Eliminate caddq intrinsics by willieyz · Pull Request #905 · pq-code-package/mldsa-native

willieyz · 2026-01-22T08:03:50Z

Resolves: AVX2: Replace intrinsics implementation of poly_caddq with assembly #491

In this PR, we replace the AVX2 intrinsics implementation of poly_caddq with a x86_64 assembly version.
To estimate the performance impact, we compare the results shown in the two tables below.
Overall, for keypair, sign, and verify (opt), the performance difference is below 1%, which is consistent with the no-opt case.

In the component-level benchmark for mld_poly_caddq, the observed performance differences are at least 17%. After unrolling the loop by a factor of 4, the differences are reduced to approximately 10%.

bench components
- Δ (%) = (asm − AVX2) / AVX2 × 100

Component	Implementation	Build	ML-DSA-44	ML-DSA-65	ML-DSA-87	Notes
mld_poly_caddq (avg)	AVX2 intrinsics	no-opt	391	393	391
	x86_64 asm	no-opt	390	393	392
	Δ (%)	no-opt	-0.26	0.00	+0.26
mld_poly_caddq (avg)	AVX2 intrinsics	opt	38	40	39
	x86_64 asm	opt	51	50	46
	x86_64 asm (unroll)	opt	42	42	42	unroll by 4
	Δ (%)	opt	+34.21	+25.00	+17.95
	Δ (%) (unroll)	opt	+10.53	+5.00	+7.69	unroll by 4

bench
- Δ (%) = (asm − AVX2) / AVX2 × 100

Component	Implementation	Build	ML-DSA-44	ML-DSA-65	ML-DSA-87	Notes
keypair cycles (avg)	AVX2	no-opt	134355	226117	377069	baseline (main)
	x86_64 asm	no-opt	133831	226345	374963
	Δ (%)	no-opt	-0.39	+0.10	-0.56
	AVX2	opt	60367	105019	166676	baseline (main)
	x86_64 asm	opt	60535	104479	165781
	x86_64 asm (unroll)	opt	59921	104367	165795	unroll by 4
	Δ (%)	opt	+0.28	-0.51	-0.54
	Δ (%) (unroll)	opt	-0.74	-0.62	-0.53	unroll by 4
sign cycles (avg)	AVX2	no-opt	473892	779091	998026	baseline (main)
	x86_64 asm	no-opt	473262	779359	993245
	Δ (%)	no-opt	-0.13	+0.03	-0.48
	AVX2	opt	179804	301077	364509	baseline (main)
	x86_64 asm	opt	180253	298598	363742
	x86_64 asm (unroll)	opt	178255	299153	363505	unroll by 4
	Δ (%)	opt	+0.25	-0.82	-0.21
	Δ (%) (unroll)	opt	-0.86	-0.64	-0.28	unroll by 4
verify cycles (avg)	AVX2	no-opt	140765	228322	379244	baseline (main)
	x86_64 asm	no-opt	140872	228255	377091
	Δ (%)	no-opt	+0.08	-0.03	-0.57
	AVX2	opt	63674	105734	164897	baseline (main)
	x86_64 asm	opt	63924	105192	164131
	x86_64 asm (unroll)	opt	62955	105111	163861	unroll by 4
	Δ (%)	opt	+0.39	-0.51	-0.46
	Δ (%) (unroll)	opt	-1.13	-0.59	-0.63	unroll by 4

oqs-bot · 2026-01-23T07:09:45Z

CBMC Results (ML-DSA-87)

Full Results (175 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2638s	2484s	+6.2%
`sign_verify_internal`	✅	320s	313s	+2%
`mld_attempt_signature_generation`	✅	229s	218s	+5%
`polyvecl_pointwise_acc_montgomery_c`	✅	200s	183s	+9%
`polyvec_matrix_expand`	✅	169s	165s	+2%
`poly_pointwise_montgomery_c`	✅	156s	133s	+17%
`rej_uniform_native`	✅	149s	144s	+3%
`mld_invntt_layer`	✅	127s	122s	+4%
`polyvec_matrix_expand_serial`	✅	111s	108s	+3%
`mld_ct_memcmp`	✅	86s	81s	+6%
`sign_signature_internal`	✅	76s	73s	+4%
`keccak_squeezeblocks_x4`	✅	45s	43s	+5%
`mld_ntt_layer`	✅	45s	44s	+2%
`polymat_permute_bitrev_to_custom`	✅	27s	27s	+0%
`mld_compute_t0_t1_tr_from_sk_components`	✅	25s	25s	+0%
`rej_uniform`	✅	21s	20s	+5%
`rej_uniform_c`	✅	21s	18s	+17%
`fqmul`	✅	20s	18s	+11%
`poly_chknorm_c`	✅	17s	18s	-6%
`poly_uniform_eta_4x`	✅	17s	14s	+21%
`polyvec_matrix_pointwise_montgomery`	✅	17s	13s	+31%
`poly_uniform_4x`	✅	16s	16s	+0%
`polyeta_unpack`	✅	15s	12s	+25%
`polyt0_unpack`	✅	15s	15s	+0%
`polyveck_add`	✅	15s	15s	+0%
`keccakf1600x4_permute_native`	✅	14s	15s	-7%
`polyveck_power2round`	✅	14s	13s	+8%
`mld_check_pct`	✅	13s	11s	+18%
`mld_ntt_butterfly_block`	✅	12s	13s	-8%
`polyveck_use_hint`	✅	12s	10s	+20%
`keccak_absorb_once_x4`	✅	11s	11s	+0%
`mld_compute_pack_z`	✅	11s	11s	+0%
`polyveck_caddq`	✅	11s	8s	+38%
`polyveck_chknorm`	✅	11s	11s	+0%
`polyveck_reduce`	✅	11s	9s	+22%
`keccakf1600_permute`	✅	10s	5s	+100%
`poly_decompose_c`	✅	10s	8s	+25%
`keccakf1600_permute_native`	✅	9s	9s	+0%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	9s	8s	+12%
`polyveck_invntt_tomont`	✅	9s	7s	+29%
`sign`	✅	9s	7s	+29%
`sign_pk_from_sk`	✅	9s	10s	-10%
`poly_invntt_tomont_c`	✅	8s	7s	+14%
`polyveck_decompose`	✅	8s	8s	+0%
`polyveck_ntt`	✅	8s	6s	+33%
`polyveck_pointwise_poly_montgomery`	✅	8s	8s	+0%
`keccak_absorb`	✅	7s	6s	+17%
`mld_prepare_domain_separation_prefix`	✅	7s	5s	+40%
`mld_sample_s1_s2`	✅	7s	8s	-12%
`poly_chknorm_native`	✅	7s	3s	+133%
`poly_uniform_gamma1_4x`	✅	7s	2s	+250%
`polyveck_sub`	✅	7s	7s	+0%
`polyvecl_uniform_gamma1`	✅	7s	3s	+133%
`sign_keypair_internal`	✅	7s	6s	+17%
`sign_open`	✅	7s	3s	+133%
`sign_verify_pre_hash_internal`	✅	7s	3s	+133%
`mld_sample_s1_s2_serial`	✅	6s	8s	-25%
`polyveck_make_hint`	✅	6s	4s	+50%
`polyveck_pack_eta`	✅	6s	3s	+100%
`polyveck_shiftl`	✅	6s	6s	+0%
`polyvecl_ntt`	✅	6s	8s	-25%
`polyvecl_uniform_gamma1_serial`	✅	6s	3s	+100%
`polyz_unpack_c`	✅	6s	5s	+20%
`sign_signature_pre_hash_internal`	✅	6s	6s	+0%
`sign_signature_pre_hash_shake256`	✅	6s	6s	+0%
`sign_verify`	✅	6s	2s	+200%
`mld_h`	✅	5s	4s	+25%
`poly_add`	✅	5s	3s	+67%
`poly_challenge`	✅	5s	4s	+25%
`poly_decompose_native`	✅	5s	3s	+67%
`poly_invntt_tomont_native`	✅	5s	2s	+150%
`poly_ntt`	✅	5s	2s	+150%
`poly_ntt_c`	✅	5s	3s	+67%
`poly_use_hint_c`	✅	5s	2s	+150%
`polyveck_unpack_t0`	✅	5s	4s	+25%
`polyz_pack`	✅	5s	2s	+150%
`rej_eta_c`	✅	5s	4s	+25%
`rej_eta_native`	✅	5s	3s	+67%
`shake256x4_absorb_once`	✅	5s	4s	+25%
`sign_signature`	✅	5s	6s	-17%
`sign_signature_extmu`	✅	5s	6s	-17%
`unpack_pk`	✅	5s	5s	+0%
`unpack_sk`	✅	5s	8s	-38%
`fqscale`	✅	4s	2s	+100%
`keccak_finalize`	✅	4s	3s	+33%
`keccakf1600x4_xor_bytes`	✅	4s	4s	+0%
`mld_ct_get_optblocker_u8`	✅	4s	2s	+100%
`mld_ct_sel_int32`	✅	4s	1s	+300%
`ntt_native_x86_64`	✅	4s	5s	-20%
`pack_sig_c_h`	✅	4s	7s	-43%
`poly_caddq_c`	✅	4s	3s	+33%
`poly_caddq_native`	✅	4s	5s	-20%
`poly_make_hint`	✅	4s	2s	+100%
`poly_pointwise_montgomery`	✅	4s	2s	+100%
`poly_shiftl`	✅	4s	5s	-20%
`poly_uniform_eta`	✅	4s	5s	-20%
`poly_use_hint_native`	✅	4s	3s	+33%
`polyt0_pack`	✅	4s	4s	+0%
`polyt1_pack`	✅	4s	4s	+0%
`polyveck_pack_w1`	✅	4s	3s	+33%
`polyvecl_permute_bitrev_to_custom`	✅	4s	2s	+100%
`polyvecl_pointwise_acc_montgomery_native`	✅	4s	4s	+0%
`polyvecl_unpack_z`	✅	4s	6s	-33%
`reduce32`	✅	4s	2s	+100%
`shake128_init`	✅	4s	2s	+100%
`shake128_squeeze`	✅	4s	3s	+33%
`shake256_init`	✅	4s	3s	+33%
`sign_verify_pre_hash_shake256`	✅	4s	3s	+33%
`unpack_hints`	✅	4s	5s	-20%
`caddq`	✅	3s	4s	-25%
`intt_native_x86_64`	✅	3s	3s	+0%
`keccak_init`	✅	3s	2s	+50%
`keccakf1600_xor_bytes (big endian)`	✅	3s	2s	+50%
`keccakf1600x4_permute`	✅	3s	2s	+50%
`make_hint`	✅	3s	3s	+0%
`mld_ct_abs_i32`	✅	3s	3s	+0%
`mld_ct_cmask_neg_i32`	✅	3s	3s	+0%
`mld_ct_cmask_nonzero_u32`	✅	3s	3s	+0%
`mld_ct_cmask_nonzero_u8`	✅	3s	3s	+0%
`mld_ct_get_optblocker_u32`	✅	3s	3s	+0%
`mld_keccakf1600_extract_bytes`	✅	3s	2s	+50%
`pack_pk`	✅	3s	2s	+50%
`poly_decompose`	✅	3s	3s	+0%
`poly_invntt_tomont`	✅	3s	4s	-25%
`poly_ntt_native`	✅	3s	3s	+0%
`poly_pointwise_montgomery_native`	✅	3s	3s	+0%
`poly_power2round`	✅	3s	1s	+200%
`poly_reduce`	✅	3s	2s	+50%
`poly_sub`	✅	3s	4s	-25%
`poly_uniform`	✅	3s	2s	+50%
`polyeta_pack`	✅	3s	2s	+50%
`polyveck_pack_t0`	✅	3s	3s	+0%
`polyvecl_chknorm`	✅	3s	4s	-25%
`polyvecl_pointwise_acc_montgomery`	✅	3s	3s	+0%
`polyvecl_unpack_eta`	✅	3s	5s	-40%
`polyw1_pack`	✅	3s	4s	-25%
`polyz_unpack`	✅	3s	3s	+0%
`power2round`	✅	3s	2s	+50%
`shake128_absorb`	✅	3s	4s	-25%
`shake128x4_absorb_once`	✅	3s	3s	+0%
`shake256_squeeze`	✅	3s	4s	-25%
`sign_keypair`	✅	3s	4s	-25%
`sign_verify_extmu`	✅	3s	4s	-25%
`sys_check_capability`	✅	3s	2s	+50%
`use_hint`	✅	3s	3s	+0%
`keccak_squeeze`	✅	2s	4s	-50%
`keccakf1600_extract_bytes (big endian)`	✅	2s	3s	-33%
`keccakf1600_xor_bytes`	✅	2s	3s	-33%
`keccakf1600x4_extract_bytes`	✅	2s	2s	+0%
`mld_ct_get_optblocker_i64`	✅	2s	2s	+0%
`mld_value_barrier_u8`	✅	2s	2s	+0%
`montgomery_reduce`	✅	2s	4s	-50%
`pack_sig_z`	✅	2s	2s	+0%
`pack_sk`	✅	2s	3s	-33%
`poly_caddq`	✅	2s	4s	-50%
`poly_caddq_native_aarch64`	✅	2s	4s	-50%
`poly_chknorm`	✅	2s	2s	+0%
`poly_uniform_gamma1`	✅	2s	4s	-50%
`poly_use_hint`	✅	2s	4s	-50%
`polyt1_unpack`	✅	2s	3s	-33%
`polyveck_unpack_eta`	✅	2s	5s	-60%
`polyvecl_pack_eta`	✅	2s	3s	-33%
`rej_eta`	✅	2s	2s	+0%
`shake128_finalize`	✅	2s	2s	+0%
`shake128x4_squeezeblocks`	✅	2s	2s	+0%
`shake256`	✅	2s	3s	-33%
`shake256_absorb`	✅	2s	2s	+0%
`shake256_finalize`	✅	2s	2s	+0%
`unpack_sig`	✅	2s	5s	-60%
`decompose`	✅	1s	3s	-67%
`mld_value_barrier_i64`	✅	1s	2s	-50%
`mld_value_barrier_u32`	✅	1s	3s	-67%
`polyz_unpack_native`	✅	1s	3s	-67%
`shake128_release`	✅	1s	3s	-67%
`shake256_release`	✅	1s	2s	-50%
`shake256x4_squeezeblocks`	✅	1s	4s	-75%

oqs-bot · 2026-01-23T07:10:01Z

CBMC Results (ML-DSA-44)

Full Results (175 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2257s	2185s	+3.3%
`mld_attempt_signature_generation`	✅	338s	330s	+2%
`polyvecl_pointwise_acc_montgomery_c`	✅	229s	217s	+6%
`sign_verify_internal`	✅	200s	192s	+4%
`rej_uniform_native`	✅	154s	151s	+2%
`poly_pointwise_montgomery_c`	✅	150s	148s	+1%
`mld_invntt_layer`	✅	132s	124s	+6%
`mld_ct_memcmp`	✅	84s	88s	-5%
`mld_ntt_layer`	✅	49s	49s	+0%
`keccak_squeezeblocks_x4`	✅	45s	43s	+5%
`rej_uniform`	✅	23s	22s	+5%
`sign_signature_internal`	✅	22s	22s	+0%
`fqmul`	✅	21s	20s	+5%
`poly_uniform_eta_4x`	✅	21s	17s	+24%
`rej_uniform_c`	✅	20s	20s	+0%
`polymat_permute_bitrev_to_custom`	✅	17s	16s	+6%
`polyt0_unpack`	✅	17s	18s	-6%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	16s	15s	+7%
`polyvec_matrix_expand`	✅	16s	16s	+0%
`mld_compute_t0_t1_tr_from_sk_components`	✅	15s	15s	+0%
`polyeta_unpack`	✅	15s	15s	+0%
`keccak_absorb_once_x4`	✅	14s	14s	+0%
`keccakf1600x4_permute_native`	✅	13s	13s	+0%
`mld_ntt_butterfly_block`	✅	13s	13s	+0%
`poly_chknorm_c`	✅	13s	14s	-7%
`poly_uniform_4x`	✅	13s	13s	+0%
`polyz_unpack_c`	✅	12s	14s	-14%
`keccakf1600_permute_native`	✅	10s	9s	+11%
`poly_invntt_tomont_c`	✅	9s	8s	+12%
`polyveck_add`	✅	9s	9s	+0%
`polyveck_decompose`	✅	9s	5s	+80%
`sign`	✅	9s	9s	+0%
`keccakf1600_permute`	✅	8s	7s	+14%
`mld_check_pct`	✅	7s	8s	-12%
`mld_sample_s1_s2`	✅	7s	4s	+75%
`polyvec_matrix_expand_serial`	✅	7s	5s	+40%
`polyveck_pointwise_poly_montgomery`	✅	7s	7s	+0%
`polyvecl_chknorm`	✅	7s	5s	+40%
`polyvecl_ntt`	✅	7s	7s	+0%
`mld_compute_pack_z`	✅	6s	7s	-14%
`mld_ct_cmask_nonzero_u32`	✅	6s	3s	+100%
`mld_sample_s1_s2_serial`	✅	6s	4s	+50%
`poly_caddq`	✅	6s	3s	+100%
`poly_challenge`	✅	6s	4s	+50%
`polyt0_pack`	✅	6s	3s	+100%
`polyt1_pack`	✅	6s	3s	+100%
`polyvec_matrix_pointwise_montgomery`	✅	6s	6s	+0%
`polyveck_caddq`	✅	6s	5s	+20%
`polyveck_ntt`	✅	6s	7s	-14%
`polyveck_reduce`	✅	6s	5s	+20%
`polyveck_shiftl`	✅	6s	5s	+20%
`polyveck_use_hint`	✅	6s	6s	+0%
`polyvecl_pack_eta`	✅	6s	3s	+100%
`polyvecl_uniform_gamma1_serial`	✅	6s	4s	+50%
`rej_eta_c`	✅	6s	4s	+50%
`sign_signature_pre_hash_shake256`	✅	6s	6s	+0%
`sign_verify_pre_hash_internal`	✅	6s	3s	+100%
`keccak_absorb`	✅	5s	6s	-17%
`keccakf1600x4_permute`	✅	5s	2s	+150%
`mld_value_barrier_i64`	✅	5s	3s	+67%
`poly_add`	✅	5s	6s	-17%
`poly_caddq_c`	✅	5s	3s	+67%
`poly_reduce`	✅	5s	2s	+150%
`poly_shiftl`	✅	5s	3s	+67%
`poly_uniform`	✅	5s	5s	+0%
`poly_use_hint_c`	✅	5s	5s	+0%
`poly_use_hint_native`	✅	5s	4s	+25%
`polyveck_chknorm`	✅	5s	3s	+67%
`polyveck_make_hint`	✅	5s	5s	+0%
`polyveck_pack_t0`	✅	5s	3s	+67%
`polyveck_power2round`	✅	5s	5s	+0%
`polyvecl_permute_bitrev_to_custom`	✅	5s	4s	+25%
`reduce32`	✅	5s	2s	+150%
`rej_eta_native`	✅	5s	4s	+25%
`sign_pk_from_sk`	✅	5s	7s	-29%
`sign_signature`	✅	5s	6s	-17%
`sign_signature_pre_hash_internal`	✅	5s	3s	+67%
`sign_verify_pre_hash_shake256`	✅	5s	3s	+67%
`unpack_sig`	✅	5s	2s	+150%
`intt_native_x86_64`	✅	4s	4s	+0%
`keccak_squeeze`	✅	4s	2s	+100%
`mld_ct_get_optblocker_u8`	✅	4s	2s	+100%
`mld_h`	✅	4s	7s	-43%
`mld_keccakf1600_extract_bytes`	✅	4s	3s	+33%
`mld_prepare_domain_separation_prefix`	✅	4s	3s	+33%
`poly_decompose_c`	✅	4s	3s	+33%
`poly_invntt_tomont_native`	✅	4s	2s	+100%
`poly_ntt`	✅	4s	3s	+33%
`poly_pointwise_montgomery_native`	✅	4s	4s	+0%
`poly_uniform_eta`	✅	4s	4s	+0%
`poly_uniform_gamma1`	✅	4s	4s	+0%
`polyveck_invntt_tomont`	✅	4s	5s	-20%
`polyvecl_pointwise_acc_montgomery_native`	✅	4s	3s	+33%
`polyvecl_unpack_eta`	✅	4s	2s	+100%
`polyvecl_unpack_z`	✅	4s	3s	+33%
`polyz_unpack`	✅	4s	4s	+0%
`rej_eta`	✅	4s	4s	+0%
`shake256`	✅	4s	3s	+33%
`shake256x4_absorb_once`	✅	4s	3s	+33%
`sign_keypair`	✅	4s	5s	-20%
`sign_keypair_internal`	✅	4s	5s	-20%
`sign_open`	✅	4s	4s	+0%
`sign_signature_extmu`	✅	4s	4s	+0%
`sign_verify`	✅	4s	7s	-43%
`unpack_hints`	✅	4s	6s	-33%
`unpack_sk`	✅	4s	3s	+33%
`caddq`	✅	3s	3s	+0%
`keccak_finalize`	✅	3s	5s	-40%
`keccakf1600_extract_bytes (big endian)`	✅	3s	2s	+50%
`keccakf1600_xor_bytes (big endian)`	✅	3s	2s	+50%
`keccakf1600x4_extract_bytes`	✅	3s	2s	+50%
`keccakf1600x4_xor_bytes`	✅	3s	3s	+0%
`make_hint`	✅	3s	3s	+0%
`mld_ct_abs_i32`	✅	3s	2s	+50%
`mld_ct_cmask_neg_i32`	✅	3s	3s	+0%
`mld_ct_cmask_nonzero_u8`	✅	3s	3s	+0%
`montgomery_reduce`	✅	3s	3s	+0%
`ntt_native_x86_64`	✅	3s	4s	-25%
`pack_pk`	✅	3s	5s	-40%
`pack_sig_z`	✅	3s	4s	-25%
`pack_sk`	✅	3s	3s	+0%
`poly_decompose`	✅	3s	2s	+50%
`poly_decompose_native`	✅	3s	4s	-25%
`poly_ntt_native`	✅	3s	3s	+0%
`poly_pointwise_montgomery`	✅	3s	1s	+200%
`poly_power2round`	✅	3s	4s	-25%
`poly_sub`	✅	3s	1s	+200%
`poly_uniform_gamma1_4x`	✅	3s	4s	-25%
`poly_use_hint`	✅	3s	5s	-40%
`polyt1_unpack`	✅	3s	4s	-25%
`polyveck_unpack_eta`	✅	3s	3s	+0%
`polyvecl_uniform_gamma1`	✅	3s	4s	-25%
`polyz_pack`	✅	3s	4s	-25%
`polyz_unpack_native`	✅	3s	3s	+0%
`power2round`	✅	3s	2s	+50%
`shake128_squeeze`	✅	3s	2s	+50%
`shake128x4_absorb_once`	✅	3s	3s	+0%
`shake128x4_squeezeblocks`	✅	3s	3s	+0%
`shake256_squeeze`	✅	3s	2s	+50%
`shake256x4_squeezeblocks`	✅	3s	2s	+50%
`sign_verify_extmu`	✅	3s	3s	+0%
`sys_check_capability`	✅	3s	3s	+0%
`unpack_pk`	✅	3s	3s	+0%
`decompose`	✅	2s	3s	-33%
`fqscale`	✅	2s	4s	-50%
`keccak_init`	✅	2s	3s	-33%
`keccakf1600_xor_bytes`	✅	2s	3s	-33%
`mld_ct_get_optblocker_i64`	✅	2s	4s	-50%
`mld_ct_get_optblocker_u32`	✅	2s	1s	+100%
`mld_ct_sel_int32`	✅	2s	2s	+0%
`mld_value_barrier_u32`	✅	2s	2s	+0%
`mld_value_barrier_u8`	✅	2s	3s	-33%
`pack_sig_c_h`	✅	2s	3s	-33%
`poly_caddq_native`	✅	2s	4s	-50%
`poly_caddq_native_aarch64`	✅	2s	4s	-50%
`poly_chknorm`	✅	2s	4s	-50%
`poly_chknorm_native`	✅	2s	2s	+0%
`poly_invntt_tomont`	✅	2s	5s	-60%
`poly_make_hint`	✅	2s	3s	-33%
`poly_ntt_c`	✅	2s	1s	+100%
`polyeta_pack`	✅	2s	2s	+0%
`polyveck_pack_eta`	✅	2s	5s	-60%
`polyveck_pack_w1`	✅	2s	2s	+0%
`polyveck_sub`	✅	2s	5s	-60%
`polyveck_unpack_t0`	✅	2s	3s	-33%
`polyvecl_pointwise_acc_montgomery`	✅	2s	5s	-60%
`polyw1_pack`	✅	2s	4s	-50%
`shake128_finalize`	✅	2s	2s	+0%
`shake128_release`	✅	2s	1s	+100%
`shake256_finalize`	✅	2s	2s	+0%
`shake256_release`	✅	2s	3s	-33%
`use_hint`	✅	2s	2s	+0%
`shake128_absorb`	✅	1s	3s	-67%
`shake128_init`	✅	1s	2s	-50%
`shake256_absorb`	✅	1s	8s	-88%
`shake256_init`	✅	1s	2s	-50%

oqs-bot · 2026-01-23T07:10:48Z

CBMC Results (ML-DSA-65)

Full Results (175 proofs)

Proof	Status	Current	Previous	Change
`TOTAL`	✅	2537s	2317s	+9.5%
`polyvecl_pointwise_acc_montgomery_c`	✅	288s	221s	+30%
`mld_attempt_signature_generation`	✅	263s	231s	+14%
`sign_verify_internal`	✅	218s	200s	+9%
`poly_pointwise_montgomery_c`	✅	171s	140s	+22%
`rej_uniform_native`	✅	159s	140s	+14%
`mld_invntt_layer`	✅	135s	119s	+13%
`polyvec_matrix_expand`	✅	123s	114s	+8%
`mld_ct_memcmp`	✅	100s	81s	+23%
`polyvec_matrix_expand_serial`	✅	69s	66s	+5%
`mld_ntt_layer`	✅	48s	45s	+7%
`keccak_squeezeblocks_x4`	✅	46s	44s	+5%
`sign_signature_internal`	✅	46s	43s	+7%
`mld_compute_t0_t1_tr_from_sk_components`	✅	27s	25s	+8%
`fqmul`	✅	21s	20s	+5%
`rej_uniform`	✅	21s	20s	+5%
`rej_uniform_c`	✅	21s	17s	+24%
`poly_chknorm_c`	✅	19s	16s	+19%
`poly_uniform_eta_4x`	✅	19s	16s	+19%
`polyveck_decompose`	✅	19s	17s	+12%
`polymat_permute_bitrev_to_custom`	✅	18s	19s	-5%
`mld_polyvecl_permute_bitrev_to_custom_native`	✅	17s	13s	+31%
`polyt0_unpack`	✅	17s	17s	+0%
`poly_uniform_4x`	✅	16s	14s	+14%
`polyvec_matrix_pointwise_montgomery`	✅	15s	13s	+15%
`mld_ntt_butterfly_block`	✅	14s	12s	+17%
`keccakf1600x4_permute_native`	✅	13s	12s	+8%
`keccak_absorb_once_x4`	✅	12s	12s	+0%
`polyveck_use_hint`	✅	12s	11s	+9%
`poly_decompose_c`	✅	10s	7s	+43%
`polyeta_unpack`	✅	10s	6s	+67%
`polyveck_reduce`	✅	10s	8s	+25%
`keccakf1600_permute`	✅	9s	10s	-10%
`poly_invntt_tomont_c`	✅	9s	7s	+29%
`polyveck_invntt_tomont`	✅	9s	9s	+0%
`polyveck_power2round`	✅	9s	8s	+12%
`sign`	✅	9s	10s	-10%
`keccakf1600_permute_native`	✅	8s	9s	-11%
`mld_sample_s1_s2_serial`	✅	8s	6s	+33%
`polyveck_add`	✅	8s	11s	-27%
`polyveck_caddq`	✅	8s	6s	+33%
`polyveck_shiftl`	✅	8s	6s	+33%
`polyveck_sub`	✅	8s	8s	+0%
`polyvecl_pointwise_acc_montgomery_native`	✅	8s	4s	+100%
`polyveck_chknorm`	✅	7s	5s	+40%
`polyveck_ntt`	✅	7s	8s	-12%
`polyveck_pointwise_poly_montgomery`	✅	7s	6s	+17%
`polyvecl_ntt`	✅	7s	8s	-12%
`sign_keypair_internal`	✅	7s	8s	-12%
`sign_verify_pre_hash_shake256`	✅	7s	4s	+75%
`mld_check_pct`	✅	6s	8s	-25%
`mld_h`	✅	6s	5s	+20%
`mld_sample_s1_s2`	✅	6s	5s	+20%
`poly_add`	✅	6s	4s	+50%
`poly_chknorm`	✅	6s	5s	+20%
`sign_pk_from_sk`	✅	6s	5s	+20%
`sign_verify_extmu`	✅	6s	6s	+0%
`caddq`	✅	5s	2s	+150%
`keccak_absorb`	✅	5s	6s	-17%
`mld_compute_pack_z`	✅	5s	5s	+0%
`ntt_native_x86_64`	✅	5s	4s	+25%
`poly_caddq_c`	✅	5s	4s	+25%
`poly_decompose_native`	✅	5s	6s	-17%
`poly_uniform_eta`	✅	5s	6s	-17%
`poly_uniform_gamma1_4x`	✅	5s	3s	+67%
`polyveck_make_hint`	✅	5s	5s	+0%
`polyvecl_chknorm`	✅	5s	5s	+0%
`polyvecl_pointwise_acc_montgomery`	✅	5s	6s	-17%
`polyvecl_uniform_gamma1`	✅	5s	3s	+67%
`shake256_release`	✅	5s	3s	+67%
`shake256x4_squeezeblocks`	✅	5s	2s	+150%
`sign_signature_pre_hash_internal`	✅	5s	4s	+25%
`sign_signature_pre_hash_shake256`	✅	5s	3s	+67%
`sign_verify`	✅	5s	2s	+150%
`sys_check_capability`	✅	5s	2s	+150%
`keccakf1600x4_extract_bytes`	✅	4s	3s	+33%
`keccakf1600x4_permute`	✅	4s	3s	+33%
`mld_ct_get_optblocker_u32`	✅	4s	2s	+100%
`mld_keccakf1600_extract_bytes`	✅	4s	2s	+100%
`poly_challenge`	✅	4s	5s	-20%
`poly_chknorm_native`	✅	4s	4s	+0%
`poly_make_hint`	✅	4s	4s	+0%
`poly_pointwise_montgomery`	✅	4s	6s	-33%
`poly_uniform_gamma1`	✅	4s	5s	-20%
`poly_use_hint_c`	✅	4s	6s	-33%
`polyeta_pack`	✅	4s	4s	+0%
`polyt0_pack`	✅	4s	5s	-20%
`polyveck_pack_eta`	✅	4s	4s	+0%
`polyveck_pack_t0`	✅	4s	7s	-43%
`polyz_unpack_c`	✅	4s	7s	-43%
`reduce32`	✅	4s	2s	+100%
`rej_eta_c`	✅	4s	5s	-20%
`shake128_init`	✅	4s	4s	+0%
`shake128x4_absorb_once`	✅	4s	3s	+33%
`sign_open`	✅	4s	4s	+0%
`sign_signature`	✅	4s	3s	+33%
`unpack_hints`	✅	4s	5s	-20%
`unpack_sk`	✅	4s	6s	-33%
`use_hint`	✅	4s	1s	+300%
`fqscale`	✅	3s	2s	+50%
`keccak_finalize`	✅	3s	2s	+50%
`make_hint`	✅	3s	4s	-25%
`mld_ct_cmask_nonzero_u32`	✅	3s	3s	+0%
`mld_ct_cmask_nonzero_u8`	✅	3s	4s	-25%
`mld_ct_get_optblocker_u8`	✅	3s	2s	+50%
`mld_prepare_domain_separation_prefix`	✅	3s	2s	+50%
`montgomery_reduce`	✅	3s	2s	+50%
`pack_pk`	✅	3s	3s	+0%
`pack_sig_c_h`	✅	3s	4s	-25%
`pack_sig_z`	✅	3s	2s	+50%
`poly_caddq`	✅	3s	4s	-25%
`poly_caddq_native`	✅	3s	2s	+50%
`poly_invntt_tomont_native`	✅	3s	3s	+0%
`poly_ntt`	✅	3s	5s	-40%
`poly_ntt_c`	✅	3s	3s	+0%
`poly_power2round`	✅	3s	6s	-50%
`poly_shiftl`	✅	3s	5s	-40%
`poly_sub`	✅	3s	3s	+0%
`poly_uniform`	✅	3s	2s	+50%
`polyt1_pack`	✅	3s	3s	+0%
`polyveck_pack_w1`	✅	3s	4s	-25%
`polyveck_unpack_eta`	✅	3s	4s	-25%
`polyvecl_pack_eta`	✅	3s	5s	-40%
`polyvecl_permute_bitrev_to_custom`	✅	3s	4s	-25%
`polyvecl_unpack_eta`	✅	3s	3s	+0%
`polyvecl_unpack_z`	✅	3s	4s	-25%
`polyw1_pack`	✅	3s	3s	+0%
`polyz_unpack`	✅	3s	6s	-50%
`polyz_unpack_native`	✅	3s	4s	-25%
`power2round`	✅	3s	3s	+0%
`rej_eta_native`	✅	3s	4s	-25%
`shake128_finalize`	✅	3s	4s	-25%
`shake128_release`	✅	3s	2s	+50%
`shake128_squeeze`	✅	3s	2s	+50%
`shake256_absorb`	✅	3s	3s	+0%
`shake256_finalize`	✅	3s	4s	-25%
`shake256x4_absorb_once`	✅	3s	4s	-25%
`sign_keypair`	✅	3s	3s	+0%
`sign_signature_extmu`	✅	3s	3s	+0%
`sign_verify_pre_hash_internal`	✅	3s	5s	-40%
`unpack_pk`	✅	3s	2s	+50%
`intt_native_x86_64`	✅	2s	4s	-50%
`keccak_init`	✅	2s	2s	+0%
`keccakf1600_extract_bytes (big endian)`	✅	2s	3s	-33%
`keccakf1600_xor_bytes`	✅	2s	4s	-50%
`keccakf1600_xor_bytes (big endian)`	✅	2s	2s	+0%
`keccakf1600x4_xor_bytes`	✅	2s	3s	-33%
`mld_ct_cmask_neg_i32`	✅	2s	2s	+0%
`mld_ct_get_optblocker_i64`	✅	2s	3s	-33%
`mld_value_barrier_i64`	✅	2s	3s	-33%
`mld_value_barrier_u32`	✅	2s	4s	-50%
`mld_value_barrier_u8`	✅	2s	1s	+100%
`pack_sk`	✅	2s	3s	-33%
`poly_caddq_native_aarch64`	✅	2s	6s	-67%
`poly_decompose`	✅	2s	4s	-50%
`poly_invntt_tomont`	✅	2s	2s	+0%
`poly_ntt_native`	✅	2s	4s	-50%
`poly_pointwise_montgomery_native`	✅	2s	3s	-33%
`poly_reduce`	✅	2s	3s	-33%
`poly_use_hint`	✅	2s	6s	-67%
`poly_use_hint_native`	✅	2s	4s	-50%
`polyt1_unpack`	✅	2s	3s	-33%
`polyveck_unpack_t0`	✅	2s	4s	-50%
`polyvecl_uniform_gamma1_serial`	✅	2s	3s	-33%
`polyz_pack`	✅	2s	3s	-33%
`shake256`	✅	2s	4s	-50%
`shake256_init`	✅	2s	3s	-33%
`shake256_squeeze`	✅	2s	2s	+0%
`unpack_sig`	✅	2s	3s	-33%
`decompose`	✅	1s	3s	-67%
`keccak_squeeze`	✅	1s	5s	-80%
`mld_ct_abs_i32`	✅	1s	2s	-50%
`mld_ct_sel_int32`	✅	1s	4s	-75%
`rej_eta`	✅	1s	3s	-67%
`shake128_absorb`	✅	1s	3s	-67%
`shake128x4_squeezeblocks`	✅	1s	3s	-67%

github-actions

Mac Mini (M1, 2020) benchmarks (opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`45681` cycles	`45685` cycles	`1.00`
`ML-DSA-44 sign`	`131153` cycles	`131164` cycles	`1.00`
`ML-DSA-44 verify`	`47527` cycles	`47530` cycles	`1.00`
`ML-DSA-65 keypair`	`80457` cycles	`80479` cycles	`1.00`
`ML-DSA-65 sign`	`215715` cycles	`215740` cycles	`1.00`
`ML-DSA-65 verify`	`79737` cycles	`79735` cycles	`1.00`
`ML-DSA-87 keypair`	`131177` cycles	`131175` cycles	`1.00`
`ML-DSA-87 sign`	`277048` cycles	`277004` cycles	`1.00`
`ML-DSA-87 verify`	`130004` cycles	`129971` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Mac Mini (M1, 2020) benchmarks (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`111983` cycles	`111979` cycles	`1.00`
`ML-DSA-44 sign`	`403592` cycles	`403622` cycles	`1.00`
`ML-DSA-44 verify`	`119886` cycles	`119876` cycles	`1.00`
`ML-DSA-65 keypair`	`192137` cycles	`192166` cycles	`1.00`
`ML-DSA-65 sign`	`657120` cycles	`657078` cycles	`1.00`
`ML-DSA-65 verify`	`193900` cycles	`193891` cycles	`1.00`
`ML-DSA-87 keypair`	`317930` cycles	`318010` cycles	`1.00`
`ML-DSA-87 sign`	`836905` cycles	`836903` cycles	`1.00`
`ML-DSA-87 verify`	`322922` cycles	`322994` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Intel Xeon 4th gen (c7i)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`34340` cycles	`34361` cycles	`1.00`
`ML-DSA-44 sign`	`119648` cycles	`120023` cycles	`1.00`
`ML-DSA-44 verify`	`37990` cycles	`38140` cycles	`1.00`
`ML-DSA-65 keypair`	`60562` cycles	`60626` cycles	`1.00`
`ML-DSA-65 sign`	`201239` cycles	`200228` cycles	`1.01`
`ML-DSA-65 verify`	`62873` cycles	`62578` cycles	`1.00`
`ML-DSA-87 keypair`	`93377` cycles	`93913` cycles	`0.99`
`ML-DSA-87 sign`	`232229` cycles	`235482` cycles	`0.99`
`ML-DSA-87 verify`	`94479` cycles	`94514` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A72 (Raspberry Pi 4) benchmarks (opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`229063` cycles	`232745` cycles	`0.98`
`ML-DSA-44 sign`	`628858` cycles	`629812` cycles	`1.00`
`ML-DSA-44 verify`	`229339` cycles	`229277` cycles	`1.00`
`ML-DSA-65 keypair`	`378941` cycles	`422090` cycles	`0.90`
`ML-DSA-65 sign`	`1007370` cycles	`1067756` cycles	`0.94`
`ML-DSA-65 verify`	`376246` cycles	`393848` cycles	`0.96`
`ML-DSA-87 keypair`	`690237` cycles	`673725` cycles	`1.02`
`ML-DSA-87 sign`	`1396068` cycles	`1405386` cycles	`0.99`
`ML-DSA-87 verify`	`663094` cycles	`657567` cycles	`1.01`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Intel Xeon 4th gen (c7i) (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`93562` cycles	`93808` cycles	`1.00`
`ML-DSA-44 sign`	`332581` cycles	`332528` cycles	`1.00`
`ML-DSA-44 verify`	`99714` cycles	`99696` cycles	`1.00`
`ML-DSA-65 keypair`	`159833` cycles	`160037` cycles	`1.00`
`ML-DSA-65 sign`	`543737` cycles	`544483` cycles	`1.00`
`ML-DSA-65 verify`	`160524` cycles	`160826` cycles	`1.00`
`ML-DSA-87 keypair`	`267186` cycles	`266702` cycles	`1.00`
`ML-DSA-87 sign`	`707232` cycles	`705628` cycles	`1.00`
`ML-DSA-87 verify`	`270355` cycles	`270568` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

AMD EPYC 3rd gen (c6a)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`68896` cycles	`69270` cycles	`0.99`
`ML-DSA-44 sign`	`187431` cycles	`187049` cycles	`1.00`
`ML-DSA-44 verify`	`68887` cycles	`69047` cycles	`1.00`
`ML-DSA-65 keypair`	`119600` cycles	`119031` cycles	`1.00`
`ML-DSA-65 sign`	`299540` cycles	`299818` cycles	`1.00`
`ML-DSA-65 verify`	`115518` cycles	`115291` cycles	`1.00`
`ML-DSA-87 keypair`	`203742` cycles	`203891` cycles	`1.00`
`ML-DSA-87 sign`	`393131` cycles	`394659` cycles	`1.00`
`ML-DSA-87 verify`	`195707` cycles	`195766` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Intel Xeon 3rd gen (c6i)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`57327` cycles	`56563` cycles	`1.01`
`ML-DSA-44 sign`	`180726` cycles	`181874` cycles	`0.99`
`ML-DSA-44 verify`	`60901` cycles	`61156` cycles	`1.00`
`ML-DSA-65 keypair`	`98660` cycles	`98757` cycles	`1.00`
`ML-DSA-65 sign`	`298138` cycles	`298537` cycles	`1.00`
`ML-DSA-65 verify`	`100095` cycles	`100518` cycles	`1.00`
`ML-DSA-87 keypair`	`152331` cycles	`152679` cycles	`1.00`
`ML-DSA-87 sign`	`355616` cycles	`355558` cycles	`1.00`
`ML-DSA-87 verify`	`154183` cycles	`152966` cycles	`1.01`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

AMD EPYC 4th gen (c7a)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`41477` cycles	`41136` cycles	`1.01`
`ML-DSA-44 sign`	`133286` cycles	`132617` cycles	`1.01`
`ML-DSA-44 verify`	`44031` cycles	`44492` cycles	`0.99`
`ML-DSA-65 keypair`	`72317` cycles	`72104` cycles	`1.00`
`ML-DSA-65 sign`	`213181` cycles	`214651` cycles	`0.99`
`ML-DSA-65 verify`	`71974` cycles	`72444` cycles	`0.99`
`ML-DSA-87 keypair`	`107833` cycles	`107657` cycles	`1.00`
`ML-DSA-87 sign`	`250476` cycles	`250266` cycles	`1.00`
`ML-DSA-87 verify`	`109230` cycles	`112595` cycles	`0.97`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'AMD EPYC 4th gen (c7a)'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.03.

Benchmark suite	Current: `3819863`	Previous: `9258ea1`	Ratio
`ML-DSA-65 keypair`	`75829` cycles	`72591` cycles	`1.04`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

AMD EPYC 3rd gen (c6a) (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`134758` cycles	`134710` cycles	`1.00`
`ML-DSA-44 sign`	`523723` cycles	`526054` cycles	`1.00`
`ML-DSA-44 verify`	`147705` cycles	`147500` cycles	`1.00`
`ML-DSA-65 keypair`	`226449` cycles	`226690` cycles	`1.00`
`ML-DSA-65 sign`	`860712` cycles	`861192` cycles	`1.00`
`ML-DSA-65 verify`	`235070` cycles	`235381` cycles	`1.00`
`ML-DSA-87 keypair`	`370974` cycles	`370668` cycles	`1.00`
`ML-DSA-87 sign`	`1079141` cycles	`1078305` cycles	`1.00`
`ML-DSA-87 verify`	`383049` cycles	`383429` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Intel Xeon 3rd gen (c6i) (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`157394` cycles	`157188` cycles	`1.00`
`ML-DSA-44 sign`	`549561` cycles	`548996` cycles	`1.00`
`ML-DSA-44 verify`	`169498` cycles	`169283` cycles	`1.00`
`ML-DSA-65 keypair`	`267800` cycles	`269077` cycles	`1.00`
`ML-DSA-65 sign`	`903011` cycles	`906033` cycles	`1.00`
`ML-DSA-65 verify`	`273909` cycles	`275229` cycles	`1.00`
`ML-DSA-87 keypair`	`449680` cycles	`448040` cycles	`1.00`
`ML-DSA-87 sign`	`1161535` cycles	`1157923` cycles	`1.00`
`ML-DSA-87 verify`	`460234` cycles	`457343` cycles	`1.01`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton4

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`67942` cycles	`68235` cycles	`1.00`
`ML-DSA-44 sign`	`201998` cycles	`201899` cycles	`1.00`
`ML-DSA-44 verify`	`70776` cycles	`70799` cycles	`1.00`
`ML-DSA-65 keypair`	`121036` cycles	`121045` cycles	`1.00`
`ML-DSA-65 sign`	`331322` cycles	`331301` cycles	`1.00`
`ML-DSA-65 verify`	`117850` cycles	`117988` cycles	`1.00`
`ML-DSA-87 keypair`	`198669` cycles	`197907` cycles	`1.00`
`ML-DSA-87 sign`	`428529` cycles	`426619` cycles	`1.00`
`ML-DSA-87 verify`	`194582` cycles	`194362` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton3

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`72254` cycles	`72298` cycles	`1.00`
`ML-DSA-44 sign`	`211942` cycles	`211862` cycles	`1.00`
`ML-DSA-44 verify`	`75645` cycles	`75651` cycles	`1.00`
`ML-DSA-65 keypair`	`127516` cycles	`127564` cycles	`1.00`
`ML-DSA-65 sign`	`350254` cycles	`350256` cycles	`1.00`
`ML-DSA-65 verify`	`125449` cycles	`125447` cycles	`1.00`
`ML-DSA-87 keypair`	`208196` cycles	`208014` cycles	`1.00`
`ML-DSA-87 sign`	`448893` cycles	`448910` cycles	`1.00`
`ML-DSA-87 verify`	`205308` cycles	`205681` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

AMD EPYC 4th gen (c7a) (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`120311` cycles	`120460` cycles	`1.00`
`ML-DSA-44 sign`	`447300` cycles	`447234` cycles	`1.00`
`ML-DSA-44 verify`	`129710` cycles	`130455` cycles	`0.99`
`ML-DSA-65 keypair`	`204437` cycles	`203981` cycles	`1.00`
`ML-DSA-65 sign`	`728421` cycles	`730686` cycles	`1.00`
`ML-DSA-65 verify`	`209421` cycles	`210398` cycles	`1.00`
`ML-DSA-87 keypair`	`337688` cycles	`337748` cycles	`1.00`
`ML-DSA-87 sign`	`926903` cycles	`922242` cycles	`1.01`
`ML-DSA-87 verify`	`346060` cycles	`347109` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton4 (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`128224` cycles	`128233` cycles	`1.00`
`ML-DSA-44 sign`	`447684` cycles	`447406` cycles	`1.00`
`ML-DSA-44 verify`	`138331` cycles	`142161` cycles	`0.97`
`ML-DSA-65 keypair`	`220728` cycles	`220585` cycles	`1.00`
`ML-DSA-65 sign`	`727613` cycles	`726570` cycles	`1.00`
`ML-DSA-65 verify`	`223172` cycles	`223096` cycles	`1.00`
`ML-DSA-87 keypair`	`365009` cycles	`365027` cycles	`1.00`
`ML-DSA-87 sign`	`926270` cycles	`926682` cycles	`1.00`
`ML-DSA-87 verify`	`372774` cycles	`372462` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton3 (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`138503` cycles	`138431` cycles	`1.00`
`ML-DSA-44 sign`	`484053` cycles	`483804` cycles	`1.00`
`ML-DSA-44 verify`	`148725` cycles	`156357` cycles	`0.95`
`ML-DSA-65 keypair`	`241276` cycles	`241178` cycles	`1.00`
`ML-DSA-65 sign`	`792427` cycles	`792015` cycles	`1.00`
`ML-DSA-65 verify`	`241215` cycles	`241086` cycles	`1.00`
`ML-DSA-87 keypair`	`396496` cycles	`396336` cycles	`1.00`
`ML-DSA-87 sign`	`1013013` cycles	`1012796` cycles	`1.00`
`ML-DSA-87 verify`	`402599` cycles	`402305` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton2

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`113820` cycles	`113345` cycles	`1.00`
`ML-DSA-44 sign`	`357341` cycles	`356055` cycles	`1.00`
`ML-DSA-44 verify`	`118529` cycles	`118038` cycles	`1.00`
`ML-DSA-65 keypair`	`196679` cycles	`196907` cycles	`1.00`
`ML-DSA-65 sign`	`588785` cycles	`590403` cycles	`1.00`
`ML-DSA-65 verify`	`194716` cycles	`195057` cycles	`1.00`
`ML-DSA-87 keypair`	`323237` cycles	`322985` cycles	`1.00`
`ML-DSA-87 sign`	`754039` cycles	`753517` cycles	`1.00`
`ML-DSA-87 verify`	`320375` cycles	`320636` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

SpacemiT K1 8 (Banana Pi F3) benchmarks (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`827476` cycles	`828088` cycles	`1.00`
`ML-DSA-44 sign`	`3238353` cycles	`3233170` cycles	`1.00`
`ML-DSA-44 verify`	`921919` cycles	`920794` cycles	`1.00`
`ML-DSA-65 keypair`	`1413613` cycles	`1413452` cycles	`1.00`
`ML-DSA-65 sign`	`5340696` cycles	`5347688` cycles	`1.00`
`ML-DSA-65 verify`	`1477470` cycles	`1477937` cycles	`1.00`
`ML-DSA-87 keypair`	`2311391` cycles	`2312894` cycles	`1.00`
`ML-DSA-87 sign`	`6659117` cycles	`6665352` cycles	`1.00`
`ML-DSA-87 verify`	`2409640` cycles	`2411069` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

Graviton2 (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`214003` cycles	`213077` cycles	`1.00`
`ML-DSA-44 sign`	`765036` cycles	`760523` cycles	`1.01`
`ML-DSA-44 verify`	`230465` cycles	`233125` cycles	`0.99`
`ML-DSA-65 keypair`	`380442` cycles	`380915` cycles	`1.00`
`ML-DSA-65 sign`	`1253729` cycles	`1251999` cycles	`1.00`
`ML-DSA-65 verify`	`371997` cycles	`372378` cycles	`1.00`
`ML-DSA-87 keypair`	`604923` cycles	`605968` cycles	`1.00`
`ML-DSA-87 sign`	`1594853` cycles	`1593941` cycles	`1.00`
`ML-DSA-87 verify`	`619102` cycles	`617894` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A72 (Raspberry Pi 4) benchmarks (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`311698` cycles	`306606` cycles	`1.02`
`ML-DSA-44 sign`	`1174058` cycles	`1166146` cycles	`1.01`
`ML-DSA-44 verify`	`333560` cycles	`335430` cycles	`0.99`
`ML-DSA-65 keypair`	`550737` cycles	`562274` cycles	`0.98`
`ML-DSA-65 sign`	`1894590` cycles	`1916493` cycles	`0.99`
`ML-DSA-65 verify`	`529438` cycles	`533535` cycles	`0.99`
`ML-DSA-87 keypair`	`872695` cycles	`865006` cycles	`1.01`
`ML-DSA-87 sign`	`2468410` cycles	`2417913` cycles	`1.02`
`ML-DSA-87 verify`	`900121` cycles	`884966` cycles	`1.02`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Arm Cortex-A72 (Raspberry Pi 4) benchmarks (no-opt)'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.03.

Benchmark suite	Current: `f9a6d30`	Previous: `9258ea1`	Ratio
`ML-DSA-44 keypair`	`309195` cycles	`299195` cycles	`1.03`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A55 (Snapdragon 888) benchmarks (opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`277182` cycles	`278160` cycles	`1.00`
`ML-DSA-44 sign`	`816109` cycles	`822535` cycles	`0.99`
`ML-DSA-44 verify`	`280990` cycles	`278070` cycles	`1.01`
`ML-DSA-65 keypair`	`477648` cycles	`476503` cycles	`1.00`
`ML-DSA-65 sign`	`1398700` cycles	`1347085` cycles	`1.04`
`ML-DSA-65 verify`	`461181` cycles	`456015` cycles	`1.01`
`ML-DSA-87 keypair`	`825204` cycles	`796551` cycles	`1.04`
`ML-DSA-87 sign`	`1886968` cycles	`1773335` cycles	`1.06`
`ML-DSA-87 verify`	`803609` cycles	`772360` cycles	`1.04`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A76 (Raspberry Pi 5) benchmarks (opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`113150` cycles	`113204` cycles	`1.00`
`ML-DSA-44 sign`	`355525` cycles	`355548` cycles	`1.00`
`ML-DSA-44 verify`	`117877` cycles	`117886` cycles	`1.00`
`ML-DSA-65 keypair`	`196192` cycles	`196406` cycles	`1.00`
`ML-DSA-65 sign`	`588774` cycles	`588666` cycles	`1.00`
`ML-DSA-65 verify`	`194576` cycles	`194481` cycles	`1.00`
`ML-DSA-87 keypair`	`322391` cycles	`321917` cycles	`1.00`
`ML-DSA-87 sign`	`751848` cycles	`752728` cycles	`1.00`
`ML-DSA-87 verify`	`319927` cycles	`320132` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A76 (Raspberry Pi 5) benchmarks (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`212745` cycles	`212659` cycles	`1.00`
`ML-DSA-44 sign`	`759537` cycles	`759393` cycles	`1.00`
`ML-DSA-44 verify`	`229014` cycles	`228980` cycles	`1.00`
`ML-DSA-65 keypair`	`380339` cycles	`380359` cycles	`1.00`
`ML-DSA-65 sign`	`1251422` cycles	`1251433` cycles	`1.00`
`ML-DSA-65 verify`	`372106` cycles	`372151` cycles	`1.00`
`ML-DSA-87 keypair`	`605932` cycles	`605385` cycles	`1.00`
`ML-DSA-87 sign`	`1591645` cycles	`1591182` cycles	`1.00`
`ML-DSA-87 verify`	`617975` cycles	`617388` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

oqs-bot

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Graviton2 (no-opt)'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.03.

Benchmark suite	Current: `f9a6d30`	Previous: `9258ea1`	Ratio
`ML-DSA-44 verify`	`241958` cycles	`229196` cycles	`1.06`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

Arm Cortex-A55 (Snapdragon 888) benchmarks (no-opt)

Details

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-44 keypair`	`464439` cycles	`465373` cycles	`1.00`
`ML-DSA-44 sign`	`2143725` cycles	`2152438` cycles	`1.00`
`ML-DSA-44 verify`	`551422` cycles	`551474` cycles	`1.00`
`ML-DSA-65 keypair`	`783189` cycles	`781420` cycles	`1.00`
`ML-DSA-65 sign`	`3519184` cycles	`3519262` cycles	`1.00`
`ML-DSA-65 verify`	`855778` cycles	`854831` cycles	`1.00`
`ML-DSA-87 keypair`	`1261568` cycles	`1263149` cycles	`1.00`
`ML-DSA-87 sign`	`4343372` cycles	`4339952` cycles	`1.00`
`ML-DSA-87 verify`	`1377900` cycles	`1379633` cycles	`1.00`

This comment was automatically generated by workflow using github-action-benchmark.

github-actions

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Arm Cortex-A72 (Raspberry Pi 4) benchmarks (opt)'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.03.

Benchmark suite	Current: `f9a6d30`	Previous: `9258ea1`	Ratio
`ML-DSA-44 keypair`	`237495` cycles	`229189` cycles	`1.04`

This comment was automatically generated by workflow using github-action-benchmark.

mkannwischer

Thanks @willieyz. Performance is looking good and I checked that the code is doing the correct thing. Here are a few stylistic comments.

mkannwischer · 2026-02-08T05:58:38Z

dev/x86_64/src/poly_caddq_avx2.S

+.balign 16
+MLD_ASM_FN_SYMBOL(poly_caddq_avx2)
+
+movabsq	$35993616950222849, %rdx


Why are you using a 64-bit constant here? It would be much easier to follow if you take a 32-bit one:

mov $8380417, %edx vmovd %edx, %xmm1 vpbroadcastd %xmm1, %ymm1

Unless that is slower, it should be prefered.

mkannwischer · 2026-02-08T05:59:00Z

dev/x86_64/src/poly_caddq_avx2.S

+addq   $128, %rdi        # advance by 128 bytes (4 vectors)
+cmpq   %rdi, %rax
+jne    poly_caddq_avx2_loop   # 8 iterations (32/4 = 8)
+vzeroupper


We never use vzeroupper in any other AVX2 files, so this should be eliminated.

mkannwischer · 2026-02-08T06:00:48Z

dev/x86_64/src/poly_caddq_avx2.S

+vpcmpgtd (%rdi), %ymm2, %ymm0
+vpand   %ymm1, %ymm0, %ymm0  
+vpaddd  (%rdi), %ymm0, %ymm0
+vmovdqa  %ymm0, (%rdi)


Please wrap this in a macro - similar to the caddq code for aarch64.

mkannwischer · 2026-02-08T06:16:28Z

dev/x86_64/src/poly_caddq_avx2.S

+vpaddd  96(%rdi), %ymm5, %ymm5
+vmovdqa  %ymm5, 96(%rdi)
+
+addq   $128, %rdi        # advance by 128 bytes (4 vectors)


We never use # comments.
Please use // comments or /* */ comments like in the other files. Also try to follow their style.

This commit adds mld_poly_caddq to the benchmark components to evaluate the performance impact of replacing the caddq AVX2 intrinsics with x86_64 assembly code. Signed-off-by: willieyz <willie.zhao@chelpis.com>

This commit replace the eurrently caddq AVX2 implementation to x86_64 assembly code. Signed-off-by: willieyz <willie.zhao@chelpis.com>

Signed-off-by: willieyz <willie.zhao@chelpis.com>

github-actions

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Arm Cortex-A55 (Snapdragon 888) benchmarks (opt)'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.03.

Benchmark suite	Current: `14097e6`	Previous: `daf271b`	Ratio
`ML-DSA-65 sign`	`1398700` cycles	`1347085` cycles	`1.04`
`ML-DSA-87 keypair`	`825204` cycles	`796551` cycles	`1.04`
`ML-DSA-87 sign`	`1886968` cycles	`1773335` cycles	`1.06`
`ML-DSA-87 verify`	`803609` cycles	`772360` cycles	`1.04`

This comment was automatically generated by workflow using github-action-benchmark.

willieyz force-pushed the eliminate-caddq-intrinsics branch from 00b155f to 3819863 Compare January 23, 2026 06:52

willieyz added the benchmark label Jan 26, 2026

github-actions bot reviewed Jan 26, 2026

View reviewed changes

oqs-bot reviewed Jan 26, 2026

View reviewed changes

github-actions bot reviewed Jan 26, 2026

View reviewed changes

oqs-bot reviewed Jan 26, 2026

View reviewed changes

github-actions bot reviewed Jan 26, 2026

View reviewed changes

oqs-bot reviewed Jan 26, 2026

View reviewed changes

github-actions bot reviewed Jan 26, 2026

View reviewed changes

willieyz removed the benchmark label Jan 26, 2026

willieyz force-pushed the eliminate-caddq-intrinsics branch 2 times, most recently from 72bc3f8 to d186f5e Compare January 26, 2026 10:12

willieyz added the benchmark label Jan 27, 2026

github-actions bot reviewed Jan 27, 2026

View reviewed changes

oqs-bot reviewed Jan 27, 2026

View reviewed changes

github-actions bot reviewed Jan 27, 2026

View reviewed changes

willieyz removed the benchmark label Jan 27, 2026

willieyz force-pushed the eliminate-caddq-intrinsics branch from 6faaac2 to 5b1b8a7 Compare January 27, 2026 10:46

willieyz added the benchmark label Jan 28, 2026

willieyz force-pushed the eliminate-caddq-intrinsics branch from 5b1b8a7 to f9a6d30 Compare January 28, 2026 03:58

willieyz removed the benchmark label Jan 28, 2026

mkannwischer added the benchmark label Jan 28, 2026

willieyz marked this pull request as ready for review January 28, 2026 06:42

willieyz requested a review from a team as a code owner January 28, 2026 06:42

mkannwischer requested changes Feb 8, 2026

View reviewed changes

mkannwischer marked this pull request as draft February 10, 2026 02:45

mkannwischer mentioned this pull request Feb 10, 2026

Eliminate use_hint 32/88 intrinsics #940

Open

willieyz force-pushed the eliminate-caddq-intrinsics branch 4 times, most recently from 7dc5f6f to 6761759 Compare February 11, 2026 03:12

willieyz marked this pull request as ready for review February 11, 2026 09:40

willieyz added 3 commits February 11, 2026 17:51

Add mld_poly_caddq to benchmark components

29e7b16

This commit adds mld_poly_caddq to the benchmark components to evaluate the performance impact of replacing the caddq AVX2 intrinsics with x86_64 assembly code. Signed-off-by: willieyz <willie.zhao@chelpis.com>

Eliminate caddq intrinsics

9d97a20

This commit replace the eurrently caddq AVX2 implementation to x86_64 assembly code. Signed-off-by: willieyz <willie.zhao@chelpis.com>

Extract the Macro caddq_vector and unroll loop by factor of 4

14097e6

Signed-off-by: willieyz <willie.zhao@chelpis.com>

mkannwischer force-pushed the eliminate-caddq-intrinsics branch from 6761759 to 14097e6 Compare February 11, 2026 09:51

mkannwischer added benchmark and removed benchmark labels Feb 11, 2026

github-actions bot reviewed Feb 11, 2026

View reviewed changes

jakemas self-requested a review February 13, 2026 03:49

Conversation

willieyz commented Jan 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oqs-bot commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-DSA-87)

Uh oh!

oqs-bot commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-DSA-44)

Uh oh!

oqs-bot commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CBMC Results (ML-DSA-65)

Uh oh!

github-actions bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Mac Mini (M1, 2020) benchmarks (opt)

Uh oh!

github-actions bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Mac Mini (M1, 2020) benchmarks (no-opt)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Intel Xeon 4th gen (c7i)

Uh oh!

github-actions bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Arm Cortex-A72 (Raspberry Pi 4) benchmarks (opt)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Intel Xeon 4th gen (c7i) (no-opt)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

AMD EPYC 3rd gen (c6a)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Intel Xeon 3rd gen (c6i)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

AMD EPYC 4th gen (c7a)

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

⚠️ Performance Alert ⚠️

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

AMD EPYC 3rd gen (c6a) (no-opt)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Intel Xeon 3rd gen (c6i) (no-opt)

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Graviton4

Uh oh!

oqs-bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Graviton3

Uh oh!

willieyz commented Jan 22, 2026 •

edited

Loading

oqs-bot commented Jan 23, 2026 •

edited

Loading

oqs-bot commented Jan 23, 2026 •

edited

Loading

oqs-bot commented Jan 23, 2026 •

edited

Loading

github-actions bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading

oqs-bot left a comment •

edited

Loading

github-actions bot left a comment •

edited

Loading