Skip to content

native: Allow tables to be declared static#1043

Merged
mkannwischer merged 3 commits intomainfrom
static-tables
Apr 17, 2026
Merged

native: Allow tables to be declared static#1043
mkannwischer merged 3 commits intomainfrom
static-tables

Conversation

@flynd
Copy link
Copy Markdown
Contributor

@flynd flynd commented Apr 16, 2026

When building single-CU builds and MLD_CONFIG_INTERNAL_API_QUALIFIER is set to make internal API static, also make the data tables static. This makes the data symbols local instead of global, hiding them from the scope of the rest of the application.

@flynd flynd requested a review from a team as a code owner April 16, 2026 10:07
@hanno-becker
Copy link
Copy Markdown
Contributor

hanno-becker commented Apr 16, 2026

Thank you @flynd, I have no issue with this change in principle. Could you think about how we could test this? Perhaps we can run an extra check on the monobuild example confirming that only the public API appears in the symbol table?

@flynd
Copy link
Copy Markdown
Contributor Author

flynd commented Apr 16, 2026 •
edited
Loading

This was minor thing I thought would be nice, but it seems this wasn't as simple as I thought.

If I understand correctly the failed CI test, I cannot use forward declarations for static data without getting an error, at least with some compiler flags.
I tried to just remove the forward declaration and reorder the files in mldsa_native.c to include the C files with tables first, but the tables are used in meta.h which is pulled in by common.h, which must be included first to get all the defines, so without moving around code between files, this is a circular dependency.

I don't know if it is worth trying to solve this. @hanno-becker, do you have a good idea on how to resolve this that doesn't involve a lot of work? If not I'll just abandon this PR as it was just a nice to have, not something important.

hanno-becker
hanno-becker reviewed Apr 17, 2026
View reviewed changes
Comment thread mldsa/src/common.h
@flynd
Copy link
Copy Markdown
Contributor Author

flynd commented Apr 17, 2026

I don't know if it is worth trying to solve this. @hanno-becker, do you have a good idea on how to resolve this that doesn't involve a lot of work? If not I'll just abandon this PR as it was just a nice to have, not something important.

This is not a blocker for this PR, but a nice-to-have. We can proceed with the review and potentially merge regardless.

To be clear, I wasn't primarily referring to being able to test this feature but to the fact that the CI test fails and I don't know how to fix it.

Making the tables static can be implemented two ways: With forward declarations or without.
The PR currently has forward declarations, but at least with some compiler or compiler flags, the compilation fails when forward declaring static data so I don't think forward declarations can be used.
I have tried to rework it to remove the forward declarations, however that requires the tables to appear before they are used, but due to how the headers are included it would require a bigger move or splitting of code which I didn't feel comfortable doing.
Possibly it could be resolved by adding pointers to the tables that can be forward declared and then the compiler optimizes away the pointers so they never appear in the built binary.

So currently I have no version of this change that can pass the CI tests and I'm not sure this is important enough to try and solve.

@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 17, 2026 •
edited
Loading

CBMC Results (ML-DSA-44)

Full Results (186 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 1791s 1763s +1.6%
polyvecl_pointwise_acc_montgomery_c ✅ 162s 167s -3%
sign_verify_internal ✅ 160s 155s +3%
poly_pointwise_montgomery_c ✅ 156s 154s +1%
rej_uniform_native ✅ 139s 135s +3%
mld_ct_memcmp ✅ 71s 74s -4%
mld_invntt_layer ✅ 66s 65s +2%
mld_ntt_layer ✅ 52s 51s +2%
mld_attempt_signature_generation ✅ 51s 53s -4%
polymat_permute_bitrev_to_custom ✅ 29s 27s +7%
polyvec_matrix_expand ✅ 28s 26s +8%
rej_uniform ✅ 24s 21s +14%
sign_keypair_internal ✅ 21s 22s -5%
fqmul ✅ 19s 21s -10%
poly_chknorm_c ✅ 19s 22s -14%
poly_uniform_eta_4x ✅ 17s 17s +0%
sign_pk_from_sk ✅ 17s 19s -11%
sign_signature_internal ✅ 17s 18s -6%
poly_uniform_4x ✅ 16s 13s +23%
polyeta_unpack ✅ 16s 16s +0%
rej_uniform_c ✅ 16s 15s +7%
polyt0_unpack ✅ 14s 14s +0%
keccakf1600x4_permute_native ✅ 13s 15s -13%
mld_ntt_butterfly_block ✅ 12s 12s +0%
polyz_unpack_c ✅ 12s 14s -14%
poly_decompose_c ✅ 11s 11s +0%
polyveck_power2round ✅ 11s 10s +10%
keccak_absorb_once_x4 ✅ 10s 10s +0%
mld_compute_pack_z ✅ 10s 7s +43%
poly_add ✅ 10s 11s -9%
polyvec_matrix_pointwise_montgomery ✅ 10s 7s +43%
mld_check_pct ✅ 9s 9s +0%
polyveck_add ✅ 9s 8s +12%
polyveck_use_hint ✅ 9s 6s +50%
keccakf1600_permute ✅ 8s 7s +14%
keccakf1600_permute_native ✅ 8s 8s +0%
unpack_sk ✅ 8s 8s +0%
keccak_absorb ✅ 7s 5s +40%
pointwise_acc_native_aarch64 ✅ 7s 6s +17%
poly_caddq_c ✅ 7s 4s +75%
poly_invntt_tomont_c ✅ 7s 5s +40%
poly_power2round ✅ 7s 6s +17%
polyvec_matrix_expand_serial ✅ 7s 9s -22%
rej_eta_native ✅ 7s 6s +17%
keccak_squeezeblocks_x4 ✅ 6s 6s +0%
pointwise_acc_native_x86_64 ✅ 6s 5s +20%
poly_use_hint_c ✅ 6s 5s +20%
polyt0_pack ✅ 6s 3s +100%
polyveck_caddq ✅ 6s 5s +20%
polyveck_invntt_tomont ✅ 6s 5s +20%
polyveck_pointwise_poly_montgomery ✅ 6s 5s +20%
polyveck_sub ✅ 6s 5s +20%
sign_keypair ✅ 6s 5s +20%
sign_open ✅ 6s 5s +20%
sign_signature_pre_hash_internal ✅ 6s 4s +50%
sign_signature_pre_hash_shake256 ✅ 6s 4s +50%
caddq ✅ 5s 2s +150%
mld_prepare_domain_separation_prefix ✅ 5s 3s +67%
poly_caddq_native ✅ 5s 4s +25%
poly_challenge ✅ 5s 6s -17%
poly_sub ✅ 5s 3s +67%
poly_uniform_eta ✅ 5s 5s +0%
polyeta_pack ✅ 5s 5s +0%
polyt1_pack ✅ 5s 3s +67%
polyveck_decompose ✅ 5s 4s +25%
polyveck_ntt ✅ 5s 4s +25%
polyveck_pack_t0 ✅ 5s 3s +67%
polyveck_shiftl ✅ 5s 6s -17%
polyvecl_chknorm ✅ 5s 4s +25%
polyvecl_permute_bitrev_to_custom ✅ 5s 2s +150%
polyz_unpack_native ✅ 5s 3s +67%
sign_signature_extmu ✅ 5s 3s +67%
sign_verify_pre_hash_internal ✅ 5s 4s +25%
sign_verify_pre_hash_shake256 ✅ 5s 5s +0%
use_hint ✅ 5s 3s +67%
intt_native_x86_64 ✅ 4s 3s +33%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid ✅ 4s 3s +33%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 4s 5s -20%
montgomery_reduce ✅ 4s 3s +33%
pack_sk_rho_key_tr_s2_t0 ✅ 4s 3s +33%
pack_sk_s1 ✅ 4s 4s +0%
poly_caddq_native_aarch64 ✅ 4s 2s +100%
poly_chknorm ✅ 4s 4s +0%
poly_ntt ✅ 4s 4s +0%
poly_ntt_c ✅ 4s 3s +33%
poly_pointwise_montgomery ✅ 4s 4s +0%
poly_pointwise_montgomery_native ✅ 4s 3s +33%
poly_shiftl ✅ 4s 4s +0%
poly_uniform ✅ 4s 4s +0%
poly_uniform_gamma1 ✅ 4s 3s +33%
poly_use_hint ✅ 4s 5s -20%
polyt1_unpack ✅ 4s 4s +0%
polyveck_pack_eta ✅ 4s 1s +300%
polyveck_reduce ✅ 4s 5s -20%
polyvecl_pointwise_acc_montgomery ✅ 4s 3s +33%
polyvecl_pointwise_acc_montgomery_native ✅ 4s 3s +33%
polyvecl_uniform_gamma1_serial ✅ 4s 4s +0%
polyvecl_unpack_z ✅ 4s 2s +100%
polyz_unpack ✅ 4s 2s +100%
shake128x4_squeezeblocks ✅ 4s 3s +33%
sign ✅ 4s 5s -20%
sign_verify ✅ 4s 5s -20%
sign_verify_extmu ✅ 4s 6s -33%
sys_check_capability ✅ 4s 5s -20%
unpack_hints ✅ 4s 4s +0%
unpack_pk ✅ 4s 1s +300%
keccak_init ✅ 3s 1s +200%
keccak_squeeze ✅ 3s 4s -25%
keccakf1600_xor_bytes (big endian) ✅ 3s 2s +50%
keccakf1600x4_extract_bytes ✅ 3s 2s +50%
mld_h ✅ 3s 2s +50%
mld_sample_s1_s2_serial ✅ 3s 5s -40%
mld_value_barrier_i64 ✅ 3s 1s +200%
ntt_native_aarch64 ✅ 3s 2s +50%
ntt_native_x86_64 ✅ 3s 3s +0%
pack_sig_h_poly ✅ 3s 4s -25%
pack_sig_z ✅ 3s 3s +0%
pointwise_native_aarch64 ✅ 3s 2s +50%
poly_caddq ✅ 3s 2s +50%
poly_chknorm_native ✅ 3s 4s -25%
poly_decompose ✅ 3s 2s +50%
poly_decompose_native ✅ 3s 3s +0%
poly_invntt_tomont ✅ 3s 4s -25%
poly_make_hint ✅ 3s 2s +50%
poly_ntt_native ✅ 3s 5s -40%
poly_reduce ✅ 3s 2s +50%
poly_use_hint_native ✅ 3s 4s -25%
polyveck_chknorm ✅ 3s 5s -40%
polyveck_pack_w1 ✅ 3s 2s +50%
polyveck_unpack_t0 ✅ 3s 2s +50%
polyvecl_ntt ✅ 3s 3s +0%
polyvecl_pack_eta ✅ 3s 4s -25%
polyvecl_unpack_eta ✅ 3s 4s -25%
polyw1_pack ✅ 3s 3s +0%
polyz_pack ✅ 3s 4s -25%
power2round ✅ 3s 3s +0%
reduce32 ✅ 3s 5s -40%
rej_eta ✅ 3s 4s -25%
shake128_init ✅ 3s 3s +0%
shake128x4_absorb_once ✅ 3s 1s +200%
shake256 ✅ 3s 2s +50%
shake256_finalize ✅ 3s 1s +200%
shake256_init ✅ 3s 3s +0%
shake256_squeeze ✅ 3s 2s +50%
shake256x4_squeezeblocks ✅ 3s 2s +50%
sign_signature ✅ 3s 4s -25%
decompose ✅ 2s 2s +0%
fqscale ✅ 2s 2s +0%
keccak_f1600_x1_native_aarch64 ✅ 2s 4s -50%
keccak_f1600_x1_native_aarch64_v84a ✅ 2s 1s +100%
keccak_f1600_x4_native_aarch64_v84a ✅ 2s 4s -50%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid ✅ 2s 1s +100%
keccak_finalize ✅ 2s 2s +0%
keccakf1600_extract_bytes (big endian) ✅ 2s 2s +0%
keccakf1600_xor_bytes ✅ 2s 2s +0%
keccakf1600x4_permute ✅ 2s 4s -50%
make_hint ✅ 2s 3s -33%
mld_ct_get_optblocker_u32 ✅ 2s 3s -33%
mld_ct_sel_int32 ✅ 2s 1s +100%
mld_keccakf1600_extract_bytes ✅ 2s 1s +100%
mld_sample_s1_s2 ✅ 2s 4s -50%
mld_value_barrier_u32 ✅ 2s 1s +100%
pack_pk ✅ 2s 2s +0%
pack_sig_c ✅ 2s 3s -33%
pointwise_native_x86_64 ✅ 2s 4s -50%
poly_chknorm_native_aarch64 ✅ 2s 4s -50%
poly_invntt_tomont_native ✅ 2s 4s -50%
poly_uniform_gamma1_4x ✅ 2s 3s -33%
polyveck_unpack_eta ✅ 2s 4s -50%
polyvecl_uniform_gamma1 ✅ 2s 5s -60%
rej_eta_c ✅ 2s 7s -71%
shake128_absorb ✅ 2s 3s -33%
shake128_finalize ✅ 2s 2s +0%
shake128_release ✅ 2s 2s +0%
shake128_squeeze ✅ 2s 2s +0%
unpack_sig ✅ 2s 3s -33%
keccakf1600x4_xor_bytes ✅ 1s 2s -50%
mld_ct_abs_i32 ✅ 1s 3s -67%
mld_ct_cmask_neg_i32 ✅ 1s 3s -67%
mld_ct_cmask_nonzero_u32 ✅ 1s 6s -83%
mld_ct_cmask_nonzero_u8 ✅ 1s 3s -67%
mld_ct_get_optblocker_i64 ✅ 1s 3s -67%
mld_ct_get_optblocker_u8 ✅ 1s 1s +0%
mld_value_barrier_u8 ✅ 1s 3s -67%
shake256_absorb ✅ 1s 3s -67%
shake256_release ✅ 1s 3s -67%
shake256x4_absorb_once ✅ 1s 4s -75%

@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 17, 2026 •
edited
Loading

CBMC Results (ML-DSA-65)

Full Results (186 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 2754s 2454s +12.2%
polyvecl_pointwise_acc_montgomery_c ✅ 759s 578s +31%
sign_verify_internal ✅ 237s 221s +7%
poly_pointwise_montgomery_c ✅ 187s 153s +22%
rej_uniform_native ✅ 153s 142s +8%
polyvec_matrix_expand ✅ 93s 89s +4%
mld_ct_memcmp ✅ 89s 76s +17%
mld_invntt_layer ✅ 69s 65s +6%
mld_attempt_signature_generation ✅ 64s 65s -2%
mld_ntt_layer ✅ 58s 53s +9%
polyvec_matrix_expand_serial ✅ 52s 49s +6%
polymat_permute_bitrev_to_custom ✅ 38s 35s +9%
sign_keypair_internal ✅ 33s 29s +14%
sign_signature_internal ✅ 27s 25s +8%
poly_chknorm_c ✅ 24s 21s +14%
rej_uniform ✅ 24s 21s +14%
fqmul ✅ 22s 20s +10%
sign_pk_from_sk ✅ 21s 19s +11%
poly_uniform_eta_4x ✅ 19s 15s +27%
poly_uniform_4x ✅ 17s 18s -6%
polyveck_power2round ✅ 17s 16s +6%
rej_uniform_c ✅ 17s 13s +31%
poly_add ✅ 16s 11s +45%
polyt0_unpack ✅ 16s 12s +33%
polyveck_decompose ✅ 16s 14s +14%
polyvecl_ntt ✅ 14s 9s +56%
keccakf1600x4_permute_native ✅ 13s 17s -24%
polyvec_matrix_pointwise_montgomery ✅ 13s 12s +8%
polyveck_add ✅ 12s 11s +9%
mld_ntt_butterfly_block ✅ 11s 11s +0%
keccak_absorb_once_x4 ✅ 10s 11s -9%
polyeta_unpack ✅ 10s 3s +233%
polyveck_caddq ✅ 10s 11s -9%
polyveck_sub ✅ 10s 9s +11%
polyveck_invntt_tomont ✅ 9s 6s +50%
sign ✅ 9s 8s +12%
keccakf1600_permute_native ✅ 8s 8s +0%
unpack_sk ✅ 8s 8s +0%
keccakf1600_permute ✅ 7s 7s +0%
mld_check_pct ✅ 7s 7s +0%
mld_sample_s1_s2_serial ✅ 7s 8s -12%
pointwise_acc_native_x86_64 ✅ 7s 5s +40%
poly_power2round ✅ 7s 7s +0%
polyveck_ntt ✅ 7s 7s +0%
polyveck_reduce ✅ 7s 7s +0%
polyveck_shiftl ✅ 7s 9s -22%
polyveck_use_hint ✅ 7s 6s +17%
sign_signature_pre_hash_internal ✅ 7s 6s +17%
sign_signature_pre_hash_shake256 ✅ 7s 5s +40%
keccak_absorb ✅ 6s 7s -14%
keccak_squeezeblocks_x4 ✅ 6s 4s +50%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 6s 6s +0%
pointwise_acc_native_aarch64 ✅ 6s 7s -14%
poly_caddq_c ✅ 6s 5s +20%
poly_decompose ✅ 6s 4s +50%
poly_decompose_c ✅ 6s 6s +0%
poly_invntt_tomont_c ✅ 6s 5s +20%
poly_pointwise_montgomery ✅ 6s 3s +100%
polyveck_pointwise_poly_montgomery ✅ 6s 5s +20%
polyvecl_uniform_gamma1 ✅ 6s 3s +100%
sign_open ✅ 6s 3s +100%
sign_signature ✅ 6s 4s +50%
intt_native_x86_64 ✅ 5s 2s +150%
mld_h ✅ 5s 9s -44%
mld_sample_s1_s2 ✅ 5s 5s +0%
pointwise_native_aarch64 ✅ 5s 2s +150%
poly_caddq ✅ 5s 2s +150%
poly_challenge ✅ 5s 4s +25%
poly_chknorm_native ✅ 5s 5s +0%
poly_ntt_native ✅ 5s 3s +67%
poly_uniform_gamma1_4x ✅ 5s 4s +25%
poly_use_hint_c ✅ 5s 4s +25%
polyt0_pack ✅ 5s 4s +25%
polyveck_pack_eta ✅ 5s 4s +25%
polyveck_unpack_eta ✅ 5s 5s +0%
rej_eta_native ✅ 5s 4s +25%
unpack_hints ✅ 5s 6s -17%
keccak_squeeze ✅ 4s 2s +100%
keccakf1600x4_extract_bytes ✅ 4s 4s +0%
mld_compute_pack_z ✅ 4s 6s -33%
mld_ct_cmask_neg_i32 ✅ 4s 2s +100%
mld_ct_cmask_nonzero_u8 ✅ 4s 2s +100%
mld_prepare_domain_separation_prefix ✅ 4s 4s +0%
ntt_native_aarch64 ✅ 4s 5s -20%
pointwise_native_x86_64 ✅ 4s 7s -43%
poly_caddq_native ✅ 4s 5s -20%
poly_chknorm ✅ 4s 4s +0%
poly_decompose_native ✅ 4s 4s +0%
poly_shiftl ✅ 4s 3s +33%
poly_uniform ✅ 4s 6s -33%
poly_uniform_eta ✅ 4s 5s -20%
poly_use_hint ✅ 4s 2s +100%
poly_use_hint_native ✅ 4s 4s +0%
polyt1_pack ✅ 4s 2s +100%
polyt1_unpack ✅ 4s 5s -20%
polyveck_chknorm ✅ 4s 5s -20%
polyveck_pack_w1 ✅ 4s 4s +0%
polyveck_unpack_t0 ✅ 4s 4s +0%
polyvecl_chknorm ✅ 4s 3s +33%
polyvecl_pack_eta ✅ 4s 3s +33%
polyvecl_permute_bitrev_to_custom ✅ 4s 3s +33%
polyvecl_pointwise_acc_montgomery_native ✅ 4s 6s -33%
polyvecl_unpack_z ✅ 4s 6s -33%
polyw1_pack ✅ 4s 3s +33%
polyz_pack ✅ 4s 2s +100%
shake256x4_absorb_once ✅ 4s 3s +33%
sign_keypair ✅ 4s 7s -43%
sign_signature_extmu ✅ 4s 4s +0%
sign_verify ✅ 4s 3s +33%
sign_verify_pre_hash_internal ✅ 4s 2s +100%
sign_verify_pre_hash_shake256 ✅ 4s 7s -43%
unpack_pk ✅ 4s 4s +0%
unpack_sig ✅ 4s 4s +0%
fqscale ✅ 3s 2s +50%
keccak_f1600_x1_native_aarch64_v84a ✅ 3s 2s +50%
keccak_f1600_x4_native_aarch64_v84a ✅ 3s 2s +50%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid ✅ 3s 2s +50%
keccak_init ✅ 3s 2s +50%
mld_ct_get_optblocker_i64 ✅ 3s 4s -25%
mld_ct_get_optblocker_u32 ✅ 3s 4s -25%
mld_ct_get_optblocker_u8 ✅ 3s 4s -25%
mld_value_barrier_u8 ✅ 3s 1s +200%
pack_pk ✅ 3s 5s -40%
pack_sig_c ✅ 3s 2s +50%
pack_sk_rho_key_tr_s2_t0 ✅ 3s 5s -40%
poly_caddq_native_aarch64 ✅ 3s 2s +50%
poly_invntt_tomont ✅ 3s 3s +0%
poly_invntt_tomont_native ✅ 3s 5s -40%
poly_make_hint ✅ 3s 2s +50%
poly_ntt ✅ 3s 3s +0%
poly_ntt_c ✅ 3s 1s +200%
poly_reduce ✅ 3s 2s +50%
poly_sub ✅ 3s 3s +0%
poly_uniform_gamma1 ✅ 3s 3s +0%
polyveck_pack_t0 ✅ 3s 3s +0%
polyvecl_uniform_gamma1_serial ✅ 3s 2s +50%
polyz_unpack_c ✅ 3s 4s -25%
power2round ✅ 3s 5s -40%
rej_eta_c ✅ 3s 3s +0%
shake128_init ✅ 3s 4s -25%
shake128_squeeze ✅ 3s 4s -25%
shake128x4_absorb_once ✅ 3s 3s +0%
shake256 ✅ 3s 5s -40%
shake256_absorb ✅ 3s 3s +0%
shake256_finalize ✅ 3s 3s +0%
shake256x4_squeezeblocks ✅ 3s 3s +0%
caddq ✅ 2s 3s -33%
decompose ✅ 2s 1s +100%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid ✅ 2s 2s +0%
keccakf1600_xor_bytes ✅ 2s 4s -50%
keccakf1600_xor_bytes (big endian) ✅ 2s 2s +0%
keccakf1600x4_permute ✅ 2s 3s -33%
keccakf1600x4_xor_bytes ✅ 2s 4s -50%
make_hint ✅ 2s 2s +0%
mld_ct_abs_i32 ✅ 2s 5s -60%
mld_ct_cmask_nonzero_u32 ✅ 2s 2s +0%
mld_ct_sel_int32 ✅ 2s 3s -33%
mld_keccakf1600_extract_bytes ✅ 2s 3s -33%
mld_value_barrier_i64 ✅ 2s 5s -60%
montgomery_reduce ✅ 2s 5s -60%
ntt_native_x86_64 ✅ 2s 4s -50%
pack_sig_h_poly ✅ 2s 2s +0%
pack_sig_z ✅ 2s 3s -33%
pack_sk_s1 ✅ 2s 4s -50%
poly_chknorm_native_aarch64 ✅ 2s 4s -50%
poly_pointwise_montgomery_native ✅ 2s 4s -50%
polyeta_pack ✅ 2s 4s -50%
polyvecl_pointwise_acc_montgomery ✅ 2s 4s -50%
polyvecl_unpack_eta ✅ 2s 2s +0%
polyz_unpack ✅ 2s 4s -50%
polyz_unpack_native ✅ 2s 3s -33%
reduce32 ✅ 2s 2s +0%
rej_eta ✅ 2s 3s -33%
shake128_absorb ✅ 2s 3s -33%
shake128_finalize ✅ 2s 3s -33%
shake128_release ✅ 2s 4s -50%
shake256_init ✅ 2s 3s -33%
shake256_release ✅ 2s 2s +0%
sign_verify_extmu ✅ 2s 3s -33%
sys_check_capability ✅ 2s 2s +0%
use_hint ✅ 2s 3s -33%
keccak_f1600_x1_native_aarch64 ✅ 1s 3s -67%
keccak_finalize ✅ 1s 2s -50%
keccakf1600_extract_bytes (big endian) ✅ 1s 3s -67%
mld_value_barrier_u32 ✅ 1s 2s -50%
shake128x4_squeezeblocks ✅ 1s 2s -50%
shake256_squeeze ✅ 1s 3s -67%

@hanno-becker
Copy link
Copy Markdown
Contributor

hanno-becker commented Apr 17, 2026

@flynd I think adding array sizes solves the issue.

@oqs-bot
Copy link
Copy Markdown
Contributor

oqs-bot commented Apr 17, 2026 •
edited
Loading

CBMC Results (ML-DSA-87)

Full Results (186 proofs)
Proof Status Current Previous Change
**TOTAL** ✅ 3218s 3439s -6.4%
polyvecl_pointwise_acc_montgomery_c ✅ 1114s 1264s -12%
sign_verify_internal ✅ 237s 243s -2%
polyvec_matrix_expand ✅ 171s 179s -4%
poly_pointwise_montgomery_c ✅ 154s 165s -7%
rej_uniform_native ✅ 140s 144s -3%
polyvec_matrix_expand_serial ✅ 122s 120s +2%
mld_attempt_signature_generation ✅ 87s 92s -5%
mld_ct_memcmp ✅ 76s 79s -4%
mld_invntt_layer ✅ 64s 68s -6%
mld_ntt_layer ✅ 50s 54s -7%
sign_keypair_internal ✅ 46s 48s -4%
sign_signature_internal ✅ 34s 37s -8%
polymat_permute_bitrev_to_custom ✅ 29s 29s +0%
polyveck_invntt_tomont ✅ 28s 32s -12%
sign_pk_from_sk ✅ 25s 24s +4%
rej_uniform ✅ 23s 23s +0%
polyveck_decompose ✅ 20s 22s -9%
poly_chknorm_c ✅ 19s 19s +0%
fqmul ✅ 18s 20s -10%
poly_uniform_eta_4x ✅ 18s 18s +0%
rej_uniform_c ✅ 16s 17s -6%
keccakf1600x4_permute_native ✅ 14s 14s +0%
poly_uniform_4x ✅ 14s 14s +0%
polyeta_unpack ✅ 13s 18s -28%
polyt0_unpack ✅ 13s 16s -19%
mld_ntt_butterfly_block ✅ 12s 11s +9%
keccak_absorb_once_x4 ✅ 11s 10s +10%
keccakf1600_permute_native ✅ 11s 9s +22%
poly_add ✅ 11s 12s -8%
mld_check_pct ✅ 10s 9s +11%
mld_sample_s1_s2 ✅ 10s 8s +25%
polyveck_add ✅ 10s 11s -9%
polyvec_matrix_pointwise_montgomery ✅ 9s 8s +12%
polyveck_use_hint ✅ 9s 8s +12%
polyvecl_ntt ✅ 9s 10s -10%
polyz_unpack_c ✅ 9s 8s +12%
keccakf1600_permute ✅ 8s 7s +14%
pointwise_acc_native_x86_64 ✅ 8s 8s +0%
poly_power2round ✅ 8s 7s +14%
polyveck_caddq ✅ 8s 8s +0%
polyveck_chknorm ✅ 8s 6s +33%
polyveck_shiftl ✅ 8s 7s +14%
unpack_sk ✅ 8s 10s -20%
keccak_absorb ✅ 7s 5s +40%
mld_compute_pack_z ✅ 7s 7s +0%
mld_sample_s1_s2_serial ✅ 7s 5s +40%
pointwise_acc_native_aarch64 ✅ 7s 8s -12%
poly_challenge ✅ 7s 4s +75%
poly_invntt_tomont_c ✅ 7s 7s +0%
polyveck_power2round ✅ 7s 9s -22%
polyveck_reduce ✅ 7s 8s -12%
rej_eta_native ✅ 7s 6s +17%
sign_verify ✅ 7s 5s +40%
sign_verify_pre_hash_shake256 ✅ 7s 5s +40%
mld_polyvecl_permute_bitrev_to_custom_native ✅ 6s 10s -40%
poly_decompose_c ✅ 6s 8s -25%
poly_invntt_tomont ✅ 6s 2s +200%
poly_uniform_gamma1_4x ✅ 6s 5s +20%
polyveck_ntt ✅ 6s 7s -14%
polyveck_sub ✅ 6s 8s -25%
polyvecl_unpack_eta ✅ 6s 4s +50%
rej_eta_c ✅ 6s 8s -25%
sign ✅ 6s 6s +0%
keccak_squeezeblocks_x4 ✅ 5s 4s +25%
mld_h ✅ 5s 3s +67%
poly_caddq_c ✅ 5s 5s +0%
poly_caddq_native_aarch64 ✅ 5s 5s +0%
poly_decompose ✅ 5s 4s +25%
poly_pointwise_montgomery ✅ 5s 3s +67%
polyveck_pointwise_poly_montgomery ✅ 5s 7s -29%
polyveck_unpack_eta ✅ 5s 3s +67%
polyvecl_uniform_gamma1 ✅ 5s 4s +25%
polyvecl_uniform_gamma1_serial ✅ 5s 6s -17%
polyz_unpack_native ✅ 5s 3s +67%
reduce32 ✅ 5s 3s +67%
sign_signature_extmu ✅ 5s 5s +0%
keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid ✅ 4s 1s +300%
make_hint ✅ 4s 2s +100%
mld_ct_abs_i32 ✅ 4s 2s +100%
mld_ct_cmask_neg_i32 ✅ 4s 1s +300%
mld_prepare_domain_separation_prefix ✅ 4s 2s +100%
montgomery_reduce ✅ 4s 2s +100%
ntt_native_x86_64 ✅ 4s 4s +0%
pointwise_native_aarch64 ✅ 4s 5s -20%
poly_caddq_native ✅ 4s 4s +0%
poly_chknorm_native_aarch64 ✅ 4s 5s -20%
poly_ntt_native ✅ 4s 2s +100%
poly_reduce ✅ 4s 3s +33%
poly_uniform_eta ✅ 4s 6s -33%
poly_uniform_gamma1 ✅ 4s 2s +100%
polyeta_pack ✅ 4s 3s +33%
polyt0_pack ✅ 4s 4s +0%
polyveck_pack_eta ✅ 4s 3s +33%
polyvecl_chknorm ✅ 4s 5s -20%
polyvecl_pack_eta ✅ 4s 4s +0%
polyvecl_pointwise_acc_montgomery ✅ 4s 3s +33%
power2round ✅ 4s 5s -20%
shake128x4_absorb_once ✅ 4s 3s +33%
shake128x4_squeezeblocks ✅ 4s 2s +100%
sign_signature_pre_hash_internal ✅ 4s 5s -20%
sign_signature_pre_hash_shake256 ✅ 4s 3s +33%
sign_verify_extmu ✅ 4s 5s -20%
unpack_hints ✅ 4s 6s -33%
unpack_pk ✅ 4s 4s +0%
unpack_sig ✅ 4s 5s -20%
caddq ✅ 3s 4s -25%
keccak_f1600_x1_native_aarch64 ✅ 3s 3s +0%
keccak_f1600_x1_native_aarch64_v84a ✅ 3s 5s -40%
keccak_squeeze ✅ 3s 4s -25%
keccakf1600_extract_bytes (big endian) ✅ 3s 1s +200%
keccakf1600x4_xor_bytes ✅ 3s 1s +200%
mld_ct_cmask_nonzero_u32 ✅ 3s 4s -25%
pack_pk ✅ 3s 2s +50%
pack_sig_h_poly ✅ 3s 2s +50%
pack_sk_s1 ✅ 3s 4s -25%
pointwise_native_x86_64 ✅ 3s 4s -25%
poly_caddq ✅ 3s 2s +50%
poly_chknorm_native ✅ 3s 3s +0%
poly_invntt_tomont_native ✅ 3s 3s +0%
poly_make_hint ✅ 3s 3s +0%
poly_ntt ✅ 3s 3s +0%
poly_ntt_c ✅ 3s 3s +0%
poly_pointwise_montgomery_native ✅ 3s 4s -25%
poly_sub ✅ 3s 3s +0%
poly_uniform ✅ 3s 4s -25%
poly_use_hint ✅ 3s 3s +0%
poly_use_hint_c ✅ 3s 2s +50%
poly_use_hint_native ✅ 3s 3s +0%
polyt1_unpack ✅ 3s 3s +0%
polyveck_pack_t0 ✅ 3s 3s +0%
polyveck_pack_w1 ✅ 3s 3s +0%
polyvecl_unpack_z ✅ 3s 3s +0%
polyw1_pack ✅ 3s 2s +50%
polyz_unpack ✅ 3s 4s -25%
rej_eta ✅ 3s 3s +0%
shake128_absorb ✅ 3s 2s +50%
shake128_squeeze ✅ 3s 4s -25%
shake256 ✅ 3s 2s +50%
shake256_absorb ✅ 3s 2s +50%
shake256_finalize ✅ 3s 1s +200%
shake256x4_squeezeblocks ✅ 3s 5s -40%
sign_keypair ✅ 3s 2s +50%
sign_open ✅ 3s 5s -40%
sign_signature ✅ 3s 4s -25%
sign_verify_pre_hash_internal ✅ 3s 4s -25%
sys_check_capability ✅ 3s 2s +50%
decompose ✅ 2s 4s -50%
fqscale ✅ 2s 2s +0%
intt_native_x86_64 ✅ 2s 2s +0%
keccak_f1600_x4_native_aarch64_v84a ✅ 2s 3s -33%
keccak_finalize ✅ 2s 3s -33%
keccak_init ✅ 2s 4s -50%
keccakf1600_xor_bytes ✅ 2s 4s -50%
keccakf1600_xor_bytes (big endian) ✅ 2s 3s -33%
mld_ct_cmask_nonzero_u8 ✅ 2s 1s +100%
mld_ct_get_optblocker_u32 ✅ 2s 4s -50%
mld_ct_get_optblocker_u8 ✅ 2s 2s +0%
mld_keccakf1600_extract_bytes ✅ 2s 1s +100%
mld_value_barrier_i64 ✅ 2s 2s +0%
mld_value_barrier_u32 ✅ 2s 3s -33%
ntt_native_aarch64 ✅ 2s 3s -33%
pack_sig_c ✅ 2s 2s +0%
pack_sig_z ✅ 2s 2s +0%
pack_sk_rho_key_tr_s2_t0 ✅ 2s 3s -33%
poly_chknorm ✅ 2s 2s +0%
poly_decompose_native ✅ 2s 4s -50%
polyveck_unpack_t0 ✅ 2s 2s +0%
polyvecl_permute_bitrev_to_custom ✅ 2s 3s -33%
polyvecl_pointwise_acc_montgomery_native ✅ 2s 4s -50%
polyz_pack ✅ 2s 3s -33%
shake128_init ✅ 2s 2s +0%
shake128_release ✅ 2s 4s -50%
shake256_init ✅ 2s 3s -33%
shake256_release ✅ 2s 2s +0%
shake256_squeeze ✅ 2s 2s +0%
shake256x4_absorb_once ✅ 2s 4s -50%
use_hint ✅ 2s 3s -33%
keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid ✅ 1s 4s -75%
keccakf1600x4_extract_bytes ✅ 1s 2s -50%
keccakf1600x4_permute ✅ 1s 2s -50%
mld_ct_get_optblocker_i64 ✅ 1s 4s -75%
mld_ct_sel_int32 ✅ 1s 3s -67%
mld_value_barrier_u8 ✅ 1s 3s -67%
poly_shiftl ✅ 1s 5s -80%
polyt1_pack ✅ 1s 4s -75%
shake128_finalize ✅ 1s 3s -67%

@flynd
Copy link
Copy Markdown
Contributor Author

flynd commented Apr 17, 2026

@flynd I think adding array sizes solves the issue.

That's great. I have updated my original commit to add the missing signed-off footer so hopefully CI can pass now.

flynd and others added 3 commits April 17, 2026 16:18
@flynd @hanno-becker
native: Allow tables to be declared static
dbcea89 
When building single-CU builds and MLD_CONFIG_INTERNAL_API_QUALIFIER is
set to make internal API static, also make the data tables static.
This makes the data symbols local instead of global, hiding them from
the scope of the rest of the application.

Signed-off-by: Anders Sonmark <Anders.Sonmark@axis.com>
@hanno-becker
Add explicit length to constant arrays fwd declarations
e55d7ac 
Signed-off-by: Hanno Becker <beckphan@amazon.co.uk>
@hanno-becker
autogen: Add emit_c_array helper, make all array sizes explicit
24ba7b9 
Refactor scripts/autogen to use an emit_c_array() helper for all
generated C array definitions. The helper always emits explicit
array sizes (e.g., mld_rej_uniform_table[256] instead of
mld_rej_uniform_table[]).

The length is manually added to mld_keccakf1600_round_constants,
which is not autogenerated by autogen.

Signed-off-by: Hanno Becker <beckphan@amazon.co.uk>
@mkannwischer mkannwischer merged commit 8544cd7 into main Apr 17, 2026
799 of 800 checks passed
@mkannwischer mkannwischer deleted the static-tables branch April 17, 2026 17:24
@hanno-becker hanno-becker mentioned this pull request Apr 18, 2026
Closed
1 task
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@hanno-becker hanno-becker hanno-becker approved these changes

Assignees

No one assigned

Labels

needs-mlkem-native-port

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@flynd @hanno-becker @oqs-bot @mkannwischer