Operands and Addressing Modes

The PowerPC processor architecture has only one addressing mode for load and store instructions: register plus displacement. The general form for address operands is:

displacement ( register )

If there is no displacement, the parentheses around the register name must still be used. For example, the first two of the following statements are legal, but the last isn't:

To specify an arbitrary 32-bit address, two instructions must be used, since all instructions are 32 bits long and can't contain both an opcode and a full address. A pair of instructions used to load or store data at an address falls into one of a small set of idioms, using the assembler operators lo16() , hi16() , and ha16() to isolate the required portion of the 32-bit address expression. The idioms themselves are discussed below

• lo16( expression ) evaluates to the low (least significant) 16 bits of expression , with a relocation type of PPC_RELOC_LO16 or PPC_RELOC_LO14, depending on the instruction the operator is used with.
• hi16( expression ) evaluates to the high (most significant) 16 bits of expression shifted right 16 bits, with a relocation type of PPC_RELOC_HI16.
• ha16( expression ) evaluates to the high (most significant) 16 bits of expression shifted right 16 bits, incremented by one if bit 15 of expression is set (that is, if the value given by lo16( expression ) is negative). This allows the address to be properly reconstituted when the low 16 bit quantity of expression is sign-extended by some operators. It has a relocation type of PPC_RELOC_HA16.

In specifying a 32-bit address, the desired result is that the 32-bit quantity be in a register. To do this, the high and low 16 bits of the address are entered separately with instructions suited to this task. Generally, the high 16 bits can be entered into a register with the addis (Add Immediate Shifted) operator. For example, this instruction:

adds the high 16 bits of expr to 0, and enters the result into the high 16 bits of register 2. The instruction that immediately follows can then combine the low 16 bits with the high 16 bits in the register and perform whatever other operation it does (if any).

For example, to load the address of the global variable spot into general register 2, the instructions below would be used. The ori instruction doesn't sign-extend the displacement, so the high 16 bits of the address needn't be adjusted, and the hi16() assembler operator is used.

In loading the data stored at spot the lwz operator is used, which does sign-extend the displacement, the adjusted high 16 bits must be given, with the ha16() assembler operator:

lwz treats the sign-extended low 16 bits as a displacement, adding it to the contents of register 2 to get a 32-bit address, and then loads the word at that address into register 3.

Previous | Contents | Next