FMLA (indexed)

Floating-point fused multiply-add by indexed elements (Zda = Zda + Zn * Zm[indexed])

Multiply all floating-point elements within each 128-bit segment of the first source vector by the specified element in the corresponding second source vector segment. The products are then destructively added without intermediate rounding to the corresponding elements of the addend and destination vector.

The elements within the second source vector are specified using an immediate index which selects the same element position within each 128-bit vector segment. The index range is from 0 to one less than the number of elements per 128-bit segment, encoded in 1 to 3 bits depending on the size of the element. This instruction is unpredicated.

It has encodings from 3 classes: Half-precision , Single-precision and Double-precision

Half-precision

313029282726252423222120191817161514131211109876543210
011001000i3h1i3lZm000000ZnZda
op

FMLA <Zda>.H, <Zn>.H, <Zm>.H[<imm>]

if !IsFeatureImplemented(FEAT_SVE) && !IsFeatureImplemented(FEAT_SME) then UNDEFINED; constant integer esize = 16; constant integer index = UInt(i3h:i3l); constant integer n = UInt(Zn); constant integer m = UInt(Zm); constant integer da = UInt(Zda); constant boolean op1_neg = FALSE; constant boolean op3_neg = FALSE;

Single-precision

313029282726252423222120191817161514131211109876543210
01100100101i2Zm000000ZnZda
size<1>size<0>op

FMLA <Zda>.S, <Zn>.S, <Zm>.S[<imm>]

if !IsFeatureImplemented(FEAT_SVE) && !IsFeatureImplemented(FEAT_SME) then UNDEFINED; constant integer esize = 32; constant integer index = UInt(i2); constant integer n = UInt(Zn); constant integer m = UInt(Zm); constant integer da = UInt(Zda); constant boolean op1_neg = FALSE; constant boolean op3_neg = FALSE;

Double-precision

313029282726252423222120191817161514131211109876543210
01100100111i1Zm000000ZnZda
size<1>size<0>op

FMLA <Zda>.D, <Zn>.D, <Zm>.D[<imm>]

if !IsFeatureImplemented(FEAT_SVE) && !IsFeatureImplemented(FEAT_SME) then UNDEFINED; constant integer esize = 64; constant integer index = UInt(i1); constant integer n = UInt(Zn); constant integer m = UInt(Zm); constant integer da = UInt(Zda); constant boolean op1_neg = FALSE; constant boolean op3_neg = FALSE;

Assembler Symbols

<Zda>

Is the name of the third source and destination scalable vector register, encoded in the "Zda" field.

<Zn>

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

<Zm>

For the half-precision and single-precision variant: is the name of the second source scalable vector register Z0-Z7, encoded in the "Zm" field.

For the double-precision variant: is the name of the second source scalable vector register Z0-Z15, encoded in the "Zm" field.

<imm>

For the half-precision variant: is the immediate index, in the range 0 to 7, encoded in the "i3h:i3l" fields.

For the single-precision variant: is the immediate index, in the range 0 to 3, encoded in the "i2" field.

For the double-precision variant: is the immediate index, in the range 0 to 1, encoded in the "i1" field.

Operation

CheckSVEEnabled(); constant integer VL = CurrentVL; constant integer PL = VL DIV 8; constant integer elements = VL DIV esize; constant integer eltspersegment = 128 DIV esize; constant bits(VL) op1 = Z[n, VL]; constant bits(VL) op2 = Z[m, VL]; bits(VL) result = Z[da, VL]; for e = 0 to elements-1 constant integer segmentbase = e - (e MOD eltspersegment); constant integer s = segmentbase + index; constant bits(esize) elem2 = Elem[op2, s, esize]; constant bits(esize) elem1 = (if op1_neg then FPNeg(Elem[op1, e, esize], FPCR) else Elem[op1, e, esize]); constant bits(esize) elem3 = (if op3_neg then FPNeg(Elem[result, e, esize], FPCR) else Elem[result, e, esize]); Elem[result, e, esize] = FPMulAdd(elem3, elem1, elem2, FPCR); Z[da, VL] = result;

Operational information

This instruction might be immediately preceded in program order by a MOVPRFX instruction. The MOVPRFX must conform to all of the following requirements, otherwise the behavior of the MOVPRFX and this instruction is constrained unpredictable:


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.