validate

Diagnose incorrect Sirius GPU query results against a DuckDB CPU baseline.

1.0k|109|Updated Jun 19, 2024
One-click install
npx skills add https://github.com/sirius-db/sirius --skill validate-sirius-db
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: validate
Source: https://github.com/sirius-db/sirius/tree/main/.claude/skills/validate
Command: npx skills add https://github.com/sirius-db/sirius --skill validate-sirius-db

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill helps diagnose and pinpoint the exact operator causing incorrect query results in Sirius GPU execution by comparing against a DuckDB CPU baseline.

Core Features & Use Cases

  • Root Cause Analysis: Identifies the specific operator responsible for query discrepancies.
  • Data Integrity Checks: Performs row count and data checksum analysis to validate results.
  • Intermittent Bug Detection: Catches race conditions and stream synchronization issues common in GPU processing.
  • Use Case: When a complex SQL query returns unexpected results on the GPU, use this Skill to systematically find the faulty operator and understand why it's failing.

Quick Start

Use the validate skill to diagnose why the attached SQL query is producing incorrect results.

Frequently Asked Questions about validate

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I debug incorrect SQL query results from GPU execution?

To debug incorrect SQL query results from GPU execution, compare the output against a DuckDB CPU baseline to pinpoint the faulty operator using row count and checksum analysis.

How does comparing GPU and CPU baselines help find query discrepancies?

Comparing GPU and CPU baselines isolates query discrepancies by analyzing per-operator row counts and data checksums, systematically pinpointing the exact GPU operator where the data divergence occurs.

Can I diagnose intermittent GPU query errors and stream synchronization issues?

Yes, you can diagnose intermittent GPU query errors and stream synchronization issues by comparing GPU execution against a CPU baseline to catch race conditions and pinpoint the faulty operator.

What is the best way to pinpoint the specific operator causing data validation failures in SQL?

The best way to pinpoint the operator causing data validation failures in SQL is to analyze per-operator row counts and data checksums between the GPU output and the DuckDB CPU baseline.

Do I need DuckDB to validate Sirius GPU query results?

Yes, DuckDB is used as the CPU baseline to validate Sirius GPU query results, allowing you to compare outputs and identify the specific faulty operator causing the discrepancies.