ARM instruction sets

Instruction sets

These guides cover AArch64 instructions. NEON supplies the base vectors; crypto, dot-product, matrix and half-precision operations have additional feature requirements. Keep the hardware feature bits distinct from Clang target bundles when admitting a function; see target selection.

• AES/PMULL/SHA: cipher rounds, polynomial products and hash transforms
• BF16: dot products and matrix accumulation
• CRC: CRC32 and CRC32C updates
• DotProd: byte dot products
• FCMA: complex arithmetic
• FHM: FP16 products with FP32 accumulation
• I8MM: byte matrix products and mixed-sign dot products
• JSCVT: JavaScript integer conversion
• NEON: saturation, shifts, bit operations and conversions
• RDM: rounded fixed-point accumulation
• SM3 and SM4: hash rounds and block-cipher transforms