H. T. Kung

Computer scientist (National Tsing Hua University, Taiwan, BS 1968; Carnegie Mellon PhD 1974) who co-invented the systolic array with Charles Leiserson in 1978 and gave the architecture its name and rationale in his own 1982 paper, "Why Systolic Architectures?" (Computer 15(1):37-46). He is now a faculty member at Harvard SEAS, having spent much of his earlier career at Carnegie Mellon, where his group's work fed CMU's Warp and iWarp systolic-hardware projects; part of the 1982 paper was written while he was on leave at ESL, Inc.'s Advanced Processor Technology Laboratory (1981).

Matters to this vault as the named origin of a metaphor that closes its own loop three decades later: Kung reached for cardiac language to describe the systolic array's rhythmic data movement (claim-systolic-array-named-after-cardiac-systole, claim-kung-1982-opening-paragraph-frames-dataflow-as-blood-circulation), and Google's TPU engineers, apparently without invoking his etymology deliberately, called their own systolic matrix unit "the heart of the TPU" in 2017 (claim-tpu-matrix-unit-called-heart-of-the-tpu). A 2026 direct read of his 1982 paper also corrected a popular misquote that had been circulating under his name — see myth-kung-1982-processor-pumps-quote.

References

• claim-systolic-array-named-after-cardiac-systole

• claim-kung-1982-opening-paragraph-frames-dataflow-as-blood-circulation

• myth-kung-1982-processor-pumps-quote

• claim-tpu-matrix-unit-called-heart-of-the-tpu (names his 1980 paper as citation [Kun80])

• claim-systolic-arrays-dormant-decades-before-tpu-revival

• claim-occ-kung-robinson-1981-detects-conflicts-after-the-fact (his other 1981 paper, a separate field entirely)

• Captures: 2026-07-11/12 (first vault mentions, secondary-sourced quote); 2026-08-12 (hub built — his 1982 paper read directly, correcting the quote)

• 2026-10-31: Three years before "Why Systolic Architectures?," Kung co-authored (with John T. Robinson) "On Optimistic Methods for Concurrency Control" (ACM TODS, 1981) — the foundational paper for what is colloquially called "optimistic locking" in database-transaction theory, a wholly separate field from the systolic-array/hardware work this hub otherwise covers. See claim-occ-kung-robinson-1981-detects-conflicts-after-the-fact.