PRESENTED BY:
Sasmita Kumari Misra

---

[attachment=11492]
Itanium Processor
 The Itanium processor family came about for several reasons, but the primary one was that the processor architecture advances of RISC were no longer growing at the rate seen in the 1980’s or the 1990’s.Yet,customers continued demand greater application performance.
 The Itanium processor family was developed as a response to address the future performance and growth needs of business, technical and scientific users with greater flexibility, better performance and a much greater ‘bang for the buck’ in the price performance arena.
 The Itanium architecture achieves a more difficult goal than a processor that could have been designed with ‘price as no object’.
Overview
History
 32 bit Processors (Pentium Pro, Pentium Xeon)
 64 bit Processors (Xeon, Itanium, Itanium 2)
 ISA
 EPIC
 Predicated Execution (Branch Prediction)
 Software Pipelining
Overview
 ISA cont.
 Register Stacking
 IA-32 Emulation
 Speculation
 Architecture
 Benchmarks
 ISA Overview
 Most Modern Processors:
 Instruction Level Parallelism (ILP)
 Processor, at runtime, decides which instructions have no dependencies
 Hardware branch prediction
 Itanium’s ISA
 IA-64 – Intel’s (first) 64-bit ISA
 Not an extension to x86(sucks) (Completely new ISA)
 Allows for speedups without engineering “tricks”
 Largely RISC
 Surrounded by patents
 IA-64
 IA-64 largely depends on software for parallelism
 VLIW – Very Long Instruction Word
 EPIC – Explicitly Parallel Instruction Computer
 IA-64
 VLIW – Overview
 RISC technique
 Bundles of instructions to be run in parallel
 Similar to superscaling
 Uses compiler instead of branch prediction hardware
 IA-64
 EPIC – Overview
 Builds on VLIW
 Redefines instruction format
 Instruction coding tells CPU how to process data
 Very compiler dependent
 Predicated execution
 IA-64
 Predicated Execution:
 Decrease need for branch prediction
 Increase number of speculative executions
 Branch conditions put into predicate registers
 Predicate registers kill results of executions from not-taken branch
 IA-64
 Software Pipelining:
 Take advantage of programming trends and large number of available registers
 Allow multiple iterations of a loop to be in flight at once
 IA-64
 EPIC – Pros:
 Compiler has more time to spend with code
 Time spent by compiler is a one-time cost
 Reduces circuit complexity
 IA-64
 EPIC – Cons:
 Runtime behavior isn’t always obvious in source code
 Runtime behavior may depend on input data
 Depends greatly on compiler performance
 IA-64
 IA-32 Support:
 Done with hardware emulation
 Uses special jump escape instructions to access
 Slow (painfully so)
 IA-64
 32 Bit Hardware Emulation - Very Poor Performance
 Software Emulation of x86 32-bit from either Microsoft or Linux can perform 50% better than Intel’s Hardware Emulation
 Less than 1% of the chip devoted to Hardware Emulation
 IA-64
 IA-32 Slowness:
 No out-of-order execution abilities
 Functional units don’t generate flags
 Multiple outstanding unaligned memory loads not supported
 IA-64
 IA-32 Support:
 Hardware emulation augmented for Itanium 2
 Software emulation (IA-32 Execution Layer) added
 Runs IA-32 code at same speed as equivalently clocked Xeon
 Architecture
 Physical Layout
 Conceptual Design Elements
 Register Stack Engine (RSE)
 Improve performance by removing latency associated with saving/restoring state for function calls
 Hardware implementation of register stack ISA functionality
 Itanium Pipeline
 10 Stage
 Instruction Pointer Generation
 Fetch
 Rotate
 Expand
 Rename
 Word-Line Decode
 Register Read
 Execute
 Exception Detect
 Write Back
 Itanium 2 Pipeline
 8 stage
 Instruction Pointer Generation
 Rotate
 Expand
 Rename
 Register Read
 Execute
 Detect
 Write Back
 Processor Abstraction Layer (PAL)
 Internal processor firmware
 External system firmware
 Itanium 2 Execution Core
 6 multimedia units
 6 integer units
 2 FPU
 3 branch units
 4 load / store units