FMAX.H

Float Maximum (Half)

FMAX.H rd, rs1, rs2

Writes the larger of two 16-bit floating-point values to rd.

Details

Returns the half-precision (16-bit) floating-point maximum of rs1 and rs2, following IEEE 754-2019 minNum/maxNum semantics. Quiet NaN inputs return the non-NaN operand; signalling NaN inputs raise invalid-operation.

Pseudocode Operation

F[rd] = max(F[rs1], F[rs2]);

Example

FMAX.H t0, a0, a1

Encoding

Binary Layout
0010100
31:25
rs2
24:20
rs1
19:15
001
14:12
rd
11:7
1010011
6:0
 
Format R-Type
Opcode 0x53
Extension Zfh

Operands

  • rd
    Destination register (integer)
  • rs1
    Source register 1 (integer)
  • rs2
    Source register 2 (integer)