[Ipsaus-adlc] Simulador de segmentación de MIPS

Jue Jun 9 04:37:21 CEST 2011

copado el simulador este para hacer pruebas con las 5 etapas de 
segmentación de MIPS.

saludos

Nico

http://www.ecs.umass.edu/ece/koren/architecture/windlx/main.html

Overview
This application is a partial simulation of a MIPS Five Stage Pipeline. 
The user can pick from a limited number of assembly instructions to be 
executed in the pipeline. Instructions in the pipeline are checked for 
"possible" data and structural hazards before execution. During 
execution, the instructions can be stepped through the stages of the 
pipeline one cycle at a time, or the final results can be displayed 
instantly. The display shows where stalls have occurred in the pipeline. 
Data forwarding can be enabled or disabled.

The simulation does not allow the user to insert specific values into 
the registers and does not compute the actual result of any of the 
instructions. This is simply a tool that should be used to detect 
possible data and structural hazards and to view instructions as they 
are executed on a five stage MIPS pipeline.

Instructions are issued "in-order" so WAR hazards can never occur while 
the instructions are being executed. However, WAR hazards are listed 
under the "Potential Hazards" for completeness.

It is assumed that there is a sperate data cache and instruction cache 
so that stalls will never occur due to an instruction being read from 
memory at the same time as a piece of data is being read from or written 
to memory.

Registers
The simulation supports 16 Floating-Point Registers (R1 - R16) and 32 
Integer Registers (F1 - F32).

Assembly Instructions
The following assembly instructions are supported:

FP_Add(dest, src1, src2)
The floating-point add instruction adds the two source registers and 
stores the result in the destination register. This instruction can take 
from 1 - 20 cycles to complete execution. The execution time is 
selectable by the user and is always the same as the FP_Sub instruction.

FP_Sub(dest, src1, src2)
The floating-point subtract instruction subtracts the two source 
registers and stores the result in the destination register. This 
instruction can take from 1 - 20 cycles to complete execution. The 
execution time is selectable by the user and is always the same as the 
FP_Add instruction.

FP_Mult(dest, src1, src2)
The floating-point multiply instruction multiplies the two source 
registers and stores the result in the destination register. This 
instruction can take from 1 - 20 cycles to complete execution. The 
execution time is selectable by the user.

FP_Div(dest, src1, src2)
The floating-point divide instruction divides the two source registers 
and stores the result in the destination register. This instruction can 
take from 1 - 20 cycles to complete execution. The execution time is 
selectable by the user.

FP_Load(dest, offset, src1)
The floating-point load instruction uses an integer value stored in 
source register # 1 and an offset to compute the memory address from 
which data will be loaded into the destination register. This 
instruction takes 1 cycle during the execution stage to compute the 
memory address. The value from memory is available after the MEM stage. 
It is assumed that the requested data is always available in the data 
cache so that the pipeline will never have to stall while data is read 
from the memory.

FP_Store(src1, offset, src2)
The floating-point store instruction uses an integer value stored in 
source register # 2 and an offset to compute the memory address to which 
the data in source register # 1 will be stored. This instruction takes 1 
cycle during the execution stage to compute the memory address. The 
value is written to memory by the end of the MEM stage. It is assumed 
that the data to be written to memory is already available in the data 
cache so that the pipeline will never have to stall while data is 
written to memory. This instruction does not go through the WB stage.

INT_Add(dest, src1, src2)
The integer add instruction adds the two source registers and stores the 
result in the destination register. This instruction takes 1 cycle to 
complete execution.

INT_Sub(dest, src1, src2)
The integer subtract instruction subtracts the two source registers and 
stores the result in the destination register. This instruction takes 1 
cycle to complete execution.

INT_Mult(dest, src1, src2)
The integer multiply instruction multiplies the two source registers and 
stores the result in the destination register. This instruction takes 1 
cycle to complete execution.

INT_Div(dest, src1, src2)
The integer divide instruction divides the two source registers and 
stores the result in the destination register. This instruction can take 
from 1 - 20 cycles to complete execution. The execution time is 
selectable by the user.

INT_Load(dest, offset, src1)
The integer load instruction uses an integer value stored in source 
register # 1 and an offset to compute the memory address from which data 
will be loaded into the destination register. This instruction takes 1 
cycle during the execution stage to compute the memory address. The 
value from memory is available after the MEM stage. It is assumed that 
the requested data is always available in the data cache so that the 
pipeline will never have to stall while data is read from the memory.

INT_Store(src1, offset, src2)
The integer store instruction uses an integer value stored in source 
register # 2 and an offset to compute the memory address to which the 
data in source register # 1 will be stored. This instruction takes 1 
cycle during the execution stage to compute the memory address. The 
value is written to memory by the end of the MEM stage. It is assumed 
that the data to be written to memory is already available in the data 
cache so that the pipeline will never have to stall while data is 
written to memory. This instruction does not go through the WB stage.

BR_Taken(offset, src1)
The branch taken instruction uses an integer value stored in source 
register #1 and an offset to compute the branch target memory address. 
This instruction assumes that the branch condition evaluates to "TRUE." 
Thus if there is an instruction immediately following this instruction 
in the pipeline, it must be canceled. It is assumed that the branch 
target address and the branch condition evaluation can be determined by 
the end if the ID stage. Thus this instruction never enters the EX stage.

BR_Untaken(offset, src1)
The branch untaken instruction uses an integer value stored in source 
register #1 and an offset to compute the branch target memory address. 
This instruction assumes that the branch condition evaluates to "FALSE." 
Thus if there is an instruction immediately following this instruction 
in the pipeline, it will continue to execute without any problems. It is 
assumed that the branch target address and the branch condition 
evaluation can be determined by the end if the ID stage. Thus this 
instruction never enters the EX stage.

Structural Units
This simulation contains 6 structural units that can cause a structural 
hazard.
1.   Floating Point Add/Subtract Unit.
2.   Floating Point Multiply Unit.
3.   Floating Point Divide Unit.
4.   Integer Add/Subtract Unit.
5.   Integer Multiply Unit.
6.   Integer Divide Unit.

Note: The load and store instructions use the Integer Add/Subtract Unit 
to compute the memory address. Therefore, a floating-point load or store 
can be stalled by a floating-point add or subtract that takes longer 
than one cycle to complete.

Using the Application
The application loads into two frames. Initially, the bottom frame will 
be empty.

Top Frame
The top frame allows the user to enter and remove instructions from the 
pipeline. It also allows the user to set the execution time for several 
of the instructions and to enable or disable data forwarding.

Adding an Instruction
To add an instruction to the pipeline, the user first selects the name 
of the instruction they would like to add from the dropdown box. Once 
the name of the instruction has been selected, the user can select the 
registers they would like to associate with this instruction following 
the Assembly Instruction format. When the registers have been selected, 
the user should click the "Insert Instruction" button. This will cause 
the current state of the pipeline to appear in the bottom frame. The 
user can continue entering instructions repeating this procedure at any 
time.

Removing an Instruction
In the event that a user makes a mistake when entering an instruction 
into the pipeline, the last instruction can always be removed from the 
pipeline by clicking on the "Remove Instruction" button.

Data Forwarding
Data forwarding can be enabled by placing a checkmark in the "Data 
Forwarding" checkbox. Data forwarding can be disabled by removing the 
checkmark from the "Data Forwarding" checkbox. Whenever this value is 
changed, the bottom frame is updated to reflect the new state of the 
pipeline.

Resetting the Application.
The application can be reset at any time by clicking on the "Reset 
Application" button.

Help
This help page can be accessed at any time by clicking on the "Help" button.

Setting the Instruction Execution Time
There are 4 functional units for which the execution time can be 
modified by the user. To change the execution time for a particular 
functional unit, select the dropdown box next to the functional until 
and select the number of cycles that the functional unit should take to 
complete an operation. Whenever an execution time is changed, the bottom 
frame is updated to reflect the new state of the pipeline.

Bottom Frame
The bottom frame is divided into two sections. The upper section shows 
the instructions in the pipeline and what stage each instruction is in 
as it is executed. The lower section is used to list the potential 
structural and data hazards.

Stepping Through the Instructions
The user can step through the instructions in the pipeline one cycle at 
a time by clicking on the "Step" button repeatedly.

Executing All Instructions
The user can execute all instruction to completion in one step by 
clicking on the "Execute All Instructions" button.