Browse files

README with installation instructions etc.

  • Loading branch information...
dybber committed Dec 2, 2011
1 parent 20ab8fb commit 9a9e8caa7bc7c2716eb345fa4d311b33acf5f1a0
Showing with 62 additions and 0 deletions.
  1. +62 −0
@@ -0,0 +1,62 @@
+This is a partially implemented OpenCL backend for the Accelerate library
+for General purpose GPU programming.
+The backend is still in a very experimental stage and currently
+completely lacks support of the following operations:
+ * Higher dimension folds
+ * Segmented folds
+ * Scan (prefix sum) operations
+ * Stencil operations
+ * Array indexing
+The accelerate-opencl depends on the following two Cabal packages
+which are not yet available on Hackage in recent enough versions.
+This library provides Haskell bindings to the OpenCL API, and is
+available from
+language-c-quote (with OpenCL extensions)
+The version of language-c-quote currently available at Hackage does
+not contain support for OpenCL C. These extensions can be found in the
+most recent version of language-c-quote on Github:
+The Accelerate team at UNSW provides a set of examples, available as a
+Cabal-package at
+Making a `cabal install --flags="-cuda"` (disabling the CUDA backend)
+of this package, will install an executable which can be used to run a
+suite of tests.
+All the examples can then be executed by:
+> accelerate-examples --opencl -n 5000
+Which will execute all examples on 5000 element arrays. See
+`accelerate-examples --help` for further instructions.
+Questions and bug reports
+Feel free to contact me by email at <> with any
+questions or bug reports.
+Alternatively, bugs can be submitted through the bug tracker at:
+The work on this backend was initially done as a graduate project at
+the University of Copenhagen, and afterwards as a paid internship
+sponsored at the HIPERFIT research center (
+I would also like to thank Manuel Chakravarty, Trevor McDonell and the
+rest of the team behind Accelerate.

0 comments on commit 9a9e8ca

Please sign in to comment.