Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[TensorIR][M1a] Introduce Block and BlockRealize #7553

Merged
merged 1 commit into from
Mar 1, 2021
Merged

[TensorIR][M1a] Introduce Block and BlockRealize #7553

merged 1 commit into from
Mar 1, 2021

Conversation

Hzfengsy
Copy link
Member

@Hzfengsy Hzfengsy commented Mar 1, 2021 •
edited
Loading

I'm happy to start upstream TensorIR from this PR.
TensorIR is a new low-level IR with full scheduling support. Each schedule primitive is a transformation from IR to IR. To get more detailed information please refer to the previous RFC: https://discuss.tvm.apache.org/t/rfc-tensorir-a-schedulable-ir-for-tvm/7872.

Since this project is large, we'd like to upstreaming with several PRs, full progress tracking should be here #7527:

  • Introduce core data structure (this PR)
  • Add tvmscript support for TensorIR data structure
  • Introduce basic schedule state
  • Add schedule primitives one by one

co-authors: @spectrometerHBH @junrushao1994 @tqchen @jinhongyii @MasterJH5574 and @vinx13

@junrushao junrushao self-assigned this Mar 1, 2021
@junrushao junrushao mentioned this pull request Mar 1, 2021
Closed
29 tasks
@Hzfengsy @spectrometerHBH
@junrushao @tqchen @MasterJH5574 @jinhongyii @vinx13
[TensorIR] introduce Block and BlockRealize (apache#312)
a61828c 
Co-authored-by: Bohan Hou <[email protected]>
Co-authored-by: Junru Shao <[email protected]>
Co-authored-by: Tianqi Chen <[email protected]>
Co-authored-by: Ruihang Lai <[email protected]>
Co-authored-by: Hongyi Jin <[email protected]>
Co-authored-by: Wuwei Lin <[email protected]>
comaniac
comaniac approved these changes Mar 1, 2021
View reviewed changes
Copy link
Contributor

@comaniac comaniac left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM in terms of the new added nodes. It might be clearer when new schedule primitives are sent.

@comaniac comaniac merged commit 057a673 into apache:main Mar 1, 2021
@comaniac
Copy link
Contributor

comaniac commented Mar 1, 2021

Thanks @Hzfengsy @MasterJH5574 @jinhongyii @tkonolige @junrushao1994 @vinx13 @tqchen

Lokiiiiii pushed a commit to Lokiiiiii/tvm that referenced this pull request Mar 2, 2021
@Hzfengsy @spectrometerHBH
@junrushao @tqchen @MasterJH5574 @jinhongyii @vinx13 @Lokiiiiii
[TensorIR] introduce Block and BlockRealize (apache#312) (apache#7553)
1ece8c9 
Co-authored-by: Bohan Hou <[email protected]>
Co-authored-by: Junru Shao <[email protected]>
Co-authored-by: Tianqi Chen <[email protected]>
Co-authored-by: Ruihang Lai <[email protected]>
Co-authored-by: Hongyi Jin <[email protected]>
Co-authored-by: Wuwei Lin <[email protected]>

Co-authored-by: Bohan Hou <[email protected]>
Co-authored-by: Junru Shao <[email protected]>
Co-authored-by: Tianqi Chen <[email protected]>
Co-authored-by: Ruihang Lai <[email protected]>
Co-authored-by: Hongyi Jin <[email protected]>
Co-authored-by: Wuwei Lin <[email protected]>
trevor-m pushed a commit to neo-ai/tvm that referenced this pull request Mar 2, 2021
@Hzfengsy @spectrometerHBH
@junrushao @tqchen @MasterJH5574 @jinhongyii @vinx13 @trevor-m
[TensorIR] introduce Block and BlockRealize (apache#312) (apache#7553)
0bcf632 
Co-authored-by: Bohan Hou <[email protected]>
Co-authored-by: Junru Shao <[email protected]>
Co-authored-by: Tianqi Chen <[email protected]>
Co-authored-by: Ruihang Lai <[email protected]>
Co-authored-by: Hongyi Jin <[email protected]>
Co-authored-by: Wuwei Lin <[email protected]>

Co-authored-by: Bohan Hou <[email protected]>
Co-authored-by: Junru Shao <[email protected]>
Co-authored-by: Tianqi Chen <[email protected]>
Co-authored-by: Ruihang Lai <[email protected]>
Co-authored-by: Hongyi Jin <[email protected]>
Co-authored-by: Wuwei Lin <[email protected]>
@junrushao
Copy link
Member

@jroesch and I thought a bit about the class name MatchBufferRegion, which (as we all know) is a bit confusing. We might want to think about potential alternatives, like:

  • AliasBufferAtRegion
  • BufferRegionAlias
  • AliasAt
  • etc

Perhaps we should keep the PR open for like one more day after we approve the PR for more discussion?

@tqchen
Copy link
Member

tqchen commented Mar 3, 2021

Happy to discuss naming a bit further, however. This particular case, this is indeed a match instead of a simple alias, and we do not plan to introduce the concept of alias/view into the IR

@junrushao junrushao linked an issue Mar 9, 2021 that may be closed by this pull request
[RFC][Tracking Issue] TensorIR Scheduling #7527
Closed
29 tasks
@junrushao junrushao changed the title [TensorIR] introduce Block and BlockRealize [TensorIR][M1a] Introduce Block and BlockRealize Apr 16, 2021
@junrushao junrushao mentioned this pull request Nov 1, 2021
Closed
@cloud-mxd cloud-mxd mentioned this pull request Mar 2, 2023
Merged
@Hzfengsy Hzfengsy deleted the tensorir branch February 2, 2024 08:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@tkonolige tkonolige tkonolige left review comments

@tqchen tqchen tqchen approved these changes

@vinx13 vinx13 vinx13 approved these changes

@comaniac comaniac comaniac approved these changes

@junrushao junrushao junrushao approved these changes

@jinhongyii jinhongyii jinhongyii approved these changes

@MasterJH5574 MasterJH5574 MasterJH5574 approved these changes

Assignees

@junrushao junrushao

Labels
status: accepted
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[RFC][Tracking Issue] TensorIR Scheduling
8 participants
@Hzfengsy @comaniac @junrushao @tqchen @tkonolige @vinx13 @jinhongyii @MasterJH5574