TensorRT support of bitwise operations for int types

AI & Data Science Deep Learning (Training & Inference) TensorRT
,
1

Description

TensorRT does not support bitwise operators AND, OR and XOR for integer types, only for bool.

Is there any technical reason for this absence or is it just due to a lack of user request?
As far as I know, most (all?) NVidia devices support it.

Environment

TensorRT Version: 10.9.0
GPU Type: RTX 2000 Ada
Nvidia Driver Version: 535
CUDA Version: 12.4
CUDNN Version: ??
Operating System + Version:
Python Version (if applicable): Ubuntu 24.04
TensorFlow Version (if applicable):
PyTorch Version (if applicable): 2.6
Baremetal or Container (if container which image + tag):

Relevant Files

3

Hi @nicolas.granger ,
There are some technical limitations associated with it.

  • Bitwise operations on integers are less frequently used in deep learning inference tasks, where operations like matrix multiplications and convolutions are prioritized. Therefore, the developers may focus on optimizing these more commonly used operations for better performance.
  • Implementing bitwise operators for integer types may require additional hardware support or specific instructions that may not be efficiently supported on NVIDIA GPUs used for deep learning inference. This could lead to suboptimal performance or compatibility issues.
    Hope this helps

Thanks