8-bit floating-point convert to half-precision (top)
This instruction converts each odd-numbered 8-bit floating-point element of the source vector to half-precision while downscaling the value, and places the results in the overlapping 16-bit elements of the destination vector. F1CVTLT scales the values by 2-UInt(FPMR.LSCALE[3:0]). F2CVTLT scales the values by 2-UInt(FPMR.LSCALE2[3:0]).
The 8-bit floating-point encoding format for F1CVTLT is selected by FPMR.F8S1. The 8-bit floating-point encoding format for F2CVTLT is selected by FPMR.F8S2.
This instruction is unpredicated.
It has encodings from 2 classes: F1CVTLT and F2CVTLT
| 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
| 0 | 1 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 1 | 0 | 0 | Zn | Zd | ||||||||
| L | opc | ||||||||||||||||||||||||||||||
if ((!IsFeatureImplemented(FEAT_SVE2) && !IsFeatureImplemented(FEAT_SME2)) || !IsFeatureImplemented(FEAT_FP8)) then EndOfDecode(Decode_UNDEF); end; let esize : integer{} = 16; let n : integer = UInt(Zn); let d : integer = UInt(Zd); let issrc2 : boolean = FALSE;
| 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
| 0 | 1 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 1 | 0 | 1 | Zn | Zd | ||||||||
| L | opc | ||||||||||||||||||||||||||||||
if ((!IsFeatureImplemented(FEAT_SVE2) && !IsFeatureImplemented(FEAT_SME2)) || !IsFeatureImplemented(FEAT_FP8)) then EndOfDecode(Decode_UNDEF); end; let esize : integer{} = 16; let n : integer = UInt(Zn); let d : integer = UInt(Zd); let issrc2 : boolean = TRUE;
| <Zd> |
Is the name of the destination scalable vector register, encoded in the "Zd" field. |
| <Zn> |
Is the name of the source scalable vector register, encoded in the "Zn" field. |
CheckFPMREnabled(); if IsFeatureImplemented(FEAT_SME2) then CheckSVEEnabled(); else CheckNonStreamingSVEEnabled(); end; let VL : integer{} = CurrentVL(); let elements : integer = VL DIV esize; let operand : bits(VL) = Z{}(n); var result : bits(VL); for e = 0 to elements-1 do let element : bits(esize DIV 2) = operand[(2*e + 1)*:(esize DIV 2)]; result[e*:esize] = FP8ConvertFP(element, issrc2, FPCR(), FPMR()); end; Z{VL}(d) = result;
2025-09_diff_tag 2025-11-28 16:29:05
Copyright © 2010-2025 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.