code wiki / _hdl_build / nx_vmvres_gate.nx
nx_vmvres_gate.nx
buildroot/runtime/_hdl_build/nx_vmvres_gate.nx
about
nx_vmvres_gate.nx -- proves MV prediction + efficient residual coding TOGETHER slash the MV bitrate, and round-trips
exactly. Same realistic pan field as nx_vmvpred_gate; codes the predicted residuals with mvr_put (1-bit zero-flag),
reads them back (exact), and compares the total against absolute varint coding -- the 65% magnitude win now lands in
the total because the common 0-residual costs 1 bit not 5.
dependencies 5 imports · 0 importers
imports: nx_syscalls.nxnx_gate_emit_lib.nxnx_vmvpred.nxnx_vmvres.nxnx_gate_verdict.nx
imported by: nobody (leaf or entry point)
structs
| none |
consts
| none |
functions
| 11 | func vbits(v: i64) -> i64 { return 5 + ve_blen(ve_zze(v)) } |
| 13 | func main() -> i64 |