FCVT (widening)

Multi-vector half-precision convert to single-precision

This instruction converts each element of the source vector from half-precision to single-precision floating-point, and places the results in the double-width destination elements of the destination vectors.

This instruction follows SME2 floating-point numerical behaviors corresponding to instructions that place their results in one or more SVE Z vectors.

This instruction is unpredicated.

ID_AA64SMFR0_EL1.F16F16 indicates whether this instruction is implemented.

SME2
(FEAT_SME_F16F16)

313029282726252423222120191817161514131211109876543210
1100000110100000111000ZnZd0
L

Encoding

FCVT { <Zd1>.S-<Zd2>.S }, <Zn>.H

Decode for this encoding

if !IsFeatureImplemented(FEAT_SME_F16F16) then EndOfDecode(Decode_UNDEF); end; let n : integer = UInt(Zn); let d : integer = UInt(Zd::'0');

Assembler Symbols

<Zd1>

Is the name of the first scalable vector register of the destination multi-vector group, encoded as "Zd" times 2.

<Zd2>

Is the name of the second scalable vector register of the destination multi-vector group, encoded as "Zd" times 2 plus 1.

<Zn>

Is the name of the source scalable vector register, encoded in the "Zn" field.

Operation

CheckStreamingSVEEnabled(); let VL : integer{} = CurrentVL(); let elements : integer = VL DIV 16; let operand : bits(VL) = Z{}(n); var result : bits(2*VL); for e = 0 to elements-1 do let element : bits(16) = operand[e*:16]; let res : bits(32) = FPConvertSVE{32, 16}(element, FPCR()); result[e*:32] = res; end; Z{VL}(d+0) = result[VL-1:0]; Z{VL}(d+1) = result[2*VL-1:VL];


2026-03_rel 2026-03-26 20:48:11

Copyright © 2010-2026 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.