Skip to content

Lowram: Share buffers with non-overlapping lifetimes in keygen#1011

Open
mkannwischer wants to merge 2 commits intomainfrom
keygen-buffer-sharing
Open

Lowram: Share buffers with non-overlapping lifetimes in keygen#1011
mkannwischer wants to merge 2 commits intomainfrom
keygen-buffer-sharing

Conversation

@mkannwischer
Copy link
Copy Markdown
Contributor

Reuse t0 as the accumulator in mld_compute_t0_t1_tr_from_sk_components, and have the caller provide s1 already in NTT form, removing two allocations (s1hat and t) from the helper.

In mld_sign_keypair_internal, share the s1 and t1 buffers via a union since s1hat is consumed before t1 is produced. Pack s1 into the secret key before the in-place NTT so the original coefficients are preserved.

Split mld_pack_sk into mld_pack_sk_s1 and mld_pack_sk_rho_key_tr_s2_t0 to support packing s1 independently before the NTT.

@mkannwischer mkannwischer force-pushed the keygen-buffer-sharing branch from 0b96d02 to 0db7b80 Compare April 2, 2026 04:00
@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 2, 2026

CBMC Results (ML-DSA-44)

⚠️ Attention Required

Proof Status Current Previous Change
mld_attempt_signature_generation ⚠️ 187s 103s +82%
Full Results (185 proofs)
Proof Status Current Previous Change
**TOTAL** 1792s 2039s -12.1%
mld_attempt_signature_generation ⚠️ 187s 103s +82%
sign_verify_internal 160s 213s -25%
polyvecl_pointwise_acc_montgomery_c 145s 336s -57%
rej_uniform_native 137s 139s -1%
poly_pointwise_montgomery_c 136s 148s -8%
mld_invntt_layer 75s 85s -12%
mld_ct_memcmp 68s 72s -6%
mld_ntt_layer 35s 52s -33%
polyvec_matrix_expand 32s 23s +39%
rej_uniform 20s 20s +0%
poly_chknorm_c 19s 20s -5%
fqmul 18s 20s -10%
sign_signature_internal 18s 19s -5%
poly_uniform_eta_4x 16s 17s -6%
mld_check_pct 15s 12s +25%
polyeta_unpack 15s 16s -6%
rej_uniform_c 15s 15s +0%
poly_uniform_4x 14s 16s -12%
polyt0_unpack 14s 14s +0%
keccakf1600x4_permute_native 13s 13s +0%
mld_ntt_butterfly_block 13s 11s +18%
mld_compute_t0_t1_tr_from_sk_components 12s 11s +9%
poly_add 11s 10s +10%
keccak_absorb_once_x4 10s 9s +11%
polyz_unpack_c 10s 12s -17%
keccak_absorb 9s 5s +80%
mld_polyvecl_permute_bitrev_to_custom_native 9s 8s +12%
polymat_permute_bitrev_to_custom 9s 26s -65%
polyvec_matrix_pointwise_montgomery 9s 6s +50%
keccakf1600_permute_native 8s 9s -11%
poly_invntt_tomont_c 8s 6s +33%
unpack_sk 8s 7s +14%
keccakf1600_permute 7s 8s -12%
polyvecl_ntt 7s 6s +17%
sign 7s 8s -12%
sign_pk_from_sk 7s 8s -12%
sign_signature 7s 5s +40%
mld_compute_pack_z 6s 7s -14%
polyveck_add 6s 8s -25%
polyveck_decompose 6s 6s +0%
polyveck_power2round 6s 6s +0%
polyvecl_pack_eta 6s 4s +50%
sign_keypair 6s 6s +0%
sign_verify_pre_hash_internal 6s 6s +0%
unpack_hints 6s 6s +0%
keccak_squeezeblocks_x4 5s 6s -17%
mld_ct_sel_int32 5s 2s +150%
mld_value_barrier_u8 5s 2s +150%
ntt_native_x86_64 5s 4s +25%
pointwise_native_x86_64 5s 2s +150%
poly_caddq_c 5s 5s +0%
poly_caddq_native_aarch64 5s 4s +25%
poly_invntt_tomont 5s 4s +25%
poly_uniform_eta 5s 5s +0%
polyt0_pack 5s 7s -29%
polyt1_pack 5s 4s +25%
polyvec_matrix_expand_serial 5s 21s -76%
polyveck_chknorm 5s 3s +67%
polyveck_pack_eta 5s 4s +25%
polyveck_sub 5s 4s +25%
polyveck_use_hint 5s 7s -29%
shake256 5s 2s +150%
sign_keypair_internal 5s 4s +25%
sign_signature_pre_hash_internal 5s 2s +150%
sign_verify_extmu 5s 4s +25%
caddq 4s 2s +100%
fqscale 4s 3s +33%
keccak_f1600_x1_native_aarch64 4s 2s +100%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid 4s 4s +0%
mld_ct_cmask_nonzero_u32 4s 2s +100%
mld_ct_get_optblocker_i64 4s 2s +100%
mld_ct_get_optblocker_u8 4s 1s +300%
mld_prepare_domain_separation_prefix 4s 6s -33%
mld_sample_s1_s2_serial 4s 4s +0%
pack_sig_h_poly 4s 5s -20%
pack_sk_rho_key_tr_s2_t0 4s - new
poly_caddq 4s 3s +33%
poly_chknorm_native 4s 2s +100%
poly_chknorm_native_aarch64 4s 4s +0%
poly_decompose 4s 6s -33%
poly_make_hint 4s 5s -20%
poly_ntt 4s 2s +100%
poly_pointwise_montgomery 4s 3s +33%
poly_pointwise_montgomery_native 4s 3s +33%
poly_power2round 4s 2s +100%
poly_reduce 4s 3s +33%
poly_uniform 4s 4s +0%
poly_uniform_gamma1_4x 4s 4s +0%
polyt1_unpack 4s 5s -20%
polyveck_invntt_tomont 4s 3s +33%
polyveck_ntt 4s 6s -33%
polyveck_reduce 4s 4s +0%
polyveck_shiftl 4s 4s +0%
polyvecl_chknorm 4s 5s -20%
polyvecl_uniform_gamma1 4s 3s +33%
polyz_pack 4s 2s +100%
rej_eta 4s 4s +0%
rej_eta_native 4s 6s -33%
shake128_release 4s 5s -20%
sign_open 4s 3s +33%
sign_verify 4s 6s -33%
sign_verify_pre_hash_shake256 4s 3s +33%
sys_check_capability 4s 4s +0%
unpack_pk 4s 2s +100%
decompose 3s 2s +50%
keccak_finalize 3s 4s -25%
keccak_squeeze 3s 6s -50%
keccakf1600x4_extract_bytes 3s 3s +0%
keccakf1600x4_xor_bytes 3s 3s +0%
mld_ct_cmask_neg_i32 3s 1s +200%
mld_h 3s 4s -25%
mld_sample_s1_s2 3s 6s -50%
mld_value_barrier_i64 3s 2s +50%
mld_value_barrier_u32 3s 1s +200%
ntt_native_aarch64 3s 5s -40%
pack_pk 3s 4s -25%
pack_sig_z 3s 5s -40%
pointwise_native_aarch64 3s 2s +50%
poly_caddq_native 3s 3s +0%
poly_challenge 3s 7s -57%
poly_invntt_tomont_native 3s 2s +50%
poly_ntt_native 3s 1s +200%
poly_sub 3s 3s +0%
poly_use_hint_c 3s 4s -25%
poly_use_hint_native 3s 6s -50%
polyeta_pack 3s 2s +50%
polyveck_caddq 3s 3s +0%
polyveck_pack_t0 3s 3s +0%
polyveck_unpack_eta 3s 4s -25%
polyvecl_pointwise_acc_montgomery 3s 3s +0%
polyvecl_pointwise_acc_montgomery_native 3s 4s -25%
polyz_unpack_native 3s 3s +0%
rej_eta_c 3s 5s -40%
shake128_finalize 3s 3s +0%
shake256_squeeze 3s 3s +0%
sign_signature_extmu 3s 6s -50%
unpack_sig 3s 3s +0%
use_hint 3s 2s +50%
intt_native_x86_64 2s 3s -33%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid 2s 1s +100%
keccakf1600_extract_bytes (big endian) 2s 5s -60%
keccakf1600_xor_bytes 2s 2s +0%
keccakf1600_xor_bytes (big endian) 2s 2s +0%
keccakf1600x4_permute 2s 2s +0%
mld_ct_get_optblocker_u32 2s 2s +0%
mld_keccakf1600_extract_bytes 2s 3s -33%
montgomery_reduce 2s 2s +0%
pack_sig_c 2s 3s -33%
pack_sk_s1 2s - new
poly_chknorm 2s 2s +0%
poly_decompose_c 2s 3s -33%
poly_decompose_native 2s 3s -33%
poly_ntt_c 2s 2s +0%
poly_shiftl 2s 3s -33%
poly_use_hint 2s 3s -33%
polyveck_pack_w1 2s 3s -33%
polyveck_pointwise_poly_montgomery 2s 3s -33%
polyveck_unpack_t0 2s 3s -33%
polyvecl_permute_bitrev_to_custom 2s 5s -60%
polyvecl_uniform_gamma1_serial 2s 3s -33%
polyvecl_unpack_eta 2s 3s -33%
polyvecl_unpack_z 2s 3s -33%
polyw1_pack 2s 2s +0%
power2round 2s 5s -60%
reduce32 2s 2s +0%
shake128_absorb 2s 3s -33%
shake128_init 2s 5s -60%
shake128_squeeze 2s 3s -33%
shake128x4_squeezeblocks 2s 2s +0%
shake256_finalize 2s 3s -33%
shake256_init 2s 2s +0%
shake256_release 2s 3s -33%
shake256x4_absorb_once 2s 4s -50%
sign_signature_pre_hash_shake256 2s 4s -50%
keccak_f1600_x1_native_aarch64_v84a 1s 3s -67%
keccak_f1600_x4_native_aarch64_v84a 1s 2s -50%
keccak_init 1s 2s -50%
make_hint 1s 1s +0%
mld_ct_abs_i32 1s 3s -67%
mld_ct_cmask_nonzero_u8 1s 3s -67%
poly_uniform_gamma1 1s 3s -67%
polyz_unpack 1s 2s -50%
shake128x4_absorb_once 1s 4s -75%
shake256_absorb 1s 2s -50%
shake256x4_squeezeblocks 1s 1s +0%

@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 2, 2026

CBMC Results (ML-DSA-87)

⚠️ Attention Required

Proof Status Current Previous Change
**TOTAL** ⚠️ 3292s 2411s +36.5%
mld_attempt_signature_generation ⚠️ 144s 96s +50%
polymat_permute_bitrev_to_custom ⚠️ 152s 27s +463%
polyveck_decompose ⚠️ 57s 11s +418%
polyvecl_pointwise_acc_montgomery_c ⚠️ 1096s 218s +403%
Full Results (185 proofs)
Proof Status Current Previous Change
**TOTAL** ⚠️ 3292s 2411s +36.5%
polyvecl_pointwise_acc_montgomery_c ⚠️ 1096s 218s +403%
sign_verify_internal 169s 147s +15%
polyvec_matrix_expand 168s 255s -34%
poly_pointwise_montgomery_c 155s 186s -17%
polymat_permute_bitrev_to_custom ⚠️ 152s 27s +463%
rej_uniform_native 145s 155s -6%
mld_attempt_signature_generation ⚠️ 144s 96s +50%
polyvec_matrix_expand_serial 82s 88s -7%
mld_ct_memcmp 75s 89s -16%
mld_invntt_layer 68s 104s -35%
polyveck_decompose ⚠️ 57s 11s +418%
mld_ntt_layer 54s 59s -8%
sign_signature_internal 35s 41s -15%
mld_compute_t0_t1_tr_from_sk_components 28s 25s +12%
rej_uniform 22s 24s -8%
polyveck_power2round 21s 20s +5%
fqmul 19s 23s -17%
poly_uniform_4x 19s 17s +12%
polyeta_unpack 19s 19s +0%
poly_chknorm_c 18s 23s -22%
poly_uniform_eta_4x 18s 16s +12%
polyt0_unpack 14s 20s -30%
keccakf1600x4_permute_native 13s 14s -7%
rej_uniform_c 13s 19s -32%
mld_ntt_butterfly_block 12s 16s -25%
mld_polyvecl_permute_bitrev_to_custom_native 12s 14s -14%
poly_add 11s 11s +0%
polyvec_matrix_pointwise_montgomery 11s 10s +10%
keccak_absorb_once_x4 10s 12s -17%
keccakf1600_permute_native 10s 9s +11%
sign_pk_from_sk 10s 7s +43%
unpack_sk 10s 8s +25%
mld_check_pct 9s 17s -47%
polyveck_ntt 9s 8s +12%
polyveck_shiftl 9s 8s +12%
polyvecl_ntt 9s 9s +0%
polyz_unpack_c 9s 12s -25%
keccak_absorb 8s 7s +14%
poly_caddq_c 8s 7s +14%
poly_invntt_tomont_c 8s 7s +14%
polyveck_add 8s 11s -27%
polyveck_invntt_tomont 8s 10s -20%
polyveck_reduce 8s 8s +0%
keccakf1600_permute 7s 6s +17%
mld_sample_s1_s2 7s 5s +40%
poly_decompose_c 7s 7s +0%
poly_power2round 7s 6s +17%
polyveck_chknorm 7s 7s +0%
polyz_pack 7s 2s +250%
sign 7s 7s +0%
sign_keypair_internal 7s 6s +17%
sign_signature_pre_hash_shake256 7s 4s +75%
sign_verify_pre_hash_internal 7s 2s +250%
sign_verify_pre_hash_shake256 7s 7s +0%
unpack_hints 7s 6s +17%
keccak_squeezeblocks_x4 6s 9s -33%
mld_h 6s 3s +100%
mld_sample_s1_s2_serial 6s 7s -14%
pack_sig_h_poly 6s 4s +50%
poly_uniform_gamma1_4x 6s 2s +200%
polyveck_pointwise_poly_montgomery 6s 6s +0%
polyveck_unpack_t0 6s 5s +20%
polyveck_use_hint 6s 12s -50%
polyvecl_chknorm 6s 8s -25%
sign_signature_pre_hash_internal 6s 4s +50%
unpack_pk 6s 5s +20%
keccak_squeeze 5s 3s +67%
mld_compute_pack_z 5s 10s -50%
poly_chknorm_native_aarch64 5s 2s +150%
poly_use_hint_c 5s 2s +150%
polyveck_pack_w1 5s 4s +25%
polyveck_sub 5s 7s -29%
rej_eta_c 5s 3s +67%
shake256_init 5s 1s +400%
sign_keypair 5s 3s +67%
caddq 4s 2s +100%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid 4s 1s +300%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid 4s 1s +300%
keccak_finalize 4s 2s +100%
keccakf1600_extract_bytes (big endian) 4s 2s +100%
keccakf1600_xor_bytes 4s 2s +100%
mld_ct_sel_int32 4s 2s +100%
mld_prepare_domain_separation_prefix 4s 4s +0%
ntt_native_aarch64 4s 4s +0%
ntt_native_x86_64 4s 3s +33%
pack_sig_c 4s 2s +100%
pack_sk_rho_key_tr_s2_t0 4s - new
pack_sk_s1 4s - new
pointwise_native_x86_64 4s 2s +100%
poly_decompose_native 4s 7s -43%
poly_ntt_native 4s 3s +33%
poly_pointwise_montgomery 4s 1s +300%
poly_pointwise_montgomery_native 4s 4s +0%
polyt0_pack 4s 5s -20%
polyveck_caddq 4s 8s -50%
polyveck_unpack_eta 4s 6s -33%
polyvecl_pack_eta 4s 3s +33%
polyvecl_permute_bitrev_to_custom 4s 4s +0%
polyvecl_unpack_eta 4s 5s -20%
polyvecl_unpack_z 4s 5s -20%
polyw1_pack 4s 3s +33%
sign_open 4s 5s -20%
sign_verify 4s 6s -33%
sign_verify_extmu 4s 4s +0%
sys_check_capability 4s 3s +33%
unpack_sig 4s 3s +33%
intt_native_x86_64 3s 1s +200%
keccak_init 3s 3s +0%
keccakf1600_xor_bytes (big endian) 3s 3s +0%
keccakf1600x4_extract_bytes 3s 2s +50%
mld_ct_get_optblocker_u32 3s 3s +0%
pointwise_native_aarch64 3s 4s -25%
poly_caddq_native 3s 2s +50%
poly_caddq_native_aarch64 3s 5s -40%
poly_challenge 3s 5s -40%
poly_decompose 3s 4s -25%
poly_invntt_tomont_native 3s 3s +0%
poly_make_hint 3s 6s -50%
poly_ntt_c 3s 6s -50%
poly_shiftl 3s 5s -40%
poly_sub 3s 3s +0%
poly_uniform 3s 5s -40%
poly_uniform_eta 3s 7s -57%
poly_use_hint 3s 3s +0%
poly_use_hint_native 3s 3s +0%
polyt1_pack 3s 3s +0%
polyveck_pack_eta 3s 4s -25%
polyvecl_pointwise_acc_montgomery 3s 4s -25%
polyvecl_pointwise_acc_montgomery_native 3s 3s +0%
polyvecl_uniform_gamma1 3s 2s +50%
polyz_unpack_native 3s 2s +50%
power2round 3s 2s +50%
rej_eta 3s 3s +0%
rej_eta_native 3s 4s -25%
shake128_absorb 3s 3s +0%
shake128_finalize 3s 2s +50%
shake128_init 3s 4s -25%
shake128_squeeze 3s 2s +50%
shake256 3s 3s +0%
shake256_release 3s 3s +0%
shake256x4_absorb_once 3s 2s +50%
sign_signature 3s 5s -40%
use_hint 3s 2s +50%
decompose 2s 2s +0%
fqscale 2s 3s -33%
keccak_f1600_x1_native_aarch64 2s 1s +100%
keccak_f1600_x1_native_aarch64_v84a 2s 2s +0%
keccakf1600x4_permute 2s 2s +0%
keccakf1600x4_xor_bytes 2s 3s -33%
make_hint 2s 4s -50%
mld_ct_abs_i32 2s 3s -33%
mld_ct_cmask_neg_i32 2s 2s +0%
mld_ct_cmask_nonzero_u8 2s 2s +0%
mld_ct_get_optblocker_i64 2s 4s -50%
mld_ct_get_optblocker_u8 2s 2s +0%
mld_value_barrier_i64 2s 2s +0%
mld_value_barrier_u32 2s 1s +100%
mld_value_barrier_u8 2s 3s -33%
montgomery_reduce 2s 3s -33%
pack_pk 2s 4s -50%
pack_sig_z 2s 4s -50%
poly_caddq 2s 4s -50%
poly_chknorm 2s 1s +100%
poly_chknorm_native 2s 4s -50%
poly_invntt_tomont 2s 4s -50%
poly_ntt 2s 4s -50%
poly_reduce 2s 4s -50%
poly_uniform_gamma1 2s 2s +0%
polyt1_unpack 2s 3s -33%
polyveck_pack_t0 2s 4s -50%
polyvecl_uniform_gamma1_serial 2s 2s +0%
polyz_unpack 2s 4s -50%
reduce32 2s 2s +0%
shake128_release 2s 3s -33%
shake128x4_absorb_once 2s 2s +0%
shake128x4_squeezeblocks 2s 2s +0%
shake256_finalize 2s 6s -67%
shake256_squeeze 2s 4s -50%
shake256x4_squeezeblocks 2s 2s +0%
sign_signature_extmu 2s 6s -67%
keccak_f1600_x4_native_aarch64_v84a 1s 1s +0%
mld_ct_cmask_nonzero_u32 1s 3s -67%
mld_keccakf1600_extract_bytes 1s 2s -50%
polyeta_pack 1s 5s -80%
shake256_absorb 1s 4s -75%

@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 2, 2026

CBMC Results (ML-DSA-65)

⚠️ Attention Required

Proof Status Current Previous Change
polyvecl_pointwise_acc_montgomery_c ⚠️ 512s 280s +83%
Full Results (185 proofs)
Proof Status Current Previous Change
**TOTAL** 2448s 2329s +5.1%
polyvecl_pointwise_acc_montgomery_c ⚠️ 512s 280s +83%
sign_verify_internal 235s 286s -18%
poly_pointwise_montgomery_c 146s 156s -6%
rej_uniform_native 142s 148s -4%
mld_attempt_signature_generation 135s 107s +26%
polyvec_matrix_expand 100s 107s -7%
polyvec_matrix_expand_serial 80s 81s -1%
mld_ct_memcmp 71s 79s -10%
mld_invntt_layer 65s 94s -31%
mld_ntt_layer 52s 55s -5%
polymat_permute_bitrev_to_custom 33s 37s -11%
sign_signature_internal 26s 25s +4%
rej_uniform 21s 22s -5%
fqmul 20s 18s +11%
mld_compute_t0_t1_tr_from_sk_components 20s 28s -29%
poly_chknorm_c 20s 19s +5%
poly_uniform_eta_4x 15s 16s -6%
polyveck_power2round 15s 12s +25%
rej_uniform_c 15s 15s +0%
mld_check_pct 14s 13s +8%
poly_uniform_4x 14s 15s -7%
polyt0_unpack 14s 15s -7%
polyvec_matrix_pointwise_montgomery 14s 15s -7%
keccakf1600x4_permute_native 13s 13s +0%
mld_ntt_butterfly_block 13s 13s +0%
polyveck_decompose 13s 16s -19%
sign_pk_from_sk 12s 9s +33%
poly_add 11s 10s +10%
polyveck_add 11s 10s +10%
keccak_absorb_once_x4 9s 10s -10%
keccakf1600_permute 9s 10s -10%
keccakf1600_permute_native 9s 8s +12%
mld_polyvecl_permute_bitrev_to_custom_native 9s 8s +12%
polyveck_caddq 9s 11s -18%
polyveck_use_hint 9s 22s -59%
mld_sample_s1_s2 8s 7s +14%
poly_invntt_tomont_c 8s 7s +14%
polyveck_chknorm 8s 8s +0%
polyveck_invntt_tomont 8s 9s -11%
polyveck_ntt 8s 7s +14%
polyveck_pointwise_poly_montgomery 8s 6s +33%
polyvecl_ntt 8s 5s +60%
keccak_squeezeblocks_x4 7s 6s +17%
keccakf1600_xor_bytes 7s 2s +250%
poly_caddq_c 7s 5s +40%
poly_decompose_c 7s 7s +0%
polyveck_shiftl 7s 7s +0%
sign 7s 7s +0%
sign_keypair_internal 7s 6s +17%
unpack_sk 7s 9s -22%
keccak_absorb 6s 5s +20%
mld_sample_s1_s2_serial 6s 3s +100%
pack_sk_s1 6s - new
poly_power2round 6s 7s -14%
poly_uniform 6s 5s +20%
polyeta_unpack 6s 6s +0%
polyveck_sub 6s 8s -25%
sign_open 6s 4s +50%
sign_signature_extmu 6s 4s +50%
sign_signature_pre_hash_internal 6s 7s -14%
caddq 5s 2s +150%
decompose 5s 4s +25%
mld_compute_pack_z 5s 6s -17%
mld_h 5s 5s +0%
poly_challenge 5s 6s -17%
poly_chknorm_native_aarch64 5s 4s +25%
poly_ntt_native 5s 3s +67%
poly_uniform_eta 5s 3s +67%
rej_eta_c 5s 3s +67%
rej_eta_native 5s 5s +0%
shake256_absorb 5s 3s +67%
sign_keypair 5s 2s +150%
sign_verify_extmu 5s 2s +150%
keccak_squeeze 4s 6s -33%
keccakf1600_xor_bytes (big endian) 4s 3s +33%
mld_ct_abs_i32 4s 2s +100%
ntt_native_aarch64 4s 3s +33%
pack_pk 4s 3s +33%
pointwise_native_x86_64 4s 1s +300%
poly_caddq_native_aarch64 4s 5s -20%
poly_decompose_native 4s 4s +0%
poly_pointwise_montgomery_native 4s 2s +100%
poly_uniform_gamma1 4s 5s -20%
poly_use_hint 4s 1s +300%
poly_use_hint_c 4s 4s +0%
polyeta_pack 4s 4s +0%
polyt1_unpack 4s 5s -20%
polyveck_reduce 4s 5s -20%
polyvecl_pack_eta 4s 3s +33%
polyvecl_uniform_gamma1 4s 3s +33%
polyvecl_unpack_z 4s 2s +100%
shake256 4s 2s +100%
sign_signature 4s 4s +0%
sys_check_capability 4s 3s +33%
fqscale 3s 4s -25%
intt_native_x86_64 3s 2s +50%
keccak_f1600_x4_native_aarch64_v84a 3s 3s +0%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid 3s 1s +200%
keccak_finalize 3s 1s +200%
keccakf1600x4_permute 3s 5s -40%
keccakf1600x4_xor_bytes 3s 3s +0%
mld_ct_cmask_neg_i32 3s 4s -25%
mld_ct_cmask_nonzero_u32 3s 3s +0%
mld_ct_cmask_nonzero_u8 3s 2s +50%
mld_ct_get_optblocker_i64 3s 2s +50%
mld_prepare_domain_separation_prefix 3s 6s -50%
montgomery_reduce 3s 5s -40%
pack_sig_c 3s 2s +50%
pack_sig_h_poly 3s 4s -25%
pack_sk_rho_key_tr_s2_t0 3s - new
poly_chknorm 3s 2s +50%
poly_chknorm_native 3s 3s +0%
poly_invntt_tomont 3s 6s -50%
poly_ntt 3s 4s -25%
poly_reduce 3s 2s +50%
poly_shiftl 3s 2s +50%
poly_sub 3s 3s +0%
poly_use_hint_native 3s 4s -25%
polyt0_pack 3s 3s +0%
polyt1_pack 3s 3s +0%
polyveck_pack_eta 3s 3s +0%
polyveck_pack_t0 3s 4s -25%
polyveck_pack_w1 3s 3s +0%
polyveck_unpack_eta 3s 2s +50%
polyvecl_chknorm 3s 5s -40%
polyvecl_permute_bitrev_to_custom 3s 4s -25%
polyvecl_pointwise_acc_montgomery 3s 4s -25%
polyvecl_pointwise_acc_montgomery_native 3s 3s +0%
polyvecl_uniform_gamma1_serial 3s 2s +50%
polyz_unpack_c 3s 2s +50%
rej_eta 3s 4s -25%
shake128_release 3s 3s +0%
shake128_squeeze 3s 2s +50%
shake128x4_absorb_once 3s 3s +0%
shake256_init 3s 2s +50%
shake256x4_absorb_once 3s 1s +200%
sign_signature_pre_hash_shake256 3s 5s -40%
sign_verify 3s 2s +50%
sign_verify_pre_hash_internal 3s 6s -50%
sign_verify_pre_hash_shake256 3s 5s -40%
unpack_hints 3s 5s -40%
use_hint 3s 3s +0%
keccak_f1600_x1_native_aarch64 2s 2s +0%
keccak_f1600_x1_native_aarch64_v84a 2s 3s -33%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid 2s 4s -50%
keccak_init 2s 2s +0%
keccakf1600_extract_bytes (big endian) 2s 4s -50%
make_hint 2s 2s +0%
mld_ct_get_optblocker_u32 2s 2s +0%
mld_ct_sel_int32 2s 4s -50%
mld_keccakf1600_extract_bytes 2s 1s +100%
mld_value_barrier_u32 2s 1s +100%
ntt_native_x86_64 2s 5s -60%
pack_sig_z 2s 2s +0%
pointwise_native_aarch64 2s 3s -33%
poly_caddq 2s 2s +0%
poly_caddq_native 2s 2s +0%
poly_decompose 2s 3s -33%
poly_invntt_tomont_native 2s 3s -33%
poly_make_hint 2s 3s -33%
poly_ntt_c 2s 2s +0%
poly_pointwise_montgomery 2s 6s -67%
poly_uniform_gamma1_4x 2s 5s -60%
polyveck_unpack_t0 2s 2s +0%
polyvecl_unpack_eta 2s 3s -33%
polyz_pack 2s 2s +0%
polyz_unpack 2s 4s -50%
polyz_unpack_native 2s 5s -60%
power2round 2s 2s +0%
shake128_absorb 2s 4s -50%
shake128_init 2s 2s +0%
shake256_finalize 2s 2s +0%
shake256_squeeze 2s 5s -60%
shake256x4_squeezeblocks 2s 3s -33%
unpack_pk 2s 4s -50%
unpack_sig 2s 2s +0%
keccakf1600x4_extract_bytes 1s 1s +0%
mld_ct_get_optblocker_u8 1s 2s -50%
mld_value_barrier_i64 1s 4s -75%
mld_value_barrier_u8 1s 4s -75%
polyw1_pack 1s 3s -67%
reduce32 1s 3s -67%
shake128_finalize 1s 1s +0%
shake128x4_squeezeblocks 1s 3s -67%
shake256_release 1s 3s -67%

@mkannwischer mkannwischer marked this pull request as ready for review April 2, 2026 04:31
@mkannwischer mkannwischer requested a review from a team as a code owner April 2, 2026 04:31
@mkannwischer mkannwischer force-pushed the keygen-buffer-sharing branch 3 times, most recently from 93856fc to 05fc00a Compare April 8, 2026 07:08
Reuse t0 as the accumulator in mld_compute_t0_t1_tr_from_sk_components,
and have the caller provide s1 already in NTT form, removing two
allocations (s1hat and t) from the helper.

In mld_sign_keypair_internal, share the s1 and t1 buffers via a union
since s1hat is consumed before t1 is produced. Pack s1 into the secret
key before the in-place NTT so the original coefficients are preserved.

Split mld_pack_sk into mld_pack_sk_s1 and mld_pack_sk_rho_key_tr_s2_t0
to support packing s1 independently before the NTT.

Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>
@mkannwischer mkannwischer force-pushed the keygen-buffer-sharing branch from 05fc00a to 035e9aa Compare April 8, 2026 09:17
Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants