TL;DR
Torchy is presented, a tracing JIT compiler for PyTorch, one of the mainstream eager-mode frameworks, that achieves similar performance as data-flow frameworks, while providing the same semantics of straight-away execution.
Datalog is widely used to build static analyzers, yet existing engines often force a tradeoff between efficiency and extensibility. In practice, static analyses are not run once and forgotten: users edit facts, tune rules, diagnose bottlenecks, and often need semantics beyond standard Datalog, leaving these tasks to ad hoc tooling or invasive engine rewrites. We demonstrate FlowLog, a Datalog compiler that turns Souffl\'e-style programs into Differential Dataflow executables for efficient and extensible static analysis. Across 24 benchmarks derived from real-world workloads, FlowLog consistently outperforms state-of-the-art engines in runtime while remaining memory-efficient and scaling better. The demonstration uses a DOOP points-to analysis. Attendees run it, switching the same program from one-shot to incremental evaluation that retracts a fact and updates results in milliseconds; tune it, inspecting per-operator costs in a browser-based profiler and repairing a bad join order; and extend it with a k-core example beyond standard Datalog.
This article presents a novel technique for reducing the size of AOT compiled programs by using fully-context-sensitive profiles to concentrate the optimizations on the hot code.
We propose PKDB, the first interactive debugger for GPU and multithreaded low-level kernels written in Python. Python is widely used in high performance computing (HPC), with frameworks such as PyKokkos translating Python-embedded domain-specific languages to native code that runs across OpenMP-threaded CPUs and various GPUs. Yet interactive debugging support for such code is absent: developers resort to print statements, framework-specific assertions, or CPU-only execution, the last of which requires altering the program or its data and can mask device-specific bugs. PKDB enables standard interactive debugging like breakpoints, stepping, and variable inspection while preserving actual on-device execution without source modification. Beyond these fundamentals, PKDB introduces two advanced capabilities that exploit the dynamic nature of Python and PyKokkos: (i) Live code evaluation, which lets developers execute arbitrary Python expressions or entire kernels in the middle of a paused kernel without restarting the process; (ii) Kernel call site substitution, which allows an actively running kernel to be updated and reloaded on the fly, so only the kernel is recompiled and re-executed without restarting the application. Our performance evaluation on Intel, AMD, and NVIDIA CPUs, and NVIDIA and AMD GPUs shows that PKDB introduces limited overhead and is practical for everyday use while introducing critical debugging features to the Python HPC ecosystem.
The entire ecosystem of open-source language models effectively relies on a single platform. What if this platform was forced to shut down tomorrow? Implementing and maintaining efficient model definitions and translating them between different training and inference regimes is a resource-heavy task that severely limits model efficiency and portability, hindering both scaling and deployment. Here, we present Axon, a strongly typed domain-specific language with Haskell-like syntax, that enables a write-once, run everywhere paradigm for LLM architectures. By basing collaboration on a language specification rather than a specific framework's vision, Axon fosters open cooperation and empowers researchers to implement highly specialized architectures without giving up optimization infrastructure or accepting deployment lock-in. Axon allows for concise, auditable specifications that can be automatically compiled to standalone implementations for leading frameworks: PyTorch, PyTorch with Triton, JAX, MLX and vLLM. In 467 inference benchmarking experiments on models ranging from 135M to 32B parameters, we demonstrate median speedups of 7% on PyTorch, 12% on PyTorch with Triton, 91% on JAX, and 107% on MLX, compared to the reference implementations from Transformers. When deployed as native vLLM architectures with PagedAttention and KV-cache, Axon models achieve a 58% median speedup over Transformers implementations.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
