U of M CS 5641 - Intermediate Code and Optimizations

Unformatted text preview:

1Intermediate Code and Optimizations We have discussed Runtime organization Simple stack machine code generation Improvements to stack machine code generation Our compiler goes directly from AST to assembly language And does not perform optimizations Most real compilers use intermediate languagesWhy Intermediate Languages ? When to perform optimizations On AST Pro: Machine independent Con: Too high level On assembly language Pro: Exposes optimization opportunities Con: Machine dependent Con: Must re-implement optimizations when re-targetting On an intermediate language Pro: Machine independent Pro: Exposes optimization opportunities2Intermediate Languages Each compiler uses its own intermediate language IL design is still an active area of research Intermediate language = high-level assembly language Uses register names, but has an unlimited number Uses control structures like assembly language Uses op-codes but some are higher level E.g., push translates to several assembly instructions Most op-codes correspond directly to assembly op-codesThree-Address Intermediate Code Each instruction is of the formx := y op z y and z can be only registers or constants Just like assembly  Common form of intermediate code The AST expression x + y * z is translated ast1:= y * z3Generating Intermediate Code Similar to assembly code generation Major difference Use any number of IL registers to hold intermediate resultsGenerating Int. Code (Cont.) Igen(e, t) function generates code to compute the value of e in register t Example:igen(e1+ e2, t) = igen(e1, t1) (t1is a fresh register)igen(e2, t2) (t2is a fresh register)t := t1+ t2 Unlimited number of registers⇒ simple code generation4An Intermediate LanguageP → S P | εS → id := id op id| id := op id| id := id| push id| id := pop| if id relop id goto L| L:| jump L• id’s are register names• Constants can replace id’s• Typical operators: +, -, *Definition: Basic Blocks A basic block is a maximal sequence of instructions with:  no labels (except at the first instruction), and  no jumps (except in the last instruction) Idea:  Cannot jump into a basic block (except at beginning) Cannot jump out of a basic block (except at end) Each instruction in a basic block is executed after all the preceding instructions have been executed5Basic Block Example Consider the basic block1. L: 2. t := 2 * x3. w := t + x4. if w > 0 goto L No way for (3) to be executed without (2) having been executed right before We know we can change (3) to w := 3 * x Can we eliminate (2) as well?Definition. Control-Flow Graphs A control-flow graph is a directed graph with Basic blocks as nodes An edge from block A to block B if the execution can flow from the last instruction in A to the first instruction in B E.g., the last instruction in A is jump LB E.g., the execution can fall-through from block A to block B6Control-Flow Graphs. Example. The body of a method (or procedure) can be represented as a control-flow graph There is one initial node All “return” nodes are terminalx := 1i := 1L:x := x * xi := i + 1if i < 10 goto LOptimization Overview Optimization seeks to improve a program’s utilization of some resource Execution time (most often) Code size Network messages sent, etc. Optimization should not alter what the program computes The answer must still be the same7A Classification of Optimizations For languages like C and Java there are three granularities of optimizations1. Local optimizations Apply to a basic block in isolation2. Global optimizations Apply to a control-flow graph (method body) in isolation3. Inter-procedural optimizations Apply across method boundaries Most compilers do (1), many do (2) and very few do (3)Cost of Optimizations In practice, a conscious decision is made not to implement the fanciest optimization known Why? Some optimizations are hard to implement Some optimizations are costly in terms of compilation time The fancy optimizations are both hard and costly The goal: maximum improvement with minimum of cost8Local Optimizations The simplest form of optimizations No need to analyze the whole procedure body Just the basic block in question Example: algebraic simplificationAlgebraic Simplification Some statements can be deletedx := x + 0x := x * 1 Some statements can be simplifiedx := x * 0 ⇒ x := 0y := y ** 2 ⇒ y := y * yx := x * 8 ⇒ x := x << 3x := x * 15 ⇒ t := x << 4; x := t - x(on some machines << is faster than *; but not on all!)9Constant Folding Operations on constants can be computed at compile time In general, if there is a statementx := y op z And y and z are constants Then y op z can be computed at compile time Example: x := 2 + 2 ⇒ x := 4 Example: if 2 < 0 jump L can be deleted When might constant folding be dangerous?Flow of Control Optimizations Eliminating unreachable code: Code that is unreachable in the control-flow graph Basic blocks that are not the target of any jump or “fall through” from a conditional Such basic blocks can be eliminated Why would such basic blocks occur? Removing unreachable code makes the program smaller And sometimes also faster Due to memory cache effects (increased spatial locality)10Single Assignment Form Some optimizations are simplified if each register occurs only once on the left-hand side of an assignment Intermediate code can be rewritten to be in single assignment formx := z + y b := z + ya := x ⇒ a := bx := 2 * x x := 2 * b(b is a fresh register) More complicated in general, due to loopsCommon Sub-expression Elimination Assume Basic block is in single assignment form A definition x := is the first use of x in a block If any assignment have the same rhs, they compute the same value Example:x := y + z x := y + z… ⇒ …w := y + z w := x(the values of x, y, and z do not change in the … code)11Copy Propagation If w := x appears in a block, all subsequent uses of w can be replaced with uses of x Example:b := z + y b := z + ya := b ⇒ a := bx := 2 * a x := 2 * b This does not make the program


View Full Document

U of M CS 5641 - Intermediate Code and Optimizations

Download Intermediate Code and Optimizations
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Intermediate Code and Optimizations and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Intermediate Code and Optimizations 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?