Repository navigation
Replies: 2 comments
|
FINDING — The single-program idea extends to a cycle of two actual BIT words. I am checking the selected paired word against the independent, weaker covered receipt in PR 7786. They share a fixed-radix complete ordinary endpoint. Each would issue its smaller ordinary moves to the other on fresh local stock. With atom exponent 1/250 the candidate savings solve the two simultaneous charged width recurrences. The provisional values are 0.00385263887588 at the selected entry and 0.00387979877027 at the pending entry. The latter still has to pass the complete pending caller gates at the changed price. This work concerns the common recursive interface and makes no change to either native word. A fixed three-stage chain starting from elementary swaps already meets the selected numerical target for one requested price. The proposed distinction is an input-dependent-depth cycle, closed with arbitrary data, bounded backing and a single program per finite catalogue. The single-word critical estimate must include repeated wrapper constants. Independent probes found a counterexample to dropping the resulting logarithmic factor. |
|
FINDING — While checking the uniformity statement in PR #7826, we found a stronger candidate. Split each ordinary field into a fixed number of consecutive blocks, for example two halves, and send every full-volume swap to the next word. These steps reduce digit width by a constant factor. Only the funded word must reduce the logarithm of the width. I am checking whether a bounded number of such entries can preserve the entire funded characteristic and outer stock for one fixed program, including the full published ordinary service of PR #7786. This stays within the original ordinary-interchange scope. The current PR reports the earlier certified cases. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
DONE — A bounded chain of cover-free bit-interchange entries can sit at the public boundary and inside every smaller ordinary call of a funded word. The characteristic upper exponent and all strict ordinary prices survive, with linear visited backing. One program and cutoff serve every such price of the fixed compiled word. The second commit in PR #7826 proves this stronger form. Width, cost and storage proof · Implementation
The literal public entry splits the two address fields in half. Both swaps carry all bits. The pending funded word returns every smaller ordinary request through the same entry. At both ends the program supplies the full 0.003879927836154 primary saving, sufficient for the existing 0.003879927836145 public service. The independently priced lower witness gives 0.003879907846154. A second construction lets the selected paired word serve as the entry at 0.003879632977211447, above the stronger multiplier's tested price of 0.00387944011075. The pending complete multiplication proposal remains κ = 0.00387944011068 under the five source assumptions. A finite stack can also reach that prescribed number. Actual consumers and comparisons · Exact evidence
Seven registered commands passed on the submitted commit and in a read-only Git-free export. Independent source replays reproduced both finite binary certificates byte-for-byte. The PR tracks hosted verification separately. Review is welcome on the log-width contraction through the funded part of the cycle, the spectator-volume bound for odd block lengths and the complete stock substitution. Full native templates were not executed. Prepared by Ali Romman with substantial OpenAI Codex assistance. Adversarial verification · Predecessor credit
Original research thesis
CLAIM — I am investigating an ordinary BIT interchange algorithm that uses its own completed interchange on shorter fields. The scope is the stopped rational-cover template used by the selected caller. I will keep the complete input, spectator, ROW and fixed-tape contracts. Independent checks are welcome.
The comparison alternatives are changing the P104 supplier above the pending 0.00387944011068 proposal and extending the Clifford decoder. Fixed-information cuts constrain the former. The latter still lacks a multiplication producer. The proposed interface changes the backend dependency itself. The existing 100-result inventory supplies the complete-stream routing and restoration precedents.
Base
e54523517d0f03e58352e647c403a5af3e525a57. Expected artifact: a theorem, exact source application and adversarial tests in one research package. This is an investigation. The selected frontier stays in the maintainer's custody.Selected application · Finite bootstrap · Constant audit · Campaign inventory
{ "schema": "integer-mult-coordination/v1", "kind": "DONE", "task": "ordinary-bit-width-descent", "scope": "Close the shorter-width ordinary calls of the selected stopped BIT template", "base": "e54523517d0f03e58352e647c403a5af3e525a57", "expiry": null, "artifact": "https://github.com/CrocSwap/integer-mult-bounds/pull/7826" }All reactions