perf(prover): device-resident rounds 2-4 and fused NTT for GPU continuations - #875

Open

ColoCarletti wants to merge 9 commits into

mainfrom

gpu-opt-round2-impl

Commits on Jul 31, 2026

new opt
ColoCarletti
committed
fix(gpu): harden round-2 residency paths after review
ColoCarletti
committed
Merge branch 'main' into gpu-opt-round2-impl
ColoCarletti
authored
fix(gpu): address round-2 review — guarded zero-inverse, retargeted fault hook, merkle root-only
ColoCarletti
committed
Merge branch 'gpu-opt-round2-impl' of github.com:yetanotherco/lambda_vm into gpu-opt-round2-impl
ColoCarletti
committed
perf(gpu): stage htod_via in fixed 64MB chunks to bound pinned footprint
ColoCarletti
committed
style: rustfmt htod_via chunk-size expression
ColoCarletti
committed
fix(gpu): gate R2 comp-tree host fallback on the parts, not host_trace_empty
ColoCarletti
committed

Commits on Aug 1, 2026

Merge branch 'main' into gpu-opt-round2-impl
MauroToscano
authored