AOCL-DLP automatically leverages available CPU features:
Feature |
Availability |
Benefit |
|---|---|---|
AVX2/FMA3 |
AMD Zen 1+ |
Vectorized floating-point operations |
AVX512 |
AMD Zen 4+ |
Wider vector operations |
AVX512_VNNI |
AMD Zen 4+ |
Accelerated integer GEMM |
AVX512_BF16 |
AMD Zen 4+ |
Native bfloat16 operations |