Performance measurements › arm_1 › kem-01 · system: x86_1 · arm_1
kem-01 Aigis-Enc+ — performance on AArch64 (system arm_1)
Independent measurement following the structure of the NICCS ARM self-assessment guide, §3.5 (1)–(7). Not a submitter self-assessment.
1. Basic information
- Category: public-key algorithm; function: key encapsulation
- Algorithm: Aigis-Enc+
- Implementation versions measured: reference
- Parameter sets:
Aigis-enc1,Aigis-enc2,Aigis-enc3 - Security evaluation: kem-01 report
- Measurement method: arm_1 method
- Submission: NICCS page
2. Assessment environment
| item | value |
|---|---|
| processor | Qualcomm Oryon (CPU 2, one core) |
| machine | ASUS Vivobook S 15 |
| clock | fixed 2.71 GHz (governor performance, minimum = maximum), boost off, SMT none |
| memory | 30562 MiB |
| OS / kernel | Ubuntu 26.04.1 LTS / 7.0.0-34-generic |
| compiler / build tool | gcc (Ubuntu 15.2.0-16ubuntu1) 15.2.0 / cmake version 4.2.3 |
| campaign start / end (UTC) | 2026-09-28T15:05:46 / 2026-09-29T08:43:52 |
3. Functional testing (KAT)
Each library was checked against the submitted KAT vectors (SHA-256 manifest kem-01/kat.sha256 in the harness) before timing.
| instance | build flags | KAT |
|---|---|---|
Aigis-enc1 |
harness-default | BUILDFAIL |
Aigis-enc2 |
harness-default | BUILDFAIL |
Aigis-enc3 |
harness-default | BUILDFAIL |
4. Performance
| instance | operation | mean cycles | mean time | ops/s | median trial | n (trials × iters) |
|---|---|---|---|---|---|---|
5. Resource consumption
Static memory is approximated by the library's loadable ELF segments; peak memory is the process high-water mark (VmHWM) including the benchmark driver and libc, so both are upper-bound proxies rather than isolated algorithm memory.
| instance | operation | static: ELF image (bytes) | baseline RSS | peak RSS |
|---|---|---|---|---|
6. Transmission and storage overhead
External public-key, ciphertext and signature sizes follow the curated performance/external_sizes.csv catalog; secret-key and shared-secret lengths remain API figures. See the size audit for disagreements.
| instance | public key | secret key | ciphertext | shared secret |
|---|---|---|---|---|
Symmetric primitives
Survey verdict: ICCS-only — own fips202.c compiled but unreachable from the API
Share of each operation spent in the ICCS placeholder hash functions and in the ICCS DRNG (reference build, measured in the same run with link-time wrappers):
| instance | operation | pseudohash/XOF/sm3 | DRNG | calls per operation |
|---|---|---|---|---|
7. Raw evidence index
Paths are relative to performance/data/arm_1/ in the harness; each JSON file records its commands, environment and trials.
| instance | item | file |
|---|---|---|
Scripts: performance/campaign.py, performance/ngcc_perf.c, performance/hashprof/ in the harness.