cust error: struct CUDA_KERNEL_NODE_PARAMS_st has no field named ctx

[hduoc2003]

I attempted to run the vecadd example but encountered the following errors:

➜  vecadd cargo run 
   Compiling rustc_codegen_nvvm v0.3.0 (https://github.com/Rust-GPU/Rust-CUDA#290d7115)
   Compiling cust v0.3.2 (https://github.com/Rust-GPU/Rust-CUDA#290d7115)
error[E0560]: struct `CUDA_KERNEL_NODE_PARAMS_st` has no field named `kern`
  --> /root/.cargo/git/checkouts/rust-cuda-f2e68423bdc84998/290d711/crates/cust/src/graph.rs:97:13
   |
97 |             kern: ptr::null_mut(),
   |             ^^^^ `CUDA_KERNEL_NODE_PARAMS_st` does not have this field
   |
   = note: all struct fields are already assigned

error[E0560]: struct `CUDA_KERNEL_NODE_PARAMS_st` has no field named `ctx`
  --> /root/.cargo/git/checkouts/rust-cuda-f2e68423bdc84998/290d711/crates/cust/src/graph.rs:98:13
   |
98 |             ctx: ptr::null_mut(),
   |             ^^^ `CUDA_KERNEL_NODE_PARAMS_st` does not have this field
   |
   = note: all struct fields are already assigned

For more information about this error, try `rustc --explain E0560`.
error: could not compile `cust` (lib) due to 2 previous errors
warning: build failed, waiting for other jobs to finish...

---

Context

I recently switched the cust, cuda_std, and cuda_builder crates from local paths to their current Git versions.

---

Environment

$ llvm-config --version
7.0.1

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

---

rust-toolchain.toml

Same as main branch.

[toolchain]
channel = "nightly-2025-03-02"
components = [
    "clippy",
    "llvm-tools-preview",
    "rust-src",
    "rustc-dev",
    "rustfmt",
    "rust-analyzer"
]

[jorge-ortega]

Hi @hduoc2003,

The error is due to changes in CUDA SDK from version 11 to version 12. During the reboot of this project, we updated our bindings (cust-raw) and the safe wrappers (cust) but only focused on supporting CUDA 12. We should be able to add support back for 11.8. In the meantime, try to use CUDA 12.1 or higher (there is another issue with 12.0 #175).