October 17, 2025: Weekly Status Update in Gluten

[GlutenPerfBot]

This weekly update is generated by LLMs. You're welcome to join our Github for in-depth discussions.

Overall Activity Summary

The past 7 days have been exceptionally productive for Apache Gluten, with 47 merged PRs and 21 open PRs. The community is actively preparing for the 1.5.0 release while making significant progress on Velox backend improvements, GPU support, and Apache maturity model compliance. Daily Velox version updates continue to keep the project current with upstream changes.

Key Ongoing Projects

GPU Acceleration Initiative - @jinchengchenghh is leading the GPU support implementation (#9098, #10621), with the first PR merged for CUDA docker images. This represents a major architectural expansion for Gluten.

Velox Backend Enhancement - Multiple contributors are actively improving the Velox backend:

• Daily version updates by @GlutenPerfBot keeping pace with upstream Velox changes
• Function support expansion by @dmsuehir ([GLUTEN-4039][VL] Enable map_from_arrays function with Velox backend #10058, [GLUTEN-4039][VL] Enable timestamp_seconds function with Velox backend #10296) adding map_from_arrays and timestamp_seconds
• Performance optimizations by @zhouyuan ([GLUTEN-10833][VL] Reduce memcpy in ColumnToRow #10824, [VL] Reduce vector copy when doing c2r #10825) reducing memory copies and improving columnar-to-row conversion

Apache Maturity Model Compliance - @zhztheplayer is driving the project toward graduation with comprehensive documentation updates (#10873, #10874, #10876, #10878, #10885) covering release processes, security policies, and community governance.

Delta Lake Native Write Support - @zhztheplayer successfully implemented native Delta write support (#10801) for Spark 3.5/Delta 3.3, a significant milestone for data lake integration.

Priority Items

Immediate Attention Needed:

• [VL] VeloxBatchResizer may cause OOM when there exists type of variable length #10892: VeloxBatchResizer OOM issue with variable-length types needs resolution
• [VL] Maybe duplicate stages are executed in tpcds-1t query 88 #10868: Duplicate stage execution in TPCDS q88 requires investigation
• TPCDS q72 uses gluten calculation much worse than spark #8417: TPCDS q72 performance regression needs optimization

Release Preparation:

• Gluten 1.5.0 Release Preparation: Patch Porting, Documentation Updates, and Final Tasks #10574: Gluten 1.5.0 release preparation is ongoing with patch porting
• [CORE] Bump Delta 3.3.x version to 3.3.2 #10887: Delta version bump to 3.3.2 for compatibility

Notable Discussions

#8763: Documentation build process and website synchronization - @zjuwangg raised important questions about maintaining consistent documentation between the repo and the official website, with community members providing guidance on the build process.

Emerging Trends

1. Performance Optimization Focus: Multiple PRs targeting memory allocation improvements and CPU efficiency
2. Cloud Storage Integration: Enhanced support for Azure, S3, and other cloud storage systems
3. Testing Infrastructure: Increased emphasis on integration testing for cloud storage connectors
4. Community Governance: Systematic approach to Apache graduation requirements

Good First Issues

ClickHouse Backend Functions - Several good first issues for contributors interested in ClickHouse backend development:

• [CH] support expression MakeYMInterval #6814: Implement MakeYMInterval expression - Great for learning expression handling
• [CH] support function date_from_unix_date  #4730: Add date_from_unix_date function - Good introduction to date/time functions
• [CH] support split_part function #6807: Support split_part function - Excellent for string manipulation practice
• [CH] support function SparkPartitionID  #6812: Implement SparkPartitionID function - Useful for understanding Spark integration
• [CH] support expression MapZipWith #6815: Add MapZipWith expression - Advanced but well-scoped for learning map operations

These issues are ideal entry points because they involve implementing specific functions in the ClickHouse backend, allowing new contributors to understand the codebase structure while making meaningful contributions. Basic C++ knowledge and familiarity with database functions are helpful but not required - the community provides excellent guidance and code reviews.