[BUG] Value 'sm_30' is not defined for option 'gpu-name'

[nbos]

Calling Data.Array.Accelerate.LLVM.PTX.run returns the following error:

*** Internal error in package accelerate ***
*** Please submit a bug report at https://github.com/AccelerateHS/accelerate/issues

ptxas - -o /home/nbos/.cache/accelerate/accelerate-llvm-1.3.0.0/accelerate-llvm-ptx-1.3.0.0/llvm-hs-9.0.1/nvptx64-nvidia-cuda/sm30/rel/morp26739b41823e82434f36247dabfa5730ddb1ad91293e95dc8b68fa7a2ee11cfb.sass -arch=sm_30 (exit 255)
ptxas fatal   : Value 'sm_30' is not defined for option 'gpu-name'


CallStack (from HasCallStack):
  internalError: Data.Array.Accelerate.LLVM.PTX.Compile:185:24
  compileCUBIN: Data.Array.Accelerate.LLVM.PTX.Compile:123:20
  compile: Data.Array.Accelerate.LLVM.PTX.Compile:88:22

I have:

• Accelerate: accelerate-1.3.0.0
• Accelerate backend(s): accelerate-llvm-ptx-1.3.0.0 (accelerate-llvm-native executes fine)
• GHC: ghc-8.10.7
• OS: Arch Linux x86_64
• GPU: NVIDIA GeForce GTX 770 with driver version 470.103.01 from https://aur.archlinux.org/packages/nvidia-470xx-dkms per https://wiki.archlinux.org/title/NVIDIA#Installation
• cuda-11.6.1-1 but seeing the ouput of $ nvidia-smi saying CUDA Version: 11.4, I downgraded to cuda-11.4.2-1, recompiled all the packages, but the problem persists.
• libffi-3.4.2-4

[dpvanbalen]

It seems that CUDA 11 dropped support for your GPU: https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/

[nbos]

Yeah that looks right, I'm downgrading to 10.2 atm.
The README should probably have a note on cuda/GPU compatibility.

[dpvanbalen]

Good idea, I added a note in the README of accelerate-llvm. Let us know whether 10.2 works; we don't actively test older versions as often as we perhaps should!

[nbos]

It works