How do I make ISPC lit tests portable across different architectures?

Avoid hardcoding vector widths or variable names. Instead, use named patterns like [[WIDTH]] and [[TYPE]] in your CHECK statements to match flexible LLVM IR outputs.

What is the standard naming convention for ISPC regression tests?

Regression tests should be named after their GitHub issue number (e.g., 1234.ispc) to provide clear context and link the test directly to the fixed bug.

Why should I use --nowrap in ISPC lit tests?

Using --nowrap prevents line wrapping in the compiler output, which ensures that FileCheck matches remain consistent and reliable across different environments.

How can I test specific optimization passes in ISPC?

You can use the --debug-phase flag to capture the output of a specific optimization pass and then run FileCheck against the resulting dumped IR file.

When should I use the 'export' keyword in a test file?

Avoid using 'export' unless you are explicitly testing export behavior. Exported functions generate both masked and unmasked IR, which doubles the maintenance effort for the test.

ISPC Lit Test Best Practices

Name: ISPC Lit Test Best Practices
Author: Activer007

byActiver007

0•

安全与测试

Guides the creation of high-quality LLVM-based lit tests for the ISPC compiler to ensure robust code generation and regression tracking.

The ispc-lit-tests skill provides Claude with domain-specific expertise for writing and maintaining Intel Implicit SPMD Program Compiler (ISPC) tests. It offers structured guidance on configuring compiler flags for consistent FileCheck matching, writing portable LLVM IR patterns that work across different versions, and managing complex target-specific dependencies. This skill is essential for compiler engineers and contributors who need to verify diagnostics, validate intermediate IR phases, and lock in fixes for regression issues with standardized naming and documentation conventions.

主要功能

010 GitHub stars

02Portable LLVM IR check strategies for architecture-independent verification

03Target-specific testing workflows with REQUIRES feature dependencies

04Intermediate IR verification using debug-phase and dump-file flags

05Standardized regression test naming and documentation templates

06Optimized test configuration patterns using --nowrap and --nostdlib

使用场景

01Creating reproducible regression tests for GitHub issues

02Verifying compiler diagnostics and error messages for new language features

03Validating cross-platform code generation and ISA-specific assembly output

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add activer007/ordinary-claude-skills ispc-lit-tests

For use in Claude.ai and ChatGPT

主要功能

010 GitHub stars

02Portable LLVM IR check strategies for architecture-independent verification

03Target-specific testing workflows with REQUIRES feature dependencies

04Intermediate IR verification using debug-phase and dump-file flags

05Standardized regression test naming and documentation templates

06Optimized test configuration patterns using --nowrap and --nostdlib

使用场景

01Creating reproducible regression tests for GitHub issues

02Verifying compiler diagnostics and error messages for new language features

03Validating cross-platform code generation and ISA-specific assembly output

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add activer007/ordinary-claude-skills ispc-lit-tests

For use in Claude.ai and ChatGPT