FCVT (narrowing, FP32 to FP8)
Multi-vector floating-point convert from single-precision to packed 8-bit floating-point format
Convert each single-precision element of the four source vectors to 8-bit floating-point while scaling the value by 2SInt(FPMR.NSCALE), and place the results in the quarter-width elements of the destination vector. The 8-bit floating-point encoding format is selected by FPMR.F8D.
This instruction is unpredicated.
Green
False
SM_1_only
1
1
0
0
0
0
0
1
0
0
1
1
0
1
0
0
1
1
1
0
0
0
0
0
FCVT <Zd>.B, { <Zn1>.S-<Zn4>.S }
if !IsFeatureImplemented(FEAT_SME2) || !IsFeatureImplemented(FEAT_FP8) then UNDEFINED;
constant integer n = UInt(Zn:'00');
constant integer d = UInt(Zd);
<Zd>
Is the name of the destination scalable vector register, encoded in the "Zd" field.
<Zn1>
Is the name of the first scalable vector register of the source multi-vector group, encoded as "Zn" times 4.
<Zn4>
Is the name of the fourth scalable vector register of the source multi-vector group, encoded as "Zn" times 4 plus 3.
CheckFPMREnabled();
CheckStreamingSVEEnabled();
constant integer VL = CurrentVL;
constant integer elements = VL DIV 32;
bits(VL) result;
constant bits(VL) operand1 = Z[n+0, VL];
constant bits(VL) operand2 = Z[n+1, VL];
constant bits(VL) operand3 = Z[n+2, VL];
constant bits(VL) operand4 = Z[n+3, VL];
for e = 0 to elements-1
constant bits(32) element1 = Elem[operand1, e, 32];
constant bits(32) element2 = Elem[operand2, e, 32];
constant bits(32) element3 = Elem[operand3, e, 32];
constant bits(32) element4 = Elem[operand4, e, 32];
Elem[result, 0*elements + e, 8] = FPConvertFP8(element1, FPCR, FPMR, 8);
Elem[result, 1*elements + e, 8] = FPConvertFP8(element2, FPCR, FPMR, 8);
Elem[result, 2*elements + e, 8] = FPConvertFP8(element3, FPCR, FPMR, 8);
Elem[result, 3*elements + e, 8] = FPConvertFP8(element4, FPCR, FPMR, 8);
Z[d, VL] = result;