cuMemAdvise_v2 and cuMemPrefetchAsync_v2 driver functions aren't available in older cuda compilation tool <12.0.20

[akhildevelops]

cuMemAdvise_v2 and cuMemPrefetchAsync_v2 driver functions aren't available for few of the older cuda compilation tools i.e, <12.0.20.

Can't use the functions memadvise and mem_prefetch_async as they depend on the driver functions that are not available on my system.

Facing compilation error as my cuda compilation tool version is 11.8

error[E0599]: no method named `cuMemAdvise_v2` found for reference `&'static driver::sys::sys_11080::Lib` in the current scope
   --> /home/akhil/.cargo/registry/src/index.crates.io-6f17d22bba15001f/cudarc-0.11.7/src/driver/result.rs:613:10
    |
612 | /     lib()
613 | |         .cuMemAdvise_v2(dptr, num_bytes, advice, location)
    | |_________-^^^^^^^^^^^^^^
    |
help: there is a method `cuMemAdvise` with a similar name
    |
613 |         .cuMemAdvise(dptr, num_bytes, advice, location)
    |          ~~~~~~~~~~~

   Compiling sqlx-macros v0.7.4
error[E0599]: no method named `cuMemPrefetchAsync_v2` found for reference `&'static driver::sys::sys_11080::Lib` in the current scope
     --> /home/akhil/.cargo/registry/src/index.crates.io-6f17d22bba15001f/cudarc-0.11.7/src/driver/result.rs:628:10
      |
627   | /     lib()
628   | |         .cuMemPrefetchAsync_v2(dptr, num_bytes, location, 0, stream)
      | |_________-^^^^^^^^^^^^^^^^^^^^^
      |
help: there is a method `cuMemPrefetchAsync` with a similar name, but with different arguments
     --> /home/akhil/.cargo/registry/src/index.crates.io-6f17d22bba15001f/cudarc-0.11.7/src/driver/sys/sys_11080.rs:12384:5
      |
12384 | /     pub unsafe fn cuMemPrefetchAsync(
12385 | |         &self,
12386 | |         devPtr: CUdeviceptr,
12387 | |         count: usize,
12388 | |         dstDevice: CUdevice,
12389 | |         hStream: CUstream,
12390 | |     ) -> CUresult {
      | |_________________^

➜  projects 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

[akhildevelops]

Can I give a PR that conditionally selects either cuMemAdvise_v2 or cuMemAdvise based on the feature flag.

[ahem]

until this is fixed, I can work around the problem by locking the cudarc version like this:

cudarc = { version = "<=0.11.6" }

[Lyn-liyuan]

I also encountered this issue. Using v0.11.6 can avoid this issue.

[coreylowman]

Fixed by #270. Just released 0.11.8 with this fix.

Sorry for delay!