FMLAL (multiple and single vector, FP8 to FP16)

Multi-vector 8-bit floating-point multiply-add long by vector to half-precision

This 8-bit floating-point multiply-add long instruction widens all 8-bit floating-point elements in the one, two, or four first source vectors and the second source vector to half-precision format and multiplies the corresponding elements. The intermediate products are scaled by 2-UInt(FPMR.LSCALE[3:0]) before being destructively added without intermediate rounding to the overlapping 16-bit half-precision elements of the ZA double-vector groups.

The double-vector group within all of, each half of, or each quarter of the ZA array is selected by the sum of the vector select register and offset range, modulo all, half, or quarter the number of ZA array vectors.

The vector group symbol, VGx2 or VGx4, indicates that the ZA operand consists of two or four ZA double-vector groups respectively. The vector group symbol is preferred for disassembly, but optional in assembler source code.

The 8-bit floating-point encoding format for the elements of the first source vector and the second source vector is selected by FPMR.F8S1 and FPMR.F8S2 respectively.

This instruction is unpredicated.

It has encodings from 3 classes: One ZA double-vector , Two ZA double-vectors and Four ZA double-vectors

One ZA double-vector
(FEAT_SME_F8F16)

313029282726252423222120191817161514131211109876543210
110000010011Zm0Rv011Zn00off3

FMLAL ZA.H[<Wv>, <offs1>:<offs2>], <Zn>.B, <Zm>.B

if !IsFeatureImplemented(FEAT_SME_F8F16) then UNDEFINED; constant integer v = UInt('010':Rv); constant integer n = UInt(Zn); constant integer m = UInt('0':Zm); constant integer offset = UInt(off3:'0'); constant integer nreg = 1;

Two ZA double-vectors
(FEAT_SME_F8F16)

313029282726252423222120191817161514131211109876543210
110000010010Zm0Rv010Zn001off2
S

FMLAL ZA.H[<Wv>, <offs1>:<offs2>{, VGx2}], { <Zn1>.B-<Zn2>.B }, <Zm>.B

if !IsFeatureImplemented(FEAT_SME_F8F16) then UNDEFINED; constant integer v = UInt('010':Rv); constant integer n = UInt(Zn); constant integer m = UInt('0':Zm); constant integer offset = UInt(off2:'0'); constant integer nreg = 2;

Four ZA double-vectors
(FEAT_SME_F8F16)

313029282726252423222120191817161514131211109876543210
110000010011Zm0Rv010Zn001off2
S

FMLAL ZA.H[<Wv>, <offs1>:<offs2>{, VGx4}], { <Zn1>.B-<Zn4>.B }, <Zm>.B

if !IsFeatureImplemented(FEAT_SME_F8F16) then UNDEFINED; constant integer v = UInt('010':Rv); constant integer n = UInt(Zn); constant integer m = UInt('0':Zm); constant integer offset = UInt(off2:'0'); constant integer nreg = 4;

Assembler Symbols

<Wv>

Is the 32-bit name of the vector select register W8-W11, encoded in the "Rv" field.

<offs1>

For the one ZA double-vector variant: is the first vector select offset, encoded as "off3" field times 2.

For the four ZA double-vectors and two ZA double-vectors variant: is the first vector select offset, encoded as "off2" field times 2.

<offs2>

For the one ZA double-vector variant: is the second vector select offset, encoded as "off3" field times 2 plus 1.

For the four ZA double-vectors and two ZA double-vectors variant: is the second vector select offset, encoded as "off2" field times 2 plus 1.

<Zn>

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

<Zn1>

Is the name of the first scalable vector register of the first source multi-vector group, encoded as "Zn".

<Zn4>

Is the name of the fourth scalable vector register of the first source multi-vector group, encoded as "Zn" plus 3 modulo 32.

<Zn2>

Is the name of the second scalable vector register of the first source multi-vector group, encoded as "Zn" plus 1 modulo 32.

<Zm>

Is the name of the second source scalable vector register Z0-Z15, encoded in the "Zm" field.

Operation

CheckFPMREnabled(); CheckStreamingSVEAndZAEnabled(); constant integer VL = CurrentVL; constant integer elements = VL DIV 16; constant integer vectors = VL DIV 8; constant integer vstride = vectors DIV nreg; constant bits(32) vbase = X[v, 32]; integer vec = (UInt(vbase) + offset) MOD vstride; bits(VL) result; vec = vec - (vec MOD 2); for r = 0 to nreg-1 constant bits(VL) op1 = Z[(n+r) MOD 32, VL]; constant bits(VL) op2 = Z[m, VL]; for i = 0 to 1 constant bits(VL) op3 = ZAvector[vec + i, VL]; for e = 0 to elements-1 constant bits(8) elem1 = Elem[op1, 2 * e + i, 8]; constant bits(8) elem2 = Elem[op2, 2 * e + i, 8]; constant bits(16) elem3 = Elem[op3, e, 16]; Elem[result, e, 16] = FP8MulAddFP(elem3, elem1, elem2, FPCR, FPMR); ZAvector[vec + i, VL] = result; vec = vec + vstride;


Internal version only: aarchmrs v2024-03_relA, pseudocode v2024-03_rel, sve v2024-03_rel ; Build timestamp: 2024-03-26T09:45

Copyright © 2010-2024 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.