perf(aiur): reduce witness memory and proving FFT costs - #635

Merged

arthurpaulino merged 7 commits into

mainfrom

Sep 17, 2026

Commits on Sep 16, 2026

perf(aiur): cache validated function claim shapes

arthurpaulino
and
johnchandlerburnham
committed
e2d7daf
perf(aiur): tune lookup groups from compiled degree costs

arthurpaulino
and
johnchandlerburnham
committed
f1534a8
perf(aiur): compact witness metadata and correct memory estimates

arthurpaulino
and
johnchandlerburnham
committed
c00a791
perf(aiur): consolidate byte lookup tables

arthurpaulino
and
johnchandlerburnham
committed
1ea7a68
perf(aiur): use u16 limbs for u32 comparisons

arthurpaulino
and
johnchandlerburnham
committed
f29f77e
perf(aiur): remove the intermediate u32 comparison carry constraint
arthurpaulino
committed
84b2396
test(aiur): refresh FFT pins for independent optimizations

arthurpaulino
and
johnchandlerburnham
committed
eaf84e0