Introduction to Pentium

Processor
 Features of Pentium Processor
• Separate instruction and Data caches.
• Dual integer pipelines i.e. U-pipeline and V-
Pipeline.
• Branch prediction using the branch target buffer
(BTB).
• Pipeliened floating point unit.(a part of a computer system specially
designed to carry out operations on floating-point numbers)

• 64- bit external data bus.

• Even-parity checking is implemented for data bus,
caches and TLBs(A translation lookaside buffer (TLB) is a memory cache that
stores the recent translations of virtual memory to physical memory) .
 Salient Features
• 32- bit Superscalar and super-pipelined
architecture CISC processor.
• 32-bit address bus can address up to 4GB of
physical memory.
• 64- bit data bus so arithmetic and logical
operation can be perform on 64-bit operand.
• Two integer pipeline U and V with two ALU’s
provide one-clock execution for core instructions
which Improved Instructions to execute Time.
• Support five stage pipeline enables multiple
instructions to execute in parallel with high
efficiency.
• Two 8 KB caches memories, one for data and
other for code.
• On- chip pipelined floating point coprocessor.
• Support power management feature i.e.
System Management mode and Clock Control.
• Enhanced branch prediction buffer.
• On-chip memory management unit.
• Support multiprogramming and multitasking.
• Internal error Detection features.
• 4MB pages for increased TLB Hit Rate.
• Support for Second level cache and Write Back
MESI (Modified, Exclusive, Shared, Invalid)
Protocol in the data cache.
• Supports Bus cycle Pipelining, Address Parity
and Internal Parity Checking, Functional
Redundancy checking, Execution Tracing,
Performance Monitoring.
 Superscalar Execution
• It supports superscalar pipeline architecture.
• The Pentium processor sends two instructions
in parallel to the two independent integer
pipeline known as U and V pipelines for
execution of multiple instructions
concurrently.
• Thus, processor capable of parallel instruction
execution of multiple instructions is known as
Superscalar Machine.
• Each of these pipeline i.e. U and V, has various
stages of execution i.e.
– Prefetch
– First Decode
– Second Decode
– Execute
– Write Back
1. Pre-fetch (PF): In this stage, instructions are
prefetched by prefetch buffer through U and
V pipelene from the on-chip instruction
cache.
2. First Decode (D1): In this stage, two decoders
decode the instructions to generate a control
word and try to pair them together so they
can run in parallel.
3. Second Decode (D2): In this stage, the CPU
decodes the control word and calculates the
address of memory operand.
4. Execute (EX): The instruction is executed in
ALU and data cache is accessed at this stage.
For both ALU and data cache access requires
more than one clock.
5. Write Back (WB): In this stage, the CPU stores
result and update the flags.
 Separate Code and Data Caches:
• Pentium has two separate 8KB caches for code
and data as the superscalar design and branch
prediction need more bandwidth than a unified
cache.
• The data cache has a dedicated TLB to translate
linear address into physical address used by data
cache.
• The code cache is responsible for getting raw
instructions into execution units of the Pentium
processor and hence instructions are fetched from
the code cache.
 Instruction & Data Cache of Pentium
• Both caches are organized as 2-way set
associative caches with 128 sets (total 256
entries)
• There are 32 bytes in a line (8K/256)
• An LRU(Least Recently Used) algorithm(LRU
algorithm to decide where to place the new
line) is used to select victims in each cache.
 Structure of 8KB instruction and data
cache
• Each entry in a set
has its own tag.
• Tags in the data
cache are triple
ported, used for
– U pipeline
– V pipeline
– Bus snooping

Bus snooping or bus sniffing is a scheme by which a coherency controller (snooper) in a

cache (a snoopy cache) monitors or snoops the bus transactions, and its goal is to
maintain a cache coherency in distributed shared memory systems.
 Instruction Cache of Pentium
• Instruction cache is write protected to
prevent self-modifying code.
• Tags in instruction cache are also triple
ported
– Two ports for split-line accesses
– Third port for bus snooping
Code Cache organization
 Split-line Access
• In Pentium (since CISC), instructions are of
variable length(1-15bytes)
• Multibyte instructions may staddle two
sequential lines stored in code cache
• Then it has to go for two sequential access
which degrades performance.
• Solution: Split line Access
Split-line Access
 Split-line Access
• It permits upper half of one line and lower
half of next to be fetched from code cache in
one clock cycle.
• When split-line is read, the information is not
correctly aligned.
• The bytes need to be rotated so that prefetch
queue receives instruction in proper order.
 Instruction & Data Cache of Pentium
• Parity bits are used to maintain data integrity
• Each tag and every byte in data cache has its
own parity bit.
• There is one parity bit for every 8 byte of data
in instruction cache.
Data Cache organization
Data Cache organization
 Advantages of having Separate Code
and Data Caches
• Separate caches efficiently execute the branch
prediction.
• Caches raise system performance i.e. an internal read
request is performed more quickly than a bus cycle to
memory.
• Separate caches also reduce the processor’s use of the
external bus when the same location are accessed
multiple times.
• Separate caches for instructions a and data allow
simultaneous cache look-up.
• Up to two data references and up to 32 bytes of raw
op-codes can be accessed In one clock.
 Branch Prediction
• Branch prediction is used to predict the most
likely set of instruction to be executed and
prefetch to make them available to the pipeline as
and when they are called.
• Hence the pentium processor incorporate a
branch target buffer (BTB), which is an associative
memory used to improve the performance if it
takes the branch instrction.
• Branch instructions of pentium processor change
the normal sequential control flow of the program
execution and may stall the pipelined execution in
the pentium system.
• Branches instruction is of two types i.e.
conditional and unconditional branch.
• During the conditional branching, the CPU has two
wait till the execution stage to determine whether
the condition is satisfied or not.
• When the condition satisfies, a branching is to be
taken using branch prediction algorithm for speed
up of the instruction execution.
• In pentium Processor, BTB can have 256 entries
which contains the branch target address for
previously executed branches.
• The BTB is four ways set associative on-chip
memory.
• Whenever the branching occurs, the CPU checks the
branch instruction address and the destination
address in the BTB.
• When the instruction is decoded, the CPU searches
the branch target buffer to decide whether any entry
exists for a corresponding branch instruction.
• If BTB is hit, i.e. if BTB exist such entries, then the CPU
use the history to decide whether the branch will be
taken or not.
• If the entry exist in its previous history in BTB to take
the branch, the CPU fetches the instruction from the
target address and decodes them.
• If the branch prediction is correct, the process
continue else the CPU clears the pipeline and fetches
from the appropriate target address.
Branch Prediction
 Floating point Unit
• The Pentium contains an on chip floating point
unit that provides significant floating point
performance advantage over previous
generations of processors.
• Providing the coprocessor onto the same chip
as the processor, Pentium allows faster
communication and quicker execution.
• Thus many floating point instructions requires
fewer clock cycles that the previous 80X87.
 Floating Point Pipeline
• The floating point pipeline of Pentium consists of
eight stages which are used to speedup the
execution of floating point unit. Hence, it is
necessary floating point pipeline.
• These are prefetch, first decode, second decode,
operand fetch, first execute, second execute, write
float and error reporting.
• Floating point unit has eight stage pipeline which
gives a single cycle execution for many of floating
point instructions such as floating adds, subtract,
multiply and compare
• The stages and their functions are given below:
1. PF (prefetch): Instructions are prefetched from
the on chip instruction cache.
2. D1 (first decode): Instruction decode to generate
control word. A single control word causes direct
execution of an instruction and complex
instruction require micro-coded control
sequence.
3. D2 (second decode): Address of memory
resident operand are calculated.
4. Ex (operand fetch): In this stage, register read,
memory read or memory write operation is
performed to access an operand as required by
the instruction.
5. X1(first execute): In this stage, the floating point
data from register or memory is written into
floating point register or memory is written into
floating point register and converted to floating
point format before loaded into the floating
point unit.
6. X2(second execute): In this stage, the floating
point operation is performed by floating point
unit.
7. WF(write float ): In this stage, results of floating
point operation are rounded and then written to
the destination floating point register.
8. ER(error reporting): In this stage, if any error is
occurred during floating point execution, then it
is reported and FPU status word is updated.
 Floating point Exception
• There are six floating point exception condition
while execution floating point instructions are
available in status word register of Pentium FPU.
• These exceptions are:
• Stack overflow or underflow
• Invalid arithmetic operation
This exception occurs when stack fault flag (SF) of the
FPU status word indicates the type of operation i.e. stack
overflow or Underflow for SF=1 and an arithmetic
instruction has encountered an invalid operand for SF=1.
• Divide by zero: This exception occurs whenever an
instruction attempts to divide a finite non-zero operand
by 0.
• De-normalized operand exception: The de-normal
operand exception occurs if an arithmetic instruction
attempts to operate on a de-normal operand or if an
attempt is made to load de-normal single or double real
value into an FPU register.
• Numeric flow exception: This exception occurs whenever
the rounded result of an arithmetic instruction is less
than the smallest possible normalized, finite value that
will fit into the real format of the destination operand.
• Inexact result (Precision) Exception: This exception
occurs if the result of an operation is not exactly
representable in the destination format.
 Cache Coherency in Multiprocessor
System
• When multiple processors are used in a single
system, there needs to be a mechanism
whereby all processors agree on the contents
of shared cache information.
• For e.g., two or more processors may utilize
data from the same memory location,X.
• Each processor may change value of X, thus
which value of X has to be considered?
 Cache coherency in Multiprocessor
Systems
• If each processor change the value of the data
item, we have different(incoherent) values of
X’s data in each cache.
• Solution : Cache Coherency Mechanism
 Cache Coherency
• Pentium’s mechanism is called MESI
(Modified/Exclusive/Shared/Invalid)Protocol.
• This protocol uses two bits stored with each
line of data to keep track of the state of cache
line.
 Cache Coherency
• The four states are defined as follows:
• Modified:
– The current line has been modified and is only
available in a single cache.
• Exclusive:
– The current line has not been modified and is only
available in a single cache
– Writing to this line changes its state to modified
 Cache Coherency
• Shared:
– Copies of the current line may exist in more than
one cache.
– A write to this line causes a writethrough to main
memory and may invalidate the copies in the
other cache
• Invalid:
– The current line is empty
– A read from this line will generate a miss
– A write will cause a writethrough to main memory
 Cache Coherency
• Only the shared and invalid states are used in
code cache.
• MESI protocol requires Pentium to monitor all
accesses to main memory in a multiprocessor
system. This is called bus snooping.
 Cache Coherency
• Inquire cycles examine the code cache as well
(as code cache supports bus snooping)
• Pentium’s address lines are used as inputs
during an inquire cycle to accomplish bus
snooping.
MESI Protocol
Comparison of 80386 and Pentium
80386 Pentium
32- bit integer core CPU with 32 bit 32 bit CPU with 64-bit data bus
Data bus
No superscalar architecture and single Superscalar architecture i.e. two
cycle execution pipelined Integer Units are capable of
2 Instructions per clock
No internal cache available for data Separate 8KB code and 8KB data
and code cache available
80386 does not support branch Advanced design feature i.e. Dynamic
prediction branch Prediction
One integer ALU Two integer ALU
Operating frequency are 20 MHz to 66 Operating frequency 60 MHz and
MHz more
FPU is non- pipelined as it is an FPU is pipelined as it is in built in
external device 80387 Pentium FPU