【KernelGen】Add fmod operator by Schopenhauer-loves-Hegel · Pull Request #1737 · flagos-ai/FlagGems

Schopenhauer-loves-Hegel · 2026-02-27T06:14:50Z

PR Category

Operator

Type of Change

New Feature

Description

Add fmod operator implementation with Triton kernel.

Implementation mode: pointwise_dynamic
Accuracy test: 90/90 passed

Issue

N/A

Progress

Change is properly reviewed (1 reviewer required, 2 recommended).
Change is responded to an issue.
Change is fully covered by a UT.

Performance

torch.bfloat16

Shape	Torch Latency (ms)	Gems Latency (ms)	Speedup
[1073741824]	4.7215	4.7127	1.002
[64, 64]	0.0074	0.0079	0.939
[4096, 4096]	0.0871	0.0873	0.997
[64, 512, 512]	0.0879	0.0870	1.011
[1024, 1024, 1024]	4.7193	4.6852	1.007
[1024, 1]	0.0083	0.0074	1.130
[1024, 16]	0.0076	0.0076	1.000
[1024, 256]	0.0092	0.0095	0.963
[1024, 4096]	0.0296	0.0287	1.031
[1024, 65536]	0.3062	0.3044	1.006
[64, 64, 1]	0.0074	0.0080	0.932
[64, 64, 16]	0.0078	0.0078	1.000
[64, 64, 256]	0.0134	0.0140	0.961
[64, 64, 4096]	0.0870	0.0872	0.997
[64, 64, 65536]	1.1894	1.1779	1.010

torch.float16

Shape	Torch Latency (ms)	Gems Latency (ms)	Speedup
[1073741824]	4.7231	4.7091	1.003
[64, 64]	0.0074	0.0084	0.885
[4096, 4096]	0.0870	0.0873	0.996
[64, 512, 512]	0.0874	0.0869	1.005
[1024, 1024, 1024]	4.7218	4.6844	1.008
[1024, 1]	0.0079	0.0083	0.946
[1024, 16]	0.0081	0.0086	0.944
[1024, 256]	0.0101	0.0100	1.006
[1024, 4096]	0.0296	0.0295	1.003
[1024, 65536]	0.3061	0.3043	1.006
[64, 64, 1]	0.0074	0.0074	0.996
[64, 64, 16]	0.0078	0.0086	0.907
[64, 64, 256]	0.0134	0.0139	0.963
[64, 64, 4096]	0.0870	0.0866	1.005
[64, 64, 65536]	1.1882	1.1823	1.005

torch.float32

Shape	Torch Latency (ms)	Gems Latency (ms)	Speedup
[1073741824]	9.4564	9.3270	1.014
[64, 64]	0.0076	0.0076	0.996
[4096, 4096]	0.1586	0.1583	1.002
[64, 512, 512]	0.1583	0.1581	1.001
[1024, 1024, 1024]	9.4667	9.3252	1.015
[1024, 1]	0.0079	0.0080	0.988
[1024, 16]	0.0076	0.0085	0.888
[1024, 256]	0.0104	0.0112	0.926
[1024, 4096]	0.0480	0.0480	1.000
[1024, 65536]	0.6022	0.5980	1.007
[64, 64, 1]	0.0076	0.0076	1.000
[64, 64, 16]	0.0086	0.0085	1.007
[64, 64, 256]	0.0180	0.0173	1.043
[64, 64, 4096]	0.1591	0.1581	1.006
[64, 64, 65536]	2.3760	2.3485	1.012

Overall: median speedup = 1.002x, mean speedup = 0.990x (45 data points)

Generated by auto_gen tool with Claude Code

Schopenhauer-loves-Hegel · 2026-02-27T06:34:48Z

Closing in favor of grouped PRs to reduce review overhead.

CLAassistant · 2026-04-15T20:29:31Z

All committers have signed the CLA.

factnn · 2026-04-17T09:01:30Z

fmod Benchmark Speedup Results

float16

Size	Torch (ms)	Gems (ms)	Speedup
[1073741824]	4.739	4.880	0.971
[4096, 4096]	0.087	0.091	0.952
[1024, 1024, 1024]	4.739	4.889	0.969
[1024, 65536]	0.306	0.312	0.984
[64, 64, 65536]	1.195	1.206	0.991

float32

Size	Torch (ms)	Gems (ms)	Speedup
[1073741824]	9.419	9.325	1.010
[4096, 4096]	0.160	0.159	1.004
[1024, 1024, 1024]	9.421	9.332	1.010
[1024, 65536]	0.600	0.599	1.001
[64, 64, 65536]	2.365	2.357	1.003

bfloat16

Size	Torch (ms)	Gems (ms)	Speedup
[1073741824]	4.740	4.880	0.971
[4096, 4096]	0.087	0.090	0.961
[1024, 1024, 1024]	4.739	4.889	0.969
[1024, 65536]	0.306	0.312	0.982
[64, 64, 65536]	1.195	1.205	0.991

tengqm

Overall lgtm now. Thanks.

factnn · 2026-05-08T08:40:55Z

UT Results

Branch rebased to upstream/master and tests passed:

python -m pytest tests/ -m fmod --ref cpu -x -q
90 passed

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com>

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: Zang Peiyu <166481866+factnn@users.noreply.github.com>

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com>

@tengqm

* feat: add fmod operator with tests and benchmark Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Apply suggestion from @tengqm Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com> * Apply suggestions from code review Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com> * Apply suggestions from code review Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: Zang Peiyu <166481866+factnn@users.noreply.github.com> * Apply suggestions from code review Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com> --------- Signed-off-by: bin913 <842884726@qq.com> Signed-off-by: Zang Peiyu <166481866+factnn@users.noreply.github.com> Co-authored-by: factnn <1050552884@qq.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> Co-authored-by: bin913 <842884726@qq.com> Co-authored-by: Qiming Teng <tengqm@outlook.com> Co-authored-by: Zang Peiyu <166481866+factnn@users.noreply.github.com>

Schopenhauer-loves-Hegel requested review from 0x45f, huangyiqun, kiddyjinjin and zhangpeiyang1 as code owners February 27, 2026 06:14

github-actions Bot added benchmark ops/aten core tests size/Medium labels Feb 27, 2026

Schopenhauer-loves-Hegel closed this Feb 27, 2026

Schopenhauer-loves-Hegel reopened this Feb 27, 2026

tengqm added tests and removed tests labels Mar 2, 2026

Schopenhauer-loves-Hegel changed the title ~~Add fmod operator~~ 【KernelGen】Add fmod operator Mar 23, 2026

factnn force-pushed the auto-gen/fmod branch from 82a13b3 to 1d8143b Compare April 15, 2026 20:29

github-actions Bot added the KernelGen label Apr 15, 2026

factnn force-pushed the auto-gen/fmod branch from 1d8143b to 13e3beb Compare April 18, 2026 00:20

github-actions Bot added the documentation Improvements or additions to documentation label Apr 18, 2026

factnn force-pushed the auto-gen/fmod branch from 13e3beb to dd5d813 Compare April 18, 2026 00:40

github-actions Bot removed the documentation Improvements or additions to documentation label Apr 18, 2026

factnn force-pushed the auto-gen/fmod branch 3 times, most recently from d363c4a to 86b8686 Compare April 23, 2026 07:57

tengqm reviewed Apr 23, 2026

View reviewed changes

Comment thread benchmark/test_fmod_perf.py Outdated

Comment thread benchmark/test_fmod_perf.py Outdated

Comment thread src/flag_gems/ops/__init__.py Outdated

Comment thread src/flag_gems/__init__.py Outdated

Comment thread src/flag_gems/__init__.py Outdated

factnn force-pushed the auto-gen/fmod branch from 86b8686 to fce942e Compare April 24, 2026 08:15

factnn requested review from bin913 and douxetpur as code owners April 24, 2026 08:15

factnn requested a review from w1120029931-bit as a code owner April 24, 2026 08:15

factnn force-pushed the auto-gen/fmod branch from fce942e to 622deaf Compare April 24, 2026 08:34

tengqm reviewed Apr 24, 2026

View reviewed changes

Comment thread benchmark/test_fmod_perf.py Outdated

tengqm reviewed Apr 24, 2026

View reviewed changes

Comment thread tests/test_fmod.py Outdated

tengqm reviewed Apr 24, 2026

View reviewed changes

tengqm reviewed Apr 26, 2026

View reviewed changes

Comment thread src/flag_gems/ops/fmod.py Outdated

Comment thread src/flag_gems/ops/fmod.py Outdated

Comment thread src/flag_gems/ops/fmod.py Outdated

Comment thread src/flag_gems/ops/fmod.py Outdated

factnn force-pushed the auto-gen/fmod branch from 13862a4 to 002d6b0 Compare April 28, 2026 10:46

factnn force-pushed the auto-gen/fmod branch from 002d6b0 to fa18f73 Compare May 8, 2026 07:21

github-actions Bot added ci/cd vendor/Kunlunxin size/XLarge and removed size/Medium labels May 8, 2026

feat: add fmod operator with tests and benchmark

59fdd32

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

factnn force-pushed the auto-gen/fmod branch from f5cbb4e to 59fdd32 Compare May 9, 2026 02:46

github-actions Bot added size/Medium and removed ci/cd vendor/Kunlunxin size/XLarge labels May 9, 2026

bin913 and others added 2 commits May 9, 2026 16:08

Apply suggestion from @tengqm

a00296b

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com>

Apply suggestions from code review

e1dc6e6

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com>

tengqm reviewed May 9, 2026

View reviewed changes

factnn and others added 3 commits May 9, 2026 16:38

Apply suggestions from code review

b0f51e5

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: Zang Peiyu <166481866+factnn@users.noreply.github.com>

Apply suggestions from code review

09b4177

Co-authored-by: Qiming Teng <tengqm@outlook.com> Signed-off-by: bin913 <842884726@qq.com>

Merge branch 'master' into auto-gen/fmod

88ebb8a

bin913 approved these changes May 9, 2026

View reviewed changes

bin913 merged commit 49654ef into flagos-ai:master May 11, 2026
32 of 33 checks passed

Uh oh!

Conversation

Schopenhauer-loves-Hegel commented Feb 27, 2026

PR Category

Type of Change

Description

Issue

Progress

Performance

Uh oh!

Schopenhauer-loves-Hegel commented Feb 27, 2026

Uh oh!

CLAassistant commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

factnn commented Apr 17, 2026

fmod Benchmark Speedup Results

float16

float32

bfloat16

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tengqm left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

factnn commented May 8, 2026

UT Results

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

CLAassistant commented Apr 15, 2026 •

edited

Loading