Pith. sign in

REVIEW 2 major objections 4 minor 28 references

On low-power error-correcting cooling codes with large distances

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper gives exact and asymptotic formulas for the maximum size of low-power error-correcting cooling codes, over a parameter range wider than any previously known.

desk verdict Solid extremal-combinatorics paper that settles the asymptotic and many exact values for LPECC codes; the proofs check out and the external dependencies are legitimate. read the letter →

arxiv 2412.06223 v1 pith:3U3QXQ4A submitted 2024-12-09 cs.IT math.COmath.IT

Yuhao Zhao , Xiande Zhang This is my paper · ORCID
classification cs.ITmath.COmath.IT MSC 94B2505B0594B65
keywords LPECCcodescoolinglow-powerpackingsSteinersystemsframesconstant-weight
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper establishes the largest possible size of a low-power error-correcting cooling (LPECC) code: a coding scheme for an $n$-wire bus that never switches the $t$ hottest wires, makes at most $w$ transitions per transmission, and corrects up to $e$ bit errors. The main results are exact and asymptotic formulas for that maximum, denoted $C(n,t,w,e)$. For large error budgets ($e = w-2$, with $w \ge t^2+2t+2$) the paper proves $C(n,t,w,w-2) \le \lfloor \binom{n+1}{2}/\binom{w+t}{2}\rfloor$, tight for large $n$ satisfying divisibility conditions; for $w=3,e=1$ it proves $C(n,t,3,1) \le \lfloor n(n+1)/(6(t+1))\rfloor$, also tight infinitely often. For general $q$-ary codes satisfying a binomial-coefficient inequality, it shows $C_q(n,t,w,e) \sim (q-1)^{w-e} \binom{n}{w-e}/\binom{w+t}{w-e}$ as $n$ grows. A sympathetic reader would care because these are the first tight bounds for most of the parameter range, giving the exact information-theoretic cost of simultaneously cooling, saving power, and correcting errors on a bus.

What carries the argument

The machinery is the set-theoretic representation of an LPECC code as a partition $P_1 \cup \cdots \cup P_b$ of blocks in $2^{[n]}$, where each block is the support of a codeword and the error-correction property forces blocks from different parts to have symmetric difference at least $2e+1$. The upper bounds come from double counting: each part must cover at least $\binom{w+t}{2}$ pairs or, in the general case, $\binom{w+t}{w-e}$ subsets of size $w-e$, while all parts together can cover at most $\binom{n+1}{2}$ pairs or $(q-1)^{w-e}\binom{n}{w-e}$ objects. The lower bounds come from a construction that turns a $q$-ary constant-weight code of weight $w+t$ and odd distance $2(e+t)+1$ into an LPECC code by cutting each support down to size $w$; the relevant distance calculation shows the error-correcting property survives after deleting $t$ coordinates from every support.

What would settle it

Fix $q=2$ and parameters $(t,w,e)$ satisfying $2\binom{w}{w-e} \ge \binom{w-t-1}{w-e}+\binom{w+t}{w-e}$, and compute $A_2(n,2(e+t)+1,w+t)$ along a sequence of $n$. If for some fixed $\varepsilon>0$ infinitely many $n$ fall below $(1-\varepsilon)\binom{n}{w-e}/\binom{w}{w-e}$, the construction's lower bound fails and $C_2(n,t,w,e)$ cannot reach the claimed asymptotic value.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that the LPECC constraint set has a precise combinatorial price: the maximum number of messages is, up to a binomial ratio, the same as the maximum size of a packing or constant-weight code. Concretely, for binary codes with $e=w-2$ and $w \ge t^2+2t+2$, every $(n,t,w,w-2)$-LPECC code has size at most $\lfloor \binom{n+1}{2}/\binom{w+t}{2}\rfloor$, and when $n$ is sufficiently large with $\binom{w+t}{2}$ dividing $\binom{n+1}{2}$ and $w+t-1$ dividing $n$, there exist codes meeting it; any such optimal code must be the image of an $(n+1,w+t,1)$-balanced incomplete block design. For $w=3,e=1$, the paper proves $C(n,t,3,1) = \lfloor n(n+1)/(6(t+1))\rfloor$ for even $n \equiv 2 \pmod{3(t+1)}$. For the general $q$-ary setting, whenever $2\binom{w}{w-e} \ge \binom{w-t-1}{w-e}+\binom{w+t}{w-e}$, it establishes $C_q(n,t,w,e) \sim (q-1)^{w-e} \binom{n}{w-e}/\binom{w+t}{w-e}$ as $n \to \infty$.

Load-bearing premise

The asymptotic lower bound rests on the unproved-in-this-paper existence of near-perfect constant-weight codes of exactly the weight and distance used in the construction; if that existence fails for some parameter choices, the claimed limit could be too high.

Editorial extensions

If this is right

  • For large $w$ relative to $t$, the exact size of the best $(n,t,w,w-2)$-LPECC code is $\lfloor \binom{n+1}{2}/\binom{w+t}{2}\rfloor$ whenever the divisibility conditions hold, and the optimal codes are exactly those induced by balanced incomplete block designs with parameters $(n+1,w+t,1)$.
  • For $w=3,e=1$, the maximum is $\lfloor n(n+1)/(6(t+1))\rfloor$ for every even $n \equiv 2 \pmod{3(t+1)}$, so the asymptotic density $n^2/(6(t+1))$ is attained on an infinite arithmetic progression.
  • For any fixed $q,t,w,e$ satisfying the binomial inequality, the $q$-ary maximum $C_q(n,t,w,e)$ grows like $(q-1)^{w-e} \binom{n}{w-e}/\binom{w+t}{w-e}$, meaning the cooling, low-power, and error-correction requirements do not reduce the asymptotic rate below the packing or constant-weight bound.
  • The constant-power variant $C'_q(n,t,w,e)$ satisfies the same asymptotic formula, with exact equality available when generalized Steiner systems exist, such as for binary codes or for $e \in \{w-1,w-2\}$ under divisibility conditions.
  • For small $w$, the paper proves $\liminf_{n\to\infty} C(n,t,4,2)/n^2 \ge 1/(12(t+1))$, and it conjectures that the small-$w$ regime has rate $1/((t+1)w(w-1))n^2$, distinct from the large-$w$ rate.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural reading the paper does not spell out is that the LPECC constraints become asymptotically free exactly on one side of the binomial inequality, while the small-$w$ arguments point to a separate regime whose rate is $1/((t+1)w(w-1))n^2$; locating the precise boundary between these regimes is a testable extension.
  • The extremal characterization says optimal large-$w$ codes are exactly BIBD-induced, so known explicit constructions of balanced incomplete block designs should yield explicit optimal LPECC encoders; the paper does not develop that algorithmic direction.
  • Because the $q$-ary bound only uses Johnson-type counting, the same asymptotic should extend to other transition-cost models, such as energy proportional to the number of changed nonzero symbols; this is a reformulation to test rather than a claim of the paper.
  • The asymptotic lower bound relies on an imported existence result for near-optimal constant-weight codes, so a targeted numerical check on the exact parameters used in the construction would indicate how robust the limit is when the binomial inequality is barely satisfied.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper studies the maximum size C(n,t,w,e) of low-power error-correcting cooling (LPECC) codes, which combine thermal constraints (no transitions on the t hottest wires), power constraints (at most w transitions per transmission), and error correction (up to e errors). The main results are: (i) for w ≥ t^2+2t+2, an exact upper bound C(n,t,w,w-2) ≤ floor(binomial(n+1,2)/binomial(w+t,2)) with divisibility-based tightness and an extremal characterization via (n+1,w+t,1)-BIBDs (Theorem 3.1, Corollary 3.6, Theorem 3.7); (ii) for small weight w=3, the upper bound C(n,t,3,1) ≤ floor(n(n+1)/(6(t+1))) and matching constructions from 3-frames, yielding the asymptotic C(n,t,3,1) ~ n^2/(6(t+1)) (Theorem 4.1, Corollary 4.10); (iii) a general q-ary asymptotic result C_q(n,t,w,e) ~ (q-1)^{w-e} binomial(n,w-e)/binomial(w+t,w-e) under the condition 2 binomial(w,w-e) ≥ binomial(w-t-1,w-e)+binomial(w+t,w-e) (Corollary 5.8), proved via a Johnson-type upper bound and a lower bound from constant-weight codes; and (iv) related bounds and exact values for constant-power error-correcting cooling codes (Theorem 5.10, Corollary 5.11).

Significance. If the results stand, the paper substantially extends earlier work, which was limited to small t and w, to arbitrary fixed t with large w and to a general q-ary asymptotic regime. The upper-bound machinery—pair counting in Lemma 3.3, the Turán-based argument in Theorem 4.1, and the Johnson-type bound in Theorem 5.1—is coherent, and the use of external tools (Rödl's nibble, Wilson's and Keevash's existence theorems, and Liu–Shangguan's near-optimal constant-weight codes) is appropriate for the fixed-parameter setting. The extremal structure theorem (Theorem 3.7) is a strong and elegant result. The paper is essentially free of fitted parameters, and the asymptotic claims are concrete and falsifiable. The main weakness is that several load-bearing counting steps are only sketched as 'greedy' arguments, and one asymptotic derivation omits a necessary monotonicity step.

major comments (2)
  1. [§3, Claims 3.4(c), 3.5(b); §5, Claim 5.3] The lower bounds on τ(P_i) in Claims 3.4(c), 3.5(b), and 5.3 are proved by a 'greedy' counting sentence that is not a complete formal argument. These claims are load-bearing for the main upper bounds in Theorem 3.1 and Theorem 5.1. Please provide an explicit ordering of the vertices in Y, and prove that after removing previously used vertices, each y_j contributes at least (w-1)-j (respectively, binomial(w-j,w-e-1) in the q-ary claim) distinct previously uncounted pairs or (w-e)-subsets. In particular, the treatment of possible overlaps between the contributions for different y_j must be stated precisely.
  2. [§4, Corollary 4.10] The statement that Corollary 4.10 follows 'immediately' from Theorem 4.1 and Theorem 4.9 is not fully justified, because Theorem 4.9 provides a lower bound only when n ≡ 2 (mod 6). To conclude the limit over all n, the authors should explicitly state and prove the zero-padding monotonicity C(n+1,t,3,1) ≥ C(n,t,3,1), or otherwise extend the construction to all sufficiently large n. This is a short argument but it is currently absent.
minor comments (4)
  1. [§3, statement of Theorem 3.1] The notation 'w ≥ t2 + 2t + 2 ≥ 5' is slightly confusing; it should be written as 'w ≥ t^2 + 2t + 2', noting that this implies w ≥ 5 for t ≥ 1.
  2. [§4, Claim 4.6] In the proof of Claim 4.6, the transition from the independence number bound to the Turán application is terse; please spell out that e(\bar{H}) ≤ e(T(v2(P_i), r)) and that the complement of T(v,r) has exactly λ binom(ceil(v/r),2) + (r-λ) binom(floor(v/r),2) edges, matching Lemma 4.4.
  3. [§5.2, paragraph after Remark 5.9] There is a typo: 'Note that here we all consider' should read 'Note that here we always consider'.
  4. [§5, Corollary 5.7] For completeness, please state explicitly that the o(1) in Lemma 5.6 is with respect to n for fixed q, t, w, e, since the application in Corollary 5.7 requires only this fixed-parameter uniformity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: upper and lower bounds are independently proved, with external packing/Steiner/CWC existence results used as benchmarks.

full rationale

Walked the derivation chain. Theorem 3.1's upper bound is a direct double-counting/pair-covering argument (Claims 3.4 and 3.5) with no target quantity used as an assumption; its asymptotic matching and exact equality use R\"odl's nibble and Wilson's BIBD existence, which are external benchmarks, and the packing-to-LPECC construction is Theorem 2.2 from Liu-Ji [17]. Theorem 4.1 is proved from Tur\'an's theorem and an independent triangle lemma; the lower bounds use frame existence theorems from the Colbourn-Dinitz handbook. Section 5's upper bound (Theorem 5.1) is proved from Property A_q(t), the Johnson-type constant-weight-code bound (5.1), and Claim 5.3; the lower bound Construction 5.4 uses an independent near-optimal constant-weight-code existence theorem of Liu-Shangguan [16] with parameters (weight w+t, odd distance 2(e+t)+1), which are not the target LPECC parameters but are transformed by the shown construction. No parameter is fitted to a subset of data, and no defining quantity of LPECC codes is used as the statement of a result. The only author self-citation is [8], used in the secondary Corollary 5.11 to cite a published generalized Steiner system existence result; it is not load-bearing for the asymptotic Corollary 5.8 and is an external theorem with its own proof, so it does not create circularity.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no free parameters and no invented entities. All numerical inputs (n,t,w,e,q) are design parameters. The central claims rest on cited design-theoretic existence theorems and on Liu-Ji's characterization and packing construction, which are treated as axioms.

assumptions (6)
  • domain assumption Liu-Ji characterization of (n,t,w,w-2)-LPECC codes via Properties (2") and (3") in Definition 2.1, and the packing construction C(n,t,w,e) >= D(n+1,w+t,w-e) from Theorem 2.2.
    Invoked throughout Sections 3 and 4 to replace Property B(w) and C(e) by block-size and disjointness conditions, and to obtain lower bounds. It is a prior theorem not reproved here.
  • standard math Rodl's nibble theorem, Wilson's BIBD existence theorem, and Keevash's existence theorem for Steiner systems.
    Used to assert that the upper bounds are asymptotically sharp and exact for large n satisfying divisibility conditions (Corollaries 3.6, 5.11).
  • standard math Existence of 3-frames of type g^m (Lemma 4.7) and 4-frames of type g^m (Lemma 4.8).
    Used in Theorems 4.9 and 4.12 to construct (n,t,3,1) and (n,t,4,2)-LPECC codes.
  • standard math Turan's theorem (Lemma 4.3).
    Used in Claim 4.6 to bound the number of edges in the auxiliary graph built from 2-blocks.
  • standard math Johnson-type bound A_q(n,2e+1,w) <= (q-1)^(w-e) C(n,w-e)/C(w,w-e), from Eq. (5.1) citing Ostergard and Svanstrom.
    Used in Theorem 5.1 to bound the number of parts containing codewords of weight below w.
  • standard math Near-optimal q-ary constant-weight codes for fixed odd distances, Lemma 5.6 citing Liu-Shangguan [16].
    Load-bearing for the lower bound that matches the asymptotic upper bound in Corollaries 5.7 and 5.8.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On low-power error-correcting cooling codes with large distances." pith.science (2026). https://pith.science/paper/3U3QXQ4A

@misc{pith2026241206223,
  author       = {Pith},
  title        = {Pith review of: On low-power error-correcting cooling codes with large distances},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3U3QXQ4A}},
  note         = {Machine review of arXiv:2412.06223}
}
abstract

A low-power error-correcting cooling (LPECC) code was introduced as a coding scheme for communication over a bus by Chee et al. to control the peak temperature, the average power consumption of on-chip buses, and error-correction for the transmitted information, simultaneously. Specifically, an $(n, t, w, e)$-LPECC code is a coding scheme over $n$ wires that avoids state transitions on the $t$ hottest wires and allows at most $w$ state transitions in each transmission, and can correct up to $e$ transmission errors. In this paper, we study the maximum possible size of an $(n, t, w, e)$-LPECC code, denoted by $C(n,t,w,e)$. When $w=e+2$ is large, we establish a general upper bound $C(n,t,w,w-2)\leq \lfloor \binom{n+1}{2}/\binom{w+t}{2}\rfloor$; when $w=e+2=3$, we prove $C(n,t,3,1) \leq \lfloor \frac{n(n+1)}{6(t+1)}\rfloor$. Both bounds are tight for large $n$ satisfying some divisibility conditions. Previously, tight bounds were known only for $w=e+2=3,4$ and $t\leq 2$. In general, when $w=e+d$ is large for a constant $d$, we determine the asymptotic value of $C(n,t,w,w-d)\sim \binom{n}{d}/\binom{w+t}{d}$ as $n$ goes to infinity, which can be extended to $q$-ary codes.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

  1. [1]

    Benini, G

    L. Benini, G. De Micheli, E. Macii, D. Sciuto, and C. Silva no. Asymptotic zero-transition activity encoding for address busses in low-power microprocessor-b ased systems. In Proceedings Great Lakes Symposium on VLSI , pages 77–82. IEEE, 1997. 1

    work page 1997

  2. [2]

    T. Beth, D. Jungnickel, and H. Lenz. Design Theory. Cambridge University Press, 1999. 4

    work page 1999

  3. [3]

    Calimera, K

    A. Calimera, K. Duraisami, A. Sathanur, P. Sithambaram, R. I. Bahar, A. Macii, E. Macii, and M. Pon- cino. Thermal-aware design techniques for nanometer CMOS c ircuits. Journal of Low Power Electronics , 4(3):374–384, 2008. 1

    work page 2008

  4. [4]

    Y. M. Chee, C. J. Colbourn, and A. C. Ling. Optimal memoryl ess encoding for low power off-chip data buses. In Proceedings of the 2006 IEEE/ACM International Conference on C omputer-Aided Design , pages 369–374, 2006. 1

    work page 2006

  5. [5]

    Y. M. Chee, S. H. Dau, A. C. Ling, and S. Ling. Linear size op timal q-ary constant-weight codes and constant-composition codes. IEEE Transactions on Information Theory , 56(1):140–151, 2010. 13, 16

    work page 2010

  6. [6]

    Y. M. Chee, T. Etzion, H. M. Kiah, and A. Vardy. Cooling cod es: Thermal-management coding for high-performance interconnects. IEEE Transactions on Information Theory , 64(4):3062–3085, 2017. 1, 2, 3, 13

    work page 2017

  7. [7]

    Y. M. Chee, T. Etzion, H. M. Kiah, A. Vardy, and H. Wei. Low- power cooling codes with efficient encoding and decoding. IEEE Transactions on Information Theory , 66(8):4804–4818, 2020. 1, 2, 15 16

    work page 2020

  8. [8]

    Y. M. Chee, F. Gao, H. M. Kiah, A. C. Hung Ling, H. Zhang, and X. Zhang. Decompositions of edge- colored digraphs: A new technique in the construction of con stant-weight codes and related families. SIAM Journal on Discrete Mathematics , 33(1):209–229, 2019. 13, 16

    work page 2019

Show all 28 references
  1. [9]

    Y. M. Chee and S. Ling. Constructions for q-ary constant-weight codes. IEEE Transactions on Infor- mation Theory, 53(1):135–146, 2007. 15

    2007

  2. [10]

    C. J. Colbourn and J. H. Dinitz. The CRC Handbook of Combinatorial Designs, 2nd Ed. CRC Press,

  3. [11]

    Erdős and H

    P. Erdős and H. Hanani. On a limit theorem in combinatori al analysis. Publ. Math. Debrecen , 10:10–13,

  4. [12]

    Erdős, T

    P. Erdős, T. Gallai, and Z. Tuza. Covering and independe nce in triangle structures. Discrete Mathe- matics, 150(1-3):89–101, 1996. 8

    1996

  5. [13]

    T. Etzion. Optimal constant weight codes over Z k and generalized designs. Discrete Mathematics , 169(1-3):55–82, 1997. 13

    1997

  6. [14]

    T. Etzion. Perfect codes and related structures . World Scientific, 2022. 13

    2022

  7. [15]

    P. Keevash. The existence of designs. arXiv preprint arXiv:1401.3665 , 2014. 4, 13, 16

    2014 arXiv

  8. [16]

    Liu and C

    M. Liu and C. Shangguan. Approximate generalized Stein er systems and near-optimal constant weight codes. Journal of Combinatorial Theory, Series A , 209:105955, 2025. 15

    2025

  9. [17]

    Liu and L

    S. Liu and L. Ji. Combinatorial constructions of optima l low-power error-correcting cooling codes. Designs, Codes and Cryptography , 92:2235–2252, 2024. 2, 3, 4, 8, 11

    2024

  10. [18]

    P. R. J. Östergård and M. Svanström. Ternary constant we ight codes. The Electronic Journal of Combinatorics, 9(1):R41, 2002. 13

    2002

  11. [19]

    Petrov and A

    P. Petrov and A. Orailoglu. Low-power instruction bus e ncoding for embedded processors. IEEE Transactions on Very Large Scale Integration (VLSI) Systems , 12(8):812–826, 2004. 1

    2004

  12. [20]

    V. Rödl. On a packing and covering problem. European Journal of Combinatorics , 6(1):69–78, 1985. 4

    1985

  13. [21]

    P. P. Sotiriadis and A. P. Chandrakasan. A bus energy mod el for deep submicron technology. IEEE Transactions on Very Large Scale Integration (VLSI) Systems , 10(3):341–350, 2002. 1

    2002

  14. [22]

    P. P. Sotiriadis and A. P. Chandrakasan. Bus energy redu ction by transition pattern coding using a detailed deep submicrometer bus model. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications , 50(10):1280–1295, 2003. 1

    2003

  15. [23]

    M. R. Stan and W. P. Burleson. Bus-invert coding for low- power I/O. IEEE Transactions on Very Large Scale Integration (VLSI) Systems , 3(1):49–58, 1995. 1

    1995

  16. [24]

    P. Turán. On an extremal problem in graph theory. Mat. Fiz. Lapok , 48:436–452, 1941. 8

    1941

  17. [25]

    F. Wang, M. De Bole, X. Wu, Y. Xie, N. Vijaykrishnan, and M . J. Irwin. On-chip bus thermal analysis and optimisation. IET Computers and Digital Techniques , 1(5):590–599, 2007. 1

    2007

  18. [26]

    R. M. Wilson. An existence theory for pairwise balanced designs I: Composition theorems and mor- phisms. Journal of Combinatorial Theory, Series A , 13(2):220–245, 1972. 4

    1972

  19. [27]

    R. M. Wilson. An existence theory for pairwise balanced designs II: The structure of PBD-closed sets and the existence conjectures. Journal of Combinatorial Theory, Series A , 13(2):246–273, 1972. 4

    1972

  20. [28]

    R. M. Wilson. An existence theory for pairwise balanced designs III: Proof of the existence conjectures. Journal of Combinatorial Theory, Series A , 18(1):71–79, 1975. 4 17

    1975

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.