Niek Kamer | WritingNotes on zero-knowledge cryptography, embedded Rust, and the gap between them.https://niek-kamer.github.io/enThe trait-method boundary that cost Plonky3 Poseidon2 27%https://niek-kamer.github.io/writing/goldilocks-poseidon2-cross-permute-batching/https://niek-kamer.github.io/writing/goldilocks-poseidon2-cross-permute-batching/permute_mut(a); permute_mut(b) on independent states leaves 27% on Zen 5. LLVM can't see across the function-call boundary. Hand-rolling both permutes in one scope reclaims it.Sun, 17 May 2026 00:00:00 GMTplonky3goldilocksperformanceposeidon2batchingilpx86aarch64The AVX-512 kernel that was 1.79× slower than the compiler'shttps://niek-kamer.github.io/writing/goldilocks-poseidon2-avx512-falsifications/https://niek-kamer.github.io/writing/goldilocks-poseidon2-avx512-falsifications/I spent six hours writing a hand-tuned AVX-512 Goldilocks Poseidon2 kernel. It came back 1.79× slower than the path it was trying to replace. Then I almost wrote a second one.Sat, 16 May 2026 00:00:00 GMTplonky3goldilocksperformanceposeidon2x86avx512methodologypost-mortemThe Poseidon2 regression my microbenchmark told me wasn't therehttps://niek-kamer.github.io/writing/goldilocks-poseidon2-merkle-vs-microbench/https://niek-kamer.github.io/writing/goldilocks-poseidon2-merkle-vs-microbench/A 3.5% per-permute win on Pi 5 hid a 16% Merkle wall-time loss in Plonky3's Goldilocks Poseidon2. Same kernels, same hardware. How phase decomposition reconciled the two.Fri, 15 May 2026 00:00:00 GMTplonky3goldilocksperformanceposeidon2aarch64neonbenchmarkingThe SIMD path that was 0.77× scalarhttps://niek-kamer.github.io/writing/goldilocks-poseidon1-avx-mds-fallback/https://niek-kamer.github.io/writing/goldilocks-poseidon1-avx-mds-fallback/How a perf-counter sanity check on Plonky3's AVX-2 packed Goldilocks Poseidon1 surfaced a missing Karatsuba back-port, and what assumption I'd been making.Wed, 13 May 2026 00:00:00 GMTplonky3goldilocksperformanceposeidon1x86avxauditbenchmarkingThe two instructions hiding in every Goldilocks Poseidon2 addhttps://niek-kamer.github.io/writing/goldilocks-poseidon2-neon-canonicalize/https://niek-kamer.github.io/writing/goldilocks-poseidon2-neon-canonicalize/How I dropped a redundant subs/csel pair from Plonky3's NEON Goldilocks addition kernel after staring at a perf annotate dump.Wed, 13 May 2026 00:00:00 GMTplonky3goldilocksperformanceposeidon2aarch64neonbenchmarking