Add native implementation for `poly_caddq` #414

mkannwischer · 2025-08-06T07:17:42Z

Resolves AArch64: Consider adding poly_caddq assembly #396
Resolves AVX2: Add native implementation of poly_caddq #327
Depends on Add native implementation of poly_decompose #411
Depends on Inline rounding.c and reduce.c #410

mldsa/native/x86_64/src/poly_caddq_avx2.c

hanno-becker · 2025-08-08T12:56:40Z

mldsa/native/aarch64/src/poly_caddq_asm.S

+        sshr tmp.4s, \inout\().4s, #31
+        mls \inout\().4s, tmp.4s, q_reg.4s


Nit: shr + mla would be a bit more intuitive?

hanno-becker

It makes sense to wait for benchmarks first, but if we want this, the avx2 code should be written in ASM, not intrinsics.

mkannwischer · 2025-08-08T13:01:36Z

It makes sense to wait for benchmarks first, but if we want this, the avx2 code should be written in ASM, not intrinsics.

In that case we will have to split this up into two PRs. I'll do AArch64, but we will need to find someone who can write AVX2 asm. Wouldn't it make more sense to get performance on-par with the pqcrystals implementation first and then rewrite it to asm later or in parallel? The integration of the existing avx2 intrinsics implementation is straightforward.

This commit is hoisted out from #325. It adds only the native implementation of poly_caddq taken from the official AVX2 implementation. Signed-off-by: Matthias J. Kannwischer <[email protected]>

Signed-off-by: Matthias J. Kannwischer <[email protected]>

github-actions

Mac Mini (M1, 2020) benchmarks (opt)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`47851` cycles	`49574` cycles	`0.97`
`ML-DSA-44 sign`	`173145` cycles	`180472` cycles	`0.96`
`ML-DSA-44 verify`	`56651` cycles	`58524` cycles	`0.97`
`ML-DSA-65 keypair`	`83660` cycles	`86323` cycles	`0.97`
`ML-DSA-65 sign`	`284598` cycles	`298570` cycles	`0.95`
`ML-DSA-65 verify`	`91148` cycles	`93819` cycles	`0.97`
`ML-DSA-87 keypair`	`136114` cycles	`139768` cycles	`0.97`
`ML-DSA-87 sign`	`353910` cycles	`368794` cycles	`0.96`
`ML-DSA-87 verify`	`145304` cycles	`148973` cycles	`0.98`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`115056` cycles	`115045` cycles	`1.00`
`ML-DSA-44 sign`	`430540` cycles	`430523` cycles	`1.00`
`ML-DSA-44 verify`	`122149` cycles	`122136` cycles	`1.00`
`ML-DSA-65 keypair`	`197455` cycles	`197476` cycles	`1.00`
`ML-DSA-65 sign`	`701054` cycles	`701059` cycles	`1.00`
`ML-DSA-65 verify`	`197629` cycles	`197658` cycles	`1.00`
`ML-DSA-87 keypair`	`334609` cycles	`334653` cycles	`1.00`
`ML-DSA-87 sign`	`883871` cycles	`883903` cycles	`1.00`
`ML-DSA-87 verify`	`328482` cycles	`328478` cycles	`1.00`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`115259` cycles	`118636` cycles	`0.97`
`ML-DSA-44 sign`	`421458` cycles	`433976` cycles	`0.97`
`ML-DSA-44 verify`	`132918` cycles	`136098` cycles	`0.98`
`ML-DSA-65 keypair`	`200140` cycles	`204644` cycles	`0.98`
`ML-DSA-65 sign`	`698510` cycles	`722863` cycles	`0.97`
`ML-DSA-65 verify`	`214500` cycles	`219020` cycles	`0.98`
`ML-DSA-87 keypair`	`328037` cycles	`333765` cycles	`0.98`
`ML-DSA-87 sign`	`878355` cycles	`902010` cycles	`0.97`
`ML-DSA-87 verify`	`346241` cycles	`354166` cycles	`0.98`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`212904` cycles	`213088` cycles	`1.00`
`ML-DSA-44 sign`	`781158` cycles	`780956` cycles	`1.00`
`ML-DSA-44 verify`	`229596` cycles	`230268` cycles	`1.00`
`ML-DSA-65 keypair`	`380703` cycles	`380342` cycles	`1.00`
`ML-DSA-65 sign`	`1283469` cycles	`1291576` cycles	`0.99`
`ML-DSA-65 verify`	`372123` cycles	`371822` cycles	`1.00`
`ML-DSA-87 keypair`	`608948` cycles	`608922` cycles	`1.00`
`ML-DSA-87 sign`	`1642584` cycles	`1642217` cycles	`1.00`
`ML-DSA-87 verify`	`621386` cycles	`621406` cycles	`1.00`

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

oqs-bot

Intel Xeon 4th gen (c7i)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`35025` cycles	`36914` cycles	`0.95`
`ML-DSA-44 sign`	`143273` cycles	`150193` cycles	`0.95`
`ML-DSA-44 verify`	`44283` cycles	`45714` cycles	`0.97`
`ML-DSA-65 keypair`	`63469` cycles	`65371` cycles	`0.97`
`ML-DSA-65 sign`	`239729` cycles	`249917` cycles	`0.96`
`ML-DSA-65 verify`	`71212` cycles	`73553` cycles	`0.97`
`ML-DSA-87 keypair`	`96766` cycles	`98755` cycles	`0.98`
`ML-DSA-87 sign`	`289753` cycles	`297831` cycles	`0.97`
`ML-DSA-87 verify`	`107039` cycles	`109480` cycles	`0.98`

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

oqs-bot

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`95141` cycles	`95116` cycles	`1.00`
`ML-DSA-44 sign`	`342977` cycles	`343293` cycles	`1.00`
`ML-DSA-44 verify`	`101248` cycles	`101310` cycles	`1.00`
`ML-DSA-65 keypair`	`164948` cycles	`164688` cycles	`1.00`
`ML-DSA-65 sign`	`572030` cycles	`571928` cycles	`1.00`
`ML-DSA-65 verify`	`163929` cycles	`163880` cycles	`1.00`
`ML-DSA-87 keypair`	`269043` cycles	`268822` cycles	`1.00`
`ML-DSA-87 sign`	`726084` cycles	`725612` cycles	`1.00`
`ML-DSA-87 verify`	`271795` cycles	`271457` cycles	`1.00`

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

oqs-bot

AMD EPYC 3rd gen (c6a)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`73678` cycles	`72429` cycles	`1.02`
`ML-DSA-44 sign`	`238042` cycles	`235885` cycles	`1.01`
`ML-DSA-44 verify`	`83424` cycles	`81935` cycles	`1.02`
`ML-DSA-65 keypair`	`122457` cycles	`127155` cycles	`0.96`
`ML-DSA-65 sign`	`372858` cycles	`392445` cycles	`0.95`
`ML-DSA-65 verify`	`131265` cycles	`135767` cycles	`0.97`
`ML-DSA-87 keypair`	`207061` cycles	`209507` cycles	`0.99`
`ML-DSA-87 sign`	`479289` cycles	`490112` cycles	`0.98`
`ML-DSA-87 verify`	`215926` cycles	`218227` cycles	`0.99`

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

oqs-bot

AMD EPYC 4th gen (c7a)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`42278` cycles	`43749` cycles	`0.97`
`ML-DSA-44 sign`	`160044` cycles	`168465` cycles	`0.95`
`ML-DSA-44 verify`	`53629` cycles	`55266` cycles	`0.97`
`ML-DSA-65 keypair`	`71741` cycles	`74727` cycles	`0.96`
`ML-DSA-65 sign`	`258565` cycles	`269433` cycles	`0.96`
`ML-DSA-65 verify`	`83358` cycles	`85290` cycles	`0.98`
`ML-DSA-87 keypair`	`111912` cycles	`114492` cycles	`0.98`
`ML-DSA-87 sign`	`309232` cycles	`319148` cycles	`0.97`
`ML-DSA-87 verify`	`128545` cycles	`129750` cycles	`0.99`

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

oqs-bot

Intel Xeon 3rd gen (c6i)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`57459` cycles	`59245` cycles	`0.97`
`ML-DSA-44 sign`	`224603` cycles	`231008` cycles	`0.97`
`ML-DSA-44 verify`	`72810` cycles	`74684` cycles	`0.97`
`ML-DSA-65 keypair`	`104183` cycles	`104580` cycles	`1.00`
`ML-DSA-65 sign`	`375420` cycles	`383421` cycles	`0.98`
`ML-DSA-65 verify`	`117504` cycles	`118758` cycles	`0.99`
`ML-DSA-87 keypair`	`157854` cycles	`162893` cycles	`0.97`
`ML-DSA-87 sign`	`443245` cycles	`464184` cycles	`0.95`
`ML-DSA-87 verify`	`176039` cycles	`180395` cycles	`0.98`

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

oqs-bot

Graviton3

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`73991` cycles	`76425` cycles	`0.97`
`ML-DSA-44 sign`	`258818` cycles	`269792` cycles	`0.96`
`ML-DSA-44 verify`	`86840` cycles	`89274` cycles	`0.97`
`ML-DSA-65 keypair`	`129767` cycles	`133670` cycles	`0.97`
`ML-DSA-65 sign`	`428121` cycles	`448377` cycles	`0.95`
`ML-DSA-65 verify`	`139976` cycles	`144053` cycles	`0.97`
`ML-DSA-87 keypair`	`210843` cycles	`216151` cycles	`0.98`
`ML-DSA-87 sign`	`537712` cycles	`558494` cycles	`0.96`
`ML-DSA-87 verify`	`225089` cycles	`230636` cycles	`0.98`

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

oqs-bot

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`135633` cycles	`135562` cycles	`1.00`
`ML-DSA-44 sign`	`544428` cycles	`542966` cycles	`1.00`
`ML-DSA-44 verify`	`148691` cycles	`148680` cycles	`1.00`
`ML-DSA-65 keypair`	`227822` cycles	`227876` cycles	`1.00`
`ML-DSA-65 sign`	`876183` cycles	`873284` cycles	`1.00`
`ML-DSA-65 verify`	`236729` cycles	`236062` cycles	`1.00`
`ML-DSA-87 keypair`	`374848` cycles	`375435` cycles	`1.00`
`ML-DSA-87 sign`	`1097932` cycles	`1099713` cycles	`1.00`
`ML-DSA-87 verify`	`386699` cycles	`387716` cycles	`1.00`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`282186` cycles	`289927` cycles	`0.97`
`ML-DSA-44 sign`	`1059056` cycles	`1093182` cycles	`0.97`
`ML-DSA-44 verify`	`333541` cycles	`342018` cycles	`0.98`
`ML-DSA-65 keypair`	`483586` cycles	`494445` cycles	`0.98`
`ML-DSA-65 sign`	`1741574` cycles	`1801085` cycles	`0.97`
`ML-DSA-65 verify`	`525769` cycles	`536077` cycles	`0.98`
`ML-DSA-87 keypair`	`813340` cycles	`841573` cycles	`0.97`
`ML-DSA-87 sign`	`2285185` cycles	`2406369` cycles	`0.95`
`ML-DSA-87 verify`	`867799` cycles	`899537` cycles	`0.96`

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

oqs-bot

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`120986` cycles	`120436` cycles	`1.00`
`ML-DSA-44 sign`	`454793` cycles	`452210` cycles	`1.01`
`ML-DSA-44 verify`	`131296` cycles	`131572` cycles	`1.00`
`ML-DSA-65 keypair`	`205285` cycles	`204813` cycles	`1.00`
`ML-DSA-65 sign`	`737169` cycles	`741495` cycles	`0.99`
`ML-DSA-65 verify`	`210869` cycles	`210092` cycles	`1.00`
`ML-DSA-87 keypair`	`340074` cycles	`345097` cycles	`0.99`
`ML-DSA-87 sign`	`946189` cycles	`954462` cycles	`0.99`
`ML-DSA-87 verify`	`351857` cycles	`353756` cycles	`0.99`

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

oqs-bot

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`157285` cycles	`157128` cycles	`1.00`
`ML-DSA-44 sign`	`564128` cycles	`563242` cycles	`1.00`
`ML-DSA-44 verify`	`169109` cycles	`168817` cycles	`1.00`
`ML-DSA-65 keypair`	`269435` cycles	`268889` cycles	`1.00`
`ML-DSA-65 sign`	`928431` cycles	`929005` cycles	`1.00`
`ML-DSA-65 verify`	`273935` cycles	`274165` cycles	`1.00`
`ML-DSA-87 keypair`	`451404` cycles	`450904` cycles	`1.00`
`ML-DSA-87 sign`	`1181071` cycles	`1182837` cycles	`1.00`
`ML-DSA-87 verify`	`460399` cycles	`460036` cycles	`1.00`

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

oqs-bot

Graviton4

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`69485` cycles	`71384` cycles	`0.97`
`ML-DSA-44 sign`	`243406` cycles	`252411` cycles	`0.96`
`ML-DSA-44 verify`	`80510` cycles	`82570` cycles	`0.98`
`ML-DSA-65 keypair`	`123278` cycles	`126207` cycles	`0.98`
`ML-DSA-65 sign`	`402262` cycles	`418729` cycles	`0.96`
`ML-DSA-65 verify`	`130784` cycles	`133572` cycles	`0.98`
`ML-DSA-87 keypair`	`200823` cycles	`204935` cycles	`0.98`
`ML-DSA-87 sign`	`508031` cycles	`524564` cycles	`0.97`
`ML-DSA-87 verify`	`211930` cycles	`215419` cycles	`0.98`

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

oqs-bot

Graviton2

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`115972` cycles	`118893` cycles	`0.98`
`ML-DSA-44 sign`	`423405` cycles	`434340` cycles	`0.97`
`ML-DSA-44 verify`	`133762` cycles	`136351` cycles	`0.98`
`ML-DSA-65 keypair`	`200328` cycles	`204738` cycles	`0.98`
`ML-DSA-65 sign`	`698660` cycles	`723474` cycles	`0.97`
`ML-DSA-65 verify`	`214506` cycles	`219065` cycles	`0.98`
`ML-DSA-87 keypair`	`328454` cycles	`334351` cycles	`0.98`
`ML-DSA-87 sign`	`879954` cycles	`903543` cycles	`0.97`
`ML-DSA-87 verify`	`346835` cycles	`354365` cycles	`0.98`

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

oqs-bot

Graviton3 (no-opt)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`138587` cycles	`138687` cycles	`1.00`
`ML-DSA-44 sign`	`494937` cycles	`495442` cycles	`1.00`
`ML-DSA-44 verify`	`148834` cycles	`148734` cycles	`1.00`
`ML-DSA-65 keypair`	`242313` cycles	`242497` cycles	`1.00`
`ML-DSA-65 sign`	`809930` cycles	`809720` cycles	`1.00`
`ML-DSA-65 verify`	`240840` cycles	`241177` cycles	`1.00`
`ML-DSA-87 keypair`	`396223` cycles	`396469` cycles	`1.00`
`ML-DSA-87 sign`	`1031619` cycles	`1031355` cycles	`1.00`
`ML-DSA-87 verify`	`402491` cycles	`402148` cycles	`1.00`

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

oqs-bot

Graviton4 (no-opt)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`133152` cycles	`133059` cycles	`1.00`
`ML-DSA-44 sign`	`498192` cycles	`498254` cycles	`1.00`
`ML-DSA-44 verify`	`144942` cycles	`145034` cycles	`1.00`
`ML-DSA-65 keypair`	`225991` cycles	`226138` cycles	`1.00`
`ML-DSA-65 sign`	`814388` cycles	`813372` cycles	`1.00`
`ML-DSA-65 verify`	`231482` cycles	`231336` cycles	`1.00`
`ML-DSA-87 keypair`	`374634` cycles	`374386` cycles	`1.00`
`ML-DSA-87 sign`	`1019911` cycles	`1020865` cycles	`1.00`
`ML-DSA-87 verify`	`383528` cycles	`383441` cycles	`1.00`

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

oqs-bot

Graviton2 (no-opt)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`213348` cycles	`213315` cycles	`1.00`
`ML-DSA-44 sign`	`782920` cycles	`793744` cycles	`0.99`
`ML-DSA-44 verify`	`230036` cycles	`229936` cycles	`1.00`
`ML-DSA-65 keypair`	`381290` cycles	`380736` cycles	`1.00`
`ML-DSA-65 sign`	`1285487` cycles	`1284870` cycles	`1.00`
`ML-DSA-65 verify`	`372653` cycles	`372305` cycles	`1.00`
`ML-DSA-87 keypair`	`609534` cycles	`609775` cycles	`1.00`
`ML-DSA-87 sign`	`1645467` cycles	`1645982` cycles	`1.00`
`ML-DSA-87 verify`	`621701` cycles	`621716` cycles	`1.00`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`460565` cycles	`461887` cycles	`1.00`
`ML-DSA-44 sign`	`2209645` cycles	`2208687` cycles	`1.00`
`ML-DSA-44 verify`	`545435` cycles	`546011` cycles	`1.00`
`ML-DSA-65 keypair`	`774964` cycles	`772961` cycles	`1.00`
`ML-DSA-65 sign`	`3633431` cycles	`3613853` cycles	`1.01`
`ML-DSA-65 verify`	`846679` cycles	`846000` cycles	`1.00`
`ML-DSA-87 keypair`	`1250314` cycles	`1252087` cycles	`1.00`
`ML-DSA-87 sign`	`4526553` cycles	`4485312` cycles	`1.01`
`ML-DSA-87 verify`	`1363008` cycles	`1363580` 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)

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`823259` cycles	`822837` cycles	`1.00`
`ML-DSA-44 sign`	`3331843` cycles	`3330856` cycles	`1.00`
`ML-DSA-44 verify`	`919486` cycles	`918759` cycles	`1.00`
`ML-DSA-65 keypair`	`1404693` cycles	`1401325` cycles	`1.00`
`ML-DSA-65 sign`	`5455368` cycles	`5449246` cycles	`1.00`
`ML-DSA-65 verify`	`1469178` cycles	`1466794` cycles	`1.00`
`ML-DSA-87 keypair`	`2297073` cycles	`2299513` cycles	`1.00`
`ML-DSA-87 sign`	`6797279` cycles	`6802332` cycles	`1.00`
`ML-DSA-87 verify`	`2399877` cycles	`2394302` cycles	`1.00`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`219607` cycles	`259879` cycles	`0.85`
`ML-DSA-44 sign`	`709012` cycles	`796628` cycles	`0.89`
`ML-DSA-44 verify`	`243564` cycles	`257672` cycles	`0.95`
`ML-DSA-65 keypair`	`392968` cycles	`418565` cycles	`0.94`
`ML-DSA-65 sign`	`1182376` cycles	`1261720` cycles	`0.94`
`ML-DSA-65 verify`	`399468` cycles	`415437` cycles	`0.96`
`ML-DSA-87 keypair`	`665448` cycles	`668630` cycles	`1.00`
`ML-DSA-87 sign`	`1574717` cycles	`1597546` cycles	`0.99`
`ML-DSA-87 verify`	`663926` cycles	`670039` cycles	`0.99`

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

github-actions

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

Benchmark suite	Current: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 keypair`	`302847` cycles	`311403` cycles	`0.97`
`ML-DSA-44 sign`	`1222085` cycles	`1172791` cycles	`1.04`
`ML-DSA-44 verify`	`338643` cycles	`325419` cycles	`1.04`
`ML-DSA-65 keypair`	`565947` cycles	`556523` cycles	`1.02`
`ML-DSA-65 sign`	`1990741` cycles	`1925569` cycles	`1.03`
`ML-DSA-65 verify`	`539294` cycles	`527641` cycles	`1.02`
`ML-DSA-87 keypair`	`872061` cycles	`867077` cycles	`1.01`
`ML-DSA-87 sign`	`2540499` cycles	`2461379` cycles	`1.03`
`ML-DSA-87 verify`	`902482` cycles	`883514` 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: `d45408b`	Previous: `ebede4d`	Ratio
`ML-DSA-44 sign`	`1222085` cycles	`1172791` cycles	`1.04`
`ML-DSA-44 verify`	`338643` cycles	`325419` cycles	`1.04`
`ML-DSA-65 sign`	`1990741` cycles	`1925569` cycles	`1.03`
`ML-DSA-87 sign`	`2540499` cycles	`2461379` cycles	`1.03`

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

mkannwischer mentioned this pull request Aug 6, 2025

AVX2: Add native implementation of poly_caddq #327

Open

hanno-becker reviewed Aug 8, 2025

View reviewed changes

mldsa/native/x86_64/src/poly_caddq_avx2.c Show resolved Hide resolved

hanno-becker reviewed Aug 8, 2025

View reviewed changes

hanno-becker requested changes Aug 8, 2025

View reviewed changes

mkannwischer force-pushed the caddq-asm branch from db2e505 to a443fd7 Compare September 23, 2025 14:19

mkannwischer added 2 commits September 23, 2025 22:48

AVX2: Add native implementation of poly_caddq

3c45d6c

This commit is hoisted out from #325. It adds only the native implementation of poly_caddq taken from the official AVX2 implementation. Signed-off-by: Matthias J. Kannwischer <[email protected]>

AArch64: Add native implementation for poly_caddq

d45408b

Signed-off-by: Matthias J. Kannwischer <[email protected]>

mkannwischer force-pushed the caddq-asm branch from a443fd7 to d45408b Compare September 23, 2025 14:48

mkannwischer added the benchmark label Sep 23, 2025

github-actions bot reviewed Sep 23, 2025

View reviewed changes

oqs-bot reviewed Sep 23, 2025

View reviewed changes

github-actions bot reviewed Sep 23, 2025

View reviewed changes

oqs-bot reviewed Sep 23, 2025

View reviewed changes

github-actions bot reviewed Sep 23, 2025

View reviewed changes

mkannwischer marked this pull request as ready for review September 23, 2025 15:04

mkannwischer requested a review from a team as a code owner September 23, 2025 15:04

github-actions bot reviewed Sep 23, 2025

View reviewed changes

		sshr tmp.4s, \inout\().4s, #31
		mls \inout\().4s, tmp.4s, q_reg.4s

Add native implementation for poly_caddq #414

Are you sure you want to change the base?

Add native implementation for poly_caddq #414

Conversation

mkannwischer commented Aug 6, 2025

Uh oh!

Uh oh!

hanno-becker Aug 8, 2025

Choose a reason for hiding this comment

Uh oh!

hanno-becker left a comment

Choose a reason for hiding this comment

Uh oh!

mkannwischer commented Aug 8, 2025

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Mac Mini (M1, 2020) benchmarks (opt)

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

Intel Xeon 4th gen (c7i)

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

AMD EPYC 3rd gen (c6a)

Uh oh!

oqs-bot left a comment

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

Intel Xeon 3rd gen (c6i)

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

Graviton3

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

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

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

Graviton4

Uh oh!

oqs-bot left a comment

Choose a reason for hiding this comment

Graviton2

Uh oh!

oqs-bot left a comment

Add native implementation for `poly_caddq` #414

Add native implementation for `poly_caddq` #414