[nvFuser] Improving bitwise ops support by zasdfgbnm · Pull Request #77158 · pytorch/pytorch

zasdfgbnm · 2022-05-10T09:32:17Z

Some renaming to better match PyTorch API:
- lshift -> bitwise_left_shift
- rshift -> bitwise_right_shift
- andOp -> bitwise_and
- orOp -> bitwise_or
- xorOp -> bitwise_xor
- notOp -> bitwise_not
Fix type inferences and type checking of these ops
Add bitwise_* to parser and python frontend
Improve test coverage

facebook-github-bot · 2022-05-10T09:32:23Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/77158
📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓Need help or want to give feedback on the CI? Visit our office hours

❌ 2 New Failures

As of commit c4b065b (more details on the Dr. CI page):

Expand to see more

2/2 failures introduced in this PR

🕵️ 2 new failures recognized by patterns

The following CI failures do not appear to be due to upstream breakages

pull / linux-xenial-py3.7-gcc5.4 / test (backwards_compat, 1, 1, linux.2xlarge) (1/2)

Step: "Test" (full log | diagnosis details | 🔁 rerun)

2022-05-17T01:42:05.0687053Z The PR is introduc...m to confirm whether this change is wanted or not.

2022-05-17T01:42:05.0673606Z processing existing schema:  text(__torch__.torch.classes.profiling.SourceRef _0) -> (str _0)
2022-05-17T01:42:05.0675168Z processing existing schema:  count(__torch__.torch.classes.profiling.InstructionStats _0) -> (int _0)
2022-05-17T01:42:05.0676204Z processing existing schema:  duration_ns(__torch__.torch.classes.profiling.InstructionStats _0) -> (int _0)
2022-05-17T01:42:05.0677450Z processing existing schema:  source(__torch__.torch.classes.profiling.SourceStats _0) -> (__torch__.torch.classes.profiling.SourceRef _0)
2022-05-17T01:42:05.0679557Z processing existing schema:  line_map(__torch__.torch.classes.profiling.SourceStats _0) -> (Dict(int, __torch__.torch.classes.profiling.InstructionStats) _0)
2022-05-17T01:42:05.0680298Z processing existing schema:  __init__(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-05-17T01:42:05.0682679Z processing existing schema:  enable(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-05-17T01:42:05.0683394Z processing existing schema:  disable(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-05-17T01:42:05.0684964Z processing existing schema:  _dump_stats(__torch__.torch.classes.profiling._ScriptProfile _0) -> (__torch__.torch.classes.profiling.SourceStats[] _0)
2022-05-17T01:42:05.0686246Z processing existing schema:  __init__(__torch__.torch.classes.dist_rpc.WorkerInfo _0, str _1, int _2) -> (NoneType _0)
2022-05-17T01:42:05.0687053Z The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not. 
2022-05-17T01:42:05.0687359Z 
2022-05-17T01:42:05.0687418Z Broken ops: [
2022-05-17T01:42:05.0687666Z 	aten::lift(Tensor self) -> (Tensor)
2022-05-17T01:42:05.0687947Z 	aten::ccol_indices(Tensor(a) self) -> (Tensor(a))
2022-05-17T01:42:05.0688239Z 	aten::ccol_indices_copy(Tensor self) -> (Tensor)
2022-05-17T01:42:05.0688637Z 	aten::index_reduce(Tensor self, int dim, Tensor index, Tensor source, str reduce, *, bool include_self=True) -> (Tensor)
2022-05-17T01:42:05.0689150Z 	aten::index_reduce.out(Tensor self, int dim, Tensor index, Tensor source, str reduce, *, bool include_self=True, Tensor(a!) out) -> (Tensor(a!))
2022-05-17T01:42:05.0689658Z 	aten::index_reduce_(Tensor(a!) self, int dim, Tensor index, Tensor source, str reduce, *, bool include_self=True) -> (Tensor(a!))
2022-05-17T01:42:05.0690050Z 	aten::glu_jvp(Tensor glu, Tensor x, Tensor dx, int dim) -> (Tensor)
2022-05-17T01:42:05.0690437Z 	aten::_sparse_addmm(Tensor self, Tensor mat1, Tensor mat2, *, Scalar beta=1, Scalar alpha=1) -> (Tensor)

pull / pytorch-xla-linux-bionic-py3.7-clang8 / test (xla, 1, 1, linux.2xlarge) (2/2)

Step: "Test" (full log | diagnosis details | 🔁 rerun)

2022-05-17T01:48:18.4963839Z TypeError: run_gen...got an unexpected keyword argument 'get_device_fn'

2022-05-17T01:48:16.4415776Z + python setup.py install
2022-05-17T01:48:17.3548963Z No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda'
2022-05-17T01:48:17.3653713Z Building torch_xla version: 1.12
2022-05-17T01:48:17.3654192Z XLA Commit ID: 90a525b5333b1a67ae22fc2df5fe01ceceb30a5e
2022-05-17T01:48:17.3654630Z PyTorch Commit ID: c4b065b3d0608708dce580abcb616a97a8049567
2022-05-17T01:48:17.3715019Z /var/lib/jenkins/workspace /var/lib/jenkins/workspace/xla
2022-05-17T01:48:18.4048165Z /var/lib/jenkins/workspace/xla
2022-05-17T01:48:18.4962304Z Traceback (most recent call last):
2022-05-17T01:48:18.4962876Z   File "/var/lib/jenkins/workspace/xla/scripts/gen_lazy_tensor.py", line 84, in <module>
2022-05-17T01:48:18.4963314Z     get_device_fn="torch_xla::bridge::GetXlaDevice")
2022-05-17T01:48:18.4963839Z TypeError: run_gen_lazy_tensor() got an unexpected keyword argument 'get_device_fn'
2022-05-17T01:48:18.5037181Z Failed to generate lazy files: ['python', '/var/lib/jenkins/workspace/xla/scripts/gen_lazy_tensor.py']
2022-05-17T01:48:18.6592827Z + cleanup
2022-05-17T01:48:18.6593097Z + retcode=1
2022-05-17T01:48:18.6593323Z + set +x
2022-05-17T01:48:18.6625182Z ##[error]Process completed with exit code 1.
2022-05-17T01:48:18.6732506Z ##[group]Run pytorch/pytorch/.github/actions/get-workflow-job-id@master
2022-05-17T01:48:18.6732743Z with:
2022-05-17T01:48:18.6733344Z   github-token: ***
2022-05-17T01:48:18.6733523Z env:
2022-05-17T01:48:18.6733865Z   IN_CI: 1

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

test/test_jit_cuda_fuser.py

zasdfgbnm · 2022-05-10T09:37:13Z

Looks like GitHub does not allow me to request review from @csarofeen either

zasdfgbnm · 2022-05-10T17:51:54Z

@pytorchbot retest this please

…tend-bitwise

kevinstephano

Looks good!

jjsjann123

Stamping but feel free to update the tests with the cache configure if you want.

test/test_jit_cuda_fuser.py

torch/csrc/jit/codegen/cuda/arith.cpp

zasdfgbnm · 2022-05-17T01:48:58Z

@ngimel I have updated this PR to match with the updated eager mode behavior in #77621

zasdfgbnm · 2022-05-18T17:19:55Z

@pytorchbot merge this please

github-actions · 2022-05-18T17:23:07Z

Hey @zasdfgbnm.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.

Summary: - Some renaming to better match PyTorch API: - `lshift` -> `bitwise_left_shift` - `rshift` -> `bitwise_right_shift` - `andOp` -> `bitwise_and` - `orOp` -> `bitwise_or` - `xorOp` -> `bitwise_xor` - `notOp` -> `bitwise_not` - Fix type inferences and type checking of these ops - Add `bitwise_*` to parser and python frontend - Improve test coverage Pull Request resolved: #77158 Approved by: https://github.com/kevinstephano, https://github.com/jjsjann123 Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/4eec865f5802f2c34eb34f9f89165f0aa4b5502f Reviewed By: seemethere Differential Revision: D36494215 Pulled By: seemethere fbshipit-source-id: 4985890b93046c53f597ccd3fce2bebcdf07b19c

- Some renaming to better match PyTorch API: - `lshift` -> `bitwise_left_shift` - `rshift` -> `bitwise_right_shift` - `andOp` -> `bitwise_and` - `orOp` -> `bitwise_or` - `xorOp` -> `bitwise_xor` - `notOp` -> `bitwise_not` - Fix type inferences and type checking of these ops - Add `bitwise_*` to parser and python frontend - Improve test coverage Pull Request resolved: pytorch#77158 Approved by: https://github.com/kevinstephano, https://github.com/jjsjann123

- Some renaming to better match PyTorch API: - `lshift` -> `bitwise_left_shift` - `rshift` -> `bitwise_right_shift` - `andOp` -> `bitwise_and` - `orOp` -> `bitwise_or` - `xorOp` -> `bitwise_xor` - `notOp` -> `bitwise_not` - Fix type inferences and type checking of these ops - Add `bitwise_*` to parser and python frontend - Improve test coverage Pull Request resolved: pytorch#77158 Approved by: https://github.com/kevinstephano, https://github.com/jjsjann123 Co-authored-by: Xiang Gao <qasdfgtyuiop@gmail.com>

Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/ Code changes includes: - TransformPropagator refactor: switched to Dijkstra instead of exhaustive enumeration on all possible paths to reduce compilation time on transform propagation; - Indexing refactor: remove reference tensor creation in all tensor indexing logic (#1690) - (more) generic grouped grid reduction kernel; - Minor parser/fuser patches: 1. zero-dim tensor reduction support 3. no-op binary removal within fused graph 4. expand supported in fusion Squashed commits to WAR github API Commits that's actually in this PR from the devel branch: ``` a054b3e Refactor TransormPropagator to allow specifying a position and propagating to part of the DAG (#1775) d67e1cd Indexing refactor stage 1: remove reference tensor creation in all tensor indexing logic (#1690) 1b65299 Issue 1770 (#1774) 35b0427 Avoid compilation errors like below: (#1773) 452c773 Ignore reductions of zero-dim tensors per PyTorch conventions (#1771) 31d6c56 TransformPropagator refactor (#1769) 570c5a8 Merge pull request #1767 from csarofeen/upstream_merge_0621 9d6c3d8 merging upstream 61305cd 0ed815f New TransformPropagator algorithm (#1763) 6c19520 no-op binary removal (#1764) ec7fa41 Proper propagation of IterType (#1762) b263562 Fix dimensionality check (#1759) 2d6343f More generic grouped grid reduction kernel (#1740) 64e2b56 [nvfuser] prevent spamming warning message (#77777) (#1758) 0c43162 [nvFuser] Improving bitwise ops support (#77158) (#1757) b93a147 Parser expand (#1754) ``` RUN_TORCHBENCH: nvfuser Pull Request resolved: #80355 Approved by: https://github.com/davidberard98

Summary: Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/ Code changes includes: - TransformPropagator refactor: switched to Dijkstra instead of exhaustive enumeration on all possible paths to reduce compilation time on transform propagation; - Indexing refactor: remove reference tensor creation in all tensor indexing logic (#1690) - (more) generic grouped grid reduction kernel; - Minor parser/fuser patches: 1. zero-dim tensor reduction support 3. no-op binary removal within fused graph 4. expand supported in fusion Squashed commits to WAR github API Commits that's actually in this PR from the devel branch: ``` a054b3e Refactor TransormPropagator to allow specifying a position and propagating to part of the DAG (#1775) d67e1cd Indexing refactor stage 1: remove reference tensor creation in all tensor indexing logic (#1690) 1b65299 Issue 1770 (#1774) 35b0427 Avoid compilation errors like below: (#1773) 452c773 Ignore reductions of zero-dim tensors per PyTorch conventions (#1771) 31d6c56 TransformPropagator refactor (#1769) 570c5a8 Merge pull request #1767 from csarofeen/upstream_merge_0621 9d6c3d8 merging upstream 61305cd 0ed815f New TransformPropagator algorithm (#1763) 6c19520 no-op binary removal (#1764) ec7fa41 Proper propagation of IterType (#1762) b263562 Fix dimensionality check (#1759) 2d6343f More generic grouped grid reduction kernel (#1740) 64e2b56 [nvfuser] prevent spamming warning message (#77777) (#1758) 0c43162 [nvFuser] Improving bitwise ops support (#77158) (#1757) b93a147 Parser expand (#1754) ``` RUN_TORCHBENCH: nvfuser Pull Request resolved: #80355 Reviewed By: qihqi Differential Revision: D37573400 Pulled By: davidberard98 fbshipit-source-id: 52ab68d89ec01ef61f69f5abeb18c9d3a312aa64

- Some renaming to better match PyTorch API: - `lshift` -> `bitwise_left_shift` - `rshift` -> `bitwise_right_shift` - `andOp` -> `bitwise_and` - `orOp` -> `bitwise_or` - `xorOp` -> `bitwise_xor` - `notOp` -> `bitwise_not` - Fix type inferences and type checking of these ops - Add `bitwise_*` to parser and python frontend - Improve test coverage Pull Request resolved: pytorch/pytorch#77158 Approved by: https://github.com/kevinstephano, https://github.com/jjsjann123

- Some renaming to better match PyTorch API: - `lshift` -> `bitwise_left_shift` - `rshift` -> `bitwise_right_shift` - `andOp` -> `bitwise_and` - `orOp` -> `bitwise_or` - `xorOp` -> `bitwise_xor` - `notOp` -> `bitwise_not` - Fix type inferences and type checking of these ops - Add `bitwise_*` to parser and python frontend - Improve test coverage Pull Request resolved: pytorch/pytorch#77158 Approved by: https://github.com/kevinstephano, https://github.com/jjsjann123 Co-authored-by: Xiang Gao <qasdfgtyuiop@gmail.com>

zasdfgbnm added 3 commits May 10, 2022 02:24

Improving bitwise ops support

a8df5f1

breakpad

b374a0a

save

5f54e07

facebook-github-bot added the cla signed label May 10, 2022

facebook-github-bot added the oncall: jit Add this issue/PR to JIT oncall triage queue label May 10, 2022

zasdfgbnm changed the title ~~Improving bitwise ops support~~ [nvFuser] Improving bitwise ops support May 10, 2022

zasdfgbnm commented May 10, 2022

View reviewed changes

test/test_jit_cuda_fuser.py Show resolved Hide resolved

zasdfgbnm commented May 10, 2022

View reviewed changes

test/test_jit_cuda_fuser.py Show resolved Hide resolved

zasdfgbnm requested review from jjsjann123, kevinstephano and mruberry May 10, 2022 09:36

lint

7f3362b

zasdfgbnm mentioned this pull request May 10, 2022

Implement atan2 and bitwise_{xor, and, or, not} operations for python frontend. csarofeen/pytorch#1632

Closed

pytorchbot added the open source label May 10, 2022

samdow added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label May 10, 2022

Merge branch 'master' of github.com:pytorch/pytorch into nvfuser-fron…

ccdf82d

…tend-bitwise

kevinstephano approved these changes May 12, 2022

View reviewed changes

jjsjann123 approved these changes May 13, 2022

View reviewed changes

ngimel reviewed May 13, 2022

View reviewed changes

test/test_jit_cuda_fuser.py Outdated Show resolved Hide resolved

torch/csrc/jit/codegen/cuda/arith.cpp Outdated Show resolved Hide resolved

torch/csrc/jit/codegen/cuda/arith.cpp Outdated Show resolved Hide resolved

torch/csrc/jit/codegen/cuda/arith.cpp Outdated Show resolved Hide resolved

fix

c4b065b

zasdfgbnm requested a review from ngimel May 17, 2022 01:50

pytorchmergebot added the Merged label May 18, 2022

pytorchmergebot closed this in 4eec865 May 18, 2022

zasdfgbnm deleted the nvfuser-frontend-bitwise branch May 18, 2022 17:28

jjsjann123 mentioned this pull request Jun 8, 2022

[nvFuser] Improving bitwise ops support (#77158) csarofeen/pytorch#1757

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[nvFuser] Improving bitwise ops support#77158

[nvFuser] Improving bitwise ops support#77158
zasdfgbnm wants to merge 6 commits intomasterfrom
nvfuser-frontend-bitwise

zasdfgbnm commented May 10, 2022 •

edited

Loading

Uh oh!

facebook-github-bot commented May 10, 2022 •

edited

Loading

🕵️ 2 new failures recognized by patterns

pull / linux-xenial-py3.7-gcc5.4 / test (backwards_compat, 1, 1, linux.2xlarge) (1/2)

pull / pytorch-xla-linux-bionic-py3.7-clang8 / test (xla, 1, 1, linux.2xlarge) (2/2)

Uh oh!

Uh oh!

Uh oh!

zasdfgbnm commented May 10, 2022

Uh oh!

zasdfgbnm commented May 10, 2022

Uh oh!

kevinstephano left a comment

Uh oh!

jjsjann123 left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

zasdfgbnm commented May 17, 2022

Uh oh!

zasdfgbnm commented May 18, 2022

Uh oh!

github-actions bot commented May 18, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

Conversation

zasdfgbnm commented May 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented May 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

❌ 2 New Failures

🕵️ 2 new failures recognized by patterns

pull / linux-xenial-py3.7-gcc5.4 / test (backwards_compat, 1, 1, linux.2xlarge) (1/2)

pull / pytorch-xla-linux-bionic-py3.7-clang8 / test (xla, 1, 1, linux.2xlarge) (2/2)

Uh oh!

Uh oh!

Uh oh!

zasdfgbnm commented May 10, 2022

Uh oh!

zasdfgbnm commented May 10, 2022

Uh oh!

kevinstephano left a comment

Choose a reason for hiding this comment

Uh oh!

jjsjann123 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

zasdfgbnm commented May 17, 2022

Uh oh!

zasdfgbnm commented May 18, 2022

Uh oh!

github-actions bot commented May 18, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

zasdfgbnm commented May 10, 2022 •

edited

Loading

facebook-github-bot commented May 10, 2022 •

edited

Loading