NVIDIA / NVIDIA/cccl

Add an option to guarantee a static and runtime `max_total_num_items` to `DeviceBatchedTopK`

Open
#9,279 0 comments 0 reactions 1 assignee Claimed by @elstehle View on GitHub
Dominant language
C++
Stars
2.5k
Forks
486
Avg merge
2d 6h
Merged PRs (30d)
295

Description

## Description

`max_total_num_items` is needed to communicate an upper bound on the total number of items processed (e.g. the combined size of all segments in `cub::DeviceBatchedTopK`), which an algorithm can use to choose offset types (e.g., tile offset size) as well as sizing temporary storage buffers. Since this bound-information may not be attachable to a specific parameter (e.g., on a `DeviceBatchedTopK` and similarly for some other segmented algorithms that do not have a `number of items` parameter), we decided that for these cases it should go into the guarantees API.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.