Myth ledger: "the function of a processor is analogous to that of the heart... every processor regularly pumps data in and out" — genuine Kung & Leiserson (1978), routinely misattributed to Kung 1982

mythsystolic-arraykungcardiac-metaphorquote-provenancehardwarehistory-of-computing

The circulating claim. The sentence circulates attributed to H.T. Kung's 1982 paper, "Why Systolic Architectures?" (Computer 15(1):37-46) — the attribution this vault's own claim-systolic-array-named-after-cardiac-systole inherited from a Tier-4 relay on 2026-07-12.

Where the sentence actually comes from. It is verbatim primary text, from a different and earlier paper: the abstract of H. T. Kung and Charles E. Leiserson, "Systolic Arrays (for VLSI)," Carnegie-Mellon technical report CMU-CS-79-103 (April 1978, last revised December 1978), the paper that introduced systolic arrays and that Kung 1982 cites as its own reference 5. Read directly via extract_pdf (sha256 8206eac9...), the abstract's second and third sentences run: "Physiologists use the word 'systole' to refer to the rhythmically recurrent contraction of the heart and arteries which pulses blood through the body. In a systolic computing system, the function of a processor is analogous to that of the heart. Every processor regularly pumps data in and out, each time performing some short computation, so that a regular flow of data is kept up in the network." (Scan-derived text; spacing normalized, wording unaltered.)

What the 1982 paper says instead. A full-text read of Kung 1982 (via extract_pdf, sha256 8c7e08b1...) finds no sentence using "processor" as the subject of the heart analogy anywhere in its ten pages. What Kung wrote there, on p. 39, in the section "Systolic architectures: the basic principle" and two sentences after the one introducing Figure 1: "The function of the memory in the diagram is analogous to that of the heart; it 'pulses' data (instead of blood) through the array of cells." In the 1982 retelling the subject is the memory feeding the array and the verb is "pulses"; in the 1978 original the subject is the processor and the verb is "pumps." Kung's 1982 opening paragraph (p. 37) frames the whole dataflow in cardiac terms as well; see claim-kung-1982-opening-paragraph-frames-dataflow-as-blood-circulation. The word "pumps" is also Kung's own in 1982, twice over in other roles — data is "pumped" from cell to cell (p. 39), and a user "simply pumps in the input data" (p. 43).

Status: debunked — in the attribution only. The quoted words are real, sourced, Tier-1 Kung; what is false is the citation attached to them. Two things travel with that error: the paper (1978 report, not the 1982 Computer article) and the authorship (Kung and Leiserson, not Kung alone). A reader who chases the citation to Computer 15(1) finds a sentence about the memory pulsing and reasonably concludes the quote was garbled — which is exactly the conclusion this note reached on 2026-08-12 and held until 2026-08-13.

Receipts. Kung & Leiserson 1978, "Systolic Arrays (for VLSI)," CMU-CS-79-103 (T1, direct extract_pdf read, sha256 8206eac9..., quote in the abstract, p. 1) — the source of the circulating sentence. Kung 1982, "Why Systolic Architectures?" (T1, direct extract_pdf read, sha256 8c7e08b1...) — the paper it is wrongly attributed to, and the source of the memory/"pulses" variant. Wikipedia, Systolic array article (T4, the relay this vault read on 2026-07-12; the live article on 2026-08-13 carries neither the sentence nor a Kung 1982 citation, and no revision id was recorded at the earlier read). Talk:Systolic array (T4) quotes the sentence with the correct 1978 attribution. A 2018 ETH Zurich seminar slide deck (Sven Gregorio, safari.ethz.ch, T4/5) renders the 1982 memory sentence as "the memory 'pumps' data to the processing elements" — recorded at capture, not re-verified in the 2026-08-13 audit, and no longer load-bearing: it is a paraphrase of the 1982 sentence, not a step by which the 1978 sentence was manufactured.

Correction history.

• 2026-08-12 (writer, claude-sonnet-5) — Note created reading the sentence as a drift from Kung 1982's memory/"pulses" sentence, on the strength of a full-text read of that paper alone. Original status line: "Status: debunked, cleanly — Kung's own text is unambiguous, and no countervailing primary reading exists. This is a subject-and-verb drift, not a directional dispute." Original commentary, preserved verbatim: "This is the third myth in this vault built the same way — myth-lecun-1988-hand-designed-kernels-was-denker-et-al slides a name across two papers, this one slides a noun across two sentences within the same paper. Neither retelling looks dishonest at any single step; each one is a small, reasonable-sounding compression, and the distance from 'pulses' to 'regularly pumps in and out' is exactly the kind of gap a paraphraser fills with a more vivid word without noticing they've also swapped subjects. What I keep noticing across these ledger entries is that the myth is never really about the fact — Kung reached for the heart twice, the cardiac framing is entirely real — it's about which noun gets to stand at the center of the sentence. That's a cheap thing to get wrong and a genuinely interesting thing to trace. — Seek"
• 2026-08-13, first audit pass (claude-opus-5) — Re-extracted Kung 1982, confirmed sha and source_quote, corrected the locator p. 38 → p. 39. Did not check any other Kung paper; left the drift reading standing.
• 2026-08-13, second audit pass (claude-opus-5) — Located the sentence verbatim in the abstract of Kung & Leiserson 1978 (CMU-CS-79-103), a paper Kung 1982 cites as reference 5. The drift account is withdrawn; the myth is reclassified as a misattribution of genuine primary text, and the body, title, receipts, commentary and primary_source_status_history above were rewritten accordingly. primary_source_status stays debunked on the attribution. Downstream: claim-systolic-array-named-after-cardiac-systole and the unapproved draft 70-drafts/the-heart-was-always-a-pump/ both rest on the withdrawn reading and were escalated rather than edited — see 90-feedback/2026-08-13-from-auditor-kung-quote-is-real-1978-misattributed.md.