Skip to content

Commit

Permalink
Remove cudaMemAccessFlagsProtRead as it is not supported by mem pools
Browse files Browse the repository at this point in the history
  • Loading branch information
abellina committed Dec 3, 2024
1 parent c50f7fa commit c2ea2cb
Showing 1 changed file with 3 additions and 3 deletions.
6 changes: 3 additions & 3 deletions include/rmm/mr/device/cuda_async_memory_resource.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -68,12 +68,12 @@ class cuda_async_memory_resource final : public device_memory_resource {
* @brief Flags for specifying the memory pool accessibility from other devices.
*
* @note The default, `prot_none`, marks the pool's memory as private to the device in
* which it was created. `prot_read` and `prot_read_write` should only be used if memory
* sharing among devices is required.
* which it was created. `prod_read_write` should only be used if memory sharing among
* devices is required. Note that there is a `cudaMemAccessFlagsProtRead` documented,
* but memory pools don't support read-only, so it has been omitted.
*/
enum class access_flags {
prot_none = 0, ///< Default, make pool not accessible. (cudaMemAccessFlagsProtNone)
prot_read = 1, ///< Make pool read accessible. (cudaMemAccessFlagsProtRead)
prot_read_write = 3 ///< Make pool read-write accessible. (cudaMemAccessFlagsProtReadWrite)
};

Expand Down

0 comments on commit c2ea2cb

Please sign in to comment.