Skip to content

Port: Add PMU cycle counting for Armv8.1-M #909

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
mkannwischer merged 3 commits into from
Jan 24, 2026
Merged

Port: Add PMU cycle counting for Armv8.1-M #909

mkannwischer merged 3 commits into from
Jan 24, 2026

Conversation

@willieyz
Copy link
Contributor

@willieyz willieyz commented Jan 22, 2026

Ported from pq-code-package/mlkem-native#1514, adding PMU-based cycle counting support for Armv8.1-M Cortex-M processors and aligning the RISC-V PMU configuration in hal.c with mlkem-native.

@willieyz willieyz force-pushed the port-add-PMU-armv81 branch from a04e63f to 28d55e1 Compare January 22, 2026 10:33
@oqs-bot
Copy link
Contributor

oqs-bot commented Jan 22, 2026 •
edited
Loading

CBMC Results (ML-DSA-87)

Full Results (173 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 2170s 2117s +2.5%
mld_attempt_signature_generation ✅ 199s 191s +4%
polyvec_matrix_expand ✅ 186s 190s -2%
polyvecl_pointwise_acc_montgomery_c ✅ 162s 156s +4%
rej_uniform_native ✅ 130s 125s +4%
poly_pointwise_montgomery_c ✅ 129s 128s +1%
sign_verify_internal ✅ 117s 116s +1%
polyvec_matrix_expand_serial ✅ 105s 103s +2%
mld_ct_memcmp ✅ 75s 76s -1%
mld_invntt_layer ✅ 57s 59s -3%
sign_signature_internal ✅ 55s 58s -5%
mld_ntt_layer ✅ 48s 42s +14%
keccak_squeezeblocks_x4 ✅ 43s 44s -2%
mld_compute_t0_t1_tr_from_sk_components ✅ 26s 26s +0%
polymat_permute_bitrev_to_custom ✅ 25s 23s +9%
fqmul ✅ 21s 18s +17%
rej_uniform ✅ 20s 20s +0%
poly_chknorm_c ✅ 18s 15s +20%
rej_uniform_c ✅ 18s 17s +6%
poly_uniform_eta_4x ✅ 17s 17s +0%
polyvec_matrix_pointwise_montgomery ✅ 15s 13s +15%
poly_uniform_4x ✅ 13s 12s +8%
polyt0_unpack ✅ 13s 14s -7%
polyveck_add ✅ 13s 13s +0%
polyveck_power2round ✅ 13s 12s +8%
polyveck_reduce ✅ 13s 11s +18%
keccak_absorb_once_x4 ✅ 12s 13s -8%
keccakf1600x4_permute_native ✅ 12s 14s -14%
mld_ntt_butterfly_block ✅ 12s 12s +0%
polyeta_unpack ✅ 12s 13s -8%
keccakf1600_permute_native ✅ 10s 6s +67%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 10s 12s -17%
polyveck_chknorm ✅ 10s 7s +43%
mld_check_pct ✅ 9s 5s +80%
polyveck_use_hint ✅ 9s 8s +12%
polyvecl_ntt ✅ 9s 10s -10%
keccakf1600_permute ✅ 8s 9s -11%
mld_sample_s1_s2_serial ✅ 8s 6s +33%
poly_decompose_c ✅ 8s 8s +0%
polyveck_caddq ✅ 8s 9s -11%
polyveck_pointwise_poly_montgomery ✅ 8s 7s +14%
polyveck_shiftl ✅ 8s 6s +33%
sign ✅ 8s 7s +14%
poly_invntt_tomont_c ✅ 7s 8s -12%
polyveck_decompose ✅ 7s 6s +17%
polyveck_invntt_tomont ✅ 7s 9s -22%
polyveck_ntt ✅ 7s 6s +17%
sign_keypair_internal ✅ 7s 6s +17%
sign_pk_from_sk ✅ 7s 9s -22%
sign_signature_pre_hash_internal ✅ 7s 5s +40%
sign_verify ✅ 7s 5s +40%
keccakf1600x4_permute ✅ 6s 3s +100%
mld_compute_pack_z ✅ 6s 5s +20%
mld_h ✅ 6s 6s +0%
poly_use_hint_c ✅ 6s 3s +100%
polyveck_sub ✅ 6s 7s -14%
polyvecl_pack_eta ✅ 6s 2s +200%
sign_keypair ✅ 6s 3s +100%
sign_open ✅ 6s 4s +50%
sign_signature_pre_hash_shake256 ✅ 6s 4s +50%
unpack_hints ✅ 6s 6s +0%
unpack_sk ✅ 6s 4s +50%
keccak_absorb ✅ 5s 8s -38%
mld_sample_s1_s2 ✅ 5s 6s -17%
poly_caddq_native ✅ 5s 4s +25%
poly_challenge ✅ 5s 5s +0%
polyveck_make_hint ✅ 5s 6s -17%
polyveck_pack_w1 ✅ 5s 2s +150%
polyvecl_chknorm ✅ 5s 5s +0%
power2round ✅ 5s 3s +67%
sign_signature_extmu ✅ 5s 3s +67%
sign_verify_pre_hash_shake256 ✅ 5s 3s +67%
use_hint ✅ 5s 3s +67%
caddq ✅ 4s 3s +33%
keccak_finalize ✅ 4s 6s -33%
keccakf1600_extract_bytes (big endian) ✅ 4s 1s +300%
keccakf1600x4_xor_bytes ✅ 4s 2s +100%
pack_sig_c_h ✅ 4s 2s +100%
pack_sk ✅ 4s 3s +33%
poly_add ✅ 4s 3s +33%
poly_caddq ✅ 4s 3s +33%
poly_chknorm_native ✅ 4s 4s +0%
poly_invntt_tomont ✅ 4s 3s +33%
poly_invntt_tomont_native ✅ 4s 2s +100%
poly_pointwise_montgomery_native ✅ 4s 2s +100%
poly_sub ✅ 4s 2s +100%
poly_uniform ✅ 4s 2s +100%
poly_uniform_eta ✅ 4s 2s +100%
poly_uniform_gamma1 ✅ 4s 2s +100%
poly_uniform_gamma1_4x ✅ 4s 4s +0%
poly_use_hint ✅ 4s 4s +0%
polyt0_pack ✅ 4s 4s +0%
polyvecl_permute_bitrev_to_custom ✅ 4s 2s +100%
polyvecl_uniform_gamma1 ✅ 4s 4s +0%
polyvecl_uniform_gamma1_serial ✅ 4s 5s -20%
polyvecl_unpack_eta ✅ 4s 2s +100%
polyvecl_unpack_z ✅ 4s 6s -33%
polyz_unpack_c ✅ 4s 5s -20%
rej_eta_c ✅ 4s 5s -20%
shake128_release ✅ 4s 2s +100%
shake256x4_absorb_once ✅ 4s 2s +100%
sign_verify_pre_hash_internal ✅ 4s 2s +100%
unpack_pk ✅ 4s 4s +0%
decompose ✅ 3s 4s -25%
fqscale ✅ 3s 2s +50%
keccak_squeeze ✅ 3s 4s -25%
keccakf1600_xor_bytes ✅ 3s 3s +0%
make_hint ✅ 3s 5s -40%
mld_ct_cmask_nonzero_u32 ✅ 3s 3s +0%
mld_ct_cmask_nonzero_u8 ✅ 3s 3s +0%
mld_ct_get_optblocker_u8 ✅ 3s 3s +0%
mld_prepare_domain_separation_prefix ✅ 3s 5s -40%
ntt_native_x86_64 ✅ 3s 4s -25%
pack_pk ✅ 3s 4s -25%
poly_caddq_c ✅ 3s 1s +200%
poly_decompose ✅ 3s 3s +0%
poly_decompose_native ✅ 3s 3s +0%
poly_make_hint ✅ 3s 4s -25%
poly_ntt ✅ 3s 3s +0%
poly_ntt_native ✅ 3s 1s +200%
poly_pointwise_montgomery ✅ 3s 5s -40%
poly_power2round ✅ 3s 2s +50%
poly_use_hint_native ✅ 3s 4s -25%
polyeta_pack ✅ 3s 3s +0%
polyt1_pack ✅ 3s 4s -25%
polyt1_unpack ✅ 3s 2s +50%
polyveck_pack_eta ✅ 3s 4s -25%
polyveck_pack_t0 ✅ 3s 4s -25%
polyveck_unpack_eta ✅ 3s 5s -40%
polyveck_unpack_t0 ✅ 3s 4s -25%
polyvecl_pointwise_acc_montgomery ✅ 3s 4s -25%
polyvecl_pointwise_acc_montgomery_native ✅ 3s 3s +0%
polyw1_pack ✅ 3s 4s -25%
polyz_unpack ✅ 3s 4s -25%
rej_eta ✅ 3s 1s +200%
rej_eta_native ✅ 3s 5s -40%
shake128_absorb ✅ 3s 2s +50%
shake128_finalize ✅ 3s 1s +200%
shake128_squeeze ✅ 3s 3s +0%
shake256_absorb ✅ 3s 4s -25%
shake256_init ✅ 3s 3s +0%
shake256_release ✅ 3s 4s -25%
shake256_squeeze ✅ 3s 2s +50%
shake256x4_squeezeblocks ✅ 3s 4s -25%
sign_signature ✅ 3s 2s +50%
sign_verify_extmu ✅ 3s 3s +0%
sys_check_capability ✅ 3s 3s +0%
keccak_init ✅ 2s 2s +0%
mld_ct_cmask_neg_i32 ✅ 2s 2s +0%
mld_ct_get_optblocker_u32 ✅ 2s 4s -50%
mld_ct_sel_int32 ✅ 2s 3s -33%
mld_keccakf1600_extract_bytes ✅ 2s 3s -33%
mld_value_barrier_i64 ✅ 2s 5s -60%
mld_value_barrier_u32 ✅ 2s 2s +0%
mld_value_barrier_u8 ✅ 2s 2s +0%
pack_sig_z ✅ 2s 3s -33%
poly_chknorm ✅ 2s 5s -60%
poly_ntt_c ✅ 2s 4s -50%
poly_reduce ✅ 2s 1s +100%
poly_shiftl ✅ 2s 2s +0%
polyz_pack ✅ 2s 3s -33%
polyz_unpack_native ✅ 2s 3s -33%
reduce32 ✅ 2s 3s -33%
shake128_init ✅ 2s 3s -33%
shake128x4_absorb_once ✅ 2s 5s -60%
shake128x4_squeezeblocks ✅ 2s 3s -33%
shake256 ✅ 2s 3s -33%
shake256_finalize ✅ 2s 3s -33%
unpack_sig ✅ 2s 4s -50%
keccakf1600_xor_bytes (big endian) ✅ 1s 2s -50%
keccakf1600x4_extract_bytes ✅ 1s 3s -67%
mld_ct_abs_i32 ✅ 1s 2s -50%
mld_ct_get_optblocker_i64 ✅ 1s 3s -67%
montgomery_reduce ✅ 1s 2s -50%

@oqs-bot
Copy link
Contributor

oqs-bot commented Jan 22, 2026 •
edited
Loading

CBMC Results (ML-DSA-44)

Full Results (173 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 1999s 1808s +10.6%
mld_attempt_signature_generation ✅ 242s 213s +14%
polyvecl_pointwise_acc_montgomery_c ✅ 227s 186s +22%
poly_pointwise_montgomery_c ✅ 147s 130s +13%
rej_uniform_native ✅ 140s 124s +13%
sign_verify_internal ✅ 128s 117s +9%
mld_ct_memcmp ✅ 86s 74s +16%
mld_invntt_layer ✅ 75s 70s +7%
mld_ntt_layer ✅ 47s 43s +9%
keccak_squeezeblocks_x4 ✅ 46s 47s -2%
sign_signature_internal ✅ 34s 34s +0%
rej_uniform ✅ 22s 21s +5%
fqmul ✅ 19s 21s -10%
rej_uniform_c ✅ 19s 17s +12%
poly_chknorm_c ✅ 18s 15s +20%
mld_compute_t0_t1_tr_from_sk_components ✅ 17s 12s +42%
polyvec_matrix_expand ✅ 17s 15s +13%
keccakf1600x4_permute_native ✅ 16s 13s +23%
poly_uniform_eta_4x ✅ 16s 16s +0%
polymat_permute_bitrev_to_custom ✅ 16s 15s +7%
polyt0_unpack ✅ 16s 13s +23%
poly_uniform_4x ✅ 14s 13s +8%
polyeta_unpack ✅ 14s 13s +8%
keccak_absorb_once_x4 ✅ 13s 13s +0%
mld_ntt_butterfly_block ✅ 12s 11s +9%
polyz_unpack_c ✅ 11s 11s +0%
poly_invntt_tomont_c ✅ 10s 11s -9%
polyvec_matrix_pointwise_montgomery ✅ 10s 7s +43%
keccakf1600_permute_native ✅ 9s 8s +12%
keccakf1600_permute ✅ 8s 6s +33%
mld_h ✅ 8s 7s +14%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 8s 7s +14%
sign ✅ 8s 7s +14%
mld_check_pct ✅ 7s 7s +0%
polyveck_add ✅ 7s 6s +17%
polyveck_pointwise_poly_montgomery ✅ 7s 5s +40%
sign_pk_from_sk ✅ 7s 6s +17%
decompose ✅ 6s 6s +0%
keccak_absorb ✅ 6s 7s -14%
poly_add ✅ 6s 2s +200%
poly_caddq ✅ 6s 3s +100%
poly_uniform ✅ 6s 4s +50%
polyvec_matrix_expand_serial ✅ 6s 6s +0%
polyveck_caddq ✅ 6s 5s +20%
polyveck_decompose ✅ 6s 6s +0%
polyveck_invntt_tomont ✅ 6s 5s +20%
polyveck_ntt ✅ 6s 5s +20%
polyveck_sub ✅ 6s 5s +20%
polyvecl_chknorm ✅ 6s 6s +0%
polyvecl_unpack_eta ✅ 6s 2s +200%
unpack_hints ✅ 6s 6s +0%
mld_compute_pack_z ✅ 5s 6s -17%
mld_prepare_domain_separation_prefix ✅ 5s 2s +150%
mld_sample_s1_s2 ✅ 5s 4s +25%
mld_sample_s1_s2_serial ✅ 5s 5s +0%
poly_decompose ✅ 5s 3s +67%
poly_decompose_native ✅ 5s 5s +0%
poly_invntt_tomont ✅ 5s 3s +67%
poly_use_hint_c ✅ 5s 3s +67%
polyveck_chknorm ✅ 5s 3s +67%
polyveck_pack_t0 ✅ 5s 4s +25%
polyveck_power2round ✅ 5s 4s +25%
polyveck_reduce ✅ 5s 3s +67%
polyveck_use_hint ✅ 5s 3s +67%
polyvecl_ntt ✅ 5s 6s -17%
polyz_unpack ✅ 5s 4s +25%
rej_eta_c ✅ 5s 5s +0%
rej_eta_native ✅ 5s 5s +0%
shake128_finalize ✅ 5s 2s +150%
sign_keypair ✅ 5s 5s +0%
sign_keypair_internal ✅ 5s 5s +0%
sign_verify ✅ 5s 6s -17%
sign_verify_extmu ✅ 5s 7s -29%
sys_check_capability ✅ 5s 4s +25%
unpack_sk ✅ 5s 3s +67%
keccakf1600_xor_bytes (big endian) ✅ 4s 3s +33%
keccakf1600x4_permute ✅ 4s 1s +300%
make_hint ✅ 4s 2s +100%
pack_sk ✅ 4s 3s +33%
poly_caddq_native ✅ 4s 3s +33%
poly_challenge ✅ 4s 4s +0%
poly_chknorm ✅ 4s 4s +0%
poly_chknorm_native ✅ 4s 2s +100%
poly_ntt_native ✅ 4s 4s +0%
poly_pointwise_montgomery_native ✅ 4s 3s +33%
poly_reduce ✅ 4s 3s +33%
poly_uniform_eta ✅ 4s 4s +0%
poly_uniform_gamma1 ✅ 4s 3s +33%
poly_uniform_gamma1_4x ✅ 4s 5s -20%
poly_use_hint_native ✅ 4s 3s +33%
polyeta_pack ✅ 4s 5s -20%
polyt0_pack ✅ 4s 4s +0%
polyveck_unpack_eta ✅ 4s 3s +33%
polyvecl_pack_eta ✅ 4s 3s +33%
polyvecl_pointwise_acc_montgomery ✅ 4s 4s +0%
polyvecl_uniform_gamma1 ✅ 4s 4s +0%
polyw1_pack ✅ 4s 5s -20%
polyz_pack ✅ 4s 2s +100%
shake128_absorb ✅ 4s 2s +100%
shake256 ✅ 4s 3s +33%
shake256_init ✅ 4s 4s +0%
shake256x4_squeezeblocks ✅ 4s 3s +33%
sign_signature ✅ 4s 3s +33%
sign_signature_extmu ✅ 4s 4s +0%
sign_signature_pre_hash_internal ✅ 4s 4s +0%
sign_verify_pre_hash_internal ✅ 4s 3s +33%
sign_verify_pre_hash_shake256 ✅ 4s 6s -33%
unpack_sig ✅ 4s 3s +33%
use_hint ✅ 4s 3s +33%
keccak_squeeze ✅ 3s 3s +0%
keccakf1600_extract_bytes (big endian) ✅ 3s 2s +50%
mld_ct_get_optblocker_i64 ✅ 3s 2s +50%
mld_ct_sel_int32 ✅ 3s 2s +50%
mld_keccakf1600_extract_bytes ✅ 3s 3s +0%
mld_value_barrier_i64 ✅ 3s 4s -25%
montgomery_reduce ✅ 3s 4s -25%
ntt_native_x86_64 ✅ 3s 4s -25%
pack_pk ✅ 3s 3s +0%
poly_caddq_c ✅ 3s 2s +50%
poly_decompose_c ✅ 3s 3s +0%
poly_ntt ✅ 3s 5s -40%
poly_ntt_c ✅ 3s 2s +50%
poly_sub ✅ 3s 2s +50%
polyt1_pack ✅ 3s 2s +50%
polyveck_make_hint ✅ 3s 5s -40%
polyveck_pack_w1 ✅ 3s 6s -50%
polyveck_shiftl ✅ 3s 3s +0%
polyveck_unpack_t0 ✅ 3s 2s +50%
polyvecl_uniform_gamma1_serial ✅ 3s 2s +50%
polyvecl_unpack_z ✅ 3s 4s -25%
power2round ✅ 3s 3s +0%
shake128_release ✅ 3s 3s +0%
shake128_squeeze ✅ 3s 2s +50%
shake128x4_squeezeblocks ✅ 3s 3s +0%
shake256_finalize ✅ 3s 2s +50%
shake256_release ✅ 3s 3s +0%
shake256_squeeze ✅ 3s 2s +50%
sign_open ✅ 3s 7s -57%
sign_signature_pre_hash_shake256 ✅ 3s 4s -25%
unpack_pk ✅ 3s 2s +50%
caddq ✅ 2s 3s -33%
fqscale ✅ 2s 3s -33%
keccak_init ✅ 2s 2s +0%
keccakf1600_xor_bytes ✅ 2s 1s +100%
keccakf1600x4_extract_bytes ✅ 2s 2s +0%
keccakf1600x4_xor_bytes ✅ 2s 4s -50%
mld_ct_abs_i32 ✅ 2s 2s +0%
mld_ct_cmask_neg_i32 ✅ 2s 2s +0%
mld_ct_cmask_nonzero_u32 ✅ 2s 2s +0%
mld_ct_cmask_nonzero_u8 ✅ 2s 2s +0%
mld_ct_get_optblocker_u32 ✅ 2s 1s +100%
mld_ct_get_optblocker_u8 ✅ 2s 1s +100%
mld_value_barrier_u32 ✅ 2s 2s +0%
mld_value_barrier_u8 ✅ 2s 3s -33%
pack_sig_c_h ✅ 2s 4s -50%
pack_sig_z ✅ 2s 4s -50%
poly_invntt_tomont_native ✅ 2s 2s +0%
poly_make_hint ✅ 2s 3s -33%
poly_pointwise_montgomery ✅ 2s 5s -60%
poly_power2round ✅ 2s 3s -33%
poly_shiftl ✅ 2s 2s +0%
poly_use_hint ✅ 2s 2s +0%
polyt1_unpack ✅ 2s 4s -50%
polyveck_pack_eta ✅ 2s 3s -33%
polyvecl_permute_bitrev_to_custom ✅ 2s 3s -33%
polyvecl_pointwise_acc_montgomery_native ✅ 2s 5s -60%
polyz_unpack_native ✅ 2s 2s +0%
reduce32 ✅ 2s 4s -50%
shake128x4_absorb_once ✅ 2s 2s +0%
shake256_absorb ✅ 2s 2s +0%
shake256x4_absorb_once ✅ 2s 3s -33%
keccak_finalize ✅ 1s 4s -75%
rej_eta ✅ 1s 5s -80%
shake128_init ✅ 1s 3s -67%

@oqs-bot
Copy link
Contributor

oqs-bot commented Jan 22, 2026 •
edited
Loading

CBMC Results (ML-DSA-65)

Full Results (173 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 2430s 2573s -5.6%
mld_attempt_signature_generation ✅ 412s 439s -6%
polyvecl_pointwise_acc_montgomery_c ✅ 241s 258s -7%
sign_verify_internal ✅ 176s 184s -4%
poly_pointwise_montgomery_c ✅ 136s 160s -15%
polyvec_matrix_expand ✅ 134s 144s -7%
rej_uniform_native ✅ 132s 136s -3%
mld_ct_memcmp ✅ 90s 94s -4%
polyvec_matrix_expand_serial ✅ 66s 71s -7%
mld_invntt_layer ✅ 65s 67s -3%
mld_ntt_layer ✅ 46s 50s -8%
sign_signature_internal ✅ 46s 50s -8%
keccak_squeezeblocks_x4 ✅ 44s 46s -4%
mld_compute_t0_t1_tr_from_sk_components ✅ 25s 27s -7%
rej_uniform ✅ 21s 23s -9%
fqmul ✅ 19s 22s -14%
polymat_permute_bitrev_to_custom ✅ 19s 20s -5%
poly_uniform_eta_4x ✅ 17s 19s -11%
polyt0_unpack ✅ 17s 20s -15%
rej_uniform_c ✅ 17s 24s -29%
polyvec_matrix_pointwise_montgomery ✅ 16s 16s +0%
polyveck_decompose ✅ 16s 20s -20%
mld_ntt_butterfly_block ✅ 14s 12s +17%
poly_uniform_4x ✅ 14s 15s -7%
poly_chknorm_c ✅ 13s 16s -19%
keccak_absorb_once_x4 ✅ 12s 15s -20%
keccakf1600x4_permute_native ✅ 12s 14s -14%
mld_check_pct ✅ 12s 12s +0%
polyveck_use_hint ✅ 12s 15s -20%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 10s 8s +25%
polyveck_add ✅ 10s 8s +25%
sign ✅ 10s 12s -17%
poly_decompose_c ✅ 9s 8s +12%
poly_invntt_tomont_c ✅ 9s 11s -18%
polyveck_ntt ✅ 9s 8s +12%
polyveck_power2round ✅ 9s 8s +12%
polyveck_reduce ✅ 9s 6s +50%
polyvecl_ntt ✅ 9s 7s +29%
sign_pk_from_sk ✅ 9s 9s +0%
keccakf1600_permute ✅ 8s 10s -20%
keccakf1600_permute_native ✅ 8s 9s -11%
polyeta_unpack ✅ 8s 7s +14%
polyveck_shiftl ✅ 8s 6s +33%
keccak_absorb ✅ 7s 6s +17%
mld_compute_pack_z ✅ 7s 4s +75%
mld_sample_s1_s2_serial ✅ 7s 4s +75%
polyveck_caddq ✅ 7s 7s +0%
polyveck_invntt_tomont ✅ 7s 11s -36%
polyveck_sub ✅ 7s 7s +0%
poly_power2round ✅ 6s 2s +200%
mld_h ✅ 5s 4s +25%
mld_sample_s1_s2 ✅ 5s 5s +0%
ntt_native_x86_64 ✅ 5s 4s +25%
poly_add ✅ 5s 5s +0%
poly_decompose_native ✅ 5s 4s +25%
poly_use_hint_c ✅ 5s 5s +0%
polyveck_chknorm ✅ 5s 7s -29%
polyveck_make_hint ✅ 5s 7s -29%
polyveck_pointwise_poly_montgomery ✅ 5s 5s +0%
polyvecl_pointwise_acc_montgomery_native ✅ 5s 4s +25%
shake128x4_squeezeblocks ✅ 5s 3s +67%
sign_keypair_internal ✅ 5s 4s +25%
sign_open ✅ 5s 4s +25%
sign_signature_extmu ✅ 5s 3s +67%
sign_signature_pre_hash_shake256 ✅ 5s 4s +25%
sign_verify_extmu ✅ 5s 2s +150%
sign_verify_pre_hash_shake256 ✅ 5s 4s +25%
unpack_hints ✅ 5s 5s +0%
unpack_pk ✅ 5s 4s +25%
caddq ✅ 4s 3s +33%
keccakf1600x4_xor_bytes ✅ 4s 2s +100%
mld_ct_sel_int32 ✅ 4s 1s +300%
mld_prepare_domain_separation_prefix ✅ 4s 5s -20%
montgomery_reduce ✅ 4s 3s +33%
poly_caddq_c ✅ 4s 4s +0%
poly_caddq_native ✅ 4s 4s +0%
poly_challenge ✅ 4s 5s -20%
poly_chknorm ✅ 4s 4s +0%
poly_chknorm_native ✅ 4s 5s -20%
poly_invntt_tomont ✅ 4s 2s +100%
poly_ntt ✅ 4s 1s +300%
poly_ntt_native ✅ 4s 6s -33%
poly_pointwise_montgomery ✅ 4s 2s +100%
poly_shiftl ✅ 4s 4s +0%
poly_uniform ✅ 4s 3s +33%
polyt0_pack ✅ 4s 6s -33%
polyt1_pack ✅ 4s 4s +0%
polyveck_pack_eta ✅ 4s 4s +0%
polyveck_unpack_t0 ✅ 4s 4s +0%
polyvecl_pack_eta ✅ 4s 3s +33%
polyvecl_uniform_gamma1 ✅ 4s 5s -20%
polyvecl_uniform_gamma1_serial ✅ 4s 4s +0%
polyvecl_unpack_eta ✅ 4s 4s +0%
polyvecl_unpack_z ✅ 4s 2s +100%
polyz_unpack_c ✅ 4s 4s +0%
polyz_unpack_native ✅ 4s 4s +0%
rej_eta_native ✅ 4s 6s -33%
shake128_absorb ✅ 4s 2s +100%
shake128_init ✅ 4s 3s +33%
shake256x4_absorb_once ✅ 4s 4s +0%
shake256x4_squeezeblocks ✅ 4s 3s +33%
sign_keypair ✅ 4s 5s -20%
sign_verify ✅ 4s 3s +33%
sys_check_capability ✅ 4s 4s +0%
keccak_finalize ✅ 3s 2s +50%
keccak_init ✅ 3s 2s +50%
keccak_squeeze ✅ 3s 5s -40%
keccakf1600_xor_bytes ✅ 3s 4s -25%
keccakf1600_xor_bytes (big endian) ✅ 3s 2s +50%
keccakf1600x4_extract_bytes ✅ 3s 2s +50%
mld_ct_cmask_nonzero_u32 ✅ 3s 2s +50%
mld_ct_cmask_nonzero_u8 ✅ 3s 2s +50%
mld_ct_get_optblocker_u8 ✅ 3s 2s +50%
mld_value_barrier_i64 ✅ 3s 1s +200%
mld_value_barrier_u32 ✅ 3s 2s +50%
pack_pk ✅ 3s 4s -25%
pack_sig_c_h ✅ 3s 4s -25%
pack_sig_z ✅ 3s 3s +0%
pack_sk ✅ 3s 4s -25%
poly_decompose ✅ 3s 2s +50%
poly_reduce ✅ 3s 4s -25%
poly_uniform_eta ✅ 3s 5s -40%
poly_uniform_gamma1 ✅ 3s 5s -40%
poly_uniform_gamma1_4x ✅ 3s 3s +0%
poly_use_hint ✅ 3s 3s +0%
poly_use_hint_native ✅ 3s 3s +0%
polyt1_unpack ✅ 3s 2s +50%
polyveck_pack_t0 ✅ 3s 2s +50%
polyveck_unpack_eta ✅ 3s 2s +50%
polyvecl_pointwise_acc_montgomery ✅ 3s 4s -25%
polyw1_pack ✅ 3s 4s -25%
polyz_pack ✅ 3s 4s -25%
polyz_unpack ✅ 3s 3s +0%
power2round ✅ 3s 1s +200%
rej_eta ✅ 3s 5s -40%
rej_eta_c ✅ 3s 3s +0%
shake128_finalize ✅ 3s 2s +50%
shake128_release ✅ 3s 3s +0%
shake128_squeeze ✅ 3s 2s +50%
shake128x4_absorb_once ✅ 3s 3s +0%
shake256_finalize ✅ 3s 3s +0%
shake256_init ✅ 3s 2s +50%
sign_signature_pre_hash_internal ✅ 3s 6s -50%
sign_verify_pre_hash_internal ✅ 3s 4s -25%
unpack_sig ✅ 3s 5s -40%
unpack_sk ✅ 3s 5s -40%
use_hint ✅ 3s 3s +0%
decompose ✅ 2s 4s -50%
fqscale ✅ 2s 2s +0%
keccakf1600x4_permute ✅ 2s 2s +0%
make_hint ✅ 2s 3s -33%
mld_ct_abs_i32 ✅ 2s 2s +0%
mld_ct_cmask_neg_i32 ✅ 2s 1s +100%
mld_ct_get_optblocker_i64 ✅ 2s 3s -33%
mld_keccakf1600_extract_bytes ✅ 2s 2s +0%
mld_value_barrier_u8 ✅ 2s 2s +0%
poly_caddq ✅ 2s 4s -50%
poly_make_hint ✅ 2s 4s -50%
poly_pointwise_montgomery_native ✅ 2s 4s -50%
polyeta_pack ✅ 2s 2s +0%
polyveck_pack_w1 ✅ 2s 2s +0%
polyvecl_chknorm ✅ 2s 6s -67%
polyvecl_permute_bitrev_to_custom ✅ 2s 5s -60%
reduce32 ✅ 2s 4s -50%
shake256 ✅ 2s 3s -33%
shake256_absorb ✅ 2s 3s -33%
shake256_release ✅ 2s 2s +0%
shake256_squeeze ✅ 2s 2s +0%
sign_signature ✅ 2s 4s -50%
keccakf1600_extract_bytes (big endian) ✅ 1s 2s -50%
mld_ct_get_optblocker_u32 ✅ 1s 3s -67%
poly_invntt_tomont_native ✅ 1s 4s -75%
poly_ntt_c ✅ 1s 4s -75%
poly_sub ✅ 1s 2s -50%

@willieyz willieyz force-pushed the port-add-PMU-armv81 branch from 28d55e1 to 7b303ef Compare January 23, 2026 01:21
@willieyz willieyz marked this pull request as ready for review January 23, 2026 02:26
@willieyz willieyz requested a review from a team as a code owner January 23, 2026 02:26
@willieyz willieyz marked this pull request as draft January 23, 2026 02:52
@willieyz willieyz force-pushed the port-add-PMU-armv81 branch from 7b303ef to 99ee701 Compare January 23, 2026 09:43
@willieyz willieyz marked this pull request as ready for review January 23, 2026 10:05
mkannwischer
mkannwischer requested changes Jan 24, 2026
View reviewed changes
Copy link
Contributor

@mkannwischer mkannwischer left a comment •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @willieyz. You unintentionally changed the default benchmarking parameters - probably because you copied them from mlkem-native - we cannot use the same paramters as for mlkem-native. Besides that two other nits.

I will fix both myself shortly.

mkannwischer
mkannwischer approved these changes Jan 24, 2026
View reviewed changes
Copy link
Contributor

@mkannwischer mkannwischer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @willieyz. I fixed it myself.

Note: The benchmarking parameters for the mps3-an547 are really small. Too small to produce cycle counts for signing that are reasonably close to the expected value due to the rejection sampling. However, otherwise the benchmarks take forever. For key generation and verification, we have verified that cycles are reasonably stable, close to what we get with larger iteration counts, and benchmarking complete in less than 30 seconds on the MPS3 FPGA. For proper benchmarks of signing one will have to use much larger values.

@willieyz @mkannwischer
Port: Rename benchmark parameters to MLD_BENCHMARK_xxx
48ea98d 
Make the benchmark parameters (NWARMUP, NITERATIONS, NTESTS)
configurable via CFLAGS by wrapping them in #ifndef guards and
renaming to MLD_BENCHMARK_NWARMUP, MLD_BENCHMARK_NITERATIONS,
and MLD_BENCHMARK_NTESTS.

Signed-off-by: willieyz <[email protected]>
@willieyz @mkannwischer
Port: Add PMU cycle counting for Armv8.1-M
14226b7 
Add PMU-based cycle counting support for Armv8.1-M Cortex-M processors.
This uses the CMSIS PMU APIs for portable cycle counter access.

Signed-off-by: willieyz <[email protected]>
@willieyz @mkannwischer
Port: CI test benchmarking binary for M55-AN547 using qemu
7b99f67 
The cycle counts will be zero, but it still tests the PMU code builds.

Signed-off-by: willieyz <[email protected]>
@mkannwischer mkannwischer merged commit 87f7a12 into main Jan 24, 2026
337 checks passed
@mkannwischer mkannwischer deleted the port-add-PMU-armv81 branch January 24, 2026 12:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@mkannwischer mkannwischer mkannwischer approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Port: Add PMU cycle counting for Armv8.1-M

4 participants

@willieyz @oqs-bot @mkannwischer