layout: page title: Velox GPU nav_order: 9 parent: Getting-Started

GPU Acceleration in Velox/Gluten

Unified execution engine leveraging CUDF for hardware-accelerated Spark SQL queries

1. Overview

  • Purpose: Accelerate Velox operators via CUDF APIs, replacing CPU execution when enabled.
  • Status: Experimental (TPC-H SF1 validated). Integrates RAPIDS ecosystem with Apache Spark via Gluten .
  • Key Benefit: Some queries achieved up to 8.1x speedup on x86 vs. Spark Java engine .

2. Prerequisites

  • CUDA Toolkit: 12.8.0 (download).
  • NVIDIA Drivers: Compatible with CUDA 12.8.
  • Container Toolkit: Install nvidia-container-toolkit (guide).
  • System Reboot: Required after driver installation.
  • Environment Setup: Use start-cudf.sh for host configuration .

3. Implementation Mechanics

  • Operator Conversion:
    • Velox PlanNodes → GPU operators when spark.gluten.sql.columnar.cudf=true.
    • Falls back to CPU operators if GPU unsupported (triggers row/columnar data conversion) .
  • Debugging: Enable spark.gluten.debug.enabled.cudf=true for operator replacement logs.
  • Memory: Global RMM memory manager, cannot align with Spark memory system.

4. Docker Deployment

docker pull apache/gluten:centos-9-jdk8-cudf  # Pre-built GPU image
docker run --name gpu_gluten_container --gpus all -it apache/gluten:centos-9-jdk8-cudf
  • Image Includes: Native build cache, Gluten dependencies, Spark 3.4 environment.

5. Build & Deployment

Dependencies

The OS, Spark version, Java version aligns with Gluten CPU.

Compilation Commands

If building in the docker image, no need to set up script and build arrow.

./dev/buildbundle-veloxbe.sh --run_setup_script=OFF --build_arrow=OFF --enable_gpu=ON

6. GPU Operator Support Status

OperatorStatusNotes
Scan❌ Not supportedIn Development
Project⚠️ PartialFunction TPCH-compatible
Filter✅ ImplementedCore operator
OrderBy✅ Implemented
Aggregation⚠️ PartialTPCH-compatible
Join⚠️ PartialTPCH-compatible
Spill❌ Not supportedIn Planning

**7. Dynamic Execution

The first stage contains TableScan operator which is IO bound stage, schedule to CPU node. The second stage that contains join which is computation intensive, schedule to GPU node.

8. Performance Validation

GPU performs better on operator HashJoin and HashAggregation. Single Operator like Hash Agg shows 5x speedup.

9. Relevant Resources

  1. CUDF Docs - GPU operator APIs.
  2. Gluten GPU Issue #9098 - Development tracker.