CODEWITHSUNDEEP
haskellcabalcabal-install
user3199023
user3199023

Reputation: 433

Cannot install accelerate-cuda in Haskell

I am on a linux box and trying to experiment with Haskell's Accelerate library but having problems installing it. I have successfully installed the accelerate package but there seems to be a dependency problem, which I have detailed below.

cabal: Error: some packages failed to install:
accelerate-cuda-0.14.0.0 depends on haskell-src-exts-1.14.0.1 which failed to
install.
cuda-0.5.1.1 failed during the configure step. The exception was:
ExitFailure 1
haskell-src-exts-1.14.0.1 failed during the configure step. The exception was:
ExitFailure 1
haskell-src-meta-0.6.0.5 depends on haskell-src-exts-1.14.0.1 which failed to
install.
language-c-quote-0.7.6 depends on haskell-src-exts-1.14.0.1 which failed to
install.

I searched SO and noticed someone else had a similiar issue installed the cuda package, and was resolved by adding the cabal bin path to the PATH; I tried this but it didn't solve this problem.

Please could someone help as I really keen to play with this fantastic library.

Upvotes: 3

Views: 519

Answers (1)

Slaus
Slaus

Reputation: 2226

I wanted to check accelerate-examples and play with them and I also didn't have CUDA GPU (AMD only) and that's how I eventually installed accelerate-examples with stack:

git clone https://github.com/AccelerateHS/accelerate-examples
cd accelerate-examples
#choose version:
ln stack-8.6.yaml stack.yaml
#build without CUDA targeting:
stack build --flag accelerate-examples:-llvm-ptx --flag accelerate-fft:-llvm-ptx

Installation will build all the examples and print the info regarding where they was put. Might also need to specify GHC libs path with something like: export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/.stack/programs/x86_64-linux/ghc-8.6.5/lib/ghc-8.6.5/rts/

Also checked that accelerate installs as well the same way but even without specifying any flags (I guess because it doesn't build any programs yet?), but examples are what's fun :)

P.S. You can move llvm-ptx flags to stack.yaml config: change # flags: {} line to:

flags:
  accelerate-fft:
    llvm-ptx: false
  accelerate-examples:
    llvm-ptx: false

Upvotes: 1

Related Questions