Multi-vector 8-bit floating-point convert to deinterleaved BFloat16
This instruction converts each 8-bit floating-point element of the source vector to BFloat16 while downscaling the value, and places the two-way deinterleaved results in the corresponding 16-bit elements of the destination vectors. BF1CVTL scales the values by 2-UInt(FPMR.LSCALE[5:0]). BF2CVTL scales the values by 2-UInt(FPMR.LSCALE2[5:0]). The 8-bit floating-point encoding format for BF1CVTL is selected by FPMR.F8S1. The 8-bit floating-point encoding format for BF2CVTL is selected by FPMR.F8S2.
This instruction is unpredicated.
It has encodings from 2 classes: BF1CVTL and BF2CVTL
| 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 |
| 1 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 1 | 1 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | Zn | Zd | 1 | |||||||
| opc | L | ||||||||||||||||||||||||||||||
if !IsFeatureImplemented(FEAT_SME2) || !IsFeatureImplemented(FEAT_FP8) then EndOfDecode(Decode_UNDEF); end; let n : integer = UInt(Zn); let d : integer = UInt(Zd:: '0'); 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 |
| 1 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 1 | 1 | 0 | 0 | 1 | 1 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | Zn | Zd | 1 | |||||||
| opc | L | ||||||||||||||||||||||||||||||
if !IsFeatureImplemented(FEAT_SME2) || !IsFeatureImplemented(FEAT_FP8) then EndOfDecode(Decode_UNDEF); end; let n : integer = UInt(Zn); let d : integer = UInt(Zd:: '0'); let issrc2 : boolean = TRUE;
| <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. |
CheckFPMREnabled(); CheckStreamingSVEEnabled(); let VL : integer{} = CurrentVL(); let pairs : integer = VL DIV 16; let operand : bits(VL) = Z{}(n); var result1 : bits(VL); var result2 : bits(VL); for p = 0 to pairs-1 do let element1 : bits(8) = operand[(2*p + 0)*:8]; let element2 : bits(8) = operand[(2*p + 1)*:8]; result1[p*:16] = FP8ConvertBF(element1, issrc2, FPCR(), FPMR()); result2[p*:16] = FP8ConvertBF(element2, issrc2, FPCR(), FPMR()); end; Z{VL}(d+0) = result1; Z{VL}(d+1) = result2;
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.