Enable Thrust algorithms from CUDA __device__ Functions #140

jaredhoberock opened this Issue Jun 4, 2012 · 1 comment


None yet
1 participant

jaredhoberock commented Jun 4, 2012

In theory, CUDA dynamic parallelism will enable calling Thrust functions from inside __device__ functions.

Some corollaries:

  • Thrust functors should also be able to invoke Thrust algorithms.
  • device_vector should become nestable

Thrust algorithms should sequentialize when a GPU does not support nested kernel launch.

Blocked by:


This comment has been minimized.

Show comment Hide comment

jaredhoberock Aug 6, 2014


Thrust 1.8 enables this.


jaredhoberock commented Aug 6, 2014

Thrust 1.8 enables this.

@jaredhoberock jaredhoberock removed the blocked label Aug 6, 2014

@jaredhoberock jaredhoberock added this to the Release1.8 milestone Aug 6, 2014

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment