Itanium Architecture for Software Developers

In the last section, we showed how to transfer floating-point data between memory and the floating-point registers, and between the general registers and floating-point registers. Once information is held in the floating-point register file, it is ready to be processed with other floating-point operations that are in the Itanium processor s floating-point instruction set. In this section, the floating-point arithmetic instructions are introduced. These instructions can perform a wide variety of arithmetic operations on floating-point numbers. For example, instructions are provided to find the negative or absolute value of a number, compare two numbers to find the minimum or maximum, and to perform arithmetic operations such as addition, subtraction, or multiplication. A floating-point division instruction is not supported in the instruction set. This arithmetic function must be performed in software.
For ease in understanding, the floating-point arithmetic instructions have been organized based on the number of source operands they employ in the computation. Table 9.15 summarizes the group of arithmetic instructions that involve one source operand. This table lists the instruction mnemonic, format, and a brief description of its operations. Note that most floating-point arithmetic instructions have two forms, one form for processing floating-point normal data and a second for processing floating-point parallel data. Since the operand combinations are identical for instructions that perform similar functions, we will just describe the operation of a typical instruction of each type.
| Mnemonic |
|---|