dmlc/dgl

The ALG2 for cusparsespmm under performed when feature size is small

Open

#4,090 opened on Jun 7, 2022

View on GitHub
 (2 comments) (0 reactions) (0 assignees)Python (2,928 forks)batch import
help wanted

Repository metrics

Stars
 (12,665 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

🐛 Bug

@yzh119 report the ALG2 seems performed badly for cusparsespmm when the feature size is small.

To Reproduce

Steps to reproduce the behavior:

Expected behavior

Environment

  • DGL Version (e.g., 1.0):
  • Backend Library & Version (e.g., PyTorch 0.4.1, MXNet/Gluon 1.3):
  • OS (e.g., Linux):
  • How you installed DGL (conda, pip, source):
  • Build command you used (if compiling from source):
  • Python version:
  • CUDA/cuDNN version (if applicable):
  • GPU models and configuration (e.g. V100):
  • Any other relevant information:

Additional context

Contributor guide

Research direction
Investigate the cusparsespmm ALG2 implementation in DGL, compare its performance with ALG1 for small feature sizes, and identify the root cause of underperformance by examining the CUDA kernel and calling code.
Tech stack
pythonpytorch
Domain
backendperformance
Issue type
Bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
PythonCUDAGPU programming
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40