8-bit floating-point convert to BFloat16 (vector)
This instruction converts each 8-bit floating-point element from the lower or upper half of the source vector to BFloat16 while downscaling the value, and places the results in the 16-bit elements of the destination vector. BF1CVTL and BF2CVTL convert the elements from the lower half of the source vector while scaling the values by 2-UInt(FPMR.LSCALE[5:0]) and 2-UInt(FPMR.LSCALE2[5:0]), respectively. BF1CVTL2 and BF2CVTL2 convert the elements from the upper half of the source vector while scaling the values by 2-UInt(FPMR.LSCALE[5:0]) and 2-UInt(FPMR.LSCALE2[5:0]), respectively.
The 8-bit floating-point encoding format for BF1CVTL and BF1CVTL2 is selected by FPMR.F8S1. The 8-bit floating-point encoding format for BF2CVTL and BF2CVTL2 is selected by FPMR.F8S2.
Variants: FEAT_FP8 (ARMv9.5)
31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
0 | 1 | 0 | 1 | 1 | 1 | 0 | 1 | x | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 1 | 1 | 1 | 1 | 0 | |||||||||||
Q | U | size | opcode | Rn | Rd |
---|
if !IsFeatureImplemented(FEAT_FP8) then EndOfDecode(Decode_UNDEF); constant integer d = UInt(Rd); constant integer n = UInt(Rn); constant integer part = UInt(Q); constant integer elements = 64 DIV 8; constant boolean issrc2 = size == '11';
CheckFPMREnabled(); CheckFPAdvSIMDEnabled64(); constant bits(64) operand = Vpart[n, part, 64]; bits(128) result; for e = 0 to elements-1 Elem[result, e, 16] = FP8ConvertBF(Elem[operand, e, 8], issrc2, FPCR, FPMR); V[d, 128] = result;