FEAT_BF16 (AI) Instructions
4 ARM instructions in this extension - click any row for encoding, pseudocode, and full documentation.
| Mnemonic | Syntax | Format | Summary |
|---|---|---|---|
| bfcvtn | BFCVTN <Vd>.<Tb>, <Vn>.<Ta> | NEON 2-Reg | Converts Float32 to BFloat16 (Lower Half). |
| bfcvtn2 | BFCVTN2 <Vd>.<Tb>, <Vn>.<Ta> | NEON 2-Reg | Converts Float32 to BFloat16 (Upper Half). |
| bfdot | BFDOT <Vd>.<T>, <Vn>.<T>, <Vm>.<T> | NEON 3-Reg | Computes dot product of BFloat16 elements, accumulating to Float32 (NEON). |
| bfmmla | BFMMLA <Vd>.<T>, <Vn>.<T>, <Vm>.<T> | NEON 3-Reg | Performs 2x2 matrix multiplication on BFloat16 tiles (NEON). |