Arm® A64 Instruction Set Architecture: Future Architecture Technologies in the A architecture profile

BFCVT

Floating-point convert from single-precision to BFloat16 format (scalar) converts the single-precision floating-point value in the 32-bit SIMD&FP source register to BFloat16 format and writes the result in the 16-bit SIMD&FP destination register.

Unlike the BFloat16 multiplication instructions, this instruction honors all the control bits in the FPCR that apply to single-precision arithmetic, including the rounding mode. This instruction can generate a floating-point exception that causes a cumulative exception bit in the FPSR to be set, or a synchronous exception to be taken, depending on the enable bits in the FPCR. ID_AA64ISAR1_EL1.BF16 indicates whether this instruction is supported.

Single-precision to BFloat16
(Armv8.6)

if !HaveBF16Ext() then UNDEFINED;
integer n = UInt(Rn);
integer d = UInt(Rd);

Assembler Symbols

CheckFPAdvSIMDEnabled64();
bits(32) operand = V[n];
bits(16) result;

result = FPConvertBF(operand, FPCR[]);
V[d] = result;