Table of Contents
Advertisement
APPENDIX B V
The V
4120A core avoids contention of its internal resources by causing a pipeline interlock in such cases as when
R
the contents of the destination register of an instruction are used as a source in the succeeding instruction. Therefore,
instructions such as NOP must not be inserted between instructions.
However, interlocks do not occur on the operations related to the CP0 registers and the TLB.
contention of internal resources should be considered when composing a program that manipulates the CP0 registers
or the TLB. The CP0 hazards define the number of NOP instructions that is required to avoid contention of internal
resources, or the number of instructions unrelated to contention. This chapter describes the CP0 hazards.
The CP0 hazards of the V
Coprocessor 0 hazards of the V
the V
4000.
R
The contents of the CP0 registers or the bits in the "Source" column of this table can be used as a source after
they are fixed.
The contents of the CP0 registers or the bits in the "Destination" column of this table can be available as a
destination after they are stored.
Based on this table, the number of NOP instructions required between instructions related to the TLB is computed
by the following formula, and so is the number of instructions unrelated to contention:
(Destination Hazard number of A) – [(Source Hazard number of B) + 1]
As an example, to compute the number of instructions required between an MTC0 and a subsequent MFC0
instruction, this is:
(5) – (3 + 1) = 1 instruction
The CP0 hazards do not generate interlocks of pipeline. Therefore, the required number of instruction must be
controlled by program.
590
4120A COPROCESSOR 0 HAZARDS
R
4120A core are as or less stringent than those of the V
R
4120A core. Code that complies with these hazards will run without modification on
R
Preliminary User's Manual S15543EJ1V0UM
Therefore,
4000. Table B-1 lists the
R
Advertisement
Table of Contents
