ciciec2026/doc/mlkem_test_results.md

ML-KEM Test Results Summary

Test Setup

• Platform: LoongArch bare-metal target
• Test mode: STATIC_KAT_COUNT_0
• Levels: k=2,3,4
• Repeat count: 100
• ML-KEM software allocator: MLK_CONFIG_CUSTOM_ALLOC_FREE
• Verification sources: KAT count 0 expected values

Correctness

All three paths passed in isolated testing:

• SW_ONLY_FAILS=0
• CPU_ONLY_FAILS=0
• DMA_ONLY_FAILS=0

Final result for the isolated diagnostic run:

• MLKEM512_RESULT=PASS
• MLKEM768_RESULT=PASS
• MLKEM1024_RESULT=PASS
• MLKEM_RESULT=PASS

Performance

The following results are from isolated per-level runs. Each level was tested for 100 iterations with software-only, CPU register path, and DMA path separated.

Level	SW avg cycles	CPU avg cycles	DMA avg cycles	CPU vs SW	DMA vs SW	DMA vs CPU
k=2 / ML-KEM-512	2,587,784	352,684	185,671	7.34x	13.94x	1.90x
k=3 / ML-KEM-768	4,056,378	523,656	279,588	7.75x	14.51x	1.87x
k=4 / ML-KEM-1024	6,025,486	718,672	388,959	8.38x	15.49x	1.85x

Notes

• CPU/DMA hardware paths were stable across 100 iterations for all levels.
• Software, CPU register path, and DMA path all passed 100 iterations for k=2, k=3, and k=4.
• The DMA path is 13.94x to 15.49x faster than software across the tested levels.
• The CPU register path is 7.34x to 8.38x faster than software across the tested levels.
• The DMA path is about 1.85x to 1.90x faster than the CPU register path.
• The hardware advantage increases slightly as k grows, because the accelerated computation grows faster than the fixed software control overhead.