Skip to content

Conversation

@MichaelLettrich
Copy link
Collaborator

The double precision overloads for GPU functions (and the respective, calling functions) were renamed with a trailing d to ensure that they are not used by accident.

@davidrohr
Copy link
Collaborator

Great, now this runs into a problem with templates in OpenCL. Tried to apply a fix, let's seee...

@davidrohr
Copy link
Collaborator

Shit, still not working, will try this locally later this week and provide a fix then. Sorry for this OpenCL mess. Have to file a bug report to clang about it. Last time they rejected a similar report because their behavior was "according to the OpenCL spec" but this behavior is a mess and in my opinion just broken.
For reference: The problem are the OpenCL address spaces, and it would in principle require different overloads depending on whether the input comes from global memory or from a local variable, and any combination thereof for the input parameters...

@davidrohr davidrohr merged commit 949d94f into AliceO2Group:dev Dec 3, 2020
@MichaelLettrich MichaelLettrich deleted the O2-1702-o2-mathutils branch February 9, 2021 09:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

2 participants