GPUReplay: A 50-KB GPU Stack for Client ML

Felix Xiaozhu Lin; Heejin Park

RECORD

GPUReplay: A 50-KB GPU Stack for Client ML

Heejin Park, Felix Xiaozhu Lin

Distributed, Parallel, and Cluster Computing · Submitted 4 May 2021

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim We show that GPUReplay works with a variety of integrated GPU hardware, GPU APIs, ML frameworks, and 33 neural network (NN) implementations for inference or training.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 6,129 input tokens.

What the abstract itself states, as probabilities: claims first 0.20 · states limitation 0.11 · human participants 0.02 · controlled comparison 0.17 · randomized assignment 0.01 · states quantitative result 0.97 · compares with prior results 0.25 · mentions code or data availability 0.98.

The abstract, sentence by sentence

arXiv listing abstract v4. Results and the main claim are underlined; show every role.

GPUReplay (GR) is a novel way for deploying GPU-accelerated computation on mobile and embedded devices. It addresses high complexity of a modern GPU stack for deployment ease and security. The idea is to record GPU executions on the full GPU stack ahead of time and replay the executions on new input at run time. We address key challenges towards making GR feasible, sound, and practical to use. The resultant replayer is a drop-in replacement of the original GPU stack. It is tiny (50 KB of executable), robust (replaying long executions without divergence), portable (running in a commodity OS, in TEE, and baremetal), and quick to launch (speeding up startup by up to two orders of magnitude). We show that GPUReplay works with a variety of integrated GPU hardware, GPU APIs, ML frameworks, and 33 neural network (NN) implementations for inference or training. The code is available at https://github.com/bakhi/GPUReplay.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	GPUReplay (GR) is a novel way for deploying GPU-accelerated computation on mobile and embedded devices.	contribution 1.00	0.47	0.10	asserted 1.00
s1	It addresses high complexity of a modern GPU stack for deployment ease and security.	background 0.70	0.27	0.05	asserted 1.00
s2	The idea is to record GPU executions on the full GPU stack ahead of time and replay the executions on new input at run time.	method 0.88	0.38	0.07	asserted 0.97
s3	We address key challenges towards making GR feasible, sound, and practical to use.	contribution 0.93	0.79	0.04	asserted 0.80
s4	The resultant replayer is a drop-in replacement of the original GPU stack.	contribution 0.51	0.81	0.14	asserted 1.00
s5	It is tiny (50 KB of executable), robust (replaying long executions without divergence), portable (running in a commodity OS, in TEE, and baremetal), and quick to launch (speeding up startup by up to two orders of magnitude).	result 0.76	0.91	0.34	asserted 1.00
s6	We show that GPUReplay works with a variety of integrated GPU hardware, GPU APIs, ML frameworks, and 33 neural network (NN) implementations for inference or training.	result 0.94	0.93	0.21	asserted 1.00
s7	The code is available at https://github.com/bakhi/GPUReplay.	contribution 0.74	0.38	0.02	asserted 0.99

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.97.

What was examined, and what was not

sources read

arXiv listing abstract v4

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

GPUReplay (GR) is a novel way for deploying GPU-accelerated computation on mobile and embedded devices. It addresses high complexity of a modern GPU stack for deployment ease and security. The idea is to record GPU executions on the full GPU stack ahead of time and replay the executions on new input at run time. We address key challenges towards making GR feasible, sound, and practical to use. The resultant replayer is a drop-in replacement of the original GPU stack. It is tiny (50 KB of executable), robust (replaying long executions without divergence), portable (running in a commodity OS, in TEE, and baremetal), and quick to launch (speeding up startup by up to two orders of magnitude). We show that GPUReplay works with a variety of integrated GPU hardware, GPU APIs, ML frameworks, and 33 neural network (NN) implementations for inference or training. The code is available at https://github.com/bakhi/GPUReplay.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

What this paper does with the arXiv papers it cites

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

What it cites · 1 sentence without a confident label

Cites as background 1 paper

Much work has transformed ML workloads rather than the GPU stack; outsourcing security-sensitive compute to TEE, they preserve data/model privacy or ensure compute integrity [ref] .

Confidential Inference via Ternary Model Partitioning · Related Work · 0.90

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.