REVIEW 18 cited by
MLIR: A Compiler Infrastructure for the End of Moore's Law
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
This work presents MLIR, a novel approach to building reusable and extensible compiler infrastructure. MLIR aims to address software fragmentation, improve compilation for heterogeneous hardware, significantly reduce the cost of building domain specific compilers, and aid in connecting existing compilers together. MLIR facilitates the design and implementation of code generators, translators and optimizers at different levels of abstraction and also across application domains, hardware targets and execution environments. The contribution of this work includes (1) discussion of MLIR as a research artifact, built for extension and evolution, and identifying the challenges and opportunities posed by this novel design point in design, semantics, optimization specification, system, and engineering. (2) evaluation of MLIR as a generalized infrastructure that reduces the cost of building compilers-describing diverse use-cases to show research and educational opportunities for future programming languages, compilers, execution environments, and computer architecture. The paper also presents the rationale for MLIR, its original design principles, structures and semantics.
Forward citations
Cited by 18 Pith papers
-
Triton for MTIA: Bridging the Programming Model Gaps for Custom AI Accelerators
A Triton compiler backend, TorchInductor adaptations, and small language extensions let Meta's MTIA-2i run Triton kernels competitively with expert-tuned C++ in production.
-
Can Large Language Models Recover Semantic Optimization Opportunities That Compilers Miss?
On the new SeGaBench benchmark, the strongest of five LLMs recovered compiler-missed optimization semantics and produced validated, speedup-delivering code changes in a large majority of cases.
-
KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta
An agentic kernel-coding system combining tree search with hardware-knowledge retrieval generated optimized Triton kernels for NVIDIA, AMD, and Meta's MTIA accelerators: 100% correctness on 480 operator-platform confi...
-
GraphMend: Code Transformations for Fixing Graph Breaks in PyTorch 2
GraphMend uses two Jaseci-based code transformations to eliminate dynamic-control-flow and side-effect graph breaks in PyTorch 2, reducing breaks to zero in six of eight Hugging Face models and yielding up to 75% late...
-
Re-thinking Memory-Bound Limitations in CGRAs
Adding a cache hierarchy, CGRA-specific runahead prefetching, and per-PE cache reconfiguration allows CGRAs to process irregular memory-access kernels at near-ideal speed with 1.27% of the storage.
-
Guided Tensor Lifting
STAGG combines LLM-generated candidate programs, a learned probabilistic grammar, and enumerative search to lift dense tensor C code into TACO, solving 76 of 77 benchmarks.
-
Rulebook: bringing co-routines to reinforcement learning environments
Rulebook compiles coroutine-based RL environment descriptions into inspectable, serializable, checkable state machines with C-comparable runtime performance.
-
Triton-distributed: Programming Overlapping Kernels on Distributed AI Systems with the Triton Compiler
Triton-distributed adds OpenSHMEM-compatible communication primitives to the Triton compiler, enabling Python-level distributed kernels with fine-grained computation-communication overlap that match or beat hand-writt...
-
Stream-HLS: Towards Automatic Dataflow Acceleration
Stream-HLS automatically converts affine multi-kernel C/C++ or PyTorch programs into streaming FPGA dataflow designs using a combined MINLP scheduler, with RTL-simulated geometric mean speedups up to 79.43x over prior...
-
DeComp2: Description Complexity aware Decomposition
Adding a description-length term to the quantum-compiler objective changes the chosen circuit on ~0.3% of tested single-qubit targets, showing gate-count-only compilation discards genuinely structured alternatives.
-
From Intent to Infrastructure: LLM-Driven Agent Compilers for ISAC Networks
An LLM-driven compilation layer for ISAC networks parses engineer intent into a policy graph and adapts configurations at three time scales.
-
ATLAS: Multi-View Code Representation Tool for C and C++ Source Programs
ATLAS claims no-build aligned AST/CFG/DFG extraction for C/C++ at 96.8%/91.7% CFG correctness, and a 34.7-point LLM line-coverage lift when fed its CFG paths.
-
Improving compiler support for SIMD offload using Arm Streaming SVE
Current LLVM and MLIR auto-vectorizers produce slow Streaming SVE code because their cost models ignore streaming-mode overheads and memory hazards.
-
A Unified Framework for Mapping and Synthesis of Approximate R-Blocks CGRAs
A CGRA design flow that maps neural network channels onto approximate DRUM multipliers and static voltage islands, reporting ~30% power reduction for MobileNetV2 with only output RMSE, not top-1 accuracy, as the quali...
-
FluidML: Fast and Memory Efficient Inference Optimization
FluidML combines graph splitting, dynamic programming, and greedy memory allocation to optimize ML inference memory layout, but its reported improvements are inconsistent across models and its headline numbers contrad...
-
An MLIR-Based Compilation Method for Large Language Models
A two-dialect MLIR compiler (TopOp→TpuOp) that statically splits each layer into prefill, prefill_kv, and decode stages achieves 78–83% DDR bandwidth on the BM1684X TPU.
-
Tensor Evolution: A Framework for Fast Evaluation of Tensor Computations using Recurrences
Tensor Evolution extends the Chain of Recurrences theory from scalars to tensors, giving rewrite rules that simplify loop-carried tensor computations into closed-form expressions.
-
Pattern-Guided Design Space Exploration for FPGA Accelerator Design
Pattern-guided design-space exploration reduces HLS evaluations from 140 to 29 on six kernels while matching the best achieved latency, but relies on hand-assigned patterns and a weak baseline.
Discussion (0). Continue with ORCID to comment.