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 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_FP16)

313029282726252423222120191817161514131211109876543210
0101111100LMRm0001H0RnRd
Usizeo2

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

if !IsFeatureImplemented(FEAT_FP16) then UNDEFINED; constant integer idxdsize = 64 << UInt(H); integer n = UInt(Rn); integer m = UInt(Rm); integer d = UInt(Rd); integer index = UInt(H:L:M); constant integer esize = 16; constant integer datasize = esize; integer elements = 1; boolean sub_op = (o2 == '1');

Scalar, single-precision and double-precision

313029282726252423222120191817161514131211109876543210
010111111szLMRm0001H0RnRd
Uo2

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

constant integer idxdsize = 64 << UInt(H); integer index; bit Rmhi = M; case sz:L of when '0x' index = UInt(H:L); when '10' index = UInt(H); when '11' UNDEFINED; integer d = UInt(Rd); integer n = UInt(Rn); integer m = UInt(Rmhi:Rm); constant integer esize = 32 << UInt(sz); constant integer datasize = esize; integer elements = 1; boolean sub_op = (o2 == '1');

Vector, half-precision
(FEAT_FP16)

313029282726252423222120191817161514131211109876543210
0Q00111100LMRm0001H0RnRd
Usizeo2

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

if !IsFeatureImplemented(FEAT_FP16) then UNDEFINED; constant integer idxdsize = 64 << UInt(H); integer n = UInt(Rn); integer m = UInt(Rm); integer d = UInt(Rd); integer index = UInt(H:L:M); constant integer esize = 16; constant integer datasize = 64 << UInt(Q); integer elements = datasize DIV esize; boolean sub_op = (o2 == '1');

Vector, single-precision and double-precision

313029282726252423222120191817161514131211109876543210
0Q0011111szLMRm0001H0RnRd
Uo2

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

constant integer idxdsize = 64 << UInt(H); integer index; bit Rmhi = M; case sz:L of when '0x' index = UInt(H:L); when '10' index = UInt(H); when '11' UNDEFINED; integer d = UInt(Rd); integer n = UInt(Rn); integer m = UInt(Rmhi:Rm); if sz:Q == '10' then UNDEFINED; constant integer esize = 32 << UInt(sz); constant integer datasize = 64 << UInt(Q); integer elements = datasize DIV esize; boolean sub_op = (o2 == '1');

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

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

<index>

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

For the single-precision and double-precision variant: 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 half-precision variant: is an arrangement specifier, encoded in Q:

Q <T>
0 4H
1 8H

For the 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

CheckFPAdvSIMDEnabled64(); bits(datasize) operand1 = V[n, datasize]; bits(idxdsize) operand2 = V[m, idxdsize]; bits(datasize) operand3 = V[d, datasize]; bits(esize) element1; bits(esize) element2 = Elem[operand2, index, esize]; boolean merge = elements == 1 && IsMerging(FPCR); bits(128) result = if merge then V[d, 128] else Zeros(128); for e = 0 to elements-1 element1 = Elem[operand1, e, esize]; if sub_op then element1 = FPNeg(element1, FPCR); Elem[result, e, esize] = FPMulAdd(Elem[operand3, e, esize], element1, element2, FPCR); V[d, 128] = result;


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.