FMLALLBB, FMLALLBT, FMLALLTB, FMLALLTT (by element)

8-bit floating-point multiply-add long-long to single-precision (vector, by element)

This instruction widens the first (bottom bottom), second (bottom top), third (top bottom), or fourth (top top) 8-bit element of each 32-bit container in the first source vector and the indexed element from the second source vector to single-precision format and multiplies the corresponding elements. The intermediate products are scaled by 2-UInt(FPMR.LSCALE), before being destructively added without intermediate rounding to the single-precision elements of the destination vector that overlap with the corresponding 8-bit floating-point elements in the first source vector.

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

Encoding: Advanced SIMD

Variants: FEAT_FP8FMA (ARMv9.5)

313029282726252423222120191817161514131211109876543210
01011110x10000
QUsizeLMRmopcodeHRnRd

FMLALLBB (Q == 0 && size == 00)

FMLALLBB <Vd>.4S, <Vn>.16B, <Vm>.B[<index>]

FMLALLBT (Q == 0 && size == 01)

FMLALLBT <Vd>.4S, <Vn>.16B, <Vm>.B[<index>]

FMLALLTB (Q == 1 && size == 00)

FMLALLTB <Vd>.4S, <Vn>.16B, <Vm>.B[<index>]

FMLALLTT (Q == 1 && size == 01)

FMLALLTT <Vd>.4S, <Vn>.16B, <Vm>.B[<index>]

Decoding algorithm

if !IsFeatureImplemented(FEAT_FP8FMA) then EndOfDecode(Decode_UNDEF);
constant integer n = UInt(Rn);
constant integer m = UInt('00':Rm<2:0>);
constant integer d = UInt(Rd);
constant integer index = UInt(H:L:M:Rm<3>);
constant integer elements = 128 DIV 32;
constant integer sel = UInt(Q:size<0>);

Operation

CheckFPMREnabled(); CheckFPAdvSIMDEnabled64();
constant bits(128) operand1 = V[n, 128];
constant bits(128) operand2 = V[m, 128];
constant bits(128) operand3 = V[d, 128];
bits(128) result;

for e = 0 to elements-1
    constant bits(8) element1  = Elem[operand1, 4 * e + sel, 8];
    constant bits(8) element2  = Elem[operand2, index, 8];
    constant bits(32) element3 = Elem[operand3, e, 32];
    Elem[result, e, 32] = FP8MulAddFP(element3, element1, element2, FPCR, FPMR);

V[d, 128] = result;

Explanations

<Vd>: Is the name of the SIMD&FP destination register, encoded in the "Rd" field.
<Vn>: Is the name of the first SIMD&FP source register, encoded in the "Rn" field.
<Vm>: Is the name of the second SIMD&FP source register, in the range V0 to V7, encoded in the "Rm<2:0>" field.
<index>: Is the element index, in the range 0 to 15, encoded in the "H:L:M:Rm<3>" fields.