Dynamic Memory Allocation

CS61, Lecture 10
Prof. Stephen Chong
October 4, 2011
 Announcements 1/2

•Assignment 4: Malloc
•Will be released today
•May work in groups of one or two
• Please go to website and enter your group by Sunday
11:59PM
‣ Every group must do this (i.e., even if you are working individually)
•Two deadlines:
• Design checkpoint: Thursday October 13, 10pm
• Final submission: Thursday October 20, 11:59pm
•Encourage you to use version control
•Assignment 3 (Buffer bomb) due Thursday
Stephen Chong, Harvard University 2
 Announcements 2/2

•Give us feedback on when you’d like office hours

•http://tinyurl.com/689seas
•Link posted on website

Stephen Chong, Harvard University 3

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 4

 Harsh Reality: Memory Matters!
•Memory is not unlimited!
•It must be allocated and managed
•Many applications are memory dominated
• Especially those based on complex, graph algorithms
•Memory referencing bugs especially pernicious
•Effects are distant in both time and space
•Memory performance is not uniform
•Cache and virtual memory effects can greatly affect program
performance
•Adapting program to characteristics of memory system can lead
to major speed improvements
Stephen Chong, Harvard University 5
 A process's view of memory
Process address space
Process not allowed to read or 0xffffffff
write this region OS memory
0xc0000000
User stack
%esp

Shared libraries
0x40000000

Dynamically allocated
memory Heap (used by malloc)
Read/write segment
Global vars .data, .bss
Read-only segment
Program code .text, .rodata
0x08048000
unused
Stephen Chong, Harvard University 0x00000000 6
 The heap
• The heap is the region of a program's memory used for dynamic
allocation.
• Program can allocate and free blocks of memory within the heap.
Heap

Free byte
Allocated block Free block
(4 bytes) (3 bytes) Allocated byte

Stephen Chong, Harvard University 7

 Free blocks, and allocated blocks
• Heap starts out as a single big “free block” of some fixed size (say a few MB)
...

• Program may request memory, which splits up the the free space.
...

• Program may free up some memory, which increases the free space
...

• Over time the heap will contain a mixture of free and allocated blocks.
...

• Heap may need to grow in size (but typically never shrinks)

• Program can grow the heap if it is too small to handle an allocation request.
• On UNIX, the sbrk() system call is used to expand the size of the heap.

... ...

Stephen Chong, Harvard University 8

 Dynamic Memory Management
• How do we decide when do expand the heap (using sbrk())?
• How do we manage allocating and freeing bytes on the heap?
• There are two broad classes of memory management schemes:
• Explicit memory management
• Application code responsible for both explicitly allocating and freeing memory.
• Example: malloc() and free()
• Implicit memory management
• Application code can allocate memory, but does not explicitly free memory
• Rather, rely on garbage collection to “clean up” memory objects no longer in
use
• Used in languages like Java, Python, OCaml

Stephen Chong, Harvard University 9

 Malloc Package
• #include <stdlib.h>
• void *malloc(size_t size)
• If successful:
• Returns a pointer to a memory block of at least size bytes
• If size == 0, returns NULL
• If unsuccessful: returns NULL.
• void free(void *p)
• Returns the block pointed at by p to pool of available memory
• p must come from a previous call to malloc or realloc.
• void *realloc(void *p, size_t size)
• Changes size of block p and returns pointer to new block.
• Contents of new block unchanged up to min of old and new size.
Stephen Chong, Harvard University 10
 Allocation Examples
Heap

p1 = malloc(4)

p2 = malloc(5)

p3 = malloc(6)

free(p2)

p4 = malloc(2)

Stephen Chong, Harvard University 11

 Constraints
• Application code must obey following constraints:
• Allowed to issue arbitrary sequence of allocation and free requests
• Free requests must correspond to an allocated block
• Memory management code must obey following constraints:
• Can’t control number or size of requested blocks
• Must respond immediately to all allocation requests
• i.e., can’t reorder or buffer requests
• Must allocate blocks from free memory
• i.e., can only place allocated blocks in free memory
• Must align blocks so they satisfy all alignment requirements
• Can’t mess around with allocated memory
• Can only manipulate and modify free memory
• Can’t move the allocated blocks once they are allocated (i.e., compaction is not allowed)

Stephen Chong, Harvard University 12

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 13

 What do we want?
•Want our memory management to be:
•Fast
• Minimize overhead of allocation and deallocation
operations.
•Efficient
• Don’t waste memory space

Stephen Chong, Harvard University 14

 Performance Goals: Allocation overhead

•Want our memory allocator to be fast!

•Minimize the overhead of both allocation and
deallocation operations.
•One useful metric is throughput:
•Given a series of allocate or free requests
•Maximize the number of completed requests per unit
time
•Example:
•5,000 malloc calls and 5,000 free calls in 10 seconds
•Throughput is 1,000 operations/second.
Stephen Chong, Harvard University 15
 Performance Goals: Memory Utilization

• Allocators rarely do a perfect job of managing memory.

• Usually there is some “waste” involved in the process.
• Examples of waste...
• Extra metadata or internal structures used by the allocator itself
(example: Keeping track of where free memory is located) Heap
• Chunks of heap memory that are unallocated (fragments)
Allocated
• We define memory utilization as... memory
• The total amount of memory allocated to the
application divided by the total heap size
Unallocated
• Ideally, we'd like utilization to be to 100% fragments

• In practice this is not possible, but want to be

as close as possible
Metadata

Stephen Chong, Harvard University 16

 Conflicting performance goals
• High throughput and good utilization are difficult to
achieve simultaneously.
• A fast allocator may not be efficient in terms of memory
utilization.
•Faster allocators tend to be “sloppier” with their memory usage.
• Likewise, a space-efficient allocator may not be very fast
•To keep track of memory waste (i.e., tracking fragments), the
allocation operations generally take longer to run.
• Trick is to balance these two conflicting goals.

Stephen Chong, Harvard University 17

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs
•Alignment
Stephen Chong, Harvard University 18
 Internal Fragmentation
• Poor memory utilization caused by fragmentation.
• Comes in two forms: internal and external fragmentation
• Internal fragmentation
• Internal fragmentation is the difference between block size and payload size.
block

Internal Internal
fragmentation payload fragmentation

• Caused by overhead of maintaining heap data structures, padding for alignment

purposes, or the policy used by the memory allocator
• Example: Say the allocator always “rounds up” to next highest power of 2 when
allocating blocks.
• So malloc(1025) will actually allocate 2048 bytes of heap space!

Stephen Chong, Harvard University 19

 External Fragmentation
• Occurs when there is enough aggregate heap memory, but no
single free block is large enough to satisfy a given request.
p1 = malloc(4)

p2 = malloc(5)

p3 = malloc(6)

free(p2)

p4 = malloc(6)
No free block big enough!
Stephen Chong, Harvard University 20
 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 21

 Free block list management

•How do we know how much memory to free just

given a pointer?
•How do we keep track of the free blocks?
•What do we do with the extra space when
allocating a memory block that is smaller than
the free block it is placed in?
•How do we pick which free block to use for
allocation?

Stephen Chong, Harvard University 22

 Knowing how much to free
•Standard method
•Keep the length of a block in a header preceding the block.
•Requires an extra word for every allocated block

p = malloc(4) p

5
data
header
(address p-1)
free(p)

Stephen Chong, Harvard University 23

 Keeping Track of Free Blocks

•One of the biggest jobs of an allocator is

knowing where the free memory is.
•Affects throughput and utilization
•Many approaches to free block management.
•Today, we will talk about three techniques
• Implicit free lists
• Explicit free lists
• Segregated free lists
•There are other approaches

Stephen Chong, Harvard University 24

 Implicit free list
• Idea: Each block contains a header with some
a size
extra information.
• Allocated bit indicates whether block is allocated or free.
payload
• Size field indicates entire size of block (including the or free space
header)
• Trick: Allocation bit is just the high-order bit of the size
word optional
padding
• For this lecture, assume header size is 1 byte.
• Makes later pictures easier to understand.
a = 1: block is allocated
• This means the block size is only 7 bits, so max. block
a = 0: block is free
size is 127 bytes (27-1).
• Clearly a real implementation would want to use a larger size: block size
header (e.g., 4 bytes).
payload: application data

Stephen Chong, Harvard University 25

 Examples
0x84

4 bytes total size 0x84 in binary: 1000 0100

payload allocated = 1
size = 000 0100 = 0x4 = 4 bytes

0x0f
0x0f in binary: 0000 1111
allocated = 0
15 bytes total size size = 000 1111 = 0xf = 15 bytes
payload

Stephen Chong, Harvard University 26

 Implicit free list

Implicit list

5 4 6 2
Free Allocated Free Allocated

• No explicit structure tracking location of free/allocated blocks.

• Rather, the size word (and allocated bit) in each block form an implicit
“block list”
• How do we find a free block in the heap?
• Start scanning from the beginning of the heap.
• Traverse each block until (a) we find a free block and (b) the block is large
enough to handle the request.
• This is called the first fit strategy.
Stephen Chong, Harvard University 27
 Implicit List: Finding a Free Block
• First fit strategy:
• Search list from beginning, choose first free block that fits
• Can take linear time in total number of blocks (allocated and free)
• In practice it can cause “splinters” at beginning of list
• Next fit strategy:
• Like first-fit, but search list from location of end of previous search
• Research suggests that fragmentation is worse than first-fit
• Best fit strategy:
• Search the list, choose the free block with the closest size that fits
• Keeps fragments small — usually helps fragmentation
• Runs slower than first- or next-fit, since the entire list must be searched
each time
Stephen Chong, Harvard University 28
 Implicit List: Allocating in Free Block

•Splitting free blocks

•Since allocated space might be smaller than free space, we
may need to split the free block that we're allocating within
• E.g., malloc(3)
2 4 6 2

p
void addblock(ptr_t p, int len) {
int oldsize = *p; // Get old size of free block
*p = len | 0x80; // Set new size + alloc bit
if (len < oldsize)
*(p+len) = oldsize - len; // Set size of remaining
} // part of free block

addblock(p, 4)
2 4 4 2 2
Stephen Chong, Harvard University 29
 Implicit List: Freeing a Block
• Simplest implementation:
• Simply clear the allocated bit in the header
/* Here, p points to the block header. */
/* This sets the high-order bit to 0. */
void free_block(ptr_t p) { *p = *p & 0x7f; }

4 4 4 2 2
p
free(p+1)

4 4 4 2 2

Stephen Chong, Harvard University 30

 Implicit List: Freeing a Block
• Simplest implementation:
• Simply clear the allocated bit in the header
/* Here, p points to the block header. */
/* This sets the high-order bit to 0. */
void free_block(ptr_t p) { *p = *p & ~0x80; }
• But can lead to “false fragmentation”

4 4 4 2 2
p
free(p+1)

4 4 4 2 2

malloc(5)

• There is enough free space, but the allocator won’t be able to find it!
Stephen Chong, Harvard University 31
 Implicit List: Coalescing
• Coalesce with adjacent block(s) if they are free
void free_block(ptr_t p) {
*p = *p & 0x7f; // Clear allocated bit
next = p + *p; // Find next block
if ((*next & 0x80) == 0)
*p = *p + *next; // Add to this block if
} // next is also free

4 4 4 2 2
p
free(p+1)

4 4 6 2

• This is coalescing with the next free block.

• How would we coalesce with the previous free block?
Stephen Chong, Harvard University 32
 Implicit List: Bidirectional Coalescing

• Boundary tags [Knuth73]

• Also maintain the size/allocated word at end of free blocks (a footer)
“Pointers” in headers

4 4 4 4 6 6 4 4

“Pointers” in footers
• Allows us to traverse the “block list” backwards, but requires extra space

Header a size a = 1: block is allocated

a = 0: block is free
payload
and size: total block size
padding
a size payload: application data
Boundary tag (footer)

• Important and general technique!

Stephen Chong, Harvard University 33
 Constant Time Coalescing

Case 1 Case 2 Case 3 Case 4

allocated allocated free free

block being
freed
allocated free allocated free

Stephen Chong, Harvard University 34

 Constant Time Coalescing (Case 1)

m1 1 m1 1

m1 1 m1 1
n 1 n 0

n 1 n 0
m2 1 m2 1

m2 1 m2 1

Stephen Chong, Harvard University 35

 Constant Time Coalescing (Case 2)

m1 1 m1 1

m1 1 m1 1
n 1 n+m2 0

n 1
m2 0

m2 0 n+m2 0

Stephen Chong, Harvard University 36

 Constant Time Coalescing (Case 3)

m1 0 n+m1 0

m1 0
n 1

n 1 n+m1 0
m2 1 m2 1

m2 1 m2 1

Stephen Chong, Harvard University 37

 Constant Time Coalescing (Case 4)

m1 0 n 0

m1 0
n 1

n 1
m2 0

m2 0 n 0

Stephen Chong, Harvard University 38

 Implicit Lists: Summary
• Implementation: Very simple.
• Allocation cost: Linear time worst case
• Free cost: Constant time, even with coalescing
• Memory usage: Depends on placement policy
• First fit, next fit, or best fit
• Not used in practice for malloc/free because of linear time
allocation.

• The concepts of splitting and boundary tag coalescing are general

to all allocators.

Stephen Chong, Harvard University 39

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 40

 Explicit Free Lists

•Use an explicit data structure to track the free blocks

•Just a doubly-linked list.
•No pointers to or from allocated blocks: Can put the
pointers into the payload!
•Still need boundary tags, in order to perform free block
coalescing.
A B C

Forward links

4 4 4 4 6 6 4 4 4 4

Back links

Stephen Chong, Harvard University 41

 Explicit Free Lists

•Note that free blocks need not be linked in the

same order
they appear in memory!
•Free blocks can be chained together in any order.

A B C

Forward links

4 4 4 4 6 6 4 4 4 4

Back links

Stephen Chong, Harvard University 42

 Allocation using an explicit free list
Before:

p = malloc(size);
After:

split 2nd free block

Stephen Chong, Harvard University 43

 Deallocation with an explicit free list
• Same idea as previously
• Step 1: Mark block as free
• Step 2: Coalesce adjacent free blocks
• Step 3: Insert free block into the free list

• Where in the free list do we put a newly freed block?

• LIFO (last-in-first-out) policy
• Insert freed block at the beginning of the free list
• Simple and constant time
• Address-ordered policy
• Insert freed blocks so that free list blocks are always in address order
• Con: requires search
• Pro: studies suggest fragmentation is lower than LIFO
Stephen Chong, Harvard University 44
 Freeing With a LIFO Policy (Case 1)
free(p)
Before:

Root

•Insert the freed block at the root of the free block list

After:

Root

Stephen Chong, Harvard University 45

 Freeing With a LIFO Policy (Case 2)
free(p)
Before:

Root

• Splice out predecessor block, coalesce both memory blocks and

insert the new block at the root of the list

After:

Root

Stephen Chong, Harvard University 46

 Freeing With a LIFO Policy (Case 3)
free(p)
Before:

Root

• Splice out successor block, coalesce both memory blocks and insert
the new block at the root of the list

After:

Root

Stephen Chong, Harvard University 47

 Freeing With a LIFO Policy (Case 4)
free(p)
Before:

Root

• Splice out predecessor and successor blocks, coalesce all 3 memory

blocks and insert the new block at the root of the list

After:

Root

Stephen Chong, Harvard University 48

 Explicit List Summary
• Comparison to implicit list:
•Allocation is linear time in number of free blocks
•Implicit list allocation is linear time in the number of total blocks
• Slightly more complicated allocate and free since need
to splice blocks in and out of the list
• Need some extra space for the links
•2 extra words needed for each free block
•But these can be stored in the payload, since only needed for free
blocks.

Stephen Chong, Harvard University 49

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 50

 Segregated List (seglist) Allocators

•Use a different free list for blocks of different sizes!

1-2

5-8

9-inf

•Often have separate size class for every small size (4,5,6,…)
•For larger sizes typically have a size class for each power of 2

Stephen Chong, Harvard University 51

 Seglist Allocator
• To allocate a block of size n:
• Determine correct free list to use
• Search that free list for block of size m ≥ n
• If an appropriate block is found:
• Split block and place fragment on appropriate list
• If no block is found, try next larger class
• Repeat until block is found
• If no free block is found:
• Request additional heap memory from OS (using sbrk() system call)
• Allocate block of n bytes from this new memory
• Place remainder as a single free block in largest size class.

Stephen Chong, Harvard University 52

 Freeing with Seglist
• To free a block:
• Mark block as free
• Coalesce (if needed)
• Place free block on appropriate sized list

Stephen Chong, Harvard University 53

 Seglist advantages

•Advantages of seglist allocators

•Higher throughput
• Faster to find appropriate sized block: Look in the right list.
•Better memory utilization
• First-fit search of segregated free list approximates a best-fit search
of entire heap.
• Extreme case: Giving each size its own segregated list is equivalent
to best-fit.

Stephen Chong, Harvard University 54

 Topics for today

•Dynamic memory allocation

•Implicit vs. explicit memory management
•Performance goals
•Fragmentation
•Free block list management
•Implicit free list
•Explicit free list
•Segregated lists
•Tradeoffs

Stephen Chong, Harvard University 55

 Allocation Policy Tradeoffs
• Data structure for free lists
• Implicit lists, explicit lists, segregated lists
• Other structures possible, e.g.explicit free blocks in binary tree, sorted
by size
• Placement policy: First fit, next fit, or best fit
• Best fit has higher overhead, but less fragmentation.
• Splitting policy: When do we split free blocks?
• Splitting leads to more internal fragmentation, since each block needs its own header.
• Coalescing policy: When do we coalesce free blocks?
• Immediate coalescing: Coalesce each time free is called
• Deferred coalescing: Improve free performance by deferring coalescing until needed.
• E.g., While scanning the free list for malloc(), or when external fragmentation
reaches some threshold.
Stephen Chong, Harvard University 56
 Topics for next time

•Allocation requirements
•Common memory bugs
•Implicit memory management: Garbage
collection

Stephen Chong, Harvard University 57