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

[CodeGen]Support Convert NVVM IR to Cubin With LibDevice Linked #10200

Merged
merged 26 commits into from
Apr 27, 2023
Merged

[CodeGen]Support Convert NVVM IR to Cubin With LibDevice Linked #10200

merged 26 commits into from
Apr 27, 2023

Conversation

howin98
Copy link
Contributor

@howin98 howin98 commented Apr 26, 2023

No description provided.

@howin98
Copy link
Contributor Author

howin98 commented Apr 26, 2023

fix #10191

@howin98 howin98 requested a review from oneflow-ci-bot April 26, 2023 03:35
@github-actions
Copy link
Contributor

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

@github-actions
Copy link
Contributor

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

@github-actions
Copy link
Contributor

CI failed when running job: Build cu116. PR label automerge has been removed

@github-actions
Copy link
Contributor

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

@github-actions
Copy link
Contributor

CI failed when running job: Build cpu. PR label automerge has been removed

@github-actions
Copy link
Contributor

CI failed when running job: Build cpu. PR label automerge has been removed

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10200/

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: NVIDIA GeForce RTX 3080 Ti 

❌ OneFlow resnet50 time: 43.2ms (= 4322.7ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 57.2ms (= 5716.1ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.32 (= 57.2ms / 43.2ms)

OneFlow resnet50 time: 26.2ms (= 2623.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 37.6ms (= 3760.9ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.43 (= 37.6ms / 26.2ms)

OneFlow resnet50 time: 19.7ms (= 3948.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 36.4ms (= 7270.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.84 (= 36.4ms / 19.7ms)

OneFlow resnet50 time: 18.9ms (= 3784.0ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 32.0ms (= 6394.5ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.69 (= 32.0ms / 18.9ms)

OneFlow resnet50 time: 18.9ms (= 3783.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 30.4ms (= 6084.3ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.61 (= 30.4ms / 18.9ms)

OneFlow swin dataloader time: 0.202s (= 40.342s / 200, num_workers=1)
PyTorch swin dataloader time: 0.129s (= 25.718s / 200, num_workers=1)
Relative speed: 0.638 (= 0.129s / 0.202s)

OneFlow swin dataloader time: 0.053s (= 10.664s / 200, num_workers=4)
PyTorch swin dataloader time: 0.032s (= 6.468s / 200, num_workers=4)
Relative speed: 0.607 (= 0.032s / 0.053s)

OneFlow swin dataloader time: 0.031s (= 6.162s / 200, num_workers=8)
PyTorch swin dataloader time: 0.017s (= 3.345s / 200, num_workers=8)
Relative speed: 0.543 (= 0.017s / 0.031s)

❌ OneFlow resnet50 time: 47.5ms (= 4749.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 65.5ms (= 6552.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.38 (= 65.5ms / 47.5ms)

OneFlow resnet50 time: 31.6ms (= 3155.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 44.2ms (= 4417.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.40 (= 44.2ms / 31.6ms)

OneFlow resnet50 time: 24.0ms (= 4793.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 40.8ms (= 8168.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.70 (= 40.8ms / 24.0ms)

OneFlow resnet50 time: 22.3ms (= 4467.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 38.7ms (= 7748.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.73 (= 38.7ms / 22.3ms)

OneFlow resnet50 time: 21.4ms (= 4280.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 35.1ms (= 7021.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.64 (= 35.1ms / 21.4ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10200/

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: NVIDIA GeForce RTX 3090 

❌ OneFlow resnet50 time: 42.7ms (= 4269.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 57.7ms (= 5766.3ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.35 (= 57.7ms / 42.7ms)

OneFlow resnet50 time: 26.4ms (= 2637.2ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 37.6ms (= 3759.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.43 (= 37.6ms / 26.4ms)

OneFlow resnet50 time: 18.4ms (= 3670.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 35.3ms (= 7067.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.93 (= 35.3ms / 18.4ms)

OneFlow resnet50 time: 17.9ms (= 3577.4ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 31.0ms (= 6191.5ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.73 (= 31.0ms / 17.9ms)

OneFlow resnet50 time: 16.1ms (= 3212.9ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 28.8ms (= 5766.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.79 (= 28.8ms / 16.1ms)

OneFlow swin dataloader time: 0.201s (= 40.172s / 200, num_workers=1)
PyTorch swin dataloader time: 0.130s (= 26.005s / 200, num_workers=1)
Relative speed: 0.647 (= 0.130s / 0.201s)

OneFlow swin dataloader time: 0.054s (= 10.857s / 200, num_workers=4)
PyTorch swin dataloader time: 0.033s (= 6.697s / 200, num_workers=4)
Relative speed: 0.617 (= 0.033s / 0.054s)

OneFlow swin dataloader time: 0.031s (= 6.127s / 200, num_workers=8)
PyTorch swin dataloader time: 0.017s (= 3.450s / 200, num_workers=8)
Relative speed: 0.563 (= 0.017s / 0.031s)

❌ OneFlow resnet50 time: 49.0ms (= 4899.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 65.0ms (= 6496.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 65.0ms / 49.0ms)

OneFlow resnet50 time: 36.7ms (= 3666.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 46.8ms (= 4684.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 46.8ms / 36.7ms)

OneFlow resnet50 time: 28.7ms (= 5741.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 39.4ms (= 7873.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.37 (= 39.4ms / 28.7ms)

OneFlow resnet50 time: 25.6ms (= 5118.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 38.5ms (= 7708.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.51 (= 38.5ms / 25.6ms)

OneFlow resnet50 time: 24.5ms (= 4907.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 36.6ms (= 7316.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.49 (= 36.6ms / 24.5ms)

@mergify mergify bot merged commit a22327b into master Apr 27, 2023
@mergify mergify bot deleted the support-nvvm-to-cubin-serial-passes branch April 27, 2023 14:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jackalcooper jackalcooper jackalcooper left review comments

@mosout mosout mosout approved these changes

@BBuf BBuf BBuf approved these changes

@daquexian daquexian Awaiting requested review from daquexian

@hjchen2 hjchen2 Awaiting requested review from hjchen2 hjchen2 is a code owner

@oneflow-ci-bot oneflow-ci-bot Awaiting requested review from oneflow-ci-bot

Assignees
No one assigned
Labels
automerge feature ir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@howin98 @jackalcooper @mosout @BBuf @oneflow-ci-bot