pith. sign in

arxiv: 2605.12959 · v3 · pith:GD2EHGZVnew · submitted 2026-05-13 · 💻 cs.AR

A detailed algorithmic study on a reuse-aware, near memory, all-digital Ising machine

Siddhartha Raman Sundara Raman , Lizy K. John , Jaydeep P. Kulkarni This is my paper

Pith reviewed 2026-06-30 21:38 UTC · model grok-4.3

classification 💻 cs.AR
keywords Ising machineprocessing-in-memoryL1 cacheSRAMoptimization acceleratorenergy efficiencyCPU integrationall-digital design
0
0 comments X

The pith

SACHI turns a CPU's L1 cache into an all-digital Ising accelerator using SRAM processing-in-memory, cutting energy use by 80x and raising performance 300x over prior designs.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that existing SRAM cache hardware can be repurposed for Ising-model optimization without adding analog circuits or dedicated accelerators. By mapping spins and interactions directly onto cache lines and performing updates in place, the design keeps everything digital and inside the CPU pipeline. A sympathetic reader would care because this removes the reliability problems of embedded memory and the overhead of off-chip data movement for problems such as traveling salesman, image segmentation, and molecular dynamics. The central mechanism is reuse-aware mapping that lets the same cache resources serve both normal processor traffic and parallel spin updates, producing reuse factors up to 4000 times.

Core claim

SACHI is an all-digital Ising architecture implemented by repurposing the L1 cache of a CPU using SRAM-based processing-in-memory techniques. It eliminates the need for ADCs/DACs, improves reliability compared to prior approaches, and enables Ising acceleration with minimal hardware overhead integrated into the CPU pipeline. The design achieves 300x performance improvement and 80x energy reduction across applications including asset allocation, molecular dynamics, image segmentation, and traveling salesman problems, with reuse factors up to 4000x.

What carries the argument

SRAM-based processing-in-memory inside the L1 cache that performs spin-state updates and interaction summations directly on cache lines while the CPU pipeline continues to run.

If this is right

  • Optimization workloads can run inside the same core that executes the rest of the program instead of requiring a separate accelerator chip.
  • High-precision Ising problems become feasible because all operations stay in the digital domain without quantization from ADCs.
  • The same cache hardware can be shared between normal memory traffic and acceleration, reducing the area cost of adding Ising capability.
  • Applications with repeated problem instances see large reuse of the mapped spin interactions stored in cache.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Future CPUs could expose a simple instruction to switch a cache slice into Ising mode, allowing software to request acceleration without driver or OS changes.
  • The approach might extend to other cache levels or to vector units if the reuse-aware mapping rules can be generalized beyond L1.
  • Designers of domain-specific accelerators might reconsider whether full custom memory is needed when existing SRAM can be reused with modest control logic.

Load-bearing premise

The SRAM structures already present in the L1 cache can perform the required Ising updates without introducing pipeline stalls, reliability failures, or process-variation problems.

What would settle it

Fabricated silicon measurements showing that the added control logic for spin updates causes more than a few percent increase in cache miss rate or produces incorrect spin values under normal voltage and temperature variation.

Figures

Figures reproduced from arXiv: 2605.12959 by Jaydeep P. Kulkarni, Lizy K. John, Siddhartha Raman Sundara Raman.

Figure 1
Figure 1. Figure 1: Ising machines provide higher accuracy and lower solution time. (i) Top - Solution accuracy under iso-performance condition, and (ii) Bottom - execution time normalized to Ising under iso-accuracy, comparing Genetic Algorithm (GA) and Ising for a) traveling salesman, and b) image segmentation problems. offering innovative approaches that leverage the principles of statistical mechanics to represent and sol… view at source ↗
Figure 3
Figure 3. Figure 3: a) 3*3 King’s Graph, commonly used for mapping Ising models b) Sparsely connected 3*3 Graph c) Fully connected 3*3 Graph. The target spin is shown in blue in all the figures COPs indicate that SACHI offers 160x/36x/286x/300x better performance and 79x/72x/80x/75x better en￾ergy over BRIM for molecular dynamics/asset allo￾cation/image segmentation/traveling salesman problems. Furthermore, SACHI offers a spe… view at source ↗
Figure 4
Figure 4. Figure 4: SACHI in comparison to to state of the art Ising architectures BRIM and Ising-CIM - SACHI is a repurposable, scalable, reconfigurable near-memory architecture with no modifications to the memory array, no ADCs/DACs, achieving better reuse/energy as compared to prior Ising accelerators like BRIM [1], Ising-CIM [49] The local spin update might result in H being trapped in a local minimum. Simulated annealing… view at source ↗
Figure 5
Figure 5. Figure 5: Real-life COPs have a wide range of problem sizes with varied graph connectivity, varied minimum resolution (4-7 bits) to achieve 90% accuracy under iso-performance condition. Using a fixed 8-bit IC for all COPs results in additional data movement cost because 8-bit COP overflows when placed in 64KB L1 cache while lower resolution (4-7bit) compute fits inside the L1 cache, motivating a reconfigurable, scal… view at source ↗
Figure 7
Figure 7. Figure 7: Overview of SACHI- SACHI’s compute array is mapped onto L1 cache, storage array onto L2 cache with minimal near-L1 logic (0.3% of CPU area), enabling easy integration into CPU pipeline. DRAM controller prefetches requests based on the number of uncomputed rows in compute array to minimize data movement cost between DRAM and storage array. 1V range, requiring an infeasibly small voltage resolution for DAC o… view at source ↗
Figure 9
Figure 9. Figure 9: Scalability to any graph size - a) Ising-CIM approach: duplicate edge cells (dark brown) onto adjacent CIM arrays. 1 indicates updated spin value computed in the array. 2 indicates writeback of the same updated spin value for non-edge cells. 3 indicates the broadcast of updated spin value in the case of edge cells (duplication) to adjacent CIM arrays. This approach minimizes interaction between compute arr… view at source ↗
Figure 10
Figure 10. Figure 10: SACHI’s reconfigurability achieved using mixed encoding scheme, with -1/+1 spins stored as 0/1, ICs encoded in 2’s complement form to enable in-memory XNOR for dot product between Jij and σj without DAC/ADC (unlike BRIM). (S XNOR IC)/((S XNOR IC)+1) is computed to enable multi-bit signed IC dot product (unlike Ising-CIM). 9-bit Jij= 135 (9’h087), -135 (9’h179) and 3-bit Jij= 3 (3’h3), -3 (3’h5) product wi… view at source ↗
Figure 11
Figure 11. Figure 11: SACHI’s use of in-memory XNOR - a) 2 8T SRAM bitcells in the same column storing complimentary values of spin/IC with the bitcells computing S’&J’, S&J b) RBL discharge c) RBL retain for XNOR compute with no memory array modifications/requirement of DAC and ADC d) In-memory XNOR test-structure in a silicon prototype e) Oscilloscope capture confirming bitline discharge for S=1 XNOR J=1 f) SACHI reuse in di… view at source ↗
Figure 12
Figure 12. Figure 12: a.1) SACHI(n1a) a.2) SACHI(n1b) for graph in Fig.2 with 3-bit Jij. SACHI(n1a) computes ithbit of all ICs before proceeding to the (i+1)th bit (reuse=1). SACHI(n1b) computes all bits of an IC before proceeding to the next IC (reuse=1). Only 1 column is highlighted to indicate reuse of 1. b) Bitwise XNOR requires XNOR queue to queue the different bits computed b.1) SACHI(n1a) requires more number of entries… view at source ↗
Figure 14
Figure 14. Figure 14: SACHI(n3): Mixed stationary design making use of reuse-aware compute to perform XNOR across all Jij bits and neighbors of target spin in a single cycle in Phase 1, improving maximum reuse to (neighbors)*(Jij resolution) (entire memory array highlighted). The decision logic and adder in Phase 3 are modified to support high throughput array, making it a mixed stationary design. σi is shared across a complet… view at source ↗
Figure 15
Figure 15. Figure 15: Software support - FIST in x86 with different secondary opcodes to indicate movement from DRAM to storage/compute array. New XNORM instruction with SRC1 = the storage array address that is mapped onto RWL, SRC2 = the compute array address, BIT = Jij resolution and DEST = register for storing XNOR 64-bit ISA, has a primary opcode (PO) of 0xDB without using a secondary opcode (SO). SO is used to indicate SA… view at source ↗
Figure 16
Figure 16. Figure 16: SACHI comparison with BRIM and Ising-CIM for reuse and designs applicability to different COPs. SACHI comparison with BRIM - b) Number of cycles c) total energy to solve COP including loading SACHI comparison with Ising-CIM - e) Number of cycles f) total energy to solve COP including loading slightly degraded energy, but still ∼75x energy improvement (Fig.16c). The major reason for SACHI’s superior perfor… view at source ↗
Figure 17
Figure 17. Figure 17: Comparison with GA/PSO/optimized solvers (OPTSolv) - Solu￾tion accuracy, normalized execution time wrt SACHI for asset allocation, image segmentation, traveling salesman and molecular dynamics as shown in (Fig.17a). This could be attributed to GA’s global-only search for selecting the best candidates in each generation. In contrast, Ising/PSO performs updates based on neighbors, resulting in faster conver… view at source ↗
Figure 18
Figure 18. Figure 18: Scalability - Cycles Per Hamiltonian iteration (I) with increasing variable size for a) Asset allocation b) Image segmentation c) Traveling salesman and d) Molecular dynamics COPs showing SACHI’s scalability to different large variable count COPs 13 [PITH_FULL_IMAGE:figures/full_fig_p013_18.png] view at source ↗
Figure 20
Figure 20. Figure 20: Time to solution/solution accuracy - a) Hamiltonian energy (H) reduction with increasing iteration count for Asset allocation, using simulated annealing (SA) to move out of local minima. b) Solution time improvement from SACHI(n1) to SACHI(n3) architecture due to reuse aware compute c) Increased iteration count with lower resolution for IC due to frequent requirement of SA steps d) Solution accuracy degra… view at source ↗
read the original abstract

Recently, nature-inspired computing approaches have gained significant attention for solving difficult optimization problems, particularly through Ising machines for NP-complete applications. Existing Ising accelerators range from quantum and optical annealers to CMOS-based von-Neumann and in-memory architectures. However, many prior designs are specialized accelerators limited to specific problem classes, rely on ADC/DAC circuits, and suffer from reliability challenges due to process-variation-sensitive embedded memory technologies. This paper presents SACHI, an all-digital Ising architecture implemented by repurposing the L1 cache of a CPU using SRAM-based processing-in-memory techniques. SACHI eliminates the need for ADCs/DACs, improves reliability compared to prior approaches such as BRIM, and enables Ising acceleration with minimal hardware overhead integrated into the CPU pipeline. The paper also provides detailed architectural analysis and pseudo-code for the proposed algorithms. The key contributions of SACHI are: (i) tight integration of the accelerator with the CPU pipeline, (ii) reuse of existing cache hardware for acceleration, (iii) higher parallelism enabled through reuse-aware computation, and (iv) improved performance and energy efficiency for large-scale, high-precision optimization problems using novel compute and mapping strategies. Compared to BRIM, SACHI achieves 300x performance improvement and 80x energy reduction across applications including asset allocation, molecular dynamics, image segmentation, and traveling salesman problems. Additionally, reuse factors up to 4000x are observed for several workloads. This work demonstrates that reliable and efficient all-digital Ising acceleration can be achieved using commodity SRAM structures tightly integrated with general-purpose processors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes SACHI, a reuse-aware all-digital Ising machine implemented via SRAM-based processing-in-memory by repurposing the L1 cache of a general-purpose CPU. It claims tight pipeline integration with negligible overhead, elimination of ADCs/DACs, improved reliability relative to prior embedded-memory designs such as BRIM, and quantitative gains of 300x performance and 80x energy reduction across asset allocation, molecular dynamics, image segmentation, and TSP workloads, together with reuse factors reaching 4000x. The paper supplies architectural analysis and pseudo-code for the mapping and compute strategies.

Significance. If the reported speedups and energy reductions are substantiated by reproducible simulations that include workload sizes, error bars, and explicit hardware-overhead accounting, and if the SRAM PIM integration is shown to avoid the process-variation and reliability problems the manuscript attributes to earlier designs, the result would constitute a meaningful step toward embedding Ising acceleration inside commodity CPUs rather than as discrete accelerators.

major comments (2)
  1. [Abstract] Abstract: the central performance claims (300x speedup, 80x energy reduction versus BRIM) are presented without any accompanying methodology, workload sizes, simulation assumptions, or error bars. Because these numbers constitute the primary evidence for the superiority of the reuse-aware PIM approach, their verification is load-bearing for the manuscript's contribution.
  2. [Abstract] Abstract / §1: the claim that L1-SRAM PIM integration incurs "minimal hardware overhead" and avoids the reliability/process-variation issues of prior embedded-memory designs is asserted but not supported by synthesis results, area/power overhead tables, or variation-tolerance analysis. This assumption directly underpins both the 300x/80x comparisons and the assertion of CPU-pipeline compatibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and indicate the revisions we will make to strengthen the presentation of our results and claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central performance claims (300x speedup, 80x energy reduction versus BRIM) are presented without any accompanying methodology, workload sizes, simulation assumptions, or error bars. Because these numbers constitute the primary evidence for the superiority of the reuse-aware PIM approach, their verification is load-bearing for the manuscript's contribution.

    Authors: We agree that the abstract would benefit from a concise reference to the evaluation methodology. The manuscript details the workloads (asset allocation, molecular dynamics, image segmentation, and TSP), simulation assumptions, workload sizes, and results including error bars and reuse factors up to 4000x in Section 4. We will revise the abstract to briefly note the simulation framework and direct readers to the evaluation section for full methodology and quantitative results. revision: yes

  2. Referee: [Abstract] Abstract / §1: the claim that L1-SRAM PIM integration incurs "minimal hardware overhead" and avoids the reliability/process-variation issues of prior embedded-memory designs is asserted but not supported by synthesis results, area/power overhead tables, or variation-tolerance analysis. This assumption directly underpins both the 300x/80x comparisons and the assertion of CPU-pipeline compatibility.

    Authors: The claims are grounded in the architectural analysis and pseudo-code presented in Sections 2 and 3, which show reuse of existing L1 SRAM with only lightweight control logic and the use of mature all-digital SRAM to avoid analog circuits and variation-sensitive memories employed in BRIM. We acknowledge that explicit synthesis results and variation-tolerance tables are not included. We will add a discussion subsection with estimated overheads from standard-cell analysis and a qualitative comparison of reliability in the revised manuscript. revision: partial

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper presents an architectural design for SACHI, an all-digital Ising machine reusing L1 SRAM, along with algorithmic pseudo-code and performance claims versus BRIM. No equations, fitted parameters, self-citations, or ansatzes are quoted that reduce any prediction or result to the inputs by construction. The 300x/80x claims are stated as outcomes of the proposed reuse-aware mapping and integration rather than derived tautologically from the design assumptions themselves. The work is self-contained against external benchmarks with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central performance claims rest on the unverified feasibility of SRAM PIM integration and the correctness of the reuse-aware algorithms; no free parameters, additional axioms, or invented physical entities are mentioned in the abstract.

axioms (1)
  • domain assumption SRAM cells can perform the required Ising spin updates reliably when repurposed for processing-in-memory
    Invoked when the abstract states that SACHI eliminates ADCs/DACs and improves reliability compared with prior embedded-memory approaches.
invented entities (1)
  • SACHI architecture no independent evidence
    purpose: All-digital Ising acceleration via L1 cache reuse
    New design introduced in the paper; no independent evidence outside the abstract is provided.

pith-pipeline@v0.9.1-grok · 5830 in / 1454 out tokens · 34583 ms · 2026-06-30T21:38:06.389832+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A comprehensive study on ILP acceleration accounting for sparsity, area, energy, data movement using near-memory architecture

    cs.AR 2026-05 unverdicted novelty 6.0

    SPARK is a sparsity-aware near-cache ILP accelerator achieving up to 15x performance and 152x energy reduction over CPUs for sparse ILPs on MIPLIB 2017 workloads.

  2. A complete discussion on fully reconfigurable, digital, scalable, graph and sparsity-aware near-memory accelerator for graph neural networks

    cs.AR 2026-05 unverdicted novelty 5.0

    NEM-GNN is a scalable DAC/ADC-less processing-in-memory architecture for GNNs that uses early compute termination, reconfigurable SoC pre-computation, and compute-as-soon-as-ready broadcast execution to deliver large ...

  3. A complete discussion on fully reconfigurable, digital, scalable, graph and sparsity-aware near-memory accelerator for graph neural networks

    cs.AR 2026-05 unverdicted novelty 5.0

    NEM-GNN proposes a scalable DAC/ADC-less PIM architecture for GNNs with early termination and CAR execution, claiming 80-230x performance and 850-1134x energy gains over prior accelerators.

  4. A comprehensive study on ILP acceleration accounting for sparsity, area, energy, data movement using near-memory architecture

    cs.AR 2026-05 unverdicted novelty 5.0

    SPARK is a sparsity-aware near-cache ILP accelerator that reuses L1 cache structures to deliver up to 15x speedup and 152x energy reduction versus CPUs on sparse MIPLIB workloads with 1.4% area overhead.

  5. Emerging memory technologies at room/cryogenic temperature

    cs.AR 2026-05 unverdicted novelty 1.0

    Overview chapter surveying volatile and non-volatile memories including SRAM, DRAM, RRAM, MRAM, FeFET and cryogenic JJFET devices, with focus on principles, tradeoffs, and challenges.

Reference graph

Works this paper leans on

49 extracted references · 10 canonical work pages · cited by 3 Pith papers · 5 internal anchors

  1. [1]

    Brim: Bistable resistively-coupled ising machine,

    R. Afoakwa, Y . Zhang, U. K. R. Vengalam, Z. Ignjatovic, and M. Huang, “Brim: Bistable resistively-coupled ising machine,” in2021 IEEE In- ternational Symposium on High-Performance Computer Architecture (HPCA), 2021, pp. 749–760

    2021

  2. [2]

    Conv-sram: An energy-efficient sram with in-memory dot-product computation for low-power convolu- tional neural networks,

    A. Biswas and A. P. Chandrakasan, “Conv-sram: An energy-efficient sram with in-memory dot-product computation for low-power convolu- tional neural networks,”IEEE Journal of Solid-State Circuits, vol. 54, no. 1, pp. 217–230, 2019

    2019

  3. [3]

    Pt/Cu:ZnO/Nb:STO memristive dual port for cache memory applications,

    P. K. R. Boppidi, S. S. Raman, H. Renuka, and S. Kundu, “Pt/Cu:ZnO/Nb:STO memristive dual port for cache memory applications,”AIP Conference Proceedings, vol. 2265, no. 1, p. 030212, 11 2020. [Online]. Available: https://doi.org/10.1063/5.0016597

    work page doi:10.1063/5.0016597 2020

  4. [4]

    History of the lenz-ising model,

    S. G. Brush, “History of the lenz-ising model,”Reviews of modern physics, vol. 39, no. 4, p. 883, 1967

    1967

  5. [5]

    An introduction to the ising model,

    B. A. Cipra, “An introduction to the ising model,”The American Mathematical Monthly, vol. 94, no. 10, pp. 937–959, 1987

    1987

  6. [6]

    Open Quantum Assembly Language

    A. W. Cross, L. S. Bishop, J. A. Smolin, and J. M. Gambetta, “Open quantum assembly language,”arXiv preprint arXiv:1707.03429, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  7. [7]

    The traveling-salesman problem,

    M. M. Flood, “The traveling-salesman problem,”Operations research, vol. 4, no. 1, pp. 61–75, 1956

    1956

  8. [8]

    Time-dependent statistics of the ising model,

    R. J. Glauber, “Time-dependent statistics of the ising model,”Journal of mathematical physics, vol. 4, no. 2, pp. 294–307, 1963

    1963

  9. [9]

    Experi- mental investigation of performance differences between coherent ising machines and a quantum annealer,

    R. Hamerly, T. Inagaki, P. L. McMahon, D. Venturelli, A. Marandi, T. Onodera, E. Ng, C. Langrock, K. Inaba, T. Honjoet al., “Experi- mental investigation of performance differences between coherent ising machines and a quantum annealer,”Science advances, vol. 5, no. 5, p. eaau0823, 2019. 14

    2019

  10. [10]

    Quantum annealing with manufactured spins,

    M. W. Johnson, M. H. Amin, S. Gildert, T. Lanting, F. Hamze, N. Dickson, R. Harris, A. J. Berkley, J. Johansson, P. Bunyket al., “Quantum annealing with manufactured spins,”Nature, vol. 473, no. 7346, pp. 194–198, 2011

    2011

  11. [11]

    Polaritonic xy-ising machine,

    K. P. Kalinin, A. Amo, J. Bloch, and N. G. Berloff, “Polaritonic xy-ising machine,”Nanophotonics, vol. 9, no. 13, pp. 4127–4138, 2020

    2020

  12. [12]

    Unconven- tional computing using ising accelerators,

    J. P. Kulkarni, S. R. Sundara Raman, S. Xie, and C.-P. Lo, “Unconven- tional computing using ising accelerators,”Computer, vol. 58, no. 6, pp. 83–86, 2025

    2025

  13. [13]

    Analogue signal and image processing with large-scale rram crossbars,

    C. Li, D. Belkin, Y . Li, P. Yan, M. Hu, N. Ge, H. Sheng, H. Chang, C. Pao, J. M. Linet al., “Analogue signal and image processing with large-scale rram crossbars,”Nature Electronics, vol. 1, no. 1, pp. 52–59, 2018

    2018

  14. [14]

    Ising formulations of many NP problems

    A. Lucas, “Ising formulations of many NP problems,”Frontiers in Physics, vol. 2, 2014. [Online]. Available: https://doi.org/10.3389% 2Ffphy.2014.00005

    work page arXiv 2014

  15. [15]

    Net- work of time-multiplexed optical parametric oscillators as a coherent ising machine,

    A. Marandi, Z. Wang, K. Takata, R. L. Byer, and Y . Yamamoto, “Net- work of time-multiplexed optical parametric oscillators as a coherent ising machine,”Nature Photonics, vol. 8, no. 12, pp. 937–942, 2014

    2014

  16. [16]

    Image segmentation using deep learning: A survey,

    S. Minaee, Y . Boykov, F. Porikli, A. Plaza, N. Kehtarnavaz, and D. Terzopoulos, “Image segmentation using deep learning: A survey,” IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 7, pp. 3523–3542, 2021

    2021

  17. [17]

    Genetic algorithm,

    S. Mirjalili and S. Mirjalili, “Genetic algorithm,”Evolutionary Algo- rithms and Neural Networks: Theory and Applications, pp. 43–55, 2019

    2019

  18. [18]

    Huang et al., ”Machine Learning for Electronic Design Automation: A Survey,” ACM Transactions on Design Automation of Electronic Systems, vol

    O. Mutlu, S. Ghose, J. G ´omez-Luna, and R. Ausavarungnirun, “A mod- ern primer on processing in memory,”arXiv preprint arXiv:2012.03112, 2020

    work page arXiv 2012

  19. [19]

    On the theory of the ising model of ferromagnetism,

    G. F. Newell and E. W. Montroll, “On the theory of the ising model of ferromagnetism,”Reviews of Modern Physics, vol. 25, no. 2, p. 353, 1953

    1953

  20. [20]

    Ultra-low-voltage utbb-soi-based, pseudo-static storage circuits for cryogenic cmos applications,

    S. S. T. Nibhanupudi, S. R. Sundara Raman, M. Cass ´e, L. Hutin, and J. P. Kulkarni, “Ultra-low-voltage utbb-soi-based, pseudo-static storage circuits for cryogenic cmos applications,”IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 7, no. 2, pp. 201– 208, 2021

    2021

  21. [21]

    Phase tran- sition material-assisted low-power sram design,

    S. T. Nibhanupudi, S. R. S. Raman, and J. P. Kulkarni, “Phase tran- sition material-assisted low-power sram design,”IEEE Transactions on Electron Devices, vol. 68, no. 5, pp. 2281–2288, 2021

    2021

  22. [22]

    Improving solutions by embedding larger subproblems in a d-wave quantum annealer,

    S. Okada, M. Ohzeki, M. Terabe, and S. Taguchi, “Improving solutions by embedding larger subproblems in a d-wave quantum annealer,” Scientific reports, vol. 9, no. 1, pp. 1–10, 2019

    2019

  23. [23]

    Large-scale photonic ising machine by spatial light modulation,

    D. Pierangeli, G. Marcucci, and C. Conti, “Large-scale photonic ising machine by spatial light modulation,”Physical review letters, vol. 122, no. 21, p. 213902, 2019

    2019

  24. [24]

    A review on non-volatile and volatile emerging memory technologies,

    S. R. S. Raman, “A review on non-volatile and volatile emerging memory technologies,” inComputer Memory and Data Storage, A. Seyedi, Ed. Rijeka: IntechOpen, 2024, ch. 3. [Online]. Available: https://doi.org/10.5772/intechopen.110617

    work page doi:10.5772/intechopen.110617 2024

  25. [25]

    Compute in edram using indium gallium zinc oxide transistors,

    S. R. S. Raman, “Compute in edram using indium gallium zinc oxide transistors,” Ph.D. dissertation, The University of Texas at Austin, 2026, available: https://repositories.lib.utexas.edu/items/4dbc7f92-c062- 4cb8-b07b-ed29761b9704. [Online]. Available: https://repositories.lib. utexas.edu/items/4dbc7f92-c062-4cb8-b07b-ed29761b9704

    2026

  26. [26]

    Emerging memory technologies at room/cryogenic temperature,

    S. R. S. Raman, “Emerging memory technologies at room/cryogenic temperature,” 2026. [Online]. Available: https://arxiv.org/abs/2605. 21912

    2026

  27. [27]

    Spark: Sparsity aware, low area, energy-efficient, near-memory architecture for accelerating linear programming problems,

    S. R. S. Raman, L. John, and J. P. Kulkarni, “Spark: Sparsity aware, low area, energy-efficient, near-memory architecture for accelerating linear programming problems,” in2025 IEEE International Symposium on High Performance Computer Architecture (HPCA), 2025, pp. 99–112

    2025

  28. [28]

    A complete discussion on fully reconfigurable, digital, scalable, graph and sparsity-aware near-memory accelerator for graph neural networks

    S. R. S. Raman, L. John, and J. P. Kulkarni, “A complete discussion on fully reconfigurable, digital, scalable, graph and sparsity-aware near-memory accelerator for graph neural networks,” 2026. [Online]. Available: https://arxiv.org/abs/2605.19405

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  29. [29]

    A comprehensive study on ILP acceleration accounting for sparsity, area, energy, data movement using near-memory architecture

    S. R. S. Raman, L. K. John, and J. P. Kulkarni, “A comprehensive study on ilp acceleration accounting for sparsity, area, energy, data movement using near-memory architecture,” 2026. [Online]. Available: https://arxiv.org/abs/2605.17158

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  30. [30]

    S. R. S. Raman and J. P. Kulkarni, “Abi: A tightly integrated, unified, sparsity-aware, reconfigurable, compute near-register file/cache gpu architecture with light-weight softmax for deep learning, linear algebra, and ising compute,” 2026. [Online]. Available: https://arxiv.org/abs/2602.14262

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  31. [31]

    A comparative study on power delivery aspects of compute-in/near-memory approaches using DRAM

    S. R. S. Raman, S. Ma, and L. K. John, “A comparative study on power delivery aspects of compute-in/near-memory approaches using dram,” arXiv preprint arXiv:2604.04773, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  32. [32]

    Threshold selector and capacitive coupled assist techniques for write voltage reduction in metal–ferroelectric–metal field-effect transistor,

    S. R. S. Raman, S. S. T. Nibhanupudi, A. K. Saha, S. Gupta, and J. P. Kulkarni, “Threshold selector and capacitive coupled assist techniques for write voltage reduction in metal–ferroelectric–metal field-effect transistor,”IEEE Transactions on Electron Devices, vol. 68, no. 12, pp. 6132–6138, 2021

    2021

  33. [33]

    High noise margin, digital logic design using josephson junction field-effect transistors for cryogenic computing,

    S. R. S. Raman, F. Wen, R. Pillarisetty, V . De, and J. P. Kulkarni, “High noise margin, digital logic design using josephson junction field-effect transistors for cryogenic computing,”IEEE Transactions on Applied Superconductivity, vol. 31, no. 5, pp. 1–5, 2021

    2021

  34. [34]

    Increasing ising machine capacity with multi-chip architectures,

    A. Sharma, R. Afoakwa, Z. Ignjatovic, and M. Huang, “Increasing ising machine capacity with multi-chip architectures,” inProceedings of the 49th Annual International Symposium on Computer Architecture, 2022, pp. 508–521

    2022

  35. [35]

    Freepdk: An open-source variation-aware design kit,

    J. E. Stine, I. Castellanos, M. Wood, J. Henson, F. Love, W. R. Davis, P. D. Franzon, M. Bucher, S. Basavarajaiah, J. Ohet al., “Freepdk: An open-source variation-aware design kit,” in2007 IEEE international conference on Microelectronic Systems Education (MSE’07). IEEE, 2007, pp. 173–174

    2007

  36. [36]

    31.2 cim-spin: A 0.5-to-1.2v scalable annealing processor using digital compute-in-memory spin operators and register-based spins for combinatorial optimization problems,

    Y . Su, H. Kim, and B. Kim, “31.2 cim-spin: A 0.5-to-1.2v scalable annealing processor using digital compute-in-memory spin operators and register-based spins for combinatorial optimization problems,” in2020 IEEE International Solid- State Circuits Conference - (ISSCC), 2020, pp. 480–482

    2020

  37. [37]

    Nem-gnn: Dac/adc-less, scalable, reconfigurable, graph and sparsity-aware near- memory accelerator for graph neural networks,

    S. R. Sundara Raman, L. John, and J. P. Kulkarni, “Nem-gnn: Dac/adc-less, scalable, reconfigurable, graph and sparsity-aware near- memory accelerator for graph neural networks,”ACM Trans. Archit. Code Optim., vol. 21, no. 2, May 2024. [Online]. Available: https://doi.org/10.1145/3652607

    work page doi:10.1145/3652607 2024

  38. [38]

    Sachi: A stationarity-aware, all-digital, near-memory, ising architecture,

    S. R. Sundara Raman, L. K. John, and J. P. Kulkarni, “Sachi: A stationarity-aware, all-digital, near-memory, ising architecture,” in2024 IEEE International Symposium on High-Performance Computer Archi- tecture (HPCA), 2024, pp. 719–731

    2024

  39. [39]

    Enabling in-memory computations in non-volatile sram designs,

    S. R. Sundara Raman, S. S. T. Nibhanupudi, and J. P. Kulkarni, “Enabling in-memory computations in non-volatile sram designs,”IEEE Journal on Emerging and Selected Topics in Circuits and Systems, vol. 12, no. 2, pp. 557–568, 2022

    2022

  40. [41]

    Igzo cim: Enabling in-memory computations using multilevel capacitorless indium–gallium–zinc–oxide-based embedded dram technology,

    S. R. Sundara Raman, S. Xie, and J. P. Kulkarni, “Igzo cim: Enabling in-memory computations using multilevel capacitorless indium–gallium–zinc–oxide-based embedded dram technology,”IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 8, no. 1, pp. 35–43, 2022

    2022

  41. [42]

    Application of ising machines and a software development for ising machines,

    K. Tanahashi, S. Takayanagi, T. Motohashi, and S. Tanaka, “Application of ising machines and a software development for ising machines,” Journal of the Physical Society of Japan, vol. 88, no. 6, p. 061010, 2019

    2019

  42. [43]

    LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales,

    A. P. Thompson, H. M. Aktulga, R. Berger, D. S. Bolintineanu, W. M. Brown, P. S. Crozier, P. J. in ’t Veld, A. Kohlmeyer, S. G. Moore, T. D. Nguyen, R. Shan, M. J. Stevens, J. Tranchida, C. Trott, and S. J. Plimpton, “LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales,”Comp. Phys. Comm., vol...

    2022

  43. [44]

    Wide-range many-core soc design in scaled cmos: Challenges and opportunities,

    S. Vangal, S. Paul, S. Hsu, A. Agarwal, S. Kumar, R. Krishnamurthy, H. Krishnamurthy, J. Tschanz, V . De, and C. H. Kim, “Wide-range many-core soc design in scaled cmos: Challenges and opportunities,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 29, no. 5, pp. 843–856, 2021

    2021

  44. [45]

    Galib: A c++ library of genetic algorithm components,

    M. Wall, “Galib: A c++ library of genetic algorithm components,” Mechanical Engineering Department, Massachusetts Institute of Tech- nology, vol. 87, p. 54, 1996

    1996

  45. [46]

    A compute-in-memory chip based on resistive random-access memory,

    W. Wan, R. Kubendran, C. Schaefer, S. Eryilmaz, W. Zhang, D. Wu, S. Deiss, P. Raina, H. Qian, B. Gao, S. Joshi, H. Wu, H.-S. Wong, and G. Cauwenberghs, “A compute-in-memory chip based on resistive random-access memory,”Nature, vol. 608, pp. 504–512, 08 2022

    2022

  46. [47]

    Oim: Oscillator-based ising machines for solving combinatorial optimisation problems,

    T. Wang and J. Roychowdhury, “Oim: Oscillator-based ising machines for solving combinatorial optimisation problems,” inUnconventional Computation and Natural Computation: 18th International Conference, 15 UCNC 2019, Tokyo, Japan, June 3–7, 2019, Proceedings 18. Springer, 2019, pp. 232–256

    2019

  47. [48]

    New computational results and hardware prototypes for oscillator-based ising machines,

    T. Wang, L. Wu, and J. Roychowdhury, “New computational results and hardware prototypes for oscillator-based ising machines,” inProceedings of the 56th Annual Design Automation Conference 2019, 2019, pp. 1–2

    2019

  48. [49]

    Ising-cim: A reconfigurable and scalable compute within memory ana- log ising accelerator for solving combinatorial optimization problems,

    S. Xie, S. R. S. Raman, C. Ni, M. Wang, M. Yang, and J. P. Kulkarni, “Ising-cim: A reconfigurable and scalable compute within memory ana- log ising accelerator for solving combinatorial optimization problems,” IEEE Journal of Solid-State Circuits, pp. 1–13, 2022

    2022

  49. [50]

    A 20k-spin ising chip to solve combinatorial optimization problems with cmos annealing,

    M. Yamaoka, C. Yoshimura, M. Hayashi, T. Okuyama, H. Aoki, and H. Mizuno, “A 20k-spin ising chip to solve combinatorial optimization problems with cmos annealing,”IEEE Journal of Solid-State Circuits, vol. 51, no. 1, pp. 303–309, 2016. 16

    2016