By: ValentinHirschi <valentin.hirschi@gmail.com>
speed joint certification and benchmark compressed symjit Start directed joint support certification at 128 bits and escalate the same predicate only on uncertainty, preserving the fixed Arb1000 proposal. Count all physical evaluator wrappers and expose disjoint canonical/other bench timings. Pin SymJIT 2.25.6 and archive the matched all-orientation 1/20-worker benchmark. Validation: fmt and core/API checks; 21 focused tests; clippy with only existing warnings. All-three backend audit: 28,632 checks and 8704 estimator comparisons; near-zero real-relative misses retained. 87 archived originals verified. Source index matches all 16 frozen build8 hashes. Later fixed-Quad source work remains unstaged. Saved-state and integration-checkpoint layouts are unchanged. User requests: Then, now that we have some evidence of things working well, we should aim at improving a bit the runtime, before doing more long expensive runs. To this end: a) The following is too inefficient: 1. every generated point and its complete sampling weight are prepared at 1000-bit precision, even for ordinary LMB channels. 2. Joint-map support checks use 2048-bit interval arithmetic. Find a way to safely avoid *always* going to higher-arithmetic precision (launch an agent on doing research on how avoid this safely). b) Let's compile the integrand now with SymJIT o2, and make sure we're using the latest symjit version (patch it in Cargo.toml if need be, but it's a dependency of Symbolica at this point), and use the compression mode. Measure how the timings compare w.r.t current eager mode for double-precision evaluation, but not just monocore but also in production with 20 concurrence cores (it make a difference because of L3 cache saturation potentially). Report what's your best measure runtime (separating integrand overhead, time spent in evaluators and time spent in advanced sampling) in this best setup you landed on. Can you give me a table with just the evaluators timing with Eager, Symjit uncompressed and Symjit compressed. Also mu_r should be kept to the z-mass at 91.188 GeV, and m_uv can be maybe lowered to 50 GeV or so (you can see if there is sensitivity on this, the central value must be independent of it, but the error may be a bit better at slightly smaller values of m_uv; to be checked).