Skip to content

Enable Global Intermediate Buffers#325

Merged
rdspring1 merged 5 commits into20_8_18_develfrom
rds_global_intermediate
Sep 2, 2020
Merged

Enable Global Intermediate Buffers#325
rdspring1 merged 5 commits into20_8_18_develfrom
rds_global_intermediate

Conversation

@rdspring1
Copy link
Copy Markdown
Collaborator

@rdspring1 rdspring1 commented Aug 26, 2020 •
edited
Loading

  • When the user designates a intermediate TensorView with the Global memory type, the fuser will pass the intermediate value as an argument to the kernel via global memory.
  • Enable support for the default schedule.

@rdspring1 rdspring1 mentioned this pull request Aug 26, 2020
Closed
@rdspring1 rdspring1 force-pushed the rds_global_intermediate branch from 453fd0a to 9ab2f8e Compare August 26, 2020 04:29
@rdspring1 rdspring1 requested a review from csarofeen August 26, 2020 06:50
@rdspring1 rdspring1 changed the title [WIP] Enable Global Intermediate Buffers Enable Global Intermediate Buffers Aug 26, 2020
@rdspring1
Copy link
Copy Markdown
Collaborator Author

rdspring1 commented Aug 26, 2020

I made changes so that local memory is the default type and the memory type matches between the TensorView and its corresponding allocate node.

@csarofeen
Copy link
Copy Markdown
Owner

csarofeen commented Aug 27, 2020

Could you also add/check there are couple things:
In addOutput/Input in fusion add a check to make sure they're in global memory
In the function to change memory type of a TV make sure we don't change a TV registered as input/output to anything besides global

@rdspring1 rdspring1 force-pushed the rds_global_intermediate branch from 408ace1 to 9a9a8d3 Compare August 27, 2020 17:07
csarofeen
csarofeen approved these changes Sep 1, 2020
View reviewed changes
Comment thread torch/csrc/jit/codegen/cuda/fusion.cpp Outdated
Comment thread torch/csrc/jit/codegen/cuda/fusion.cpp Outdated
@rdspring1 rdspring1 force-pushed the rds_global_intermediate branch from 9a9a8d3 to 243cb52 Compare September 2, 2020 01:57
@rdspring1 rdspring1 merged commit 4ec6d5a into 20_8_18_devel Sep 2, 2020
@csarofeen csarofeen deleted the rds_global_intermediate branch June 9, 2021 13:49
jjsjann123 pushed a commit that referenced this pull request Jun 8, 2022
@swang392 @pytorchmergebot
Add functionality for rebasebot to rebase onto viable/strict branch (p…
bbb2964 
…ytorch#78276)

Fixes #325
**Summary**: Currently, the pytorchbot only allows for rebasing to the master branch. These modifications add functionality for rebasing to the 'viable/strict' branch of pytorch/pytorch by adding a flag to the comment.
**Test Plan:** tested manually on personal fork ([#1](swang392#1)), and included a test case in test_tryrebase.py that checks if rebasing to viable/strict branch was successful.
Pull Request resolved: pytorch#78276
Approved by: https://github.com/clee2000, https://github.com/janeyx99
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@csarofeen csarofeen csarofeen approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@rdspring1 @csarofeen