PolyUFC: Polyhedral Compilation Meets Roofline Analysis for Uncore Frequency Capping

We present PolyUFC, an MLIR based compilation flow for uncore frequency capping that combines (performance and power) roofline analyses and polyhedral compilation-based static analysis for characterization of affine programs. We introduce a parametric mathematical model that links operational intensity and uncore frequency to derive frequency caps, validated through empirical evaluation on real hardware. By embedding these caps into Pluto optimized code generated by Polygeist, we achieve improvements in Energy Delay Product (EDP) up to 42% on compute-bound, and up to 54% on bandwidth-bound programs—carefully selected from ML-models from vision/NLP domains and PolyBench—over Intel UFS driver. Our framework is retargetable across multiple micro-architectures; and can handle multiple optimization goals like performance, energy and EDP, and is applicable across inter/intra dialects.

• Publication year

  2026

• Venue

  IEEE/ACM International Symposium on Code Generation and Optimization

• Publication date

  2026-01-31

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/CGO68049.2026.11395211
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Using Analytical Performance/Power Model and Fine-Grained DVFS to Enhance AI Accelerator Energy Efficiency

  2025cited by this paper
• Advancements of PAPI for the exascale generation

  2024cited by this paper
• Gemma 2: Improving Open Language Models at a Practical Size

  2024cited by this paper
• Falcon: A Scalable Analytical Cache Model

  2024cited by this paper
• Characterizing Power Management Opportunities for LLMs in the Cloud

  2024cited by this paper
• Expanding Datacenter Capacity with DVFS Boosting: A safe and scalable deployment experience

  2024cited by this paper
• PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation

  2024cited by this paper
• FCUFS: Core-Level Frequency Tuning for Energy Optimization on Intel Processors

  2024cited by this paper
• SYnergy: Fine-Grained Energy-Efficient Heterogeneous Computing for Scalable Energy Saving

  2023cited by this paper
• Llama 2: Open Foundation and Fine-Tuned Chat Models

  2023cited by this paper
• A ConvNet for the 2020s

  2022cited by this paper
• Combining Uncore Frequency and Dynamic Power Capping to Improve Power Savings

  2022cited by this paper
• BullsEye : Scalable and Accurate Approximation Framework for Cache Miss Calculation

  2022cited by this paper
• MLIR: Scaling Compiler Infrastructure for Domain Specific Computation

  2021cited by this paper
• PolyBench/Python: benchmarking Python environments with polyhedral optimizations

  2021cited by this paper
• (When) Do Multiple Passes Save Energy?

  2021cited by this paper
• Polygeist: Raising C to Polyhedral MLIR

  2021influential reference
• IOOpt: automatic derivation of I/O complexity bounds for affine programs

  2021cited by this paper
• Flex: High-Availability Datacenters With Zero Reserved Power

  2021cited by this paper
• Fine-Grained Powercap Allocation for Power-Constrained Systems Based on Multi-Objective Machine Learning

  2021cited by this paper
• SysScale: Exploiting Multi-domain Dynamic Voltage and Frequency Scaling for Energy Efficient Mobile Processors

  2020cited by this paper
• duf: Dynamic uncore frequency scaling to reduce power consumption

  2020cited by this paper
• RAPL in Action: Experiences in Using RAPL for Power Measurements

  2020cited by this paper
• Coordinated Priority-aware Charging of Distributed Batteries in Oversubscribed Data Centers

  2020cited by this paper
• Data Center Power Oversubscription with a Medium Voltage Power Plane and Priority-Aware Capping

  2020cited by this paper
• A fast analytical model of fully associative caches

  2019cited by this paper
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

  2019cited by this paper
• Automated derivation of parametric data movement lower bounds for affine programs

  2019cited by this paper
• nanoBench: A Low-Overhead Tool for Running Microbenchmarks on x86 Systems

  2019cited by this paper
• Uncore Power Scavenger: A Runtime for Uncore Power Conservation on HPC Systems

  2019cited by this paper
• Understanding the Impact of Dynamic Power Capping on Application Progress

  2019cited by this paper
• Language Models are Unsupervised Multitask Learners

  2019cited by this paper
• Tiramisu: A Polyhedral Compiler for Expressing Fast and Portable Code

  2018cited by this paper
• RAPL in Action

  2018cited by this paper
• Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions

  2018cited by this paper
• SWOOP: software-hardware co-design for non-speculative, execute-ahead, in-order cores

  2018cited by this paper
• Polyhedral auto-transformation with no integer linear programming

  2018cited by this paper
• A survival guide to presburger arithmetic

  2018cited by this paper
• Core and Uncore Joint Frequency Scaling Strategy

  2018cited by this paper
• The Computational Complexity of Presburger Arithmetic

  2018cited by this paper
• Synergistic CPU-GPU Frequency Capping for Energy-Efficient Mobile Games

  2017cited by this paper
• Clairvoyance: Look-ahead compile-time scheduling

  2017cited by this paper
• Towards Fine-grained Dynamic Tuning of HPC Applications on Modern Multi-Core Architectures

  2017cited by this paper
• Analytical modeling of cache behavior for affine programs

  2017influential reference
• Multiversioned decoupled access-execute: the key to energy-efficient compilation of general-purpose programs

  2016cited by this paper
• Static and Dynamic Frequency Scaling on Multicore CPUs

  2016influential reference
• Computer Architecture A Quantitative Approach

  2016cited by this paper
• The Pluto+ Algorithm

  2016cited by this paper
• Maximizing Performance Under a Power Cap: A Comparison of Hardware, Software, and Hybrid Techniques

  2016cited by this paper
• Wide Residual Networks

  2016cited by this paper
• On Using the Roofline Model with Lower Bounds on Data Movement

  2015cited by this paper
• An Energy Efficiency Feature Survey of the Intel Haswell Processor

  2015cited by this paper
• Core vs. uncore: The heart of darkness

  2015cited by this paper
• On characterizing the data movement complexity of computational DAGs for parallel execution

  2014cited by this paper
• Fix the code. Don't tweak the hardware: A new compiler approach to Voltage-Frequency scaling

  2014cited by this paper
• Towards energy-proportional computing for enterprise-class server workloads

  2013cited by this paper
• Folklore Confirmed: Compiling for Speed = Compiling for Energy

  2013cited by this paper
• SmartCap: User experience-oriented power adaptation for smartphone's application processor

  2013cited by this paper
• Polyhedral parallel code generation for CUDA

  2013cited by this paper
• A Roofline Model of Energy

  2013cited by this paper
• Polyhedral Extraction Tool

  2012cited by this paper
• Sniper: scalable and accurate parallel multi-core simulation

  2012cited by this paper
• ImageNet classification with deep convolutional neural networks

  2012cited by this paper
• The Forgotten 'Uncore': On the Energy-Efficiency of Heterogeneous Cores

  2012cited by this paper
• How much power oversubscription is safe and allowed in data centers

  2011cited by this paper
• Green governors: A framework for Continuously Adaptive DVFS

  2011cited by this paper
• RAPL: Memory power estimation and capping

  2010cited by this paper
• isl: An Integer Set Library for the Polyhedral Model

  2010influential reference
• Advanced Configuration and Power Interface

  2009cited by this paper
• Roofline: an insightful visual performance model for multicore architectures

  2009cited by this paper
• Collecting Performance Data with PAPI-C

  2009cited by this paper
• Iterative optimization in the polyhedral model

  2008cited by this paper
• A practical automatic polyhedral parallelizer and locality optimizer

  2008influential reference
• Iterative optimization in the polyhedral model: part ii, multidimensional time

  2008cited by this paper
• Iterative Optimization in the Polyhedral Model: Part I, One-Dimensional Time

  2007cited by this paper
• Counting Integer Points in Parametric Polytopes Using Barvinok's Rational Functions

  2007influential reference
• Generating cache hints for improved program efficiency

  2005cited by this paper
• A dynamic compilation framework for controlling microprocessor energy and performance

  2005cited by this paper
• Compile-time dynamic voltage scaling settings: opportunities and limits

  2003cited by this paper
• Reuse Distance as a Metric for Cache Behavior.

  2001cited by this paper
• Exact analysis of the cache behavior of nested loops

  2001cited by this paper
• A static power model for architects

  2000cited by this paper
• Cache miss equations: a compiler framework for analyzing and tuning memory behavior

  1999cited by this paper
• PAPI: A Portable Interface to Hardware Performance Counters

  1999cited by this paper
• OpenMP: an industry standard API for shared-memory programming

  1998cited by this paper
• Nonlinear Programming

  1997cited by this paper
• Some efficient solutions to the affine scheduling problem. Part II. Multidimensional time

  1992cited by this paper
• Some efficient solutions to the affine scheduling problem. I. One-dimensional time

  1992cited by this paper
• SUPER-EXPONENTIAL COMPLEXITY OF PRESBURGER ARITHMETIC

  1974cited by this paper
• Representation of Events in Nerve Nets and Finite Automata

  1951cited by this paper
• P ARM : Adaptive Resource Allocation for Datacenter Power Capping

  year unknowncited by this paper
• Languages and

  year unknowncited by this paper

• No citing papers are available for this paper.