add Cortex-M33 MPS3-AN524 platform support by jricardini · Pull Request #950 · pq-code-package/mldsa-native

jricardini · 2026-02-05T14:06:12Z

Add bare-metal platform support for ARM Cortex-M33 on MPS3-AN524 FPGA. Works on both QEMU (qemu-system-arm -M mps3-an524) and real hardware.

CMSIS patches for AN524 board configuration
UART driver at 0x41303000 (32 MHz, 115200 baud)
Memory: 512KB Flash @ 0x10000000, 128KB SRAM @ 0x20000000
96KB stack for ML-DSA-87 support
Reduced RAM mode enabled by default
CI integration for automated testing

oqs-bot · 2026-02-05T14:47:18Z

CBMC Results (ML-DSA-87)

Full Results (174 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2450s	2207s	+11.0%
`mld_attempt_signature_generation`	✅	256s	231s	+11%
`sign_verify_internal`	✅	235s	213s	+10%
`polyvecl_pointwise_acc_montgomery_c`	✅	197s	162s	+22%
`poly_pointwise_montgomery_c`	✅	151s	130s	+16%
`polyvec_matrix_expand`	✅	148s	128s	+16%
`rej_uniform_native`	✅	138s	120s	+15%
`polyvec_matrix_expand_serial`	✅	114s	101s	+13%
`mld_ct_memcmp`	✅	87s	78s	+12%
`mld_invntt_layer`	✅	74s	71s	+4%
`mld_ntt_layer`	✅	57s	49s	+16%
`keccak_squeezeblocks_x4`	✅	47s	44s	+7%
`sign_signature_internal`	✅	42s	40s	+5%
`mld_compute_t0_t1_tr_from_sk_components`	✅	27s	25s	+8%
`polymat_permute_bitrev_to_custom`	✅	25s	24s	+4%
`fqmul`	✅	24s	19s	+26%
`rej_uniform`	✅	24s	22s	+9%
`rej_uniform_c`	✅	20s	18s	+11%
`poly_chknorm_c`	✅	19s	17s	+12%
`poly_uniform_eta_4x`	✅	18s	16s	+12%
`keccakf1600x4_permute_native`	✅	15s	13s	+15%
`poly_uniform_4x`	✅	15s	15s	+0%
`polyvec_matrix_pointwise_montgomery`	✅	15s	14s	+7%
`polyveck_add`	✅	15s	14s	+7%
`polyt0_unpack`	✅	14s	14s	+0%
`polyveck_power2round`	✅	14s	15s	-7%
`keccak_absorb_once_x4`	✅	13s	12s	+8%
`polyeta_unpack`	✅	13s	13s	+0%
`sign`	✅	13s	8s	+62%
`poly_invntt_tomont_c`	✅	12s	8s	+50%
`mld_ntt_butterfly_block`	✅	11s	11s	+0%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	11s	9s	+22%
`polyveck_caddq`	✅	11s	8s	+38%
`polyveck_invntt_tomont`	✅	11s	9s	+22%
`polyveck_chknorm`	✅	10s	6s	+67%
`polyveck_reduce`	✅	10s	10s	+0%
`polyveck_shiftl`	✅	10s	7s	+43%
`poly_decompose_c`	✅	9s	7s	+29%
`sign_pk_from_sk`	✅	9s	9s	+0%
`keccakf1600_permute`	✅	8s	9s	-11%
`keccakf1600_permute_native`	✅	8s	9s	-11%
`mld_check_pct`	✅	8s	7s	+14%
`mld_sample_s1_s2`	✅	8s	8s	+0%
`polyveck_pointwise_poly_montgomery`	✅	8s	6s	+33%
`sign_keypair_internal`	✅	8s	5s	+60%
`polyveck_sub`	✅	7s	7s	+0%
`polyveck_use_hint`	✅	7s	6s	+17%
`polyvecl_ntt`	✅	7s	9s	-22%
`polyvecl_pointwise_acc_montgomery`	✅	7s	2s	+250%
`keccak_absorb`	✅	6s	4s	+50%
`keccak_squeeze`	✅	6s	2s	+200%
`mld_keccakf1600_extract_bytes`	✅	6s	2s	+200%
`mld_sample_s1_s2_serial`	✅	6s	6s	+0%
`poly_caddq`	✅	6s	3s	+100%
`poly_power2round`	✅	6s	5s	+20%
`poly_uniform_gamma1_4x`	✅	6s	5s	+20%
`polyt0_pack`	✅	6s	4s	+50%
`polyveck_decompose`	✅	6s	9s	-33%
`polyveck_ntt`	✅	6s	7s	-14%
`sign_signature`	✅	6s	4s	+50%
`sign_signature_pre_hash_shake256`	✅	6s	7s	-14%
`sign_verify_pre_hash_internal`	✅	6s	3s	+100%
`keccakf1600x4_xor_bytes`	✅	5s	3s	+67%
`mld_compute_pack_z`	✅	5s	6s	-17%
`mld_ct_get_optblocker_u32`	✅	5s	2s	+150%
`mld_h`	✅	5s	6s	-17%
`montgomery_reduce`	✅	5s	4s	+25%
`poly_challenge`	✅	5s	4s	+25%
`poly_make_hint`	✅	5s	3s	+67%
`polyvecl_chknorm`	✅	5s	4s	+25%
`polyvecl_uniform_gamma1_serial`	✅	5s	3s	+67%
`polyz_unpack_c`	✅	5s	6s	-17%
`rej_eta_native`	✅	5s	3s	+67%
`sign_signature_extmu`	✅	5s	5s	+0%
`sign_signature_pre_hash_internal`	✅	5s	4s	+25%
`unpack_hints`	✅	5s	5s	+0%
`unpack_sk`	✅	5s	5s	+0%
`keccak_finalize`	✅	4s	5s	-20%
`mld_ct_cmask_neg_i32`	✅	4s	2s	+100%
`mld_prepare_domain_separation_prefix`	✅	4s	5s	-20%
`mld_value_barrier_u32`	✅	4s	3s	+33%
`pack_sk`	✅	4s	2s	+100%
`poly_caddq_native`	✅	4s	2s	+100%
`poly_decompose_native`	✅	4s	4s	+0%
`poly_ntt`	✅	4s	2s	+100%
`poly_ntt_native`	✅	4s	3s	+33%
`poly_shiftl`	✅	4s	2s	+100%
`poly_uniform`	✅	4s	3s	+33%
`poly_uniform_eta`	✅	4s	4s	+0%
`poly_uniform_gamma1`	✅	4s	2s	+100%
`polyveck_pack_w1`	✅	4s	1s	+300%
`polyveck_unpack_eta`	✅	4s	7s	-43%
`polyveck_unpack_t0`	✅	4s	4s	+0%
`polyvecl_pack_eta`	✅	4s	2s	+100%
`polyvecl_pointwise_acc_montgomery_native`	✅	4s	4s	+0%
`polyvecl_uniform_gamma1`	✅	4s	4s	+0%
`polyvecl_unpack_eta`	✅	4s	4s	+0%
`polyvecl_unpack_z`	✅	4s	6s	-33%
`polyz_pack`	✅	4s	3s	+33%
`shake256_squeeze`	✅	4s	2s	+100%
`shake256x4_absorb_once`	✅	4s	2s	+100%
`sign_keypair`	✅	4s	3s	+33%
`sign_verify`	✅	4s	5s	-20%
`sys_check_capability`	✅	4s	3s	+33%
`unpack_pk`	✅	4s	3s	+33%
`decompose`	✅	3s	4s	-25%
`keccak_init`	✅	3s	2s	+50%
`keccakf1600_extract_bytes (big endian)`	✅	3s	3s	+0%
`keccakf1600_xor_bytes`	✅	3s	2s	+50%
`keccakf1600x4_permute`	✅	3s	3s	+0%
`make_hint`	✅	3s	3s	+0%
`mld_ct_cmask_nonzero_u8`	✅	3s	4s	-25%
`mld_ct_get_optblocker_u8`	✅	3s	2s	+50%
`mld_value_barrier_u8`	✅	3s	2s	+50%
`pack_pk`	✅	3s	2s	+50%
`pack_sig_c_h`	✅	3s	2s	+50%
`poly_add`	✅	3s	5s	-40%
`poly_caddq_native_aarch64`	✅	3s	4s	-25%
`poly_invntt_tomont`	✅	3s	5s	-40%
`poly_invntt_tomont_native`	✅	3s	2s	+50%
`poly_ntt_c`	✅	3s	2s	+50%
`poly_pointwise_montgomery_native`	✅	3s	2s	+50%
`poly_reduce`	✅	3s	3s	+0%
`poly_sub`	✅	3s	4s	-25%
`poly_use_hint_c`	✅	3s	3s	+0%
`polyeta_pack`	✅	3s	4s	-25%
`polyveck_make_hint`	✅	3s	5s	-40%
`polyveck_pack_eta`	✅	3s	2s	+50%
`polyw1_pack`	✅	3s	3s	+0%
`polyz_unpack`	✅	3s	4s	-25%
`polyz_unpack_native`	✅	3s	2s	+50%
`reduce32`	✅	3s	2s	+50%
`rej_eta`	✅	3s	2s	+50%
`rej_eta_c`	✅	3s	5s	-40%
`shake128_absorb`	✅	3s	2s	+50%
`shake128_finalize`	✅	3s	3s	+0%
`shake128_init`	✅	3s	5s	-40%
`shake128_release`	✅	3s	2s	+50%
`shake128_squeeze`	✅	3s	2s	+50%
`shake128x4_absorb_once`	✅	3s	1s	+200%
`shake256_init`	✅	3s	2s	+50%
`sign_open`	✅	3s	4s	-25%
`sign_verify_pre_hash_shake256`	✅	3s	4s	-25%
`use_hint`	✅	3s	5s	-40%
`caddq`	✅	2s	4s	-50%
`fqscale`	✅	2s	3s	-33%
`keccakf1600_xor_bytes (big endian)`	✅	2s	3s	-33%
`keccakf1600x4_extract_bytes`	✅	2s	5s	-60%
`mld_ct_abs_i32`	✅	2s	2s	+0%
`mld_ct_cmask_nonzero_u32`	✅	2s	2s	+0%
`mld_ct_get_optblocker_i64`	✅	2s	1s	+100%
`mld_value_barrier_i64`	✅	2s	3s	-33%
`ntt_native_x86_64`	✅	2s	3s	-33%
`pack_sig_z`	✅	2s	3s	-33%
`poly_caddq_c`	✅	2s	5s	-60%
`poly_chknorm`	✅	2s	5s	-60%
`poly_chknorm_native`	✅	2s	2s	+0%
`poly_decompose`	✅	2s	2s	+0%
`poly_pointwise_montgomery`	✅	2s	2s	+0%
`poly_use_hint`	✅	2s	2s	+0%
`polyt1_pack`	✅	2s	2s	+0%
`polyt1_unpack`	✅	2s	5s	-60%
`polyveck_pack_t0`	✅	2s	4s	-50%
`polyvecl_permute_bitrev_to_custom`	✅	2s	4s	-50%
`power2round`	✅	2s	3s	-33%
`shake128x4_squeezeblocks`	✅	2s	4s	-50%
`shake256`	✅	2s	4s	-50%
`shake256_absorb`	✅	2s	1s	+100%
`shake256_release`	✅	2s	5s	-60%
`shake256x4_squeezeblocks`	✅	2s	2s	+0%
`sign_verify_extmu`	✅	2s	4s	-50%
`mld_ct_sel_int32`	✅	1s	1s	+0%
`poly_use_hint_native`	✅	1s	3s	-67%
`shake256_finalize`	✅	1s	2s	-50%
`unpack_sig`	✅	1s	4s	-75%

oqs-bot · 2026-02-05T14:48:43Z

CBMC Results (ML-DSA-44)

Full Results (174 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2190s	2088s	+4.9%
`sign_verify_internal`	✅	264s	249s	+6%
`polyvecl_pointwise_acc_montgomery_c`	✅	224s	208s	+8%
`mld_attempt_signature_generation`	✅	223s	213s	+5%
`poly_pointwise_montgomery_c`	✅	159s	140s	+14%
`rej_uniform_native`	✅	132s	130s	+2%
`mld_invntt_layer`	✅	124s	116s	+7%
`mld_ct_memcmp`	✅	86s	84s	+2%
`mld_ntt_layer`	✅	47s	45s	+4%
`sign_signature_internal`	✅	46s	44s	+5%
`keccak_squeezeblocks_x4`	✅	45s	43s	+5%
`polyvec_matrix_expand`	✅	30s	29s	+3%
`rej_uniform`	✅	21s	20s	+5%
`rej_uniform_c`	✅	20s	19s	+5%
`fqmul`	✅	19s	19s	+0%
`polymat_permute_bitrev_to_custom`	✅	18s	15s	+20%
`poly_uniform_eta_4x`	✅	17s	15s	+13%
`mld_compute_t0_t1_tr_from_sk_components`	✅	16s	12s	+33%
`poly_chknorm_c`	✅	16s	17s	-6%
`polyeta_unpack`	✅	16s	12s	+33%
`poly_uniform_4x`	✅	15s	15s	+0%
`polyt0_unpack`	✅	15s	16s	-6%
`keccak_absorb_once_x4`	✅	13s	13s	+0%
`keccakf1600x4_permute_native`	✅	12s	12s	+0%
`mld_ntt_butterfly_block`	✅	12s	11s	+9%
`polyz_unpack_c`	✅	12s	11s	+9%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	9s	12s	-25%
`polyveck_decompose`	✅	9s	7s	+29%
`keccakf1600_permute`	✅	8s	7s	+14%
`keccakf1600_permute_native`	✅	8s	9s	-11%
`mld_check_pct`	✅	8s	10s	-20%
`poly_chknorm_native`	✅	8s	4s	+100%
`poly_invntt_tomont_c`	✅	8s	7s	+14%
`polyveck_add`	✅	8s	8s	+0%
`sign_pk_from_sk`	✅	8s	5s	+60%
`polyvecl_chknorm`	✅	7s	4s	+75%
`sign`	✅	7s	6s	+17%
`keccak_absorb`	✅	6s	6s	+0%
`pack_sk`	✅	6s	3s	+100%
`poly_uniform_eta`	✅	6s	5s	+20%
`polyvec_matrix_pointwise_montgomery`	✅	6s	8s	-25%
`polyveck_power2round`	✅	6s	6s	+0%
`polyveck_unpack_eta`	✅	6s	2s	+200%
`polyveck_use_hint`	✅	6s	3s	+100%
`polyvecl_ntt`	✅	6s	4s	+50%
`polyvecl_pack_eta`	✅	6s	4s	+50%
`polyvecl_unpack_z`	✅	6s	6s	+0%
`rej_eta`	✅	6s	5s	+20%
`sign_signature_pre_hash_shake256`	✅	6s	5s	+20%
`sign_verify_extmu`	✅	6s	6s	+0%
`caddq`	✅	5s	5s	+0%
`decompose`	✅	5s	4s	+25%
`keccak_squeeze`	✅	5s	2s	+150%
`pack_sig_c_h`	✅	5s	3s	+67%
`poly_add`	✅	5s	5s	+0%
`poly_decompose_c`	✅	5s	4s	+25%
`poly_sub`	✅	5s	2s	+150%
`poly_uniform_gamma1_4x`	✅	5s	4s	+25%
`polyvec_matrix_expand_serial`	✅	5s	6s	-17%
`polyveck_caddq`	✅	5s	6s	-17%
`polyveck_chknorm`	✅	5s	5s	+0%
`polyveck_invntt_tomont`	✅	5s	5s	+0%
`polyveck_ntt`	✅	5s	6s	-17%
`polyveck_pack_t0`	✅	5s	2s	+150%
`polyveck_reduce`	✅	5s	4s	+25%
`polyveck_shiftl`	✅	5s	5s	+0%
`polyvecl_permute_bitrev_to_custom`	✅	5s	3s	+67%
`rej_eta_native`	✅	5s	5s	+0%
`sign_keypair`	✅	5s	4s	+25%
`sign_keypair_internal`	✅	5s	8s	-38%
`sign_signature_pre_hash_internal`	✅	5s	3s	+67%
`sign_verify`	✅	5s	3s	+67%
`sign_verify_pre_hash_shake256`	✅	5s	6s	-17%
`unpack_pk`	✅	5s	4s	+25%
`make_hint`	✅	4s	3s	+33%
`mld_compute_pack_z`	✅	4s	4s	+0%
`mld_ct_sel_int32`	✅	4s	3s	+33%
`mld_prepare_domain_separation_prefix`	✅	4s	5s	-20%
`mld_sample_s1_s2_serial`	✅	4s	3s	+33%
`montgomery_reduce`	✅	4s	2s	+100%
`ntt_native_x86_64`	✅	4s	2s	+100%
`poly_caddq_c`	✅	4s	2s	+100%
`poly_challenge`	✅	4s	4s	+0%
`poly_chknorm`	✅	4s	2s	+100%
`poly_decompose_native`	✅	4s	5s	-20%
`poly_ntt_native`	✅	4s	4s	+0%
`poly_reduce`	✅	4s	2s	+100%
`poly_use_hint_c`	✅	4s	5s	-20%
`poly_use_hint_native`	✅	4s	4s	+0%
`polyeta_pack`	✅	4s	5s	-20%
`polyt0_pack`	✅	4s	3s	+33%
`polyveck_pack_eta`	✅	4s	4s	+0%
`polyveck_pointwise_poly_montgomery`	✅	4s	5s	-20%
`polyveck_sub`	✅	4s	5s	-20%
`polyveck_unpack_t0`	✅	4s	3s	+33%
`polyvecl_pointwise_acc_montgomery`	✅	4s	3s	+33%
`polyvecl_pointwise_acc_montgomery_native`	✅	4s	2s	+100%
`polyvecl_uniform_gamma1_serial`	✅	4s	4s	+0%
`polyw1_pack`	✅	4s	4s	+0%
`polyz_unpack_native`	✅	4s	3s	+33%
`reduce32`	✅	4s	2s	+100%
`rej_eta_c`	✅	4s	3s	+33%
`shake256_finalize`	✅	4s	5s	-20%
`shake256x4_absorb_once`	✅	4s	3s	+33%
`shake256x4_squeezeblocks`	✅	4s	5s	-20%
`sign_verify_pre_hash_internal`	✅	4s	4s	+0%
`unpack_sk`	✅	4s	3s	+33%
`fqscale`	✅	3s	2s	+50%
`keccakf1600_extract_bytes (big endian)`	✅	3s	4s	-25%
`mld_ct_cmask_nonzero_u32`	✅	3s	3s	+0%
`mld_ct_get_optblocker_i64`	✅	3s	2s	+50%
`mld_ct_get_optblocker_u32`	✅	3s	4s	-25%
`mld_ct_get_optblocker_u8`	✅	3s	2s	+50%
`mld_h`	✅	3s	5s	-40%
`mld_sample_s1_s2`	✅	3s	6s	-50%
`pack_sig_z`	✅	3s	5s	-40%
`poly_caddq_native`	✅	3s	2s	+50%
`poly_caddq_native_aarch64`	✅	3s	3s	+0%
`poly_invntt_tomont_native`	✅	3s	4s	-25%
`poly_make_hint`	✅	3s	5s	-40%
`poly_ntt`	✅	3s	4s	-25%
`poly_ntt_c`	✅	3s	2s	+50%
`poly_pointwise_montgomery`	✅	3s	4s	-25%
`poly_power2round`	✅	3s	2s	+50%
`poly_shiftl`	✅	3s	4s	-25%
`poly_uniform`	✅	3s	5s	-40%
`poly_uniform_gamma1`	✅	3s	3s	+0%
`poly_use_hint`	✅	3s	1s	+200%
`polyveck_pack_w1`	✅	3s	3s	+0%
`polyvecl_uniform_gamma1`	✅	3s	3s	+0%
`polyvecl_unpack_eta`	✅	3s	2s	+50%
`polyz_pack`	✅	3s	3s	+0%
`polyz_unpack`	✅	3s	2s	+50%
`shake128_init`	✅	3s	2s	+50%
`shake128_squeeze`	✅	3s	2s	+50%
`shake128x4_absorb_once`	✅	3s	4s	-25%
`shake128x4_squeezeblocks`	✅	3s	3s	+0%
`shake256_init`	✅	3s	2s	+50%
`shake256_release`	✅	3s	2s	+50%
`shake256_squeeze`	✅	3s	4s	-25%
`sign_signature`	✅	3s	3s	+0%
`sign_signature_extmu`	✅	3s	6s	-50%
`sys_check_capability`	✅	3s	3s	+0%
`unpack_hints`	✅	3s	6s	-50%
`unpack_sig`	✅	3s	3s	+0%
`keccak_init`	✅	2s	2s	+0%
`keccakf1600_xor_bytes (big endian)`	✅	2s	2s	+0%
`keccakf1600x4_extract_bytes`	✅	2s	2s	+0%
`keccakf1600x4_permute`	✅	2s	2s	+0%
`keccakf1600x4_xor_bytes`	✅	2s	2s	+0%
`mld_ct_cmask_neg_i32`	✅	2s	3s	-33%
`mld_ct_cmask_nonzero_u8`	✅	2s	4s	-50%
`mld_keccakf1600_extract_bytes`	✅	2s	3s	-33%
`mld_value_barrier_i64`	✅	2s	3s	-33%
`mld_value_barrier_u32`	✅	2s	2s	+0%
`mld_value_barrier_u8`	✅	2s	3s	-33%
`pack_pk`	✅	2s	4s	-50%
`poly_caddq`	✅	2s	4s	-50%
`poly_decompose`	✅	2s	3s	-33%
`poly_invntt_tomont`	✅	2s	5s	-60%
`poly_pointwise_montgomery_native`	✅	2s	4s	-50%
`polyt1_pack`	✅	2s	2s	+0%
`polyt1_unpack`	✅	2s	5s	-60%
`polyveck_make_hint`	✅	2s	4s	-50%
`power2round`	✅	2s	2s	+0%
`shake128_absorb`	✅	2s	3s	-33%
`shake128_release`	✅	2s	2s	+0%
`shake256`	✅	2s	2s	+0%
`shake256_absorb`	✅	2s	2s	+0%
`sign_open`	✅	2s	8s	-75%
`use_hint`	✅	2s	3s	-33%
`keccak_finalize`	✅	1s	3s	-67%
`keccakf1600_xor_bytes`	✅	1s	4s	-75%
`mld_ct_abs_i32`	✅	1s	3s	-67%
`shake128_finalize`	✅	1s	2s	-50%

oqs-bot · 2026-02-05T14:49:47Z

CBMC Results (ML-DSA-65)

Full Results (174 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2253s	2319s	-2.8%
`sign_verify_internal`	✅	337s	357s	-6%
`polyvecl_pointwise_acc_montgomery_c`	✅	219s	229s	-4%
`mld_attempt_signature_generation`	✅	196s	202s	-3%
`poly_pointwise_montgomery_c`	✅	131s	136s	-4%
`rej_uniform_native`	✅	121s	130s	-7%
`polyvec_matrix_expand`	✅	91s	93s	-2%
`mld_ct_memcmp`	✅	79s	79s	+0%
`mld_invntt_layer`	✅	69s	72s	-4%
`polyvec_matrix_expand_serial`	✅	64s	63s	+2%
`mld_ntt_layer`	✅	52s	55s	-5%
`keccak_squeezeblocks_x4`	✅	43s	44s	-2%
`sign_signature_internal`	✅	42s	44s	-5%
`mld_compute_t0_t1_tr_from_sk_components`	✅	23s	24s	-4%
`rej_uniform`	✅	19s	20s	-5%
`fqmul`	✅	18s	19s	-5%
`poly_chknorm_c`	✅	17s	18s	-6%
`polymat_permute_bitrev_to_custom`	✅	17s	20s	-15%
`poly_uniform_eta_4x`	✅	16s	17s	-6%
`polyveck_decompose`	✅	16s	16s	+0%
`rej_uniform_c`	✅	16s	17s	-6%
`poly_uniform_4x`	✅	14s	18s	-22%
`polyt0_unpack`	✅	14s	15s	-7%
`keccakf1600x4_permute_native`	✅	13s	13s	+0%
`polyvec_matrix_pointwise_montgomery`	✅	13s	14s	-7%
`keccak_absorb_once_x4`	✅	11s	12s	-8%
`polyveck_invntt_tomont`	✅	11s	9s	+22%
`polyveck_power2round`	✅	11s	10s	+10%
`mld_ntt_butterfly_block`	✅	10s	12s	-17%
`polyveck_add`	✅	10s	10s	+0%
`polyveck_ntt`	✅	10s	9s	+11%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	9s	10s	-10%
`polyvecl_ntt`	✅	9s	8s	+12%
`sign`	✅	9s	9s	+0%
`poly_decompose_c`	✅	8s	8s	+0%
`poly_invntt_tomont_c`	✅	8s	9s	-11%
`polyveck_caddq`	✅	8s	6s	+33%
`unpack_sk`	✅	8s	3s	+167%
`keccakf1600_permute`	✅	7s	7s	+0%
`keccakf1600_permute_native`	✅	7s	7s	+0%
`mld_check_pct`	✅	7s	5s	+40%
`mld_compute_pack_z`	✅	7s	6s	+17%
`poly_use_hint_c`	✅	7s	4s	+75%
`polyeta_unpack`	✅	7s	5s	+40%
`polyveck_sub`	✅	7s	9s	-22%
`polyveck_use_hint`	✅	7s	7s	+0%
`sign_pk_from_sk`	✅	7s	8s	-12%
`keccak_absorb`	✅	6s	5s	+20%
`mld_ct_cmask_nonzero_u8`	✅	6s	2s	+200%
`poly_make_hint`	✅	6s	4s	+50%
`polyveck_pack_eta`	✅	6s	3s	+100%
`polyveck_reduce`	✅	6s	6s	+0%
`polyveck_shiftl`	✅	6s	12s	-50%
`sign_verify_pre_hash_internal`	✅	6s	2s	+200%
`mld_sample_s1_s2`	✅	5s	5s	+0%
`poly_chknorm`	✅	5s	3s	+67%
`poly_power2round`	✅	5s	6s	-17%
`poly_uniform_eta`	✅	5s	4s	+25%
`poly_use_hint`	✅	5s	2s	+150%
`polyt0_pack`	✅	5s	6s	-17%
`polyveck_make_hint`	✅	5s	3s	+67%
`polyveck_pack_t0`	✅	5s	2s	+150%
`polyveck_pointwise_poly_montgomery`	✅	5s	4s	+25%
`polyvecl_chknorm`	✅	5s	6s	-17%
`polyvecl_pack_eta`	✅	5s	4s	+25%
`polyz_unpack_native`	✅	5s	2s	+150%
`rej_eta_native`	✅	5s	4s	+25%
`shake128_init`	✅	5s	2s	+150%
`shake256_squeeze`	✅	5s	2s	+150%
`sign_keypair_internal`	✅	5s	5s	+0%
`sign_open`	✅	5s	7s	-29%
`sign_signature_pre_hash_internal`	✅	5s	7s	-29%
`sign_verify`	✅	5s	3s	+67%
`unpack_hints`	✅	5s	6s	-17%
`keccakf1600x4_extract_bytes`	✅	4s	2s	+100%
`keccakf1600x4_permute`	✅	4s	4s	+0%
`mld_ct_abs_i32`	✅	4s	2s	+100%
`mld_ct_get_optblocker_u32`	✅	4s	1s	+300%
`mld_h`	✅	4s	2s	+100%
`mld_prepare_domain_separation_prefix`	✅	4s	3s	+33%
`mld_sample_s1_s2_serial`	✅	4s	5s	-20%
`ntt_native_x86_64`	✅	4s	3s	+33%
`pack_pk`	✅	4s	6s	-33%
`pack_sig_c_h`	✅	4s	1s	+300%
`poly_invntt_tomont`	✅	4s	2s	+100%
`poly_uniform`	✅	4s	4s	+0%
`poly_uniform_gamma1`	✅	4s	3s	+33%
`polyt1_pack`	✅	4s	3s	+33%
`polyt1_unpack`	✅	4s	3s	+33%
`polyveck_chknorm`	✅	4s	6s	-33%
`polyveck_unpack_eta`	✅	4s	4s	+0%
`polyvecl_pointwise_acc_montgomery_native`	✅	4s	3s	+33%
`polyvecl_uniform_gamma1`	✅	4s	5s	-20%
`polyvecl_unpack_eta`	✅	4s	4s	+0%
`polyz_unpack_c`	✅	4s	5s	-20%
`reduce32`	✅	4s	3s	+33%
`rej_eta_c`	✅	4s	4s	+0%
`sign_signature`	✅	4s	5s	-20%
`sign_signature_extmu`	✅	4s	5s	-20%
`sign_verify_extmu`	✅	4s	3s	+33%
`sign_verify_pre_hash_shake256`	✅	4s	2s	+100%
`unpack_sig`	✅	4s	3s	+33%
`caddq`	✅	3s	3s	+0%
`fqscale`	✅	3s	3s	+0%
`keccak_finalize`	✅	3s	3s	+0%
`keccakf1600_extract_bytes (big endian)`	✅	3s	2s	+50%
`keccakf1600_xor_bytes`	✅	3s	3s	+0%
`keccakf1600x4_xor_bytes`	✅	3s	1s	+200%
`make_hint`	✅	3s	4s	-25%
`mld_ct_sel_int32`	✅	3s	2s	+50%
`mld_value_barrier_i64`	✅	3s	1s	+200%
`mld_value_barrier_u8`	✅	3s	3s	+0%
`pack_sig_z`	✅	3s	4s	-25%
`pack_sk`	✅	3s	2s	+50%
`poly_add`	✅	3s	4s	-25%
`poly_caddq`	✅	3s	3s	+0%
`poly_caddq_native`	✅	3s	4s	-25%
`poly_caddq_native_aarch64`	✅	3s	2s	+50%
`poly_challenge`	✅	3s	3s	+0%
`poly_chknorm_native`	✅	3s	5s	-40%
`poly_ntt`	✅	3s	2s	+50%
`poly_ntt_c`	✅	3s	5s	-40%
`poly_pointwise_montgomery`	✅	3s	5s	-40%
`poly_pointwise_montgomery_native`	✅	3s	3s	+0%
`poly_shiftl`	✅	3s	4s	-25%
`poly_sub`	✅	3s	5s	-40%
`poly_uniform_gamma1_4x`	✅	3s	4s	-25%
`poly_use_hint_native`	✅	3s	3s	+0%
`polyveck_unpack_t0`	✅	3s	3s	+0%
`polyvecl_permute_bitrev_to_custom`	✅	3s	2s	+50%
`polyvecl_pointwise_acc_montgomery`	✅	3s	3s	+0%
`polyvecl_uniform_gamma1_serial`	✅	3s	4s	-25%
`polyz_unpack`	✅	3s	3s	+0%
`shake128_absorb`	✅	3s	1s	+200%
`shake256`	✅	3s	3s	+0%
`sign_keypair`	✅	3s	2s	+50%
`sys_check_capability`	✅	3s	4s	-25%
`keccak_squeeze`	✅	2s	3s	-33%
`keccakf1600_xor_bytes (big endian)`	✅	2s	3s	-33%
`mld_ct_cmask_neg_i32`	✅	2s	1s	+100%
`mld_ct_cmask_nonzero_u32`	✅	2s	3s	-33%
`mld_ct_get_optblocker_i64`	✅	2s	3s	-33%
`mld_ct_get_optblocker_u8`	✅	2s	3s	-33%
`mld_value_barrier_u32`	✅	2s	4s	-50%
`montgomery_reduce`	✅	2s	3s	-33%
`poly_caddq_c`	✅	2s	4s	-50%
`poly_decompose`	✅	2s	4s	-50%
`poly_decompose_native`	✅	2s	4s	-50%
`poly_invntt_tomont_native`	✅	2s	3s	-33%
`poly_ntt_native`	✅	2s	2s	+0%
`poly_reduce`	✅	2s	2s	+0%
`polyvecl_unpack_z`	✅	2s	3s	-33%
`polyw1_pack`	✅	2s	3s	-33%
`polyz_pack`	✅	2s	1s	+100%
`power2round`	✅	2s	4s	-50%
`rej_eta`	✅	2s	4s	-50%
`shake128_finalize`	✅	2s	1s	+100%
`shake128_squeeze`	✅	2s	2s	+0%
`shake128x4_absorb_once`	✅	2s	3s	-33%
`shake128x4_squeezeblocks`	✅	2s	2s	+0%
`shake256_absorb`	✅	2s	1s	+100%
`shake256_finalize`	✅	2s	4s	-50%
`shake256_init`	✅	2s	3s	-33%
`shake256_release`	✅	2s	2s	+0%
`shake256x4_absorb_once`	✅	2s	3s	-33%
`sign_signature_pre_hash_shake256`	✅	2s	4s	-50%
`unpack_pk`	✅	2s	4s	-50%
`use_hint`	✅	2s	3s	-33%
`decompose`	✅	1s	4s	-75%
`keccak_init`	✅	1s	1s	+0%
`mld_keccakf1600_extract_bytes`	✅	1s	3s	-67%
`polyeta_pack`	✅	1s	3s	-67%
`polyveck_pack_w1`	✅	1s	4s	-75%
`shake128_release`	✅	1s	3s	-67%
`shake256x4_squeezeblocks`	✅	1s	3s	-67%

nix/m33-an524-cmsis/default.nix

test/baremetal/platform/m33-an524/patches/cmsis5/m33-an524.ld.patch

test/baremetal/platform/m33-an524/README.md

nix/m33-an524-cmsis/default.nix

hanno-becker

This looks very close, thank you @jricardini. There's a bit more tidying up to do, see comments. I also still need to test that the benchmarking code works; will do that and report back.

Add bare-metal platform support for ARM Cortex-M33 on MPS3-AN524 FPGA. Works on both QEMU (qemu-system-arm -M mps3-an524) and real hardware. - CMSIS patches for AN524 board configuration - UART driver at 0x41303000 (32 MHz, 115200 baud) - Memory: 512KB Flash @ 0x10000000, 128KB SRAM @ 0x20000000 - 96KB stack for ML-DSA-87 support - Reduced RAM mode enabled by default - CI integration for automated testing Signed-off-by: Jefferson Ricardini <jeffricardini@gmail.com>

- Remove separate m33-an524-cmsis package and integrate CMSIS fetching into m33-an524 - Delete nix/m33-an524-cmsis/default.nix and remove package export from flake.nix - Move CMSIS_5 repository fetch directly into m33-an524/default.nix build phase - Consolidate all platform file assembly (pqmx + CMSIS) into single derivation Signed-off-by: Jefferson Ricardini <jeffricardini@gmail.com>

jricardini requested a review from a team as a code owner February 5, 2026 14:06

mkannwischer added the needs-mlkem-native-port label Feb 5, 2026