I am using libtorch, c++ frontend, downloaded from the main page of https://pytorch.org/ and trying to profile layer by layer.
I've learn that in python i can use torch.autograd.profiler.profile. And i've read some website, including https://github.com/pytorch/pytorch/pull/16580/files#diff-81d29c29bbf9fb05edfa7fd70d189374 and https://caffe2.ai/doxygen-c/html/structtorch_1_1autograd_1_1profiler_1_1_record_profile.html
But when i use CLion to construct my code, use torch::autograd::profiler::RecordProfile , it report an error that
no member named 'profiler' in namespace 'torch::autograd
I turned to https://pytorch.org/cppdocs/api/library_root.html, and cannot find something about profiler, so how should i do to use the profiler in my project?
Related
I am trying to port an existing C++ application running on linux currently to freertos.
My application has various custom libraries ( prj_lib1, prj_lib2, … prj_lib9) which I wanna link to one .exe(Project_main.exe)
I just need a bit of clarification on how to proceed in this scenario.
Do I need to create freertos static library for each library to link in the Project_main.exe or normal c++ static library should be good enough?
Currently what I am doing is, I have generated Project_main.exe by modifying an existing demo application. And then created normal C++ static library for different libraries which I link in the Project_main.exe.
Is this the correct way to go ahead? Or should I modify existing demo application to generate libraries too?
If there’s some other way please do let me know.
I'm working on a project which is an open-source cloud, similar to Nextcloud. It also provides a file upload and manage feature.
I'd like to have a shell namespace extension which displays those files and folders in the Windows Explorer. Moving, deleting, renaming and editing files / folders should be possible.
Data is available from a REST API or my Electron Client.
I know NodeJS and only some C#.
Is it possible whith NodeJS / Electron to create a shell namespace extension?
If not, are there any free libraries which make this task easier?
Windows Explorer uses 'COM' as its extension model. Therefore, technically speaking, you can use whatever you like, as long as it supports COM.
For NodeJS, you'd be dependent on the runtime to provide the COM support for you and to wrap the appropriate APIs. I'm not aware of any project that does this for the Windows Shell APIs. Here is one for the WinRT APIs.
For C# (.NET), again you'd be dependent on the runtime for COM support. .NET does expose that to you; although it is messy. There are a few projects which wrap the shell APIs and try to hide the COM interop goo as best they can for you.
Here is an open source one.
Here is a commercial one.
Be aware that Microsoft advises against building shell extensions which require a runtime (EG: .NET). The runtime may be loaded into processes that aren't expecting it. Therefore, in practice, Microsoft expects you to use C/C++ to build shell extensions. Here is the Microsoft supplied sample project. It is written in C++.
I estimate possibility to write Report Desktop Application using React NodeGUI. There is some questions what I need to find answer before starting to develop by NodeGUI. Could you help me to figure out.
What way is to load program extension dynamicaly in React NodeGUI. For example: If I want to change business logic and visualization of my program without compile code. Is it possible?
What way is to load program extension dynamicaly in React NodeGUI. For example: If I want to change business logic and visualization of program adding some program module code of other developers. How way is it possible to do?
What way is to work with XML files in React NodeGui?
Is there possibility to run other program code written by C/C++ or other languages? What restriction of running non-native JS modules?
What way is to connect to SQL or NoSQL DB (PostgreSQL, Oracle, Mongo and etc)?
What are reasons yet to write or not write Desktop Applications by React NodeGUI?
React NodeGui is just another npm module that you would install in a nodejs program. It is a pure javascript library. It depends on NodeGui which is another regular npm module. NodeGui contains C++ code which acts as a nodejs addon. But for all purposes think of nodegui and react nodegui as regular npm modules. An app built with these libraries is a regular nodejs app.
What way is to load program extension dynamicaly in React NodeGUI. For example: If I want to change business logic and visualization of my program without compile code. Is it possible?
All business logic is written in javascript so any change in those would follow the same flow of how you would in a nodejs app.
What way is to load program extension dynamicaly in React NodeGUI. For example: If I want to change business logic and visualization of program adding some program module code of other developers. How way is it possible to do?
Yes you can use all npm modules available in npm that works with nodejs.
What way is to work with XML files in React NodeGui?
You can find a lot of npm modules you can use to parse and work with xml files in the npm registry.
Is there possibility to run other program code written by C/C++ or other languages? What restriction of running non-native JS modules?
Yes it is possible to run non js code by building them as NodeJs addons.
What way is to connect to SQL or NoSQL DB (PostgreSQL, Oracle, Mongo and etc)?
You have a lot of modules available in the npm registry for these which work with NodeJs.
What are reasons yet to write or not write Desktop Applications by React NodeGUI?
Reasons to use:
- NodeGui apps are native apps so they get all the benefits of native apps. They are snappy, they are accessible and run with low memory and cpu usage.
Reasons not to use:
Its under development so is only recommended for hobby projects as of Dec 2019.
PS: I am the author of the library
I have a Win8 app that is purely native (c++) and I've already used a library that is written using managed code AFAIK. No issues there, I created objects and addressed them using C++/CX with ref counted pointers etc.
I need to add a new library, I referenced it as I did previously, but when I declare and object and try to address it I get:
error C3624: 'System::Object': use of this type requires a reference to assembly 'mscorlib'
Using #using <mscorlib.dll> is not a solution obviously, because WinRT does not support #using of a managed assembly.
When I look up the definition of the class I'm using in Object Browser I see that it's inherited from System::Object. The previous library had a class that was inherited from Platform::Object which is valid for C++/CX.
I already contacted the developers of current library I'm trying to use, but it takes a lot of time for them to respond.
Can I work around this issue? What are the possible courses of action?
UPD: Can I ask the developers to rebuild a library for C++/CX?
I do not believe this is going to work in the general case, unless the C# library is a PCL (portable class library). If it leverages anything that is not in the WinRT .NET client profile, it simply will not work.
If it is a PCL, what you can do is write a C# Windows Runtime Component that itself has a reference to this third-party library and wraps the necessary functionality. Then you reference that C# Windows Runtime Component from your C++/Cx application.
What is this and what is it purpose. I notice that when i compile my console apps in c++ is dynamically links it. I'm just curious what purpose this file serves and is it exploitable in the sense that if it has to be on every ones machine to run c code , is there some way for Microsoft or some other entity to exploit it for malicious ends.
That is the Standard C++ Library for native code, here you can read more about these libraries: http://msdn.microsoft.com/en-us/library/8kche8ah%28v=vs.90%29.aspx