Translate DSL operator descriptions into AscendC code via a four-pass lowering workflow.
npx skills add https://github.com/Just-it/AscendOpGenAgent --skill dsl-lowering
Please help me install this Agent Skill. Skill: dsl-lowering Source: https://github.com/Just-it/AscendOpGenAgent/tree/main/skills/dsl_lowering Command: npx skills add https://github.com/Just-it/AscendOpGenAgent --skill dsl-lowering
Translate a DSL-based operator description into AscendC code through a four-pass lowering workflow that enables automated code generation and integration with existing ASCEND C APIs.
Run the DSL lowering workflow on a sample DSL kernel to generate an AscendC implementation and automatically repair errors as needed.
High-intent search queries and answers about installing and using this skill.
You can automate DSL lowering to AscendC by running a four-pass workflow that translates operator descriptions through tiling_pass, init_pass, process_pass, and nonaligned_pass to generate kernel code.
DSL lowering is a four-pass code generation process that transforms DSL-based operator descriptions into AscendC kernel code using tiling_pass, init_pass, process_pass, and nonaligned_pass execution stages.
Automated error recovery uses sub-agent orchestration to repair compilation or runtime issues during DSL lowering, applying reference-guided templates and error-correction guidance with up to three repair attempts per pass.
Yes, the DSL lowering workflow autonomously repairs compilation and runtime issues during the translation process, using reference-guided error-correction guidance to attempt up to three fixes per pass.
The main limitation is that error recovery is capped at three repair attempts per pass, meaning complex compilation or runtime issues in the DSL to AscendC translation may require manual intervention.
Yes, the DSL lowering workflow is designed to generate AscendC code that integrates with existing AscendC APIs, requiring a valid DSL-based operator description as input for the automated four-pass transformation.
ScavieFaeDetect the host platform, install dependencies, and configure Dolphin, Slippi, and Phillip.
AscendGenerate Triton kernels from Ascend NPU operator designs with smoke tests.
majiayu000Quantify AI-generated code quality using pass@k and pass^k metrics.
AscendGenerate AscendC operator host/kernel code from design docs and integrate into PyTorch.
