arXiv subjects

Hanfeng Chen

Publications and source records attributed to Hanfeng Chen.

7 recordsLinked to original sources

arXiv2026

DLCB: Ahead-of-Time Compilation for Dynamic Deep Learning

Deep learning workloads are increasingly deployed in settings where tensor shapes are not fully known at compile time. Batch sizes vary across requests, sequence lengths differ between inputs, and model architectures admit a range of spatial resolutions. This dynamism creates a fundamental tension: ahead-of-time compiled GPU kernels deliver peak performance but traditionally require fully static tensor shapes, while just-in-time compilation supports dynamic shapes at the cost of runtime compilation overhead. We present DLCB (Deep Learning Compiler Backend), a deep learning compiler that resolves this tension through a unified three-tier compilation strategy. For programs with fully static tensor shapes, DLCB generates static CUDA C++ kernels ahead of time. For programs whose tensor rank is statically known but whose dimension sizes are dynamic, DLCB generates AOT kernels parameterized by those dimensions, compiling once and executing across a range of shapes. For programs with tensors of unknown rank, DLCB falls back to JIT compilation at runtime. For AOT compilation with dynamic shape tensors, our approach builds a system of shape constraints from the input program, and uses a solver to reduce these to either a) fixed constraints that produce hard-coded constants in the generated kernel; b) unresolved constraints which become kernel parameters passed at launch time, along with interpreted host code to check that the constraint holds for a given set of inputs and to compute the values to pass. Our compiler and runtime are embedded in PyTorch, and use an automatic shape generalization pass to enable AOT compilation with dynamic shapes for supported subsets of PyTorch programs

cs.PL↗

arXiv2026

CAKE: Compiler-Agent Co-Design for Frontier Kernel Evolution

GPU kernel agents and GPU programming languages have advanced separately, leaving expert kernels difficult to reproduce. Agents usually treat the compiler as a fixed black box and receive only errors, correctness outcomes, and timing, while existing DSLs either hide critical scheduling decisions or expose them through difficult layout abstractions. We present CAKE, a compiler-agent co-design in which agents author CAKE IR, a typed, hardware-explicit schedule representation. CAKE exposes warp roles, memory movement, synchronization, and pipelines while supporting verification, cost modeling, and localized diagnostics. The harness itself evolves: recurring failures become verifier rules, IR primitives, model calibrations, and reusable optimization tactics. In matched implementation-hidden Flash-KMeans clean starts on B200, the best CAKE IR candidate at an 80-million-token budget runs at 1.144x the tuned FlashML baseline, compared with 0.928x for direct CUDA/PTX. Beyond this benchmark, agent-generated Kimi Delta Attention achieves a 2.05x geometric-mean speedup over official FlashKDA and passes end-to-end serving validation. Dispatcher-backed KNN and KMeans improve performance by 1.42x to 2.12x across more than 400 shapes, and four kernel changes are available as upstream PRs. CAKE targets NVIDIA GPUs from Ampere through Blackwell and separates single-shape evolution from library generalization and dispatch.

cs.LG↗

arXiv2026

AVO: Agentic Variation Operators for Autonomous Evolutionary Search

Agentic Variation Operators (AVO) are a new family of evolutionary variation operators that replace the fixed mutation, crossover, and hand-designed heuristics of classical evolutionary search with autonomous coding agents. Rather than confining a language model to candidate generation within a prescribed pipeline, AVO instantiates variation as a self-directed agent loop that can consult the current lineage, a domain-specific knowledge base, and execution feedback to propose, repair, critique, and verify implementation edits. We evaluate AVO on attention, among the most aggressively optimized kernel targets in AI, on NVIDIA Blackwell (B200) GPUs. Over 7 days of continuous autonomous evolution on multi-head attention, AVO discovers kernels that outperform cuDNN by up to 3.5% and FlashAttention-4 by up to 10.5% across the evaluated configurations. The discovered optimizations transfer readily to grouped-query attention, requiring only 30 minutes of additional autonomous adaptation and yielding gains of up to 7.0% over cuDNN and 9.3% over FlashAttention-4. Together, these results show that agentic variation operators move beyond prior LLM-in-the-loop evolutionary pipelines by elevating the agent from candidate generator to variation operator, and can discover performance-critical micro-architectural optimizations that produce kernels surpassing state-of-the-art expert-engineered attention implementations on today's most advanced GPU hardware.

cs.LG↗

arXiv2026

SOL-ExecBench: Speed-of-Light Benchmarking for Real-World GPU Kernels Against Hardware Limits

As agentic AI systems become increasingly capable of generating and optimizing GPU kernels, progress is constrained by benchmarks that reward speedup over software baselines rather than proximity to hardware-efficient execution. We present SOL-ExecBench, a benchmark of 235 CUDA kernel optimization problems extracted from 124 production and emerging AI models spanning language, diffusion, vision, audio, video, and hybrid architectures, targeting NVIDIA Blackwell GPUs. The benchmark covers forward and backward workloads across BF16, FP8, and NVFP4, including kernels whose best performance is expected to rely on Blackwell-specific capabilities. Unlike prior benchmarks that evaluate kernels primarily relative to software implementations, SOL-ExecBench measures performance against analytically derived Speed-of-Light (SOL) bounds computed by SOLAR, our pipeline for deriving hardware-grounded SOL bounds, yielding a fixed target for hardware-efficient optimization. We report a SOL Score that quantifies how much of the gap between a release-defined scoring baseline and the hardware SOL bound a candidate kernel closes. To support robust evaluation of agentic optimizers, we additionally provide a sandboxed harness with GPU clock locking, L2 cache clearing, isolated subprocess execution, and static analysis based checks against common reward-hacking strategies. SOL-ExecBench reframes GPU kernel benchmarking from beating a mutable software baseline to closing the remaining gap to hardware Speed-of-Light.

cs.LG↗

arXiv2026

Central Bank Digital Currency: Demand Shocks and Optimal Monetary Policy

We study the implications of a central bank digital currency (CBDC) for the transmission of household preference shocks and for welfare in a New Keynesian framework where the CBDC competes with bank deposits for household resources and banks have market power. We show that an increase in the perceived benefit of CBDC has a mildly expansionary effect, weakening bank market power and significantly reducing the deposit spread. As households economize on liquid asset holdings, they reduce both CBDC and deposit balances. However, the degree of bank disintermediation is low, as deposit outflows remain modest. We then examine the welfare implications of CBDC rate setting and find that, compared to a non-interest-bearing CBDC, the gains with standard coefficients for a CBDC interest rate Taylor rule are modest, but they become considerable when the coefficients are optimized. Welfare gains increase with the CBDC benefit, and the optimal policy responses vary with the banking market structure.

econ.TH↗

arXiv2025

Central Bank Digital Currency with Collateral-constrained Banks

We analyze the risks to bank intermediation following the introduction of a central bank digital currency (CBDC) competing with commercial bank deposits as households' source of liquidity. We revisit the result in the literature regarding the equivalence of payment systems introducing a collateral constraint on banks borrowing from the central bank. Comparing two equilibria with and without the CBDC, we find that even with this constraint, the central bank can ensure the same equilibrium allocation and price system by offering loans to banks. However, to access loans, banks must hold collateral at the expense of extending credit to firms. Thus, while the CBDC introduction has no real effects on the economy, it does not guarantee full neutrality as it affects banks' business models. In a dynamic model extension, we examine the effects of an increase in the CBDC and show that the CBDC does not cause bank disintermediation or crowd out deposits but may foster an expansion of bank credit to firms.

econ.TH↗

arXiv2024

Scaling Deep Learning Training with MPMD Pipeline Parallelism

We present JaxPP, a system for efficiently scaling the training of large deep learning models with flexible pipeline parallelism. We introduce a seamless programming model that allows implementing user-defined pipeline schedules for gradient accumulation. JaxPP automatically distributes tasks, corresponding to pipeline stages, over a cluster of nodes and automatically infers the communication among them. We implement a MPMD runtime for asynchronous execution of SPMD tasks. The pipeline parallelism implementation of JaxPP improves hardware utilization by up to $1.11\times$ with respect to the best performing SPMD configuration.

cs.DC↗