Instruction level parallelism

Instruction-level parallelism (ILP) is a measure of how many of the operations in a computer program can be performed simultaneously. Consider the following program:

1. e = a + b 2. f = c + d 3. g = e * f

Operation 3 depends on the results of operations 1 and 2, so it cannot be calculated until both of them are completed. However, operations 1 and 2 do not depend on any other operation, so they can be calculated simultaneously. (See also: "Data dependency") If we assume that each operation can be completed in one unit of time then these three instructions can be completed in a total of two units of time, giving an ILP of 3/2.

A goal of compiler and processor designers is to identify and take advantage of as much ILP as possible. Ordinary programs are typically written under a sequential execution model where instructions execute one after the other and in the order specified by the programmer. ILP allows the compiler and the processor to overlap the execution of multiple instructions or even to change the order in which instructions are executed.

How much ILP exists in programs is very application specific. In certain fields, such as graphics and scientific computing the amount can be very large. However, workloads such as cryptography exhibit much less parallelism.

Micro-architectural techniques that are used to exploit ILP include:
*Instruction pipelining where the execution of multiple instructions can be partially overlapped.
*Superscalar execution in which multiple execution units are used to execute multiple instructions in parallel. In typical superscalar processors, the instructions executing simultaneously are adjacent in the original program order.
*Out-of-order execution where instructions execute in any order that does not violate data dependencies. Note that this technique is independent of both pipelining and superscalar.
*Register renaming which refers to a technique used to avoid unnecessary serialization of program operations imposed by the reuse of registers by those operations, used to enable out-of-order execution.
*Speculative execution which allow the execution of complete instructions or parts of instructions before being certain whether this execution should take place. A commonly used form of speculative execution is control flow speculation where instructions past a control flow instruction (e.g., a branch) are executed before the target of the control flow instruction is determined. Several other forms of speculative execution have been proposed and are in use including speculative execution driven by value prediction, memory dependence prediction and cache latency prediction.
*Branch prediction which is used to avoid stalling for control dependencies to be resolved. Branch prediction is used with speculative execution.

Current implementations of out-of-order execution dynamically (i.e., while the program is executing and without any help from the compiler) extract ILP from ordinary programs. An alternative is to extract this parallelism at compile time and somehow convey this information to the hardware. Due to the complexity of scaling the out-of-order execution technique, the industry has re-examined instruction sets which explicitly encode multiple independent operations per instruction. These instruction set types include:

*VLIW and the closely related Explicitly Parallel Instruction Computing concepts

Dataflow architectures are another class of architectures where ILP is explicitly specified, but they have not been actively researched since the 1980s.

In recent years, ILP techniques have been used to provide performance improvements in spite of the growing disparity between processor operating frequencies and memory access times (early ILP designs such as the IBM 360 used ILP techniques to overcome the limitations imposed by a relatively small register file). Presently, a cache miss penalty to main memory costs several hundreds of CPU cycles. While in principle it is possible to use ILP to tolerate even such memory latencies the associated resource and power dissipation costs are disproportionate. Moreover, the complexity and often the latency of the underlying hardware structures results in reduced operating frequency further reducing any benefits. Hence, the aforementioned techniques prove inadequate to keep the CPU from stalling for the off-chip data. Instead, the industry is heading towards exploiting higher levels of parallelism that can be exploited through techniques such as multiprocessing and multithreading. [ [http://www.csl.cornell.edu/~sam/papers/cf04.pdf Reflections of the Memory Wall] ]

ee also

*Data dependency
*Memory level parallelism, MLP.

References

External links

* [http://www.wired.com/wired/archive/4.08/geek.html Wired magazine article that refers to the above paper]
* [http://www.itee.uq.edu.au/~philip/Publications/Techreports/2002/Reports/memory-wall-survey.pdf Approaches to addressing the Memory Wall]