FMLA (by element)

Floating-point fused multiply-add to accumulator (by element)

This instruction multiplies the vector elements in the first source SIMD&FP register by the specified value in the second source SIMD&FP register, and accumulates the results in the vector elements of the destination SIMD&FP register. All the values in this instruction are floating-point values.

This instruction can generate a floating-point exception. Depending on the settings in FPCR, the exception results in either a flag being set in FPSR or a synchronous exception being generated. For more information, see Floating-point exceptions and exception traps.

Depending on the settings in the CPACR_EL1, CPTR_EL2, and CPTR_EL3 registers, and the current Security state and Exception level, an attempt to execute the instruction might be trapped.

It has encodings from 4 classes: Scalar, half-precision , Scalar, single-precision and double-precision , Vector, half-precision and Vector, single-precision and double-precision

Scalar, half-precision
(FEAT_AdvSIMD && FEAT_FP16)

313029282726252423222120191817161514131211109876543210
0101111100LMRm0001H0RnRd
Usizeo2

Encoding

FMLA <Hd>, <Hn>, <Vm>.H[<index>]

Decode for this encoding

if !IsFeatureImplemented(FEAT_AdvSIMD) || !IsFeatureImplemented(FEAT_FP16) then EndOfDecode(Decode_UNDEF); end; let idxdsize : integer{} = 64 << UInt(H); let n : integer = UInt(Rn); let m : integer = UInt(Rm); let d : integer = UInt(Rd); let index : integer = UInt(H::L::M); let esize : integer{} = 16; let datasize : integer{} = esize; let elements : integer = 1;

Scalar, single-precision and double-precision
(FEAT_AdvSIMD)

313029282726252423222120191817161514131211109876543210
010111111szLMRm0001H0RnRd
Uo2

Encoding

FMLA <V><d>, <V><n>, <Vm>.<Ts>[<index>]

Decode for this encoding

if !IsFeatureImplemented(FEAT_AdvSIMD) then EndOfDecode(Decode_UNDEF); end; let idxdsize : integer{} = 64 << UInt(H); var index : integer; let Rmhi : bit = M; case sz::L of when '0x' => index = UInt(H::L); when '10' => index = UInt(H); when '11' => EndOfDecode(Decode_UNDEF); end; let d : integer = UInt(Rd); let n : integer = UInt(Rn); let m : integer = UInt(Rmhi::Rm); let esize : integer{} = 32 << UInt(sz); let datasize : integer{} = esize; let elements : integer = 1;

Vector, half-precision
(FEAT_AdvSIMD && FEAT_FP16)

313029282726252423222120191817161514131211109876543210
0Q00111100LMRm0001H0RnRd
Usizeo2

Encoding

FMLA <Vd>.<T>, <Vn>.<T>, <Vm>.H[<index>]

Decode for this encoding

if !IsFeatureImplemented(FEAT_AdvSIMD) || !IsFeatureImplemented(FEAT_FP16) then EndOfDecode(Decode_UNDEF); end; let idxdsize : integer{} = 64 << UInt(H); let n : integer = UInt(Rn); let m : integer = UInt(Rm); let d : integer = UInt(Rd); let index : integer = UInt(H::L::M); let esize : integer{} = 16; let datasize : integer{} = 64 << UInt(Q); let elements : integer = datasize DIV esize;

Vector, single-precision and double-precision
(FEAT_AdvSIMD)

313029282726252423222120191817161514131211109876543210
0Q0011111szLMRm0001H0RnRd
Uo2

Encoding

FMLA <Vd>.<T>, <Vn>.<T>, <Vm>.<Ts>[<index>]

Decode for this encoding

if !IsFeatureImplemented(FEAT_AdvSIMD) then EndOfDecode(Decode_UNDEF); end; if sz::Q == '10' then EndOfDecode(Decode_UNDEF); end; let idxdsize : integer{} = 64 << UInt(H); var index : integer; let Rmhi : bit = M; case sz::L of when '0x' => index = UInt(H::L); when '10' => index = UInt(H); when '11' => EndOfDecode(Decode_UNDEF); end; let d : integer = UInt(Rd); let n : integer = UInt(Rn); let m : integer = UInt(Rmhi::Rm); let esize : integer{} = 32 << UInt(sz); let datasize : integer{} = 64 << UInt(Q); let elements : integer = datasize DIV esize;

Assembler Symbols

<Hd>

Is the 16-bit name of the SIMD&FP destination register, encoded in the "Rd" field.

<Hn>

Is the 16-bit name of the first SIMD&FP source register, encoded in the "Rn" field.

<Vm>

For the "Scalar, half-precision" and "Vector, half-precision" variants: is the name of the second SIMD&FP source register, in the range V0 to V15, encoded in the "Rm" field.

For the "Scalar, single-precision and double-precision" and "Vector, single-precision and double-precision" variants: is the name of the second SIMD&FP source register, encoded in the "M:Rm" fields.

<index>

For the "Scalar, half-precision" and "Vector, half-precision" variants: is the element index, in the range 0 to 7, encoded in the "H:L:M" fields.

For the "Scalar, single-precision and double-precision" and "Vector, single-precision and double-precision" variants: is the element index, encoded in (sz :: L :: H):

sz L <index>
0 x UInt(H :: L)
1 0 UInt(H)
1 1 RESERVED
<V>

Is a width specifier, encoded in sz:

sz <V>
0 S
1 D
<d>

Is the number of the SIMD&FP destination register, encoded in the "Rd" field.

<n>

Is the number of the first SIMD&FP source register, encoded in the "Rn" field.

<Ts>

Is an element size specifier, encoded in sz:

sz <Ts>
0 S
1 D
<Vd>

Is the name of the SIMD&FP destination register, encoded in the "Rd" field.

<T>

For the "Vector, half-precision" variant: is an arrangement specifier, encoded in Q:

Q <T>
0 4H
1 8H

For the "Vector, single-precision and double-precision" variant: is an arrangement specifier, encoded in (Q :: sz):

Q sz <T>
0 0 2S
0 1 RESERVED
1 0 4S
1 1 2D
<Vn>

Is the name of the first SIMD&FP source register, encoded in the "Rn" field.

Operation

AArch64_CheckFPAdvSIMDEnabled(); let operand1 : bits(datasize) = V{}(n); let operand2 : bits(idxdsize) = V{}(m); let operand3 : bits(datasize) = V{}(d); var element1 : bits(esize); let element2 : bits(esize) = operand2[index*:esize]; let merge : boolean = elements == 1 && IsMerging(FPCR()); var result : bits(128) = if merge then V{128}(d) else Zeros{128}; for e = 0 to elements-1 do element1 = operand1[e*:esize]; result[e*:esize] = FPMulAdd{esize}(operand3[e*:esize], element1, element2, FPCR()); end; V{128}(d) = result;


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.