Research map / Architectures and efficiency / CNN design
High Performance Zero-Memory Overhead Direct Convolutions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace SGEMM plus im2col packing with direct convolution | Convolution layers in AlexNet, GoogLeNet, and VGG · Intel, AMD, and ARM architectures | performance: between 10% and 400% | Helped |
| Replace OpenBLAS-based SGEMM convolution with direct convolution | Convolution layers in AlexNet, GoogLeNet, and VGG · 4 threads on OpenBLAS architectures | performance: minimum of 1.5 times performance gain | Helped |
| Use direct convolution instead of NNPACK FFT-based convolution | Convolution layers · ARM | performance | Helped |
| Use direct convolution instead of NNPACK FFT-based convolution | Convolution layers · Intel | performance | Helped |
| Increase thread count up to the physical core count | High-performance direct convolution · Intel, AMD, and ARM | performance per core | No clear effect |
| Increase thread count beyond the physical core count | High-performance direct convolution · Intel, AMD, and ARM | performance per core | Hurt |
| Increase thread count | SGEMM-based convolution · Intel, AMD, and ARM | performance per core | Hurt |
Closest papers by meaning
- Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks (CVPR 2023 · similarity 0.87)
- MEC: Memory-efficient Convolution for Deep Neural Network (ICML 2017 · similarity 0.87)
- Fast and Accurate Model Scaling (CVPR 2021 · similarity 0.86)
- InceptionNeXt: When Inception Meets ConvNeXt (CVPR 2024 · similarity 0.85)
- MEET: Towards Memory-Efficient Temporal Sparse Deep Neural Networks (CVPR 2025 · similarity 0.85)
- Accelerating Linear Recurrent Neural Networks for the Edge with Unstructured Sparsity (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.