Ptx Assembly

Use when the user asks about PTX, inline asm volatile in CUDA, nvcc -ptx, ptxas -v register spills, state spaces, cp.async, mma.sync or bar.sync. Produces explained PTX listings or inline-asm snippets with constraints and SASS verification steps. Not for native machine code and stall analysis (use sass-machine-code) or kernel design (use cuda-programming).

45 stars 6 forksSoftware Development

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add j4flmao/agent-skills --skill ptx-assembly
Manual install
git clone --depth 1 https://github.com/j4flmao/agent-skills.git /tmp/agent-skills
mkdir -p ~/.claude/skills
cp -r /tmp/agent-skills/skills/hardware-acceleration/ptx-assembly ~/.claude/skills/ptx-assembly

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# PTX Assembly

## Purpose
PTX (Parallel Thread Execution) is NVIDIA's virtual, forward-compatible ISA. nvcc compiles CUDA C++ to PTX, and `ptxas` (or the driver JIT) compiles PTX to SASS for a specific GPU. Engineers read PTX to check what the front end generated and write small amounts of inline PTX to reach instructions not exposed by CUDA C++ intrinsics. Real performance is decided after `ptxas`, so PTX work always ends with a SASS and profiler check.

## Agent Protocol

### Trigger
"ptx", "inline ptx", "asm volatile", "nvcc -ptx", "ptxas -v", "state space", "ld.global.nc", "bar.sync", "cp.async", "mma.sync", "register spill", "predicate".

### Input Context
- Target architecture (`sm_80`, `sm_90a`, ...), CUDA version, kernel source.
- `ptxas -v` output or Nsight Compute report if diagnosing performance.

### Output Artifact
PTX listing explanation or inline-asm snippet + verification commands. No file unless requested.

### Response Format
```
Goal: {instruction/behavior needed}
PTX: {snippet}
Constraints: {operand constraints, clobbers, arch requirement}
Verify: {cuobjdump -sass / ncu check}
```

### Completion Criteria
- [ ] Prefer a CUDA intrinsic or library (CUTLASS/CuTe, cuda::pipeline) if one exists.
- [ ] Inline asm has correct constraints, `volatile` and `"memory"` clobber where it touches memory.
- [ ] SASS inspected to confirm the intended instruction was emitted.
- [ ] Architecture guard (`__CUDA_ARCH__`) for arch-specific instructions.

## Toolchain
Read the full skill on GitHub

Details

Skill name
ptx-assembly
Path
skills/hardware-acceleration/ptx-assembly/SKILL.md
License
MIT
Repo updated
17 November 2026

FAQ

What does the Ptx Assembly skill do?

Use when the user asks about PTX, inline asm volatile in CUDA, nvcc -ptx, ptxas -v register spills, state spaces, cp.async, mma.sync or bar.sync. Produces explained PTX listings or inline-asm snippets with constraints and SASS verification steps. Not for native machine code and stall analysis (use sass-machine-code) or kernel design (use cuda-programming).

How do I install the Ptx Assembly skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add j4flmao/agent-skills --skill ptx-assembly. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Ptx Assembly popular?

It lives in j4flmao/agent-skills, which has 45 GitHub stars and 6 forks.

Related skills

Agent Observability

45

j4flmao/agent-skills

Use when the user asks about agent observability, tracing agent reasoning, debugging an agent loop, OpenTelemetry GenAI spans for LLM calls, LLM cost tracking per agent, agent audit logs or hallucination/eval monitoring in production. Produces instrumentation config, trace schema, dashboards, alerts and cost attribution. Not for generic service tracing without LLMs (use devops-opentelemetry).

DevOps & Cloud

AI Agent Architectures

45

j4flmao/agent-skills

Use when the user asks to build an AI agent, design an agent that calls tools and keeps memory, write an agent system prompt or pick an agent framework (Claude Agent SDK, OpenAI Agents SDK, LangGraph). Produces an architecture brief covering loop, tools, memory, guardrails and evals, routing depth to child skills. Not for workflow patterns (use ai-agentic-workflows).

AI, LLMs & Agents

AI Agent Core Loop

45

j4flmao/agent-skills

Use when the user asks about the agent loop, ReAct vs plan-and-solve, re-planning, Reflexion, an agent stuck in a loop, max iterations or termination conditions. Produces a loop spec or patch with state, success predicate, caps and no-progress detection. Not for overall agent design (use ai-agent-architectures) or tool schemas (use tool-calling-principles).

AI, LLMs & Agents

Python Django Architecture

45

j4flmao/agent-skills

Use when the user asks about Django, Django ORM queries that are slow, select_related and prefetch_related, Django REST framework (DRF), Django apps structure or migrations. Produces Django code with a service layer and N+1-free queries. Not for FastAPI (use python-fastapi-architecture).

Software Development

Platform Abstraction

45

j4flmao/agent-skills

Use when the user asks about a platform abstraction layer, cross-platform engine code, windowing, engine filesystem, dynamic library loading, high-resolution timers or platform ifdefs. Produces a PAL for Windows, Linux, macOS and consoles. Not for mobile app platform APIs (use mobile-patterns).

Software Development

Python Fastapi Architecture

45

j4flmao/agent-skills

Use when the user asks about FastAPI, setting up a FastAPI project with dependency injection and Pydantic models, Depends(), APIRouter structure, lifespan events or async SQLAlchemy. Produces a layered FastAPI project. Not for Django (use python-django-architecture) or Flask (use flask-backend).

Software Development

Browse the most popular Claude skills, all software development skills, or write your own with the skills generator.