# ================================================================== # 64-bit Architecture ## ================================================================= # If you want to specifically test negative normal numbers, # negative subnormal numbers, or negative zero, you can use the # class instruction. # a. instruction: fclass.s rd, rs1 # b. how it works: It inspects rs1 and writes a 10-bit bitmask to # integer register rd. # c. bit positions for negative values in rd: # Bit 0: negative infinity (-inf) # Bit 1: negative normal number (-normal) # Bit 2: negative subnormal number (-subnormal) # Bit 3: negative zero (-0.0) (-0) # d. example assembly code # fclass.s x5, fa0 # get classification bitmask of fa0 # andi x5, x5, 15 # check if any of the last 4 bits are set # bnez x5, is_negative # branch if negative # ================================================================== .data pi: .double 3.141592653589793 .text .global main main: # ---- load pi into a register la t0, pi # load address of pi fld ft0, 0(t0) # load pi into register ft0 # ---- test if (double) is negative fclass.d t0, ft0 # test register ft0 andi t0, t0, 15 # mask test bits (0b1111) bnez t0, is_negative # branch if negative