Having problems with CUDA installation, VS, Win11

[jeff.lawton]

I’m having enormous trouble trying to simply run the code for the first sample (using CUDA) shown in the eBook “Hands-on GPU Computing using Python” (which I understand is already a bit old). The author gives an example of PTX execution using nvcc directly, this is from the chapter “Understanding how CUDA-C/C++ works via a simple example”, he hasn’t yet demonstrated the use of cmake or anything so I don’t have any options at this point. No matter what I try adding to the PATH when I run the example I keep getting “nvcc fatal, unable to find compiler cl.exe” and there’s no obvious way to fix it (yes I’m running in a command box and changing directories all over etc.). Everything I’ve run tells me the Visual Studio (and VSC) installations are fine and I’ve installed the VC++ “workload” OK. One problem I suspect could be present is I’m on v12.6 and I’ve read this is high enough that it may be incompatible with some CUDA tools, in other words it’s not that it doesn’t FIND A compiler, it’s just not one whose parameters it’s completely happy with. Also there don’t seem to be a lot of ways to verify that the elements of the individual installations are compatible with one another. I’m looking for either 1) a way to make this install work with the instructions I have, 2) a way to “roll something back” to a point where everything is compatible and works anyway, or 3) the set of instructions thatI need in order to understand how to do this with cmake (although why I need to use that for just one file baffles me). Or maybe 4) there’s a more recent book for me to study that doesn’t assume 20 times the level of knowledge I currently have? Can I get some help here, I’d like to start moving forward again rather quickly…