Skip to content

[inductor] Refactor LoopBody.memory_usage#135286

Closed
jansel wants to merge 6 commits intogh/jansel/388/basefrom
gh/jansel/388/head
Closed

[inductor] Refactor LoopBody.memory_usage#135286
jansel wants to merge 6 commits intogh/jansel/388/basefrom
gh/jansel/388/head

Conversation

[ghstack-poisoned]
@pytorch-bot
Copy link

pytorch-bot bot commented Sep 6, 2024

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/135286

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 Cancelled Job

As of commit 02b614a with merge base e72ed47 (image):

CANCELLED JOB - The following job was cancelled. Please retry:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

jansel added a commit that referenced this pull request Sep 6, 2024
ghstack-source-id: ef9a9f2
Pull Request resolved: #135286
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
pytorchmergebot pushed a commit that referenced this pull request Sep 8, 2024
This was (almost) unused.

Pull Request resolved: #135306
Approved by: https://github.com/oulgen
ghstack dependencies: #135286
pytorchmergebot pushed a commit that referenced this pull request Sep 8, 2024
Might be a small compile time improvement since we remove a call to extract_read_writes().

Pull Request resolved: #135400
Approved by: https://github.com/oulgen
ghstack dependencies: #135286, #135306, #135377
pytorchmergebot pushed a commit that referenced this pull request Sep 8, 2024
Before this we would take multiple passes over the body of each IRNode as we did lowering.  This combines most analysis into `OpCounterCSE` so it can be done in a single pass.

Before:
![image](https://github.com/user-attachments/assets/0047db09-4258-4491-a9a6-b078e183092a)

After:
![image](https://github.com/user-attachments/assets/1e03adcb-8303-4bb1-8bbb-cc42dacd44d7)

This stack:
![image](https://github.com/user-attachments/assets/d6b50b24-c30c-4d23-8b1a-344b3ba65d7a)

Pull Request resolved: #135412
Approved by: https://github.com/oulgen
ghstack dependencies: #135286, #135306, #135377, #135400
Chao1Han pushed a commit to Chao1Han/pytorch that referenced this pull request Sep 20, 2024
This is preparing for some other changes where I speed up extract_read_writes tracing.

Pull Request resolved: pytorch#135286
Approved by: https://github.com/oulgen
Chao1Han pushed a commit to Chao1Han/pytorch that referenced this pull request Sep 20, 2024
This was (almost) unused.

Pull Request resolved: pytorch#135306
Approved by: https://github.com/oulgen
ghstack dependencies: pytorch#135286
Chao1Han pushed a commit to Chao1Han/pytorch that referenced this pull request Sep 20, 2024
Might be a small compile time improvement since we remove a call to extract_read_writes().

Pull Request resolved: pytorch#135400
Approved by: https://github.com/oulgen
ghstack dependencies: pytorch#135286, pytorch#135306, pytorch#135377
Chao1Han pushed a commit to Chao1Han/pytorch that referenced this pull request Sep 20, 2024
Before this we would take multiple passes over the body of each IRNode as we did lowering.  This combines most analysis into `OpCounterCSE` so it can be done in a single pass.

Before:
![image](https://github.com/user-attachments/assets/0047db09-4258-4491-a9a6-b078e183092a)

After:
![image](https://github.com/user-attachments/assets/1e03adcb-8303-4bb1-8bbb-cc42dacd44d7)

This stack:
![image](https://github.com/user-attachments/assets/d6b50b24-c30c-4d23-8b1a-344b3ba65d7a)

Pull Request resolved: pytorch#135412
Approved by: https://github.com/oulgen
ghstack dependencies: pytorch#135286, pytorch#135306, pytorch#135377, pytorch#135400
@github-actions github-actions bot deleted the gh/jansel/388/head branch October 9, 2024 02:05
KnAwnime pushed a commit to KnAwnime/Biblioteka that referenced this pull request Oct 16, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants