NVIDIA / NVIDIA/cutlass

[QST] Is it possible to create non-contiguous counting tensor?

Open
#1,747 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

? - Needs Triage inactive-30d inactive-90d question
Dominant language
C++
Stars
10.5k
Forks
2.1k
Avg merge
3d 11h
Merged PRs (30d)
7

Description

What is your question?

First of all, thank you for the awesome project!

Regarding the make_counting_tensor, is it possible to create non-contiguous iterator?

For example,

auto tensor = make_tensor(make_inttuple_iter(15, 0), make_layout(make_shape(4, 3), make_stride(E<1>{}, E<0>{})));

creates

ArithTuple(15,0) o (4,3):(_1@1,_1@0):
  (15,0)  (16,0)  (17,0)
  (15,1)  (16,1)  (17,1)
  (15,2)  (16,2)  (17,2)
  (15,3)  (16,3)  (17,3)

I am trying to find whether is there any way to extend an iterator like

  (15,0)  (16,0)  (17,0)  (20, 0)  (21, 0)  (38, 0), ...
  (15,1)  (16,1)  (17,1)  (20, 1)  (21, 1)  (38, 1), ...
  (15,2)  (16,2)  (17,2)  (20, 2)  (21, 2)  (38, 2), ...
  (15,3)  (16,3)  (17,3)  (20, 3)  (21, 3)  (38, 3), ...

, but I haven't figured out yet.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing the named make_counting_tensor, make_tensor, make_inttuple_iter, make_layout, make_shape, and make_stride APIs to determine whether they support the requested non-contiguous sequence. Confirm the intended layout and iterator semantics with existing tests or examples, then establish whether a supported construction or a new feature is needed.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
backend
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.