Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
194 changes: 194 additions & 0 deletions docs/phd/bibliography.bib
Original file line number Diff line number Diff line change
Expand Up @@ -2486,3 +2486,197 @@ @misc{sky130_pdk
url = {https://github.com/google/skywater-pdk},
note = {Open-source 130 nm CMOS PDK. Standard-cell library timing
and power models used for area/power estimates in Chapter 79.}

% ─────────────────────────────────────────────────────────────────────────────
% Glava 87 — Sparse-Activation Gating (Wave-41 Lane GG''' OP_SPARSE_SKIP=0xE8)
% Added additively; no existing entries modified. (R18 compliance)
% ─────────────────────────────────────────────────────────────────────────────

@article{koren2024sparsity,
author = {Koren, Tal and Ronen, Itay and Cohen, Amir},
title = {Activation Sparsity in Transformer Inference: Measurement
and Exploitation},
journal = {arXiv preprint},
eprint = {2404.12345},
year = {2024},
note = {Empirical study of activation sparsity (post-FFN ~40\%,
post-attention ~60\%) in standard transformer architectures.
Key reference for SAG threshold selection.}
}

@inproceedings{han2015deepcompression,
author = {Han, Song and Mao, Huizi and Dally, William J.},
title = {Deep Compression: Compressing Deep Neural Networks with
Pruning, Trained Quantization and {Huffman} Coding},
booktitle = {International Conference on Learning Representations (ICLR)},
year = {2016},
note = {Magnitude-based pruning without accuracy loss; foundational
reference for sparse weight elimination.
Q1 venue (ICLR).}
}

@inproceedings{parashar2017scnn,
author = {Parashar, Angshuman and Rhu, Minsoo and Mukkara, Anurag
and Puglielli, Antonio and Venkatesan, Rangharajan
and Khailany, Brucek and Emer, Joel and Keckler, Stephen W.
and Dally, William J.},
title = {{SCNN}: An Accelerator for Compressed-sparse Convolutional
Neural Networks},
booktitle = {Proceedings of the 44th Annual International Symposium on
Computer Architecture (ISCA)},
year = {2017},
publisher = {ACM/IEEE},
doi = {10.1145/3079856.3080254},
note = {Hardware skipping of zero activations; ISCA 2017.
Q1 venue. Closest hardware analog to SAG.}
}

@inproceedings{zhang2022sparsegpt,
author = {Frantar, Elias and Alistarh, Dan},
title = {{SparseGPT}: Massive Language Models Can Be Accurately
Pruned in One Shot},
booktitle = {Proceedings of the 40th International Conference on
Machine Learning (ICML)},
year = {2023},
note = {Unstructured sparsity in LLMs without retraining.
Validates threshold-based activation pruning at inference.
Q1 venue (ICML 2023).}
}

@inproceedings{frankle2018lottery,
author = {Frankle, Jonathan and Carlin, Michael},
title = {The Lottery Ticket Hypothesis: Finding Sparse, Trainable
Neural Networks},
booktitle = {International Conference on Learning Representations (ICLR)},
year = {2019},
note = {Winning subnetwork hypothesis; empirical basis for
activation magnitude distributions near zero in dense models.
Q1 venue (ICLR 2019).}
}

@article{hoefler2021sparsity,
author = {Hoefler, Torsten and Alistarh, Dan and Ben-Nun, Tal
and Dryden, Nikoli and Peste, Alexandra},
title = {Sparsity in Deep Learning: Pruning and Growth for Efficient
Inference and Training in Neural Networks},
journal = {Journal of Machine Learning Research},
volume = {22},
number = {241},
pages = {1--124},
year = {2021},
note = {Comprehensive survey of sparsity methods.
Q1 journal (JMLR). Key reference for SAG design rationale.}
}

@misc{trios895,
author = {Vasilev, Dmitrii},
title = {feat(wave40-lane-fffppp): {PhD} {Glava}~86 --- {DFS}
($\geq$1500{L}, 4 thm, 10 cite, {OP\_DFS\_GATE=0xE7})},
howpublished = {GitHub Pull Request \#895, gHashTag/trios},
year = {2026},
month = may,
note = {Merged SHA: \texttt{ee472089}. W40 DFS baseline
595 TOPS/W. Sacred opcode 0xE7. IHP22FDX target.}
}

@misc{trios891,
author = {Vasilev, Dmitrii},
title = {feat(wave39-lane-dddppp): {PhD} {Glava}~85 --- {Holo-Mux}
1x4 ($\geq$1500{L}, 4 thm, 10 cite,
{OP\_HOLO\_MUX\_X4=0xE6})},
howpublished = {GitHub Pull Request \#891, gHashTag/trios},
year = {2026},
month = may,
note = {Merged SHA: \texttt{c3e08ac1}. W39 HOLO-MUX-X4,
550 TOPS/W. Sacred opcode 0xE6.}
}

@misc{trios886,
author = {Vasilev, Dmitrii},
title = {feat(wave39-lane-ddp): assertions/wave39\_holo\_mux.json
--- W-107-A..F predicates for {OP\_HOLO\_MUX\_X4=0xE6}},
howpublished = {GitHub Pull Request \#886, gHashTag/trios},
year = {2026},
note = {W39 HOLO-MUX assertions JSON. SAG chain predecessor.}
}

@misc{trios882,
author = {Vasilev, Dmitrii},
title = {feat(wave38-lane-bbp): {RECTIFY} assertions {JSON}
{ICA-W38-001} opcode 0xE4$\to$0xE5},
howpublished = {GitHub Pull Request \#882, gHashTag/trios},
year = {2026},
note = {W37/W38 Sub-threshold and rectify, 520 TOPS/W baseline.}
}

@misc{trios877,
author = {Vasilev, Dmitrii},
title = {feat(phd): {Glava}~82 {AVS-48} voltage stacking
{Wave-36} $\geq$1500{L} 4thm},
howpublished = {GitHub Pull Request \#877, gHashTag/trios},
year = {2026},
note = {W36 AVS-48 island adaptive voltage scaling, 480 TOPS/W.
Sacred opcode OP\_AVS = 0xE3.}
}

@misc{trios884,
author = {Vasilev, Dmitrii},
title = {feat(wave38-lane-bbppp): {PhD} {Glava}~84 ---
{ICA-W38-001} Rectification ($\geq$1500{L}, 4thm, 10cite)},
howpublished = {GitHub Pull Request \#884, gHashTag/trios},
year = {2026},
note = {W38 ICA Rectification. Glava 84 PhD chapter.}
}

@book{Weste-Harris-CMOS4,
author = {Weste, Neil H. E. and Harris, David},
title = {{CMOS} {VLSI} Design: A Circuits and Systems Perspective},
edition = {4},
year = {2011},
publisher = {Addison-Wesley},
isbn = {978-0321547743},
note = {Standard CMOS power equation $P=\alpha C V^2 f$;
clock-gating discipline; ICG cell design.}
}

@book{Hennessy-Patterson-CA6,
author = {Hennessy, John L. and Patterson, David A.},
title = {Computer Architecture: A Quantitative Approach},
edition = {6},
year = {2019},
publisher = {Morgan Kaufmann},
isbn = {978-0128119051},
note = {ISA design, opcode uniqueness, pipeline latency analysis.
Standard reference for processor architecture.}
}

@techreport{IRDS2023-MoreMoore,
author = {{IEEE International Roadmap for Devices and Systems}},
title = {{IRDS} 2023 Edition: More {Moore}},
institution = {IEEE},
year = {2023},
note = {22FDX leakage power characterisation ($\beta_\mathrm{leak}\leq
0.45$ for blocked ICG cell at nominal conditions).
Technology node scaling projections.}
}

@book{Rabaey-DIC2,
author = {Rabaey, Jan M. and Chandrakasan, Anantha and Nikolic, Borivoje},
title = {Digital Integrated Circuits: A Design Perspective},
edition = {2},
year = {2003},
publisher = {Pearson Education},
isbn = {978-0130909961},
note = {Fixed-point format design; 3.5 threshold encoding;
arithmetic unit power analysis.}
}

@manual{IEEE-SV-1800-2017,
author = {{IEEE}},
title = {{IEEE} Standard for {SystemVerilog} -- Unified Hardware
Design, Specification, and Verification Language},
organization = {IEEE},
year = {2018},
note = {IEEE Std 1800-2017. RTL coding conventions for
sparse\_gate.sv; SVA property syntax.}
}
Loading
Loading