Create Nucleo N657x0-Q Platform by bremoran · Pull Request #1547 · pq-code-package/mlkem-native

bremoran · 2026-02-05T13:21:25Z

This change introduces a new bare-metal test platform for the Nucleo N657x0-Q board and integrates it into the project’s Nix devshell and test build system.

Summary

Adds a dedicated Nix devshell target for nucleo-n657x0-q.
Introduces a full bare-metal platform under test/baremetal/platform/nucleo-n657x0-q/, including build rules, execution wrapper, argument packing utility, and platform sources.
Updates HAL and test make components to integrate the new platform.

Changes

flake.nix: Add .#nucleo-n657x0-q devshell.
nix/nucleo-n657x0-q/default.nix: Define board devshell and tooling.
nix/util.nix: Utility update for new platform.
test/baremetal/platform/nucleo-n657x0-q/README.md: Platform documentation.
test/baremetal/platform/nucleo-n657x0-q/platform.mk: Makefile integration for the platform.
test/baremetal/platform/nucleo-n657x0-q/exec_wrapper.py: Execution wrapper for running tests on target.
test/baremetal/platform/nucleo-n657x0-q/make_argv_bin.py: Helper to pack argv for target.
test/baremetal/platform/nucleo-n657x0-q/src/*.c, *.h: Platform sources (cmdline, semihosting syscall, helpers).
test/hal/hal.c, test/hal/pmu_armv8.h: Minor HAL updates for compatibility.
test/mk/components.mk: Include the new platform components.

Why

Enables running and validating ML-KEM tests on the Nucleo N657x0-Q board.
Establishes a consistent devshell and build workflow for this platform.

Usage

Devshell: nix develop .#nucleo-n657x0-q
Build tests: make test EXTRA_MAKEFILE=test/baremetal/platform/nucleo-n657x0-q/platform.mk

Notes

No breaking changes expected to other platforms.
Documentation included under the platform’s README.md.

hanno-becker · 2026-03-19T05:12:44Z

@bremoran What's your plan for this?

bremoran · 2026-03-20T14:02:38Z

I am hoping to finalise this but I'm not able to spend time on it right now.

mkannwischer · 2026-03-21T05:30:09Z

I am hoping to finalise this but I'm not able to spend time on it right now.

Thanks for the update. Marking it as draft for now. Please mark it as ready when you have finished it.

bremoran · 2026-05-05T15:00:09Z

Thanks for the update. Marking it as draft for now. Please mark it as ready when you have finished it.

Done!

oqs-bot · 2026-05-05T15:41:30Z

CBMC Results (ML-KEM-512)

Full Results (191 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	1667s	1504s	+10.8%
`mlk_indcpa_keypair_derand`	✅	265s	250s	+6%
`mlk_poly_rej_uniform`	✅	188s	155s	+21%
`mlk_indcpa_enc`	✅	185s	173s	+7%
`mlk_rej_uniform_c`	✅	165s	143s	+15%
`mlk_polyvec_basemul_acc_montgomery_cached_c`	✅	62s	54s	+15%
`poly_ntt_native`	✅	50s	42s	+19%
`mlk_ntt_layer`	✅	42s	36s	+17%
`mlk_poly_reduce_native`	✅	40s	39s	+3%
`mlk_keccak_squeezeblocks_x4`	✅	34s	27s	+26%
`mlk_fqmul`	✅	20s	16s	+25%
`keccakf1600x4_permute_native_x4`	✅	17s	16s	+6%
`mlk_poly_decompress_d10_native`	✅	17s	16s	+6%
`mlk_poly_decompress_d4_native`	✅	17s	13s	+31%
`mlk_indcpa_dec`	✅	16s	14s	+14%
`mlk_poly_frommsg`	✅	14s	11s	+27%
`mlk_polyvec_add`	✅	12s	12s	+0%
`mlk_keccak_squeezeblocks`	✅	11s	8s	+38%
`mlk_poly_frombytes_native`	✅	11s	11s	+0%
`mlk_ntt_butterfly_block`	✅	10s	7s	+43%
`mlk_poly_ntt`	✅	10s	7s	+43%
`mlk_keccak_squeeze_once`	✅	8s	9s	-11%
`mlk_poly_cbd_eta2`	✅	8s	7s	+14%
`poly_frombytes_native_x86_64`	✅	8s	5s	+60%
`mlk_invntt_layer`	✅	7s	6s	+17%
`mlk_keccak_absorb_once_x4`	✅	7s	6s	+17%
`polyvec_basemul_acc_montgomery_cached_native`	✅	7s	8s	-12%
`rej_uniform_native_x86_64`	✅	7s	9s	-22%
`mlk_keccakf1600_permute_c`	✅	6s	4s	+50%
`mlk_poly_rej_uniform_x4`	✅	6s	7s	-14%
`mlk_gen_matrix_serial`	✅	5s	1s	+400%
`mlk_keccak_absorb_once`	✅	5s	4s	+25%
`mlk_poly_ntt_c`	✅	5s	4s	+25%
`mlk_polyvec_basemul_acc_montgomery_cached`	✅	5s	2s	+150%
`mlk_shake256x4`	✅	5s	4s	+25%
`nttunpack_native_x86_64`	✅	5s	3s	+67%
`poly_decompress_d10_native_x86_64`	✅	5s	8s	-38%
`poly_decompress_d4_native_x86_64`	✅	5s	5s	+0%
`polyvec_basemul_acc_montgomery_cached_k3_native_aarch64`	✅	5s	1s	+400%
`polyvec_basemul_acc_montgomery_cached_k4_native_x86_64`	✅	5s	3s	+67%
`rej_uniform_native`	✅	5s	1s	+400%
`rej_uniform_native_aarch64`	✅	5s	3s	+67%
`kem_dec`	✅	4s	3s	+33%
`mlk_barrett_reduce`	✅	4s	3s	+33%
`mlk_ct_cmask_nonzero_u16`	✅	4s	3s	+33%
`mlk_ct_memcmp`	✅	4s	2s	+100%
`mlk_matvec_mul`	✅	4s	2s	+100%
`mlk_poly_compress_d10_c`	✅	4s	3s	+33%
`mlk_poly_compress_d11_native`	✅	4s	3s	+33%
`mlk_poly_decompress_d10`	✅	4s	3s	+33%
`mlk_poly_decompress_d11`	✅	4s	2s	+100%
`mlk_poly_getnoise_eta1_4x`	✅	4s	4s	+0%
`mlk_poly_getnoise_eta1_4x_native`	✅	4s	5s	-20%
`mlk_poly_tomsg`	✅	4s	3s	+33%
`mlk_polyvec_reduce`	✅	4s	3s	+33%
`mlk_scalar_compress_d4`	✅	4s	2s	+100%
`mlk_shake128_squeezeblocks`	✅	4s	2s	+100%
`mlk_value_barrier_u32`	✅	4s	2s	+100%
`poly_compress_d4_native_x86_64`	✅	4s	2s	+100%
`poly_mulcache_compute_native_aarch64`	✅	4s	4s	+0%
`poly_tobytes_native_x86_64`	✅	4s	2s	+100%
`sys_check_capability`	✅	4s	4s	+0%
`intt_native_aarch64`	✅	3s	2s	+50%
`intt_native_x86_64`	✅	3s	3s	+0%
`kem_enc`	✅	3s	2s	+50%
`kem_enc_derand`	✅	3s	2s	+50%
`kem_keypair`	✅	3s	3s	+0%
`mlk_check_pct`	✅	3s	4s	-25%
`mlk_ct_cmask_nonzero_u8`	✅	3s	1s	+200%
`mlk_ct_cmov_zero`	✅	3s	4s	-25%
`mlk_enc_getnoise_eta1_eta2`	✅	3s	2s	+50%
`mlk_keccakf1600_permute`	✅	3s	1s	+200%
`mlk_keccakf1600_xor_bytes`	✅	3s	3s	+0%
`mlk_keypair_getnoise_eta1`	✅	3s	4s	-25%
`mlk_poly_add`	✅	3s	3s	+0%
`mlk_poly_cbd_eta1`	✅	3s	2s	+50%
`mlk_poly_compress_d10`	✅	3s	2s	+50%
`mlk_poly_compress_d10_native`	✅	3s	3s	+0%
`mlk_poly_compress_d11_c`	✅	3s	3s	+0%
`mlk_poly_compress_d4_c`	✅	3s	2s	+50%
`mlk_poly_compress_d5`	✅	3s	2s	+50%
`mlk_poly_compress_du`	✅	3s	2s	+50%
`mlk_poly_compress_dv`	✅	3s	2s	+50%
`mlk_poly_decompress_d11_c`	✅	3s	2s	+50%
`mlk_poly_decompress_d4_c`	✅	3s	2s	+50%
`mlk_poly_decompress_dv`	✅	3s	3s	+0%
`mlk_poly_invntt_tomont_c`	✅	3s	2s	+50%
`mlk_poly_mulcache_compute_c`	✅	3s	3s	+0%
`mlk_poly_tobytes`	✅	3s	2s	+50%
`mlk_poly_tobytes_native`	✅	3s	2s	+50%
`mlk_poly_tomont`	✅	3s	4s	-25%
`mlk_poly_tomont_c`	✅	3s	3s	+0%
`mlk_poly_tomont_native`	✅	3s	5s	-40%
`mlk_polyvec_invntt_tomont`	✅	3s	1s	+200%
`mlk_polyvec_permute_bitrev_to_custom_native`	✅	3s	1s	+200%
`mlk_polyvec_tomont`	✅	3s	2s	+50%
`mlk_scalar_compress_d11`	✅	3s	1s	+200%
`ntt_native_aarch64`	✅	3s	3s	+0%
`poly_compress_d5_native_x86_64`	✅	3s	2s	+50%
`poly_decompress_d11_native_x86_64`	✅	3s	1s	+200%
`poly_mulcache_compute_native_x86_64`	✅	3s	1s	+200%
`poly_tomont_native_x86_64`	✅	3s	2s	+50%
`polyvec_basemul_acc_montgomery_cached_k2_native_aarch64`	✅	3s	2s	+50%
`keccak_f1600_x1_native_aarch64`	✅	2s	1s	+100%
`keccak_f1600_x1_native_aarch64_v84a`	✅	2s	1s	+100%
`keccak_f1600_x4_native_aarch64_v84a`	✅	2s	1s	+100%
`keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid`	✅	2s	2s	+0%
`keccak_f1600_x4_native_avx2`	✅	2s	1s	+100%
`keccakf1600_permute_native`	✅	2s	3s	-33%
`keccakf1600x4_xor_bytes_native`	✅	2s	2s	+0%
`kem_check_pk`	✅	2s	3s	-33%
`kem_check_sk`	✅	2s	2s	+0%
`kem_keypair_derand`	✅	2s	2s	+0%
`mlk_ct_get_optblocker_i32`	✅	2s	3s	-33%
`mlk_ct_get_optblocker_u32`	✅	2s	2s	+0%
`mlk_ct_get_optblocker_u8`	✅	2s	2s	+0%
`mlk_ct_sel_int16`	✅	2s	1s	+100%
`mlk_ct_sel_uint8`	✅	2s	2s	+0%
`mlk_gen_matrix`	✅	2s	3s	-33%
`mlk_keccakf1600_extract_bytes`	✅	2s	3s	-33%
`mlk_keccakf1600_extract_bytes (big endian)`	✅	2s	1s	+100%
`mlk_keccakf1600x4_extract_bytes`	✅	2s	2s	+0%
`mlk_keccakf1600x4_permute`	✅	2s	1s	+100%
`mlk_keccakf1600x4_xor_bytes`	✅	2s	3s	-33%
`mlk_poly_compress_d11`	✅	2s	3s	-33%
`mlk_poly_compress_d4`	✅	2s	2s	+0%
`mlk_poly_compress_d4_native`	✅	2s	2s	+0%
`mlk_poly_compress_d5_native`	✅	2s	3s	-33%
`mlk_poly_decompress_d10_c`	✅	2s	1s	+100%
`mlk_poly_decompress_d11_native`	✅	2s	2s	+0%
`mlk_poly_decompress_d5`	✅	2s	2s	+0%
`mlk_poly_decompress_du`	✅	2s	2s	+0%
`mlk_poly_frombytes`	✅	2s	2s	+0%
`mlk_poly_frombytes_c`	✅	2s	2s	+0%
`mlk_poly_getnoise_eta1122_4x`	✅	2s	4s	-50%
`mlk_poly_getnoise_eta2`	✅	2s	2s	+0%
`mlk_poly_mulcache_compute`	✅	2s	3s	-33%
`mlk_poly_mulcache_compute_native`	✅	2s	3s	-33%
`mlk_poly_reduce`	✅	2s	2s	+0%
`mlk_poly_reduce_c`	✅	2s	4s	-50%
`mlk_poly_tobytes_c`	✅	2s	2s	+0%
`mlk_polyvec_compress_du`	✅	2s	3s	-33%
`mlk_polyvec_decompress_du`	✅	2s	3s	-33%
`mlk_polyvec_frombytes`	✅	2s	2s	+0%
`mlk_polyvec_tobytes`	✅	2s	2s	+0%
`mlk_rej_uniform`	✅	2s	2s	+0%
`mlk_scalar_compress_d1`	✅	2s	2s	+0%
`mlk_scalar_compress_d10`	✅	2s	1s	+100%
`mlk_scalar_decompress_d10`	✅	2s	1s	+100%
`mlk_scalar_decompress_d4`	✅	2s	2s	+0%
`mlk_scalar_signed_to_unsigned_q`	✅	2s	2s	+0%
`mlk_shake128_absorb_once`	✅	2s	1s	+100%
`mlk_shake128x4_absorb_once`	✅	2s	2s	+0%
`mlk_shake128x4_squeezeblocks`	✅	2s	3s	-33%
`mlk_shake256`	✅	2s	2s	+0%
`mlk_value_barrier_i32`	✅	2s	3s	-33%
`mlk_value_barrier_u8`	✅	2s	2s	+0%
`ntt_native_x86_64`	✅	2s	2s	+0%
`poly_compress_d10_native_x86_64`	✅	2s	1s	+100%
`poly_compress_d11_native_x86_64`	✅	2s	3s	-33%
`poly_getnoise_eta1122_4x_native`	✅	2s	5s	-60%
`poly_reduce_native_aarch64`	✅	2s	4s	-50%
`poly_reduce_native_x86_64`	✅	2s	3s	-33%
`poly_tobytes_native_aarch64`	✅	2s	1s	+100%
`poly_tomont_native_aarch64`	✅	2s	3s	-33%
`polyvec_basemul_acc_montgomery_cached_k2_native_x86_64`	✅	2s	2s	+0%
`polyvec_basemul_acc_montgomery_cached_k3_native_x86_64`	✅	2s	1s	+100%
`polyvec_basemul_acc_montgomery_cached_k4_native_aarch64`	✅	2s	4s	-50%
`keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid`	✅	1s	2s	-50%
`keccakf1600x4_extract_bytes_native`	✅	1s	1s	+0%
`mlk_ct_cmask_neg_i16`	✅	1s	2s	-50%
`mlk_keccakf1600_xor_bytes (big endian)`	✅	1s	1s	+0%
`mlk_keccakf1600x4_extract_bytes_c`	✅	1s	1s	+0%
`mlk_keccakf1600x4_xor_bytes_c`	✅	1s	2s	-50%
`mlk_montgomery_reduce`	✅	1s	3s	-67%
`mlk_poly_compress_d5_c`	✅	1s	4s	-75%
`mlk_poly_decompress_d4`	✅	1s	2s	-50%
`mlk_poly_decompress_d5_c`	✅	1s	5s	-80%
`mlk_poly_decompress_d5_native`	✅	1s	2s	-50%
`mlk_poly_invntt_tomont`	✅	1s	1s	+0%
`mlk_poly_sub`	✅	1s	3s	-67%
`mlk_polymat_permute_bitrev_to_custom`	✅	1s	3s	-67%
`mlk_polyvec_mulcache_compute`	✅	1s	2s	-50%
`mlk_polyvec_ntt`	✅	1s	5s	-80%
`mlk_polyvec_permute_bitrev_to_custom`	✅	1s	2s	-50%
`mlk_scalar_compress_d5`	✅	1s	2s	-50%
`mlk_scalar_decompress_d11`	✅	1s	2s	-50%
`mlk_scalar_decompress_d5`	✅	1s	4s	-75%
`mlk_sha3_256`	✅	1s	2s	-50%
`mlk_sha3_512`	✅	1s	2s	-50%
`poly_decompress_d5_native_x86_64`	✅	1s	1s	+0%
`poly_invntt_tomont_native`	✅	1s	3s	-67%

oqs-bot · 2026-05-05T15:42:12Z

CBMC Results (ML-KEM-1024)

Full Results (191 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	1413s	1271s	+11.2%
`mlk_poly_rej_uniform`	✅	153s	127s	+20%
`mlk_indcpa_enc`	✅	142s	125s	+14%
`mlk_rej_uniform_c`	✅	134s	114s	+18%
`mlk_indcpa_keypair_derand`	✅	128s	120s	+7%
`mlk_polyvec_basemul_acc_montgomery_cached_c`	✅	80s	73s	+10%
`poly_ntt_native`	✅	42s	34s	+24%
`mlk_ntt_layer`	✅	37s	26s	+42%
`mlk_poly_reduce_native`	✅	35s	34s	+3%
`polyvec_basemul_acc_montgomery_cached_native`	✅	34s	33s	+3%
`mlk_keccak_squeezeblocks_x4`	✅	26s	23s	+13%
`keccakf1600x4_permute_native_x4`	✅	17s	18s	-6%
`mlk_fqmul`	✅	16s	14s	+14%
`mlk_poly_decompress_d11_native`	✅	16s	12s	+33%
`mlk_poly_decompress_d5_native`	✅	16s	13s	+23%
`mlk_poly_frommsg`	✅	14s	11s	+27%
`mlk_indcpa_dec`	✅	12s	11s	+9%
`mlk_polyvec_add`	✅	11s	10s	+10%
`mlk_ntt_butterfly_block`	✅	9s	8s	+12%
`mlk_poly_frombytes_native`	✅	9s	10s	-10%
`mlk_keccak_squeezeblocks`	✅	8s	9s	-11%
`mlk_poly_ntt`	✅	8s	8s	+0%
`mlk_keccak_squeeze_once`	✅	7s	8s	-12%
`mlk_keccakf1600_permute_c`	✅	7s	4s	+75%
`poly_frombytes_native_x86_64`	✅	7s	4s	+75%
`mlk_gen_matrix`	✅	6s	6s	+0%
`mlk_poly_compress_d11_c`	✅	6s	6s	+0%
`mlk_poly_compress_d4_c`	✅	6s	2s	+200%
`mlk_poly_rej_uniform_x4`	✅	6s	7s	-14%
`mlk_polymat_permute_bitrev_to_custom`	✅	6s	7s	-14%
`poly_decompress_d11_native_x86_64`	✅	6s	4s	+50%
`kem_enc`	✅	5s	2s	+150%
`mlk_gen_matrix_serial`	✅	5s	4s	+25%
`mlk_invntt_layer`	✅	5s	5s	+0%
`mlk_keccak_absorb_once_x4`	✅	5s	5s	+0%
`mlk_poly_compress_d10_native`	✅	5s	2s	+150%
`mlk_poly_decompress_d5_c`	✅	5s	2s	+150%
`mlk_poly_frombytes_c`	✅	5s	1s	+400%
`mlk_rej_uniform`	✅	5s	2s	+150%
`poly_decompress_d5_native_x86_64`	✅	5s	4s	+25%
`polyvec_basemul_acc_montgomery_cached_k4_native_x86_64`	✅	5s	2s	+150%
`rej_uniform_native_x86_64`	✅	5s	6s	-17%
`keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid`	✅	4s	2s	+100%
`keccakf1600x4_xor_bytes_native`	✅	4s	1s	+300%
`kem_check_pk`	✅	4s	5s	-20%
`kem_dec`	✅	4s	4s	+0%
`mlk_poly_compress_d5_native`	✅	4s	3s	+33%
`mlk_poly_mulcache_compute_c`	✅	4s	3s	+33%
`mlk_polyvec_invntt_tomont`	✅	4s	3s	+33%
`mlk_polyvec_mulcache_compute`	✅	4s	1s	+300%
`mlk_polyvec_permute_bitrev_to_custom_native`	✅	4s	2s	+100%
`mlk_polyvec_tobytes`	✅	4s	2s	+100%
`ntt_native_x86_64`	✅	4s	3s	+33%
`nttunpack_native_x86_64`	✅	4s	1s	+300%
`poly_compress_d5_native_x86_64`	✅	4s	5s	-20%
`poly_invntt_tomont_native`	✅	4s	3s	+33%
`poly_tobytes_native_aarch64`	✅	4s	2s	+100%
`intt_native_x86_64`	✅	3s	3s	+0%
`keccak_f1600_x4_native_aarch64_v84a`	✅	3s	1s	+200%
`mlk_barrett_reduce`	✅	3s	2s	+50%
`mlk_ct_cmask_nonzero_u16`	✅	3s	1s	+200%
`mlk_ct_get_optblocker_u32`	✅	3s	2s	+50%
`mlk_ct_sel_uint8`	✅	3s	1s	+200%
`mlk_keccak_absorb_once`	✅	3s	5s	-40%
`mlk_keccakf1600_extract_bytes (big endian)`	✅	3s	3s	+0%
`mlk_keccakf1600_xor_bytes (big endian)`	✅	3s	1s	+200%
`mlk_keccakf1600x4_extract_bytes`	✅	3s	3s	+0%
`mlk_keccakf1600x4_permute`	✅	3s	2s	+50%
`mlk_keccakf1600x4_xor_bytes`	✅	3s	2s	+50%
`mlk_keccakf1600x4_xor_bytes_c`	✅	3s	2s	+50%
`mlk_matvec_mul`	✅	3s	2s	+50%
`mlk_montgomery_reduce`	✅	3s	2s	+50%
`mlk_poly_cbd_eta1`	✅	3s	2s	+50%
`mlk_poly_cbd_eta2`	✅	3s	2s	+50%
`mlk_poly_compress_d10`	✅	3s	1s	+200%
`mlk_poly_compress_d10_c`	✅	3s	1s	+200%
`mlk_poly_decompress_d10_native`	✅	3s	2s	+50%
`mlk_poly_decompress_d4_c`	✅	3s	2s	+50%
`mlk_poly_decompress_d5`	✅	3s	2s	+50%
`mlk_poly_decompress_dv`	✅	3s	1s	+200%
`mlk_poly_getnoise_eta1_4x_native`	✅	3s	2s	+50%
`mlk_poly_ntt_c`	✅	3s	5s	-40%
`mlk_polyvec_basemul_acc_montgomery_cached`	✅	3s	2s	+50%
`mlk_polyvec_decompress_du`	✅	3s	3s	+0%
`mlk_polyvec_reduce`	✅	3s	1s	+200%
`mlk_polyvec_tomont`	✅	3s	2s	+50%
`mlk_scalar_compress_d10`	✅	3s	2s	+50%
`mlk_scalar_compress_d4`	✅	3s	2s	+50%
`mlk_scalar_decompress_d5`	✅	3s	2s	+50%
`mlk_sha3_512`	✅	3s	3s	+0%
`mlk_shake128_squeezeblocks`	✅	3s	2s	+50%
`mlk_value_barrier_u8`	✅	3s	2s	+50%
`poly_compress_d10_native_x86_64`	✅	3s	3s	+0%
`poly_decompress_d4_native_x86_64`	✅	3s	1s	+200%
`poly_getnoise_eta1122_4x_native`	✅	3s	1s	+200%
`poly_mulcache_compute_native_x86_64`	✅	3s	2s	+50%
`poly_tomont_native_aarch64`	✅	3s	4s	-25%
`polyvec_basemul_acc_montgomery_cached_k2_native_aarch64`	✅	3s	3s	+0%
`polyvec_basemul_acc_montgomery_cached_k2_native_x86_64`	✅	3s	5s	-40%
`polyvec_basemul_acc_montgomery_cached_k4_native_aarch64`	✅	3s	3s	+0%
`intt_native_aarch64`	✅	2s	4s	-50%
`keccak_f1600_x1_native_aarch64`	✅	2s	1s	+100%
`keccak_f1600_x1_native_aarch64_v84a`	✅	2s	2s	+0%
`keccakf1600_permute_native`	✅	2s	2s	+0%
`kem_check_sk`	✅	2s	2s	+0%
`kem_enc_derand`	✅	2s	4s	-50%
`kem_keypair`	✅	2s	2s	+0%
`kem_keypair_derand`	✅	2s	3s	-33%
`mlk_check_pct`	✅	2s	3s	-33%
`mlk_ct_cmask_nonzero_u8`	✅	2s	3s	-33%
`mlk_ct_cmov_zero`	✅	2s	4s	-50%
`mlk_ct_get_optblocker_i32`	✅	2s	1s	+100%
`mlk_ct_get_optblocker_u8`	✅	2s	1s	+100%
`mlk_ct_sel_int16`	✅	2s	2s	+0%
`mlk_enc_getnoise_eta1_eta2`	✅	2s	2s	+0%
`mlk_keccakf1600_extract_bytes`	✅	2s	3s	-33%
`mlk_keccakf1600_permute`	✅	2s	3s	-33%
`mlk_keccakf1600_xor_bytes`	✅	2s	2s	+0%
`mlk_keypair_getnoise_eta1`	✅	2s	2s	+0%
`mlk_poly_add`	✅	2s	2s	+0%
`mlk_poly_compress_d11`	✅	2s	2s	+0%
`mlk_poly_compress_d11_native`	✅	2s	1s	+100%
`mlk_poly_compress_d4`	✅	2s	2s	+0%
`mlk_poly_compress_d4_native`	✅	2s	1s	+100%
`mlk_poly_compress_d5`	✅	2s	2s	+0%
`mlk_poly_compress_d5_c`	✅	2s	2s	+0%
`mlk_poly_compress_du`	✅	2s	3s	-33%
`mlk_poly_compress_dv`	✅	2s	2s	+0%
`mlk_poly_decompress_d10_c`	✅	2s	2s	+0%
`mlk_poly_decompress_d11_c`	✅	2s	2s	+0%
`mlk_poly_decompress_d4`	✅	2s	2s	+0%
`mlk_poly_decompress_d4_native`	✅	2s	2s	+0%
`mlk_poly_decompress_du`	✅	2s	2s	+0%
`mlk_poly_getnoise_eta1_4x`	✅	2s	2s	+0%
`mlk_poly_getnoise_eta2`	✅	2s	2s	+0%
`mlk_poly_invntt_tomont`	✅	2s	3s	-33%
`mlk_poly_invntt_tomont_c`	✅	2s	2s	+0%
`mlk_poly_mulcache_compute`	✅	2s	3s	-33%
`mlk_poly_reduce`	✅	2s	2s	+0%
`mlk_poly_reduce_c`	✅	2s	2s	+0%
`mlk_poly_tobytes_c`	✅	2s	5s	-60%
`mlk_poly_tomont`	✅	2s	2s	+0%
`mlk_poly_tomont_c`	✅	2s	2s	+0%
`mlk_polyvec_compress_du`	✅	2s	2s	+0%
`mlk_polyvec_frombytes`	✅	2s	5s	-60%
`mlk_polyvec_ntt`	✅	2s	3s	-33%
`mlk_polyvec_permute_bitrev_to_custom`	✅	2s	4s	-50%
`mlk_scalar_compress_d1`	✅	2s	1s	+100%
`mlk_scalar_compress_d11`	✅	2s	2s	+0%
`mlk_scalar_compress_d5`	✅	2s	1s	+100%
`mlk_scalar_decompress_d10`	✅	2s	2s	+0%
`mlk_scalar_decompress_d11`	✅	2s	2s	+0%
`mlk_scalar_decompress_d4`	✅	2s	3s	-33%
`mlk_scalar_signed_to_unsigned_q`	✅	2s	2s	+0%
`mlk_sha3_256`	✅	2s	2s	+0%
`mlk_shake128_absorb_once`	✅	2s	2s	+0%
`mlk_shake128x4_squeezeblocks`	✅	2s	1s	+100%
`mlk_shake256`	✅	2s	4s	-50%
`mlk_shake256x4`	✅	2s	3s	-33%
`mlk_value_barrier_i32`	✅	2s	3s	-33%
`mlk_value_barrier_u32`	✅	2s	5s	-60%
`ntt_native_aarch64`	✅	2s	3s	-33%
`poly_compress_d4_native_x86_64`	✅	2s	2s	+0%
`poly_decompress_d10_native_x86_64`	✅	2s	2s	+0%
`poly_reduce_native_aarch64`	✅	2s	2s	+0%
`poly_reduce_native_x86_64`	✅	2s	2s	+0%
`poly_tobytes_native_x86_64`	✅	2s	1s	+100%
`poly_tomont_native_x86_64`	✅	2s	5s	-60%
`polyvec_basemul_acc_montgomery_cached_k3_native_aarch64`	✅	2s	2s	+0%
`polyvec_basemul_acc_montgomery_cached_k3_native_x86_64`	✅	2s	4s	-50%
`rej_uniform_native`	✅	2s	3s	-33%
`rej_uniform_native_aarch64`	✅	2s	3s	-33%
`sys_check_capability`	✅	2s	2s	+0%
`keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid`	✅	1s	1s	+0%
`keccak_f1600_x4_native_avx2`	✅	1s	3s	-67%
`keccakf1600x4_extract_bytes_native`	✅	1s	3s	-67%
`mlk_ct_cmask_neg_i16`	✅	1s	1s	+0%
`mlk_ct_memcmp`	✅	1s	2s	-50%
`mlk_keccakf1600x4_extract_bytes_c`	✅	1s	1s	+0%
`mlk_poly_decompress_d10`	✅	1s	2s	-50%
`mlk_poly_decompress_d11`	✅	1s	5s	-80%
`mlk_poly_frombytes`	✅	1s	2s	-50%
`mlk_poly_getnoise_eta1122_4x`	✅	1s	2s	-50%
`mlk_poly_mulcache_compute_native`	✅	1s	4s	-75%
`mlk_poly_sub`	✅	1s	3s	-67%
`mlk_poly_tobytes`	✅	1s	2s	-50%
`mlk_poly_tobytes_native`	✅	1s	2s	-50%
`mlk_poly_tomont_native`	✅	1s	3s	-67%
`mlk_poly_tomsg`	✅	1s	1s	+0%
`mlk_shake128x4_absorb_once`	✅	1s	2s	-50%
`poly_compress_d11_native_x86_64`	✅	1s	3s	-67%
`poly_mulcache_compute_native_aarch64`	✅	1s	2s	-50%

oqs-bot · 2026-05-05T15:43:02Z

CBMC Results (ML-KEM-768)

⚠️ Attention Required

Proof	Status	Current	Previous	Change
`mlk_ntt_layer`	⚠️	41s	27s	+52%

Full Results (191 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	1628s	1336s	+21.9%
`mlk_indcpa_keypair_derand`	✅	215s	190s	+13%
`mlk_indcpa_enc`	✅	188s	149s	+26%
`mlk_poly_rej_uniform`	✅	184s	139s	+32%
`mlk_rej_uniform_c`	✅	166s	126s	+32%
`mlk_polyvec_basemul_acc_montgomery_cached_c`	✅	58s	45s	+29%
`poly_ntt_native`	✅	47s	39s	+21%
`mlk_poly_reduce_native`	✅	44s	31s	+42%
`mlk_ntt_layer`	⚠️	41s	27s	+52%
`mlk_keccak_squeezeblocks_x4`	✅	30s	24s	+25%
`mlk_fqmul`	✅	22s	18s	+22%
`keccakf1600x4_permute_native_x4`	✅	19s	17s	+12%
`mlk_poly_decompress_d10_native`	✅	19s	15s	+27%
`polyvec_basemul_acc_montgomery_cached_native`	✅	19s	17s	+12%
`mlk_poly_decompress_d4_native`	✅	16s	11s	+45%
`mlk_indcpa_dec`	✅	13s	11s	+18%
`mlk_poly_frommsg`	✅	13s	11s	+18%
`mlk_keccak_squeezeblocks`	✅	11s	9s	+22%
`mlk_ntt_butterfly_block`	✅	10s	10s	+0%
`mlk_poly_frombytes_native`	✅	10s	6s	+67%
`mlk_polyvec_add`	✅	10s	10s	+0%
`mlk_poly_ntt`	✅	9s	7s	+29%
`mlk_gen_matrix_serial`	✅	8s	4s	+100%
`mlk_keccak_absorb_once_x4`	✅	8s	7s	+14%
`mlk_poly_rej_uniform_x4`	✅	8s	6s	+33%
`poly_frombytes_native_x86_64`	✅	8s	4s	+100%
`kem_dec`	✅	7s	4s	+75%
`mlk_keccak_squeeze_once`	✅	7s	9s	-22%
`rej_uniform_native_x86_64`	✅	7s	4s	+75%
`kem_enc_derand`	✅	6s	3s	+100%
`mlk_invntt_layer`	✅	6s	4s	+50%
`mlk_keccakf1600_permute_c`	✅	6s	3s	+100%
`nttunpack_native_x86_64`	✅	6s	3s	+100%
`poly_decompress_d10_native_x86_64`	✅	6s	4s	+50%
`intt_native_aarch64`	✅	5s	3s	+67%
`mlk_gen_matrix`	✅	5s	3s	+67%
`mlk_poly_ntt_c`	✅	5s	2s	+150%
`mlk_poly_reduce_c`	✅	5s	3s	+67%
`poly_decompress_d4_native_x86_64`	✅	5s	3s	+67%
`polyvec_basemul_acc_montgomery_cached_k3_native_x86_64`	✅	5s	4s	+25%
`kem_keypair`	✅	4s	3s	+33%
`mlk_ct_sel_int16`	✅	4s	2s	+100%
`mlk_keccakf1600x4_xor_bytes_c`	✅	4s	3s	+33%
`mlk_poly_cbd_eta2`	✅	4s	3s	+33%
`mlk_poly_compress_d10_c`	✅	4s	3s	+33%
`mlk_poly_compress_d5_c`	✅	4s	4s	+0%
`mlk_poly_decompress_d11_c`	✅	4s	2s	+100%
`mlk_poly_decompress_d5`	✅	4s	4s	+0%
`mlk_poly_frombytes_c`	✅	4s	1s	+300%
`mlk_poly_getnoise_eta1122_4x`	✅	4s	1s	+300%
`mlk_poly_invntt_tomont_c`	✅	4s	4s	+0%
`mlk_poly_tomsg`	✅	4s	3s	+33%
`mlk_scalar_decompress_d4`	✅	4s	2s	+100%
`mlk_shake128x4_absorb_once`	✅	4s	2s	+100%
`poly_invntt_tomont_native`	✅	4s	2s	+100%
`poly_mulcache_compute_native_aarch64`	✅	4s	3s	+33%
`poly_tobytes_native_aarch64`	✅	4s	2s	+100%
`poly_tomont_native_aarch64`	✅	4s	3s	+33%
`polyvec_basemul_acc_montgomery_cached_k2_native_aarch64`	✅	4s	4s	+0%
`keccak_f1600_x1_native_aarch64`	✅	3s	5s	-40%
`keccak_f1600_x4_native_aarch64_v8a_v84a_scalar_hybrid`	✅	3s	2s	+50%
`keccakf1600x4_extract_bytes_native`	✅	3s	1s	+200%
`keccakf1600x4_xor_bytes_native`	✅	3s	3s	+0%
`kem_check_pk`	✅	3s	2s	+50%
`kem_check_sk`	✅	3s	3s	+0%
`kem_enc`	✅	3s	3s	+0%
`mlk_ct_cmask_neg_i16`	✅	3s	2s	+50%
`mlk_ct_cmask_nonzero_u16`	✅	3s	3s	+0%
`mlk_ct_cmask_nonzero_u8`	✅	3s	2s	+50%
`mlk_ct_cmov_zero`	✅	3s	2s	+50%
`mlk_ct_get_optblocker_i32`	✅	3s	2s	+50%
`mlk_ct_memcmp`	✅	3s	3s	+0%
`mlk_keccak_absorb_once`	✅	3s	4s	-25%
`mlk_keccakf1600_extract_bytes`	✅	3s	1s	+200%
`mlk_keccakf1600_extract_bytes (big endian)`	✅	3s	1s	+200%
`mlk_keccakf1600_permute`	✅	3s	1s	+200%
`mlk_matvec_mul`	✅	3s	3s	+0%
`mlk_poly_add`	✅	3s	2s	+50%
`mlk_poly_cbd_eta1`	✅	3s	2s	+50%
`mlk_poly_compress_d11_native`	✅	3s	1s	+200%
`mlk_poly_compress_dv`	✅	3s	5s	-40%
`mlk_poly_decompress_d10`	✅	3s	2s	+50%
`mlk_poly_decompress_d10_c`	✅	3s	1s	+200%
`mlk_poly_decompress_dv`	✅	3s	1s	+200%
`mlk_poly_getnoise_eta1_4x`	✅	3s	3s	+0%
`mlk_poly_getnoise_eta1_4x_native`	✅	3s	1s	+200%
`mlk_poly_getnoise_eta2`	✅	3s	2s	+50%
`mlk_poly_mulcache_compute_c`	✅	3s	1s	+200%
`mlk_poly_sub`	✅	3s	2s	+50%
`mlk_poly_tobytes`	✅	3s	2s	+50%
`mlk_poly_tobytes_c`	✅	3s	2s	+50%
`mlk_poly_tomont`	✅	3s	3s	+0%
`mlk_poly_tomont_c`	✅	3s	3s	+0%
`mlk_polymat_permute_bitrev_to_custom`	✅	3s	4s	-25%
`mlk_polyvec_mulcache_compute`	✅	3s	2s	+50%
`mlk_polyvec_permute_bitrev_to_custom`	✅	3s	2s	+50%
`mlk_polyvec_permute_bitrev_to_custom_native`	✅	3s	3s	+0%
`mlk_scalar_compress_d10`	✅	3s	2s	+50%
`mlk_scalar_decompress_d11`	✅	3s	1s	+200%
`mlk_scalar_decompress_d5`	✅	3s	3s	+0%
`mlk_shake256`	✅	3s	2s	+50%
`mlk_value_barrier_i32`	✅	3s	2s	+50%
`ntt_native_x86_64`	✅	3s	1s	+200%
`poly_compress_d10_native_x86_64`	✅	3s	2s	+50%
`poly_compress_d11_native_x86_64`	✅	3s	3s	+0%
`poly_compress_d4_native_x86_64`	✅	3s	2s	+50%
`poly_reduce_native_x86_64`	✅	3s	2s	+50%
`poly_tobytes_native_x86_64`	✅	3s	4s	-25%
`polyvec_basemul_acc_montgomery_cached_k4_native_aarch64`	✅	3s	2s	+50%
`rej_uniform_native`	✅	3s	2s	+50%
`intt_native_x86_64`	✅	2s	1s	+100%
`keccak_f1600_x1_native_aarch64_v84a`	✅	2s	2s	+0%
`keccak_f1600_x4_native_aarch64_v84a`	✅	2s	3s	-33%
`keccak_f1600_x4_native_aarch64_v8a_scalar_hybrid`	✅	2s	2s	+0%
`keccakf1600_permute_native`	✅	2s	2s	+0%
`mlk_barrett_reduce`	✅	2s	2s	+0%
`mlk_keccakf1600_xor_bytes`	✅	2s	3s	-33%
`mlk_keccakf1600_xor_bytes (big endian)`	✅	2s	2s	+0%
`mlk_keccakf1600x4_permute`	✅	2s	2s	+0%
`mlk_keccakf1600x4_xor_bytes`	✅	2s	3s	-33%
`mlk_keypair_getnoise_eta1`	✅	2s	1s	+100%
`mlk_montgomery_reduce`	✅	2s	3s	-33%
`mlk_poly_compress_d10_native`	✅	2s	4s	-50%
`mlk_poly_compress_d4`	✅	2s	4s	-50%
`mlk_poly_compress_d5_native`	✅	2s	1s	+100%
`mlk_poly_compress_du`	✅	2s	3s	-33%
`mlk_poly_decompress_d11`	✅	2s	2s	+0%
`mlk_poly_decompress_d11_native`	✅	2s	2s	+0%
`mlk_poly_decompress_d4`	✅	2s	2s	+0%
`mlk_poly_decompress_d4_c`	✅	2s	2s	+0%
`mlk_poly_decompress_d5_native`	✅	2s	2s	+0%
`mlk_poly_decompress_du`	✅	2s	3s	-33%
`mlk_poly_frombytes`	✅	2s	2s	+0%
`mlk_poly_invntt_tomont`	✅	2s	3s	-33%
`mlk_poly_mulcache_compute_native`	✅	2s	2s	+0%
`mlk_poly_reduce`	✅	2s	3s	-33%
`mlk_poly_tobytes_native`	✅	2s	1s	+100%
`mlk_poly_tomont_native`	✅	2s	3s	-33%
`mlk_polyvec_basemul_acc_montgomery_cached`	✅	2s	2s	+0%
`mlk_polyvec_compress_du`	✅	2s	2s	+0%
`mlk_polyvec_decompress_du`	✅	2s	2s	+0%
`mlk_polyvec_frombytes`	✅	2s	2s	+0%
`mlk_polyvec_invntt_tomont`	✅	2s	3s	-33%
`mlk_polyvec_reduce`	✅	2s	2s	+0%
`mlk_polyvec_tobytes`	✅	2s	4s	-50%
`mlk_rej_uniform`	✅	2s	2s	+0%
`mlk_scalar_compress_d1`	✅	2s	1s	+100%
`mlk_scalar_compress_d11`	✅	2s	2s	+0%
`mlk_scalar_compress_d4`	✅	2s	1s	+100%
`mlk_scalar_compress_d5`	✅	2s	1s	+100%
`mlk_scalar_decompress_d10`	✅	2s	1s	+100%
`mlk_sha3_256`	✅	2s	3s	-33%
`mlk_sha3_512`	✅	2s	2s	+0%
`mlk_shake128_absorb_once`	✅	2s	4s	-50%
`mlk_shake128_squeezeblocks`	✅	2s	3s	-33%
`mlk_shake128x4_squeezeblocks`	✅	2s	2s	+0%
`mlk_shake256x4`	✅	2s	3s	-33%
`mlk_value_barrier_u8`	✅	2s	1s	+100%
`poly_compress_d5_native_x86_64`	✅	2s	2s	+0%
`poly_decompress_d11_native_x86_64`	✅	2s	2s	+0%
`poly_decompress_d5_native_x86_64`	✅	2s	3s	-33%
`poly_getnoise_eta1122_4x_native`	✅	2s	2s	+0%
`poly_mulcache_compute_native_x86_64`	✅	2s	1s	+100%
`poly_reduce_native_aarch64`	✅	2s	1s	+100%
`poly_tomont_native_x86_64`	✅	2s	3s	-33%
`polyvec_basemul_acc_montgomery_cached_k2_native_x86_64`	✅	2s	1s	+100%
`polyvec_basemul_acc_montgomery_cached_k3_native_aarch64`	✅	2s	2s	+0%
`polyvec_basemul_acc_montgomery_cached_k4_native_x86_64`	✅	2s	3s	-33%
`rej_uniform_native_aarch64`	✅	2s	4s	-50%
`sys_check_capability`	✅	2s	2s	+0%
`keccak_f1600_x4_native_avx2`	✅	1s	2s	-50%
`kem_keypair_derand`	✅	1s	2s	-50%
`mlk_check_pct`	✅	1s	5s	-80%
`mlk_ct_get_optblocker_u32`	✅	1s	2s	-50%
`mlk_ct_get_optblocker_u8`	✅	1s	3s	-67%
`mlk_ct_sel_uint8`	✅	1s	1s	+0%
`mlk_enc_getnoise_eta1_eta2`	✅	1s	4s	-75%
`mlk_keccakf1600x4_extract_bytes`	✅	1s	2s	-50%
`mlk_keccakf1600x4_extract_bytes_c`	✅	1s	3s	-67%
`mlk_poly_compress_d10`	✅	1s	2s	-50%
`mlk_poly_compress_d11`	✅	1s	2s	-50%
`mlk_poly_compress_d11_c`	✅	1s	1s	+0%
`mlk_poly_compress_d4_c`	✅	1s	3s	-67%
`mlk_poly_compress_d4_native`	✅	1s	2s	-50%
`mlk_poly_compress_d5`	✅	1s	1s	+0%
`mlk_poly_decompress_d5_c`	✅	1s	1s	+0%
`mlk_poly_mulcache_compute`	✅	1s	1s	+0%
`mlk_polyvec_ntt`	✅	1s	2s	-50%
`mlk_polyvec_tomont`	✅	1s	3s	-67%
`mlk_scalar_signed_to_unsigned_q`	✅	1s	3s	-67%
`mlk_value_barrier_u32`	✅	1s	2s	-50%
`ntt_native_aarch64`	✅	1s	4s	-75%

hanno-becker · 2026-06-04T15:28:43Z

          nix-shell: linter
          gh_token: ${{ secrets.GITHUB_TOKEN }}
          cross-prefix: "aarch64-unknown-linux-gnu-"
+  ubuntu_arm:


Is this not subsumed by the quickcheck job?

Yes, I suspect it is. I think this is because of a bad rebase; I'll prune it.

hanno-becker · 2026-06-04T15:32:30Z

+    outp="$out/platform/nucleo-n657x0-q/src/platform"
+    mkdir -p "$outp"
+
+    tpl="Projects/NUCLEO-N657X0-Q/Templates/Template"
+
+    if [ ! -d "$tpl" ]; then
+      echo "ERROR: expected Cube template at $tpl"
+      exit 1
+    fi
+
+    # Copy CMSIS headers needed by system/startup from the Cube tree
+    mkdir -p "$outp/Drivers/CMSIS/Core"
+    mkdir -p "$outp/Drivers/CMSIS/Device"


The install phase seems pretty long. Could you provide some overview in the beginning as to what is happening here?

Yes, I have added some comments that will hopefully make things clearer. I have also simplified the nix file.

hanno-becker · 2026-06-11T10:47:42Z

+        env:
+          EXTRA_MAKEFILE: ${{ matrix.target.system == 'nucleo-n657x0' && 'test/baremetal/platform/nucleo-n657x0-q/platform.mk' || '' }}


Can this be part of the matrix?

Yes, agreed. I’ll add an extra_makefile field to the benchmark matrix and pass that through as EXTRA_MAKEFILE, with an empty value for the existing non-baremetal targets. That removes the target-specific expression from the step.

hanno-becker · 2026-06-11T10:48:11Z

          nix-shell: ''
          gh_token: ${{ secrets.GITHUB_TOKEN }}
          cflags: "-DMLKEM_DEBUG -DMLK_FORCE_PPC64LE"
+  nucleo_n657x0_q_tests:


Why is this done here and not alongside the existing baremetal tests?

Good point. This should live with the existing baremetal workflow. I’ll move the NUCLEO-N657X0-Q job into .github/workflows/baremetal.yml as a matrix target, while keeping the self-hosted/fork guard because this one needs attached hardware rather than QEMU.

This still seems open

hanno-becker · 2026-06-11T10:50:14Z

+
+# NUCLEO-N657X0-Q Baremetal Platform
+
+This platform runs ML-KEM tests on the ST NUCLEO-N657X0-Q board using the


Does QEMU support this board, so one could, alternatively, emulate the final binary?

I don’t think QEMU currently supports the STM32N657/NUCLEO-N657X0-Q machine model. QEMU’s Arm docs list only a small set of STM32 boards, and this board is not among them; they also note that firmware images generally need an exact matching board model to boot. I’ll add a short note to the README saying this platform currently requires hardware. (qemu.org)

That said, I have just stumbled across https://renode.io which may be relevant to simulating hardware targets outside the baseline supported by QEMU.

hanno-becker · 2026-06-11T10:52:04Z

@@ -0,0 +1,179 @@
+#!/usr/bin/env python3


Shallow comment: Can we keep the top-level of the platform clean and ideally confined to exec_wrapper.py, platform.mk, README.md?

Yes. I’ll move host-side helpers such as flexmem_configure.py, run_test_after_flexmem.py, make_argv_bin.py, and the host tests under the existing nucleo_host/ area, leaving only exec_wrapper.py, platform.mk, and README.md at the platform top level.

I have a fix staged, but I don't have my nucleo with me today so I cannot test it until later in the day.

hanno-becker · 2026-06-11T10:53:04Z

+# Build hal.c without pedantic diagnostics (overrides global -pedantic/-Wpedantic)
+$(MLKEM512_DIR)/test/hal/hal.c.o: CFLAGS := $(filter-out -pedantic -Wpedantic -Werror -Wconversion -Wsign-conversion,$(CFLAGS)) -Wno-pedantic -Wno-conversion -Wno-sign-conversion -Wno-error
+$(MLKEM768_DIR)/test/hal/hal.c.o: CFLAGS := $(filter-out -pedantic -Wpedantic -Werror -Wconversion -Wsign-conversion,$(CFLAGS)) -Wno-pedantic -Wno-conversion -Wno-sign-conversion -Wno-error
+$(MLKEM1024_DIR)/test/hal/hal.c.o: CFLAGS := $(filter-out -pedantic -Wpedantic -Werror -Wconversion -Wsign-conversion,$(CFLAGS)) -Wno-pedantic -Wno-conversion -Wno-sign-conversion -Wno-error


hal.c seems small enough that one should be able to adjust the code instead of silencing the warnings?

Agreed. I’ll remove this per-target warning suppression and fix the warnings in test/hal/hal.c directly. The warning relaxation should stay limited to vendor/Cube sources in the platform makefile.

hanno-becker · 2026-06-11T10:55:34Z

+            return
+
+        if len(self.failed) > 0:
+            self.check_fail()


What's this?

@bremoran ping

This block exists because failed benchmark runs return booleans instead of output strings, and the code was bailing out before parsing those booleans. I’ll make that explicit in control flow.

mkannwischer · 2026-06-14T04:42:08Z

+# nucleo-n657x0-q benchmarks. The package copies only the required Cube headers
+# and sources, then patches selected files where needed to get the intended
+# RAM-loaded benchmark behavior.
+stdenvNoCC.mkDerivation {


Would it be an option to just move all the patched files to pqmx and consume them from there?

The issue with this is the license here. If this is acceptable in pqmx, then we can move the files.

https://github.com/STMicroelectronics/STM32CubeN6/blob/main/Projects/NUCLEO-N657X0-Q/Templates/LICENSE.md

hanno-becker · 2026-06-14T06:01:11Z

@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) The mldsa-native project authors


Is this file needed? It doesn't seem to be listed in the extra sources.

hanno-becker · 2026-06-14T06:03:08Z

@@ -0,0 +1,635 @@
+# Copyright (c) The mlkem-native project authors


This file is not exercised, it seems. Can it be removed?

hanno-becker · 2026-06-14T06:05:22Z

+- `.data`, `.bss`: expanded 256 KiB DTCM at `0x30000000`
+- argv block: expanded 256 KiB ITCM at `0x00000000`
+- stdout capture: AXI SRAM at `0x34080000`
+- stack: top 192 KiB of DTCM, with `_estack = 0x30040000` and


This seems out of date with the linker script which uses 240 KiB. Can you sync the README with the current state of the PR?

hanno-becker · 2026-06-14T06:06:46Z

@@ -0,0 +1,22 @@
+#!/usr/bin/env python3


(Where) is this exercised? Drop it or add to CI?

mkannwischer · 2026-06-14T14:17:35Z

+openocd.overrideAttrs (old: rec {
+  pname = "st-openocd";
+  version = "unstable-2026-05-01";
+


I have moved this directly into nixpkgs and I'll try to upstream it so we do not have to compile thsi from source: 47c5987

opened a PR here: NixOS/nixpkgs#531668

Build the mlkem-native tests as Zephyr applications so they run on QEMU-emulated Arm MPS boards spanning Cortex-M3/M4/M7/M33/M55. This greatly simplifies supporting more hardware without the need for dedicated hardware abstraction layers. Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>

The Zephyr platform now covers the QEMU MPS boards, so the baremetal m55-an547 and m33-an524 platforms and the pqmx package they used are no longer needed. Also drops the cross-arm-embedded dev shell and the two baremetal.yml jobs. Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>

Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>

Add support for running bare-metal tests and benchmarks on the STM32 NUCLEO-N657X0-Q. This adds the platform linker scripts, semihosting glue, argv/result helpers, FLEXMEM configuration and recovery tooling, and an OpenOCD-based execution backend with host-side tests. Package the required ST OpenOCD tooling in Nix and expose a nucleo-n657x0-q development shell. Wire the platform into functional tests, component benchmarks, CI workflows, and shared bare-metal result handling, including Cortex-M cycle counter support. Signed-off-by: Brendan Moran <brendan.moran@arm.com>

Signed-off-by: Brendan Moran <brendan.moran@arm.com>

Route NUCLEO-N657X0-Q CI and benchmarks through the Zephyr platform makefile, add the Zephyr board config/overlay/shim, and keep the FLEXMEM/OpenOCD hardware helpers needed by that path. Remove the obsolete baremetal platform makefile, linker script, command-line shim, semihosting runtime, and unused host package marker. Signed-off-by: Brendan Moran <brendan.moran@arm.com>

Remove the standalone NUCLEO-N657X0-Q bare-metal platform and fold the board into the Zephyr test platform with board-specific config, overlay, and shim handling. Move the NUCLEO host helpers under test/zephyr, configure FLEXMEM with direct OpenOCD register writes before each run, load RAM-resident Zephyr ELFs with GDB, inject argv after Zephyr startup, and capture stdout through ITM/SWO. Add the OpenOCD package and Zephyr shell/module wiring needed for STM32N6, update CI and benchmark flows for the Zephyr hardware target, and keep host regression coverage for the runner helpers. Signed-off-by: Brendan Moran <brendan.moran@arm.com>

Signed-off-by: Brendan Moran <brendan.moran@arm.com>

bremoran requested a review from a team as a code owner February 5, 2026 13:21

bremoran force-pushed the platform-nucleo-n657x0-q branch from c71f1bc to 32c4136 Compare February 5, 2026 13:28

mkannwischer marked this pull request as draft March 21, 2026 05:29

bremoran force-pushed the platform-nucleo-n657x0-q branch from e3776fa to c95d72a Compare May 5, 2026 14:01

bremoran marked this pull request as ready for review May 5, 2026 14:59

bremoran force-pushed the platform-nucleo-n657x0-q branch from 54df290 to dd9798a Compare May 12, 2026 12:53

mkannwischer added the benchmark this PR should be benchmarked in CI label May 12, 2026

bremoran force-pushed the platform-nucleo-n657x0-q branch from dd9798a to 1ad95e2 Compare May 12, 2026 14:49

bremoran force-pushed the platform-nucleo-n657x0-q branch 2 times, most recently from 8cd8188 to 0d5bd05 Compare June 4, 2026 08:47

bremoran added benchmark this PR should be benchmarked in CI and removed benchmark this PR should be benchmarked in CI labels Jun 4, 2026

bremoran force-pushed the platform-nucleo-n657x0-q branch from 3f291fd to 6b8a791 Compare June 4, 2026 09:51

bremoran added benchmark this PR should be benchmarked in CI and removed benchmark this PR should be benchmarked in CI labels Jun 4, 2026

hanno-becker reviewed Jun 4, 2026

View reviewed changes

hanno-becker reviewed Jun 11, 2026

View reviewed changes

bremoran force-pushed the platform-nucleo-n657x0-q branch from 199f5d7 to 3890d1d Compare June 11, 2026 11:20

mkannwischer reviewed Jun 14, 2026

View reviewed changes

hanno-becker reviewed Jun 14, 2026

View reviewed changes

mkannwischer reviewed Jun 14, 2026

View reviewed changes

mkannwischer mentioned this pull request Jun 14, 2026

openocd-unstable: init at 0.12.0-unstable-2026-06-11 NixOS/nixpkgs#531668

Closed

14 tasks

mkannwischer and others added 4 commits June 17, 2026 17:58

TEMP: Trim down CI

ebf3391

Signed-off-by: Matthias J. Kannwischer <matthias@zerorisc.com>

bremoran force-pushed the platform-nucleo-n657x0-q branch from 3890d1d to f97c538 Compare June 17, 2026 13:01

bremoran added 2 commits June 17, 2026 17:18

NUCLEO-N657X0-Q: use Zephyr platform files

b62a6a6

Signed-off-by: Brendan Moran <brendan.moran@arm.com>

Lint: Make tab checks portable

7a0b873

Signed-off-by: Brendan Moran <brendan.moran@arm.com>

bremoran force-pushed the platform-nucleo-n657x0-q branch from f97c538 to 7a0b873 Compare June 17, 2026 16:25

bremoran added 3 commits June 18, 2026 11:55

Address NUCLEO review comments

005bc99

Signed-off-by: Brendan Moran <brendan.moran@arm.com>

		env:
		EXTRA_MAKEFILE: ${{ matrix.target.system == 'nucleo-n657x0' && 'test/baremetal/platform/nucleo-n657x0-q/platform.mk' \|\| '' }}


		# NUCLEO-N657X0-Q Baremetal Platform

		This platform runs ML-KEM tests on the ST NUCLEO-N657X0-Q board using the

		@@ -0,0 +1,60 @@
		/*
		* Copyright (c) The mldsa-native project authors

		@@ -0,0 +1,635 @@
		# Copyright (c) The mlkem-native project authors

Uh oh!

Conversation

bremoran commented Feb 5, 2026

Summary

Changes

Why

Usage

Notes

Uh oh!

hanno-becker commented Mar 19, 2026

Uh oh!

bremoran commented Mar 20, 2026

Uh oh!

mkannwischer commented Mar 21, 2026

Uh oh!

bremoran commented May 5, 2026

Uh oh!

oqs-bot commented May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-KEM-512)

Uh oh!

oqs-bot commented May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-KEM-1024)

Uh oh!

oqs-bot commented May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-KEM-768)

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

hanno-becker Jun 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

oqs-bot commented May 5, 2026 •

edited

Loading

oqs-bot commented May 5, 2026 •

edited

Loading

oqs-bot commented May 5, 2026 •

edited

Loading

hanno-becker Jun 21, 2026 •

edited

Loading