Scribd
Upload
51 views4 pages

Intermediate Representations ControlFlow

The document discusses Intermediate Representations (IR) and Control-Flow Analysis, focusing on Static Single Assignment (SSA) form, which ensures each variable is assigned once and simplifies analysis and optimizations. It explains Control Flow Graph (CFG) construction, dominance relations, and the differences between imperative and functional IR designs. Additionally, it covers def-use chains in SSA, highlighting their role in optimizations and the process of lowering SSA for register allocation.

Uploaded by

DhineshKings
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
51 views4 pages

Intermediate Representations ControlFlow

The document discusses Intermediate Representations (IR) and Control-Flow Analysis, focusing on Static Single Assignment (SSA) form, which ensures each variable is assigned once and simplifies analysis and optimizations. It explains Control Flow Graph (CFG) construction, dominance relations, and the differences between imperative and functional IR designs. Additionally, it covers def-use chains in SSA, highlighting their role in optimizations and the process of lowering SSA for register allocation.

Uploaded by

DhineshKings
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

Intermediate Representations &

Control-Flow Analysis
1. Static Single Assignment (SSA) form — concept & properties
Definition:

SSA (Static Single Assignment) requires that each variable be assigned exactly once,

and every use of a variable refers to a unique definition. Practically this is done

by renaming variables and inserting φ (phi) functions at control-flow merge points.

Key Points:

- Each value has a unique definition → simplifies analysis.

- φ functions select correct incoming values at join points.

- Simplifies optimizations: constant propagation, dead code elimination, etc.

Example (simplified SSA):

Original:

a=1

b=2

c=a+b

if c > 2 goto L1 else goto L2

L1: a = c + 1; goto L3

L2: b = c - 1; goto L3

L3: d = a + b

SSA:

a0 = 1

b0 = 2

c0 = a0 + b0

if c0 > 2 goto L1 else goto L2


L1: a1 = c0 + 1; goto L3

L2: b1 = c0 - 1; goto L3

L3: a2 = φ(a1, a0)

b2 = φ(b0, b1)

d0 = a2 + b2

2. Control Flow Graph (CFG) construction


Definition:

A Control-Flow Graph (CFG) is a directed graph where nodes are basic blocks and edges

represent possible control transfers.

Steps to build CFG:

1. Identify leaders (first instruction, jump targets, instruction after jumps).

2. Form basic blocks.

3. Add edges for fallthroughs and jumps.

Notes:

- Reducible CFGs: loops with single entry (dominance-based detection possible).

- Irreducible CFGs: multiple-entry loops (harder for optimization).

Example (CFG from the SSA example):

B0 → B1, B0 → B2

B1 → B3, B2 → B3

3. Dominance relations & dominance frontiers


Dominance:

- Node D dominates N if every path from entry to N passes through D.

- Immediate dominator: closest strict dominator of a node.

- Dominator tree: hierarchy based on immediate dominators.

Dominance Frontier (DF):


- DF(n) = { m | n dominates a predecessor of m but does not strictly dominate m }.

- Used for φ placement in SSA.

Algorithm:

- Compute dominators (e.g., Lengauer–Tarjan algorithm).

- Place φ nodes at blocks in the DF of definition sites of variables.

4. IR design for functional and imperative languages


Imperative IR:

- Three-Address Code (TAC), CFG-based.

- Explicit loads, stores, mutable variables.

- Examples: GCC GIMPLE, LLVM IR.

Functional IR:

- Based on immutability and expressions.

- Often represented in CPS (Continuation-Passing Style) or ANF (A-Normal Form).

- Natural fit for SSA because values are single-assignment.

Tradeoffs:

- Imperative IRs must model mutable memory and aliasing.

- Functional IRs emphasize closures, higher-order functions, and inlining.

5. SSA & def-use chains


Def-Use Chains:

- Map each definition to its uses (def→use).

- In SSA, every use maps to exactly one definition → simplifies analysis.

Benefits:

- Enables constant propagation, dead code elimination, global value numbering.

- Sparse Conditional Constant Propagation (SCCP) is efficient in SSA.

Maintaining Def-Use Chains:


- Update when inserting/removing/replacing instructions.

- Used for optimizations and transformations.

Lowering SSA:

- Translate φ functions into parallel copies on incoming edges.

- Register allocation resolves virtual registers to physical ones.

You might also like