github/tfhe-rs
1
1
Fork 0
mirror of https://github.com/zama-ai/tfhe-rs.git synced 2026-01-09 14:47:56 -05:00
Code Issues Packages Projects Releases Wiki Activity

chore(bench): add suitable heuristic for zk throughput

Browse Source 
Heuristic based on PBS count was flawed since a ZK verification operation will eat up to 32 threads on the machine. The previous heuristic could generate an input data vector way bigger than the total of threads divided by 32. This in turn lead to long execution time for benchmark and generate bad results.
This commit is contained in:
David Testé
2025-05-06 10:13:35 +02:00
committed by David Testé
parent 8caff604ed
commit e29d615b9d
1 changed files with 9 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
tfhe-benchmark/benches/integer/zk_pke.rs
View File

@@ -24,6 +24,13 @@ fn write_result(file: &mut File, name: &str, value: usize) {
file.write_all(line.as_bytes()).expect(&error_message);
}
fn zk_throughput_num_elements() -> u64 {
// Number of usable threads for a verification is limited to 32 in the lib.
let max_threads = 32;
// We add 1 to be sure we saturate the target machine.
(rayon::current_num_threads() as u64 / max_threads).max(1) + 1
}
fn pke_zk_proof(c: &mut Criterion) {
let bench_name = "zk::pke_zk_proof";
let mut bench_group = c.benchmark_group(bench_name);
@@ -333,26 +340,7 @@ fn cpu_pke_zk_verify(c: &mut Criterion, results_file: &Path) {
BenchmarkType::Throughput => {
// In throughput mode object sizes are not recorded.
// Execute the operation once to know its cost.
let input_msg = rng.gen::<u64>();
let messages = vec![input_msg; fhe_uint_count];
let ct1 = tfhe::integer::ProvenCompactCiphertextList::builder(&pk)
.extend(messages.iter().copied())
.build_with_proof_packed(&crs, &metadata, compute_load)
.unwrap();
reset_pbs_count();
let _ = ct1.verify_and_expand(
&crs,
&pk,
&metadata,
IntegerCompactCiphertextListExpansionMode::CastAndUnpackIfNecessary(
casting_key.as_view(),
),
);
let pbs_count = max(get_pbs_count(), 1); // Operation might not perform any PBS, so we take 1 as default
let elements = throughput_num_threads(num_block, pbs_count);
let elements = zk_throughput_num_elements();
bench_group.throughput(Throughput::Elements(elements));
bench_id_verify = format!(
@@ -624,24 +612,7 @@ mod cuda {
let gpu_sks_vec = cuda_local_keys(&cks);
let gpu_count = get_number_of_gpus() as usize;
// Execute the operation once to know its cost.
let input_msg = rng.gen::<u64>();
let messages = vec![input_msg; fhe_uint_count];
let ct1 = tfhe::integer::ProvenCompactCiphertextList::builder(&pk)
.extend(messages.iter().copied())
.build_with_proof_packed(&crs, &metadata, compute_load)
.unwrap();
let gpu_ct1 =
CudaProvenCompactCiphertextList::from_proven_compact_ciphertext_list(
&ct1, &streams,
);
reset_pbs_count();
let _ =
gpu_ct1.verify_and_expand(&crs, &pk, &metadata, &d_ksk, &streams);
let pbs_count = max(get_pbs_count(), 1); // Operation might not perform any PBS, so we take 1 as default
let elements = throughput_num_threads(num_block, pbs_count);
let elements = zk_throughput_num_elements();
bench_group.throughput(Throughput::Elements(elements));
bench_id_verify = format!(