOptions {
retainEtwFiles: true
}
Processes {
HowToAttach: LaunchAnother
Command: “F:\gpu_cloth\build\cloth_sim.exe”
WorkingDirectory: “F:\gpu_cloth\build”
UserName: “Administrator”
CollectCudaTrace: true
CudaFlushPeriodically: true
CudaFlushPeriod: 10000000000
CudaSkipSomeApiCalls: true
CollectGPUMemoryUsage: true
}
StartDelayMilliseconds: 10000

[Dark_Photon]

@xiaowei.liu.april

The latest version I see here is 2020.5 https://developer.nvidia.com/nsight-systems.

Sorry, my mistake. I was crossing latest version numbers for Nsight Systems (2020.5) and Nsight Graphics (2021.1).

[yimodo.siki]

I have the same error. I fix it by changing “Working directory” to be the directory containing the .exe file.
For example, if “Command line with arguments” is “D:\code\test.exe”, I set “Working directory” as “D:\code”.