does it support windows guest vm？

[nooodles2023]

I want to hook guest cuda api on windows and transform them to host linux.
Does it support hook cuda on windows?

[coldfunction]

You can try git checkout windows branch, but this branch has not been maintained for a long time.

[nooodles2023]

Thank you,buddy,I will try this

[nooodles2023]

I'm trying to hook cuLaunchKernel, but I have some trouble to handle the kernel params, because the guest app did not call setup args before. I also confused that if guest app use a pointer to buffer as a kernel param, and its kernel implement try to write something to the buffer, would qcuda work well in this situation?😆

[coldfunction]

I remember that we used to leverage the virtio framework to pass parameters in the past, and in cases where a large amount of data was being transmitted (such as cudaMemcpy), zero-copy was achieved by transferring through memory address conversion. However, this project has not been maintained for a long time because the main authors are currently busy with other businesses, and the related technology is also very outdated. I apologize for this. I suggest that we can learn from its spirit and apply it to newer applications. Any API function that requires hooking or recording of execution history can refer to the previous approach for designing related wrapper functions. Many thanks. Roy ***@***.***> 於 2023年3月26日 週日 下午2:16寫道：

…

I'm trying to hook cuLaunchKernel, but I have some trouble to handle the kernel params, because the guest app did not call setup args before. I also confused that if guest app use a pointer to buffer as a kernel param, and its kernel implement try to write something to the buffer, would qcuda work well in this situation?😆 — Reply to this email directly, view it on GitHub <#8 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABBUDC6S4D7DSLU3AYMZJE3W57NKZANCNFSM6AAAAAAVUT5OY4> . You are receiving this because you commented.Message ID: ***@***.***>

-- Best Regards, Yu-Shiang

[nooodles2023]

Hi, buddy, I have implemented this now, thank you for your help!!!😆
https://github.com/tenclass/clink
I just transport cuda driver api through the network