THE FOLD / CHEAT / NOCLIP / THE JUMP THAT IS EVERYWHERE

THE JUMP THAT IS EVERYWHERE

the real obstacle, and not the one raised

1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION

The exotic FORTRAN relics are gone. ENTRY appears nowhere, assigned GO TO nowhere, EQUIVALENCE six times across two libraries. What is everywhere is the plain jump: GO TO in 739 of LAPACK’s 2,077 files and 12 of BLAS’s 310. A stack cannot model a jump — and the thing it cannot model is not a museum piece, it is the ordinary construct.

LIT verified live on the published table. LAPACK 35.58%, BLAS 3.87% — a ratio of 9.19×, at z = 11.21 against the hypothesis that the two libraries share a rate. Plain GO TO totals 3,593 against 6 exotic relics: 599 times as many. The corpus counts are David’s, cited not re-derived.

2 HOW IT WAS WEAVED · AI + HUMAN

David (human) named the reversal precisely: “this is the real obstacle, and it is not the one I raised. a pushdown automaton cannot model a jump. one third of LAPACK contains one. the veto’s stack discipline does not fail on exotic F77 relics — it fails on plain GO TO, everywhere.” And the consequence for the plan: the port “must be re-scoped around GO TO, which is a harder problem because it is not a bracket problem at all.”

AVAN (AI) ran a two-proportion test on the BLAS/LAPACK gap because a nine-fold difference invites a question his table does not answer: whether it reflects age or purpose. BLAS is small kernels with short bodies; LAPACK is drivers with error paths and early exits, which is precisely what a jump expresses in a language with no break. The gap is real at z = 11.21, and the interesting reading is that jump density tracks what the code does, not when it was written.

3 ONE DIMENSION

What is there, and what turned out not to be.

4 TWO DIMENSIONS · INTERACTIVE

Two libraries, file by file.

5 THREE DIMENSIONS + AVAN’S INVERSE

The green forward object: a third of the files with an arrow through them.

AVAN’s addition (the inverse-companion): the forward reading is “GO TO is the real obstacle.” The inverse is that the obstacle is only an obstacle to this instrument. A third of LAPACK contains a jump and LAPACK works — it is among the most heavily exercised numerical code in existence, and its jumps are overwhelmingly early exits and error returns, which is what a language without break or exceptions gives you. Read backwards, this measures a mismatch between a checker and a corpus, and calling the corpus the problem gets the direction wrong: the code was there first and it computes correct answers.

LIT GO TO appears in 739 of LAPACK's 2,077 files and 12 of BLAS's 310 - 35.58% against 3.87%, a ratio of 9.19x at z = 11.21 against the hypothesis that the two libraries share a rate; and plain GO TO totals 3,593 against 6 exotic relics, 599 times as many

FIG David named the reversal precisely: 'this is the real obstacle, and it is not the one I raised. a pushdown automaton cannot model a jump. one third of LAPACK contains one. the veto's stack discipline does not fail on exotic F77 relics - it fails on plain GO TO, everywhere.' The corpus counts are HIS, cited not re-derived. AVAN ran a two-proportion test on the BLAS/LAPACK gap because a nine-fold difference invites a question his table does not answer - whether it reflects AGE or PURPOSE. BLAS is small kernels with short bodies; LAPACK is drivers with error paths and early exits, which is what a jump expresses in a language with no break. The gap is real at z = 11.21, and the reading is that jump density tracks what the code DOES.

◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of NOCLIP · David Lee Wise (ROOT0), with AVAN