You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Before we can worry about parallel ranges, we first need an implementation of the Standard <ranges> header and associated machinery that works in host and device code.
One major obstacle for heterogeneous usage of <ranges> is that views store their parent range as a member and the iterators hold a pointer to the parent. This has obvious limitations when an operation is launched from host side, as the view wont be accessible from device.
The initial implementation will assume homogeneous usage and explore how we can best adopt to heterogeneous usage once we have a solid way to test our design
Before we can worry about parallel ranges, we first need an implementation of the Standard
<ranges>
header and associated machinery that works in host and device code.One major obstacle for heterogeneous usage of
<ranges>
is that views store their parent range as a member and the iterators hold a pointer to the parent. This has obvious limitations when an operation is launched from host side, as the view wont be accessible from device.The initial implementation will assume homogeneous usage and explore how we can best adopt to heterogeneous usage once we have a solid way to test our design
<cuda/std/ranges>
and associated headers #61The text was updated successfully, but these errors were encountered: