[Performance] Performance regression in Mod operator for float32 with fmod=1 between v1.18.0 and v1.19.0

[junghyunpark2001]

Describe the issue

Description

We observed a performance regression in the Mod operator when using float32 data type with fmod=1 attribute between ONNXRuntime v1.18.0 and v1.19.0. This is a separate regression from the larger one introduced in v1.20.0 → v1.21.0.

Affected Operator

Mod

• Opset Version: 13
• Data Type: float32
• Attribute: fmod=1
• Regression: +9.3% kernel slowdown

Test Case Details

Test Case: mod_mod_13_mod_fmod1_float32_negative_divisor

Inputs:

• X tensor:

  • Data type: float32 (type=1)
  • Shape: [8, 128] (1,024 elements)

• Y tensor:

  • Data type: float32 (type=1)
  • Shape: [8, 128]

Attributes:

• fmod: 1 (C-style fmod semantics)

Output:

• Data type: float32
• Shape: [8, 128]

Performance:

• v1.18.0: 0.0049 ms (kernel time)
• v1.19.0: 0.0053 ms (kernel time)
• Kernel regression: +9.3% slowdown
• Confirmation: 4/10 validation runs confirmed

Regression Characteristics

Affected Configuration

• Data type: float32
• Attribute: fmod=1 (C-style floating-point modulo)
• Tensor size: Small to medium (1K elements)

To reproduce

python script_profiling.py mod_mod_13_mod_fmod1_float32_negative_divisor 1.18.0 1.19.0

Archive.zip

Urgency

No response

Platform

Linux

OS Version

Ubuntu 24.04.3 LTS

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.19.0

ONNX Runtime API

Python

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

Model File

No response

Is this a quantized model?

Yes