move-result-wide

Purpose

Move the double-word result of the most recent invoke-kind into the indicated register pair. This must be done as the instruction immediately after an invoke-kind whose (double-word) result is not to be ignored; anywhere else is invalid.

Details

Constraints

• A+1 must be a valid register index in the current stack frame (which includes A itself being valid).
• The instruction must be immediately preceded (in the code array) by an invoke-kind instruction.
• The instruction must be immediately reached (in the actual control flow) through returning from this invoke-kind instruction (it must not be jumped to).
• The result delivered by the invoke-kind instruction must be either a long or a double value.

Behavior

• The upper 32 bits of the result delivered by the invoke-kind instruction are moved to register vA, that is, vA' = result >> 0x20.
• The lower 32 bits of the result delivered by the invoke-kind instruction are moved to register v(A+1), that is, v(A+1)' = result & 0xffffffff.
• If register v(A-1) is the lower half of a register pair, register v(A-1)' becomes undefined.
• If register v(A+2) is the upper half of a register pair, register v(A+2)' becomes undefined.

Exceptions

None.

Notes

This instruction can also be thought of as reading the contents of a special "result" register that is made valid and defined by executing a non-void return instruction or a filled-new-array instruction. The execution of any other instruction (including this one) renders this special register invalid.