Paper Citation Record · LEDGER

Nova: An End-to-End MLIR Compiler for Deep Learning

As of 9 August 2026, this Paper Citation Record lists 39 of 39 outbound references and 0 inbound Pith citation observations for arXiv:2608.00029.

A citation records a reference. It does not transfer a finding from one paper to another.

pith.paper-citation-record.v1
2608.00029 v1

Coverage vector

measured 39 of 39 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-04T01:44:49.125497Z

measured 39 of 39 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-09T06:31:02.800959+00:00

measured 0 of 0 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links

measured 0 of 1 external citation measurements

A source-named dated measurement, never combined with another source.

Source: cited_works

Reference resolution

39 of 39 outbound references displayed

verified exact0
verified fuzzy0
unresolved39
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 6b2a5a3a-3979-4bc0-b77a-c8372603e17d · outbound

This paper cites By capturing both forward activations and backward gradients into a unified graph, it unlocks the global visibility required for whole-step optimizations.

Nova: An End-to-End MLIR Compiler for Deep Learning By capturing both forward activations and backward gradients into a unified graph, it unlocks the global visibility required for whole-step optimizations

Reference 1

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.077041Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b1c872c9-cac1-4853-a601-eadb69a4f79b · outbound

This paper cites This represents the full training step natively within 2 MLIR, serving as the definitive foundation for both cross- operator fusion and compiler-native gradient synchronization.

Nova: An End-to-End MLIR Compiler for Deep Learning This represents the full training step natively within 2 MLIR, serving as the definitive foundation for both cross- operator fusion and compiler-native gradient synchronization

Reference 2

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.147871Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d7843cfd-5aa9-45a1-8e58-da6c754cde85 · outbound

This paper cites an unresolved cited work.

Nova: An End-to-End MLIR Compiler for Deep Learning Unresolved cited work

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.223514Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f0bf3f6f-4cef-4031-b89b-07129caa6fe8 · outbound

This paper cites parallel.

Nova: An End-to-End MLIR Compiler for Deep Learning parallel

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.296596Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 06d9497a-79c6-43bb-a284-a1b162ae5acc · outbound

This paper cites Nova time-shifts nvgpu.device_async_copy operations by depth − 1 iterations, guaranteeing that itera- tion i+2 prefetches from HBM while iteration i computes on resident memory.

Nova: An End-to-End MLIR Compiler for Deep Learning Nova time-shifts nvgpu.device_async_copy operations by depth − 1 iterations, guaranteeing that itera- tion i+2 prefetches from HBM while iteration i computes on resident memory

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.371972Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1df89411-c041-4bcb-afde-0217c0de178e · outbound

This paper cites an unresolved cited work.

Nova: An End-to-End MLIR Compiler for Deep Learning Unresolved cited work

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.444133Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 369a9268-8a61-445b-a7e9-19aa1300da7c · outbound

This paper cites The hardware ldmatrix instruction is highly rigid, and its ldmatrix.trans variant is strictly limited to 16-bit granularity.

Nova: An End-to-End MLIR Compiler for Deep Learning The hardware ldmatrix instruction is highly rigid, and its ldmatrix.trans variant is strictly limited to 16-bit granularity

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.516366Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6cf57b15-4b9f-48c3-ac56-8a522ccbac7b · outbound

This paper cites To optimize them, Nova pins the accumulators directly in registers and expands clustered subgroup reductions into gpu.shuffle 6 butterfly trees.

Nova: An End-to-End MLIR Compiler for Deep Learning To optimize them, Nova pins the accumulators directly in registers and expands clustered subgroup reductions into gpu.shuffle 6 butterfly trees

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.617036Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 64605d85-231c-43dd-a636-47844e5f6724 · outbound

This paper cites All tested implementations remained within the 10−4 relative error band.

Nova: An End-to-End MLIR Compiler for Deep Learning All tested implementations remained within the 10−4 relative error band

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.680645Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 43519325-4a2c-44a9-bc93-366efba83ca6 · outbound

This paper cites At 144M parameters, both PyTorch and the eager baseline run out of memory (OOM) on the 12 GB card, whereas Nova and XLA successfully fit and continue training.

Nova: An End-to-End MLIR Compiler for Deep Learning At 144M parameters, both PyTorch and the eager baseline run out of memory (OOM) on the 12 GB card, whereas Nova and XLA successfully fit and continue training

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.787792Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4b9d8b8e-fd9b-423e-8e20-44028903e59e · outbound

This paper cites By operating en- tirely on a single fused graph, Nova proves that on a memory- constrained GPU, it is possible to lower the memory floor and raise throughput simultaneously.

Nova: An End-to-End MLIR Compiler for Deep Learning By operating en- tirely on a single fused graph, Nova proves that on a memory- constrained GPU, it is possible to lower the memory floor and raise throughput simultaneously

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.863106Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ad12ddab-edad-416b-b221-e01cbc77b68c · outbound

This paper cites At 144M parame- ters, PyTorch and XLA both encounter OOM errors, whereas Nova successfully fits the DDP step.

Nova: An End-to-End MLIR Compiler for Deep Learning At 144M parame- ters, PyTorch and XLA both encounter OOM errors, whereas Nova successfully fits the DDP step

Reference 12

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:46.960003Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 908ece37-c54d-47e7-9b1b-63c6b324242e · outbound

This paper cites It efficiently overlaps gradient synchronization with computation.

Nova: An End-to-End MLIR Compiler for Deep Learning It efficiently overlaps gradient synchronization with computation

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.078424Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation dc046a27-db3f-41ad-8649-5baed81d7b1b · outbound

This paper cites Figure 10:Device memory consumption (MB) in Mixed preci- sion training.

Nova: An End-to-End MLIR Compiler for Deep Learning Figure 10:Device memory consumption (MB) in Mixed preci- sion training

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.146700Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 829c65fe-32b9-4938-8fc5-8f8289a86c6c · outbound

This paper cites 7 Discussions and limitations In this section, we analyze Nova’s architectural trade-offs, evaluating the limits of its performance characteristics and open challenges for scaling.

Nova: An End-to-End MLIR Compiler for Deep Learning 7 Discussions and limitations In this section, we analyze Nova’s architectural trade-offs, evaluating the limits of its performance characteristics and open challenges for scaling

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.207169Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a7f9ee6c-ba27-457e-beff-8eb9217e85e9 · outbound

This paper cites IREE: Intermediate Representation Execution En- vironment.

Nova: An End-to-End MLIR Compiler for Deep Learning IREE: Intermediate Representation Execution En- vironment

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.286458Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 18086e39-76ce-4999-9d8b-5358eb890999 · outbound

This paper cites OpenXLA: A compiler for machine learning.

Nova: An End-to-End MLIR Compiler for Deep Learning OpenXLA: A compiler for machine learning

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.343829Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4f824d65-e7f2-4028-9527-992194256e61 · outbound

This paper cites Operator Fusion in XLA: Analysis and Evaluation.

Nova: An End-to-End MLIR Compiler for Deep Learning Operator Fusion in XLA: Analysis and Evaluation

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.428288Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 85f1d730-47f1-4e87-8d4c-8d0eecf020fc · outbound

This paper cites MLIR: Scaling Compiler Infrastructure for Domain- Specific Computation.

Nova: An End-to-End MLIR Compiler for Deep Learning MLIR: Scaling Compiler Infrastructure for Domain- Specific Computation

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.511502Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f6d47976-55fb-41d3-bf8f-cb9b0d1bc3ed · outbound

This paper cites Halide: A Language and Compiler for Opti- mizing Parallelism, Locality, and Recomputation in Image Processing Pipelines.

Nova: An End-to-End MLIR Compiler for Deep Learning Halide: A Language and Compiler for Opti- mizing Parallelism, Locality, and Recomputation in Image Processing Pipelines

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.581293Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 23ad3ed3-a8e8-40fd-b9d4-17e31e457ec1 · outbound

This paper cites Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions.

Nova: An End-to-End MLIR Compiler for Deep Learning Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.637724Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2be29e31-f76a-4cb9-b1f7-95e54b9dc1b8 · outbound

This paper cites cuDNN: Efficient Primitives for Deep Learning.

Nova: An End-to-End MLIR Compiler for Deep Learning cuDNN: Efficient Primitives for Deep Learning

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.730009Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b15599c9-198f-46f8-bef1-b8d8db21aa32 · outbound

This paper cites TVM: An Automated End-to-End Optimizing Compiler for Deep Learning.

Nova: An End-to-End MLIR Compiler for Deep Learning TVM: An Automated End-to-End Optimizing Compiler for Deep Learning

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.841074Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6bd8cc2d-458e-4acb-9ce5-dd77b9ddddbc · outbound

This paper cites PyTorch: An Imperative Style, High-Performance Deep Learning Library.

Nova: An End-to-End MLIR Compiler for Deep Learning PyTorch: An Imperative Style, High-Performance Deep Learning Library

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:47.962392Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a23b4fa5-9f23-4d6d-8f9e-e7eea7f192fb · outbound

This paper cites TensorFlow: A System for Large-Scale Machine Learning.

Nova: An End-to-End MLIR Compiler for Deep Learning TensorFlow: A System for Large-Scale Machine Learning

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.076306Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 342254c8-20b7-42e1-9864-85bebc0e2e46 · outbound

This paper cites Composable and Modular Code Generation in MLIR: A Structured and Retargetable Approach to Tensor Compiler Construction.

Nova: An End-to-End MLIR Compiler for Deep Learning Composable and Modular Code Generation in MLIR: A Structured and Retargetable Approach to Tensor Compiler Construction

Reference 26

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.229027Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 77df85a4-e998-4474-ac03-aeec1cfae9dc · outbound

This paper cites Triton: An Intermediate Language and Compiler for Tiled Neural Network Computations.

Nova: An End-to-End MLIR Compiler for Deep Learning Triton: An Intermediate Language and Compiler for Tiled Neural Network Computations

Reference 27

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.297577Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 81cfc8be-1b58-4ecc-8623-cd31d6781211 · outbound

This paper cites Data Movement Is All You Need: A Case Study on Optimizing Transformers.

Nova: An End-to-End MLIR Compiler for Deep Learning Data Movement Is All You Need: A Case Study on Optimizing Transformers

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.387900Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation da0409ad-d884-4f62-86fe-dcd626631e0a · outbound

This paper cites The MLIR Transform Dialect: Your Compiler Is More Powerful Than You Think.

Nova: An End-to-End MLIR Compiler for Deep Learning The MLIR Transform Dialect: Your Compiler Is More Powerful Than You Think

Reference 29

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.447876Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c1ba18a3-f3fb-422b-8070-f11b82a985b9 · outbound

This paper cites Ansor: Generating High-Performance Tensor Programs for Deep Learning.

Nova: An End-to-End MLIR Compiler for Deep Learning Ansor: Generating High-Performance Tensor Programs for Deep Learning

Reference 30

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.511066Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4315f462-1de6-4588-8fea-3b7d05cb5278 · outbound

This paper cites cuBLAS Library User Guide.

Nova: An End-to-End MLIR Compiler for Deep Learning cuBLAS Library User Guide

Reference 31

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.599965Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8fb5fe0b-d5fe-4d91-9cf2-f575e5be191b · outbound

This paper cites CUTLASS: CUDA Templates for Linear Algebra Subroutines.

Nova: An End-to-End MLIR Compiler for Deep Learning CUTLASS: CUDA Templates for Linear Algebra Subroutines

Reference 32

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.668457Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 122a1c7e-d264-42f3-b883-9a37679ff6c4 · outbound

This paper cites FlashAttention: Fast and Memory-Efficient Exact Atten- tion with IO-Awareness.

Nova: An End-to-End MLIR Compiler for Deep Learning FlashAttention: Fast and Memory-Efficient Exact Atten- tion with IO-Awareness

Reference 33

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.754989Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1ac5a38b-9c92-439c-ba88-2e6fa9d0bfa4 · outbound

This paper cites PyTorch Distributed: Experiences on Accelerating Data Parallel Training.

Nova: An End-to-End MLIR Compiler for Deep Learning PyTorch Distributed: Experiences on Accelerating Data Parallel Training

Reference 34

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.811955Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b067abe4-8107-4802-9038-5fff65b18e09 · outbound

This paper cites Automatic Differentiation in Machine Learning: a Survey.

Nova: An End-to-End MLIR Compiler for Deep Learning Automatic Differentiation in Machine Learning: a Survey

Reference 35

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.866741Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0b9c0e5b-2f4c-4c9d-8da3-e88fcbfe0c56 · outbound

This paper cites JAX: Composable Transformations of Python+NumPy Programs.

Nova: An End-to-End MLIR Compiler for Deep Learning JAX: Composable Transformations of Python+NumPy Programs

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:48.965402Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9fbb5862-e225-4f48-9d5a-65bfe239036c · outbound

This paper cites Roofline: An Insightful Visual Performance Model for Multicore Architectures.

Nova: An End-to-End MLIR Compiler for Deep Learning Roofline: An Insightful Visual Performance Model for Multicore Architectures

Reference 37

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:49.008467Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ca78294d-e0b0-4632-8395-45e14830b622 · outbound

This paper cites CUDA C++ Programming Guide.

Nova: An End-to-End MLIR Compiler for Deep Learning CUDA C++ Programming Guide

Reference 38

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:49.066748Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3d7001f9-e9de-479c-bc51-5c1f2c5211be · outbound

This paper cites LLVM/MLIR 21.1.6; CUDA Toolkit 13.0.

Nova: An End-to-End MLIR Compiler for Deep Learning LLVM/MLIR 21.1.6; CUDA Toolkit 13.0

Reference 39

Resolution: unresolved; no resolver link, observed 2026-08-04T01:44:49.125497Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

No inbound Pith citation observations are available.