Support Tan operator

[jikechao]

PR Category

Operator

Type of Change

New Feature

Description

Support a new operator Tan

Progress

• Change is properly reviewed (1 reviewer required, 2 recommended).
• Change is fully covered by a UT.

Performance

[gemini-code-assist]

Summary of Changes

Hello @jikechao, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request enhances the available mathematical operations by adding a new tan operator. The implementation leverages Triton's JIT capabilities to ensure efficient, dynamic pointwise computation of the tangent function, offering both a standard return and an in-place modification variant.

Highlights

• New Operator Support: Introduced support for the tan (tangent) mathematical operator.
• Triton Integration: Implemented the tan operator using Triton's JIT compilation for efficient, dynamic pointwise computation.
• In-place Operation: Provided both out-of-place (tan) and in-place (tan_) versions of the tangent function.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request adds support for the tan operator. The implementation is straightforward. I've suggested a simplification to use triton.language.tan directly, which should be more efficient and maintainable. As indicated in the pull request description, please remember to add unit tests for the new operator to ensure its correctness.

[jikechao]

cc @0x45f @kiddyjinjin @StrongSpoon @iclementine

[jikechao]

@0x45f @kiddyjinjin @StrongSpoon @iclementine @zhzhcookie
Hi all, could you help me review this PR? It seems that the error in the CI test is not related to this PR.

[0x45f]

@0x45f @kiddyjinjin @StrongSpoon @iclementine @zhzhcookie Hi all, could you help me review this PR? It seems that the error in the CI test is not related to this PR.

please merge master to fix code-format-check ci and add benchmark result in Performance

[sgjzfzzf]

Thank you for your contribution! Could u please also put the performance report in the PR comment? It would be helpful to reviewers.

[jikechao]

@0x45f @kiddyjinjin @StrongSpoon @iclementine @zhzhcookie Hi all, could you help me review this PR? It seems that the error in the CI test is not related to this PR.

please merge master to fix code-format-check ci and add benchmark result in Performance

@0x45f Thanks for your fixing. I have merged into the master branch

[jikechao]

Thank you for your contribution! Could u please also put the performance report in the PR comment? It would be helpful to reviewers.

@sgjzfzzf Sure, I'll provide the performance comparison sooner.

[CLAassistant]

All committers have signed the CLA.

[jikechao]

Operator: tan  Performance Test (dtype=torch.float16, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS               5.644288            6.661120               0.847               0.161          [torch.Size([1073741824])]
SUCCESS               0.004096            0.005120               0.800               0.001          [torch.Size([64, 64])]
SUCCESS               0.086016            0.086016               1.000               0.195          [torch.Size([4096, 4096])]
SUCCESS               0.086016            0.084992               1.012               0.197          [torch.Size([64, 512, 512])]
SUCCESS               5.646336            6.397952               0.883               0.168          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.005120            0.005120               1.000               0.000          [torch.Size([1024, 1])]
SUCCESS               0.005120            0.005120               1.000               0.003          [torch.Size([1024, 16])]
SUCCESS               0.006144            0.006144               1.000               0.043          [torch.Size([1024, 256])]
SUCCESS               0.025600            0.025600               1.000               0.164          [torch.Size([1024, 4096])]
SUCCESS               0.323584            0.330752               0.978               0.203          [torch.Size([1024, 65536])]
SUCCESS               0.004096            0.005120               0.800               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.009216            0.009216               1.000               0.114          [torch.Size([64, 64, 256])]
SUCCESS               0.086016            0.086016               1.000               0.195          [torch.Size([64, 64, 4096])]


Operator: tan  Performance Test (dtype=torch.float32, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS              10.129408           10.323968               0.981               0.104          [torch.Size([1073741824])]
SUCCESS               0.005120            0.004096               1.250               0.001          [torch.Size([64, 64])]
SUCCESS               0.163840            0.164864               0.994               0.102          [torch.Size([4096, 4096])]
SUCCESS               0.163840            0.165888               0.988               0.101          [torch.Size([64, 512, 512])]
SUCCESS              10.122240           10.384384               0.975               0.103          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.005120            0.005120               1.000               0.000          [torch.Size([1024, 1])]
SUCCESS               0.004096            0.005120               0.800               0.003          [torch.Size([1024, 16])]
SUCCESS               0.007168            0.007168               1.000               0.037          [torch.Size([1024, 256])]
SUCCESS               0.044032            0.045056               0.977               0.093          [torch.Size([1024, 4096])]
SUCCESS               0.641024            0.642048               0.998               0.105          [torch.Size([1024, 65536])]
SUCCESS               0.005120            0.004096               1.250               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.014336            0.014336               1.000               0.073          [torch.Size([64, 64, 256])]
SUCCESS               0.163840            0.165888               0.988               0.101          [torch.Size([64, 64, 4096])]


Operator: tan  Performance Test (dtype=torch.bfloat16, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS               5.520384            6.246400               0.884               0.172          [torch.Size([1073741824])]
SUCCESS               0.004096            0.004096               1.000               0.001          [torch.Size([64, 64])]
SUCCESS               0.084992            0.086016               0.988               0.195          [torch.Size([4096, 4096])]
SUCCESS               0.084992            0.084992               1.000               0.197          [torch.Size([64, 512, 512])]
SUCCESS               5.441536            6.245376               0.871               0.172          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.004096            0.004096               1.000               0.000          [torch.Size([1024, 1])]
SUCCESS               0.005120            0.005120               1.000               0.003          [torch.Size([1024, 16])]
SUCCESS               0.006144            0.006144               1.000               0.043          [torch.Size([1024, 256])]
SUCCESS               0.025600            0.025600               1.000               0.164          [torch.Size([1024, 4096])]
SUCCESS               0.323584            0.324608               0.997               0.207          [torch.Size([1024, 65536])]
SUCCESS               0.005120            0.004096               1.250               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.009216            0.010240               0.900               0.102          [torch.Size([64, 64, 256])]
SUCCESS               0.084992            0.086016               0.988               0.195          [torch.Size([64, 64, 4096])]

[jikechao]

@sgjzfzzf @0x45f I have uploaded the performance results. Could you help me review it?

[jikechao]

benchmark/test_unary_pointwise_perf.py
Operator: tan  Performance Test (dtype=torch.float16, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS               5.746688            6.739968               0.853               0.159          [torch.Size([1073741824])]
SUCCESS               0.005120            0.004096               1.250               0.001          [torch.Size([64, 64])]
SUCCESS               0.084992            0.086016               0.988               0.195          [torch.Size([4096, 4096])]
SUCCESS               0.084992            0.086016               0.988               0.195          [torch.Size([64, 512, 512])]
SUCCESS               5.646336            5.809152               0.972               0.185          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.005120            0.005120               1.000               0.000          [torch.Size([1024, 1])]
SUCCESS               0.004096            0.005120               0.800               0.003          [torch.Size([1024, 16])]
SUCCESS               0.006112            0.006144               0.995               0.043          [torch.Size([1024, 256])]
SUCCESS               0.025600            0.025600               1.000               0.164          [torch.Size([1024, 4096])]
SUCCESS               0.323584            0.324608               0.997               0.207          [torch.Size([1024, 65536])]
SUCCESS               0.004096            0.004096               1.000               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.009216            0.010240               0.900               0.102          [torch.Size([64, 64, 256])]
SUCCESS               0.084992            0.084992               1.000               0.197          [torch.Size([64, 64, 4096])]


Operator: tan  Performance Test (dtype=torch.float32, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS              10.116096           10.195968               0.992               0.105          [torch.Size([1073741824])]
SUCCESS               0.005120            0.005120               1.000               0.001          [torch.Size([64, 64])]
SUCCESS               0.163840            0.164864               0.994               0.102          [torch.Size([4096, 4096])]
SUCCESS               0.163840            0.164864               0.994               0.102          [torch.Size([64, 512, 512])]
SUCCESS              10.507264           10.196992               1.030               0.105          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.005120            0.004096               1.250               0.000          [torch.Size([1024, 1])]
SUCCESS               0.005120            0.004096               1.250               0.004          [torch.Size([1024, 16])]
SUCCESS               0.007168            0.007168               1.000               0.037          [torch.Size([1024, 256])]
SUCCESS               0.044032            0.044032               1.000               0.095          [torch.Size([1024, 4096])]
SUCCESS               0.641024            0.640000               1.002               0.105          [torch.Size([1024, 65536])]
SUCCESS               0.005120            0.005120               1.000               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.014336            0.014336               1.000               0.073          [torch.Size([64, 64, 256])]
SUCCESS               0.169984            0.165888               1.025               0.101          [torch.Size([64, 64, 4096])]


Operator: tan  Performance Test (dtype=torch.bfloat16, mode=kernel,level=comprehensive)
Status       Torch Latency (ms)    Gems Latency (ms)         Gems Speedup               TFLOPS          Size Detail
--------------------------------------------------------------------------------------------------------------------
SUCCESS               5.439488            5.735424               0.948               0.187          [torch.Size([1073741824])]
SUCCESS               0.004096            0.005120               0.800               0.001          [torch.Size([64, 64])]
SUCCESS               0.084992            0.084992               1.000               0.197          [torch.Size([4096, 4096])]
SUCCESS               0.084992            0.084992               1.000               0.197          [torch.Size([64, 512, 512])]
SUCCESS               5.441536            5.733376               0.949               0.187          [torch.Size([1024, 1024, 1024])]
SUCCESS               0.005120            0.005120               1.000               0.000          [torch.Size([1024, 1])]
SUCCESS               0.005120            0.004096               1.250               0.004          [torch.Size([1024, 16])]
SUCCESS               0.006144            0.006144               1.000               0.043          [torch.Size([1024, 256])]
SUCCESS               0.025600            0.024576               1.042               0.171          [torch.Size([1024, 4096])]
SUCCESS               0.323584            0.324608               0.997               0.207          [torch.Size([1024, 65536])]
SUCCESS               0.004096            0.004096               1.000               0.001          [torch.Size([64, 64, 1])]
SUCCESS               0.005120            0.005120               1.000               0.013          [torch.Size([64, 64, 16])]
SUCCESS               0.009216            0.009216               1.000               0.114          [torch.Size([64, 64, 256])]
SUCCESS               0.084992            0.084992               1.000               0.197          [torch.Size([64, 64, 4096])]