Home > Articles

CPU: Pitfalls and Techniques

Software Engineer Marcel Weiher discusses how Objective-C can achieve both best-of-breed performance and high levels of expressiveness and productivity in the iOS and macOS platforms.

This chapter is from the book

Having had a look at the parameters driving performance and techniques for identifying slow code, let’s now turn to actual techniques for making code run fast. We will look at efficient object representations and ways for those objects to communicate and access data. We will also examine streamlining computation. In all this, the objective will typically be to effectively combine the “Objective” and the “C” parts of Objective-C to achieve the desired balance between performance and encapsulation.

In general, the basic idea is for objects to have C on the inside and messages on the outside, and for the objects themselves to be fairly coarse-grained, mostly static entities. When following these principles, it is possible to start with a fully object-oriented implementation without worries, but with the knowledge that it will be possible to later optimize away any inefficiencies. It has been my experience that it is quite possible to achieve the performance of plain C, and sometimes even beyond.

However, there are pitfalls that not only make an Objective-C program slow (slower than so-called scripting languages), but even worse can be major obstacles to later optimization efforts. These pitfalls usually lie in library constructs that are easy to use but have hidden performance costs, costs that are not localized within a single object where they could be eliminated, but present in interfaces and therefore spread throughout the system and much harder to expunge.

The following will show different options for data representation, communication, and computation, along with their respective trade-offs in terms of coupling, cohesion, and performance.


One of the primary tasks of a program, especially an object-oriented program, is to represent data. Due to the hybrid nature of the language, an Objective-C programmer has many options available for this task.

Without any claims of completeness, structured data can be represented using a C struct, Objective-C object, or various forms of key-value stores, most prominently Foundation’s NSDictionary and CoreFoundation’s CFDictionary, which are both getting more and more use. Simple scalars can be represented as C float, double, or int and their multitude of variations, Foundation NSInteger and CoreGraphics CGFloat typedefs, and finally Foundation NSNumber and CoreFoundation CFNumber objects. Note that the naming conventions are a bit confusing here: The names NSInteger and NSNumber strongly suggest that these two types are related—for example, with NSInteger being a specific subclass of NSNumber—but in fact they are completely unrelated. NSInteger is a typedef that resolves to a 32-bit int on 32-bit architectures and to a 64-bit long on 64-bit architectures, whereas int is 32 bits in both cases. Similar with CGFloat, which turns into a 32-bit float on 32-bit architectures and a 64-bit double on 64-bit architectures. Example 3.1 shows a few of the possible number representations.

Example 3.1 Numbers as primitives and objects

#import <Foundation/Foundation.h>

int main()
  int a=1;
  float b=2.0;
  NSNumber *c=[NSNumber numberWithInt:3];
  CFNumberRef d=CFNumberCreate(kCFAllocatorDefault,
                    kCFNumberFloatType, (const void*)&b );
  NSNumber *e=@(5);
  NSLog(@"a=%d b=%g c=%@ d=%@ e=%@",a,b,c,d,e);
  return 0;

In order to come to a good solution, the programmer must weigh trade-offs between decoupling and encapsulation on one hand and performance on the other hand, ideally getting as much decoupling and encapsulation without compromising performance, or conversely maximizing performance while minimizing coupling.

Primitive Types

Possibly the easiest call to make is in the representation of simple scalar types like characters/bytes, integers, and floating point numbers: use the built-in C primitive types whenever possible, and avoid object wrappers whenever possible.

With the language supporting them natively, scalars are convenient to use and perform anywhere from 10 to more than 100 times better than their corresponding Foundation object NSNumber or its CoreFoundation equivalent CFNumber. Table 3.1 gives the details: the first three columns are times for different arithmetic operations on scalar types. The differences in timings for 32- and 64-bit addition and multiplication are probably measuring artifacts, though they were stable when preparing these measurements and it is important to report actual results as measured, not what we think the results should be.

Table 3.1 Primitive operations in 32- and 64-bit architectures









64-bit (ns)








32-bit (ns)








Division is slower than the other arithmetic operations because dividers in CPUs usually only handle a few bits at a time, rather than a full word, which also explains why 64-bit division is significantly slower than 32-bit division.

Compared to entities that can usually be stored in registers and manipulated in a single clock cycle (or less on superscalar designs), any object representation has excessive overhead, and Objective-C’s fairly heavyweight objects are doubly so. Foundation and CoreFoundation make this overhead even worse by providing only immutable number objects, meaning any manipulation must create new objects. Finally, scalars like numbers and characters tend to be at the leaves of any object graph and therefore are the most numerous entities in a program, with every object containing at least one but more likely many instances of them.

On the flip side, there is little variation or private data that would benefit from the encapsulation and polymorphism that are made possible by an object representation, and number objects are in many ways even less capable than primitive types, for example, by not providing any arithmetic capabilities. This could change in the future if Foundation or another framework provided a number and magnitudes hierarchy similar to that of Smalltalk or LISP, where small integers automatically morph into infinite precision integers, fractions, floating point, or even complex numbers as needed. Alas, Foundation provides none of these capabilities, though the introduction of tagged integers in the 64-bit runtime on OS X 10.7 along with the addition of number literals in 10.8 could be a sign of improvements in the future.

Of course, there are times when an object is required by some other interface, for example, when adding content to an NSArray or NSDictionary. In this case, you must either use NSNumber or an equivalent or provide alternatives to those interfaces—an option we will explore more later in the chapter.

One wrinkle of Table 3.1 is that although most times are similar between 32 and 64 bits, two numbers are different. The division result is about twice as slow on 64 bit, whereas the creation of integer NSNumber objects is six times faster. The division result is easily explained by the fact that the integer division hardware on the particular CPU used processes a fixed number of bits per cycle, and 64-bit operands simply have twice as many bits. The multiply and add circuits, on the other hand, operate on full 64-bit words at once.

The difference in allocation speeds for integer objects on the other hand has nothing to do with the CPU differences and everything with the fact that Apple introduced tagged integers in OS X, but only in the modern runtime, and only for the 64-bit version of that runtime. Tagged integers are a technique taken from old LISP and Smalltalk systems where the value of an integer object is encoded not in an allocated structure pointed to by the object, as usual, but rather in the object pointer itself. This saves the pointer indirection when accessing and especially the memory allocation when creating or destroying the data (integers in this case). This representation takes advantage of the fact that object pointers are at least word aligned, so the lower 2 or 3 bits of a valid object pointer are always 0 on 32-bit and 64-bit systems, respectively. Table 3.2 shows how the tagged pointer representation puts a “1” in the low bit to distinguish tagged pointers from regular pointers, another 7 bits for typing the value, and the remaining 24 or 56 bits to store a value.

Table 3.2 Tagged and regular pointers








Regular pointer

upper address bits


Tagged pointer



tag id


In fact, it is puzzling that the performance for integer NSNumber creation isn’t much better than it is, since all it takes is the bit-shift and arithmetic OR shown in the makeInt() function of Example 3.2, possibly with some tests depending on the source and target number type—operations that should be in the 1 to 2 ns total range.

Example 3.2 Summing manually created tagged NSNumber objects

#import <Foundation/Foundation.h>

#define kCFTaggedObjectID_Integer ((3 << 1) + 1)
#define kCFNumberSInt32Type 3
#define kCFTaggedIntTypeOffset 6
#define kCFTaggedOffset 2
#define kCFTaggedIntValueOffset (kCFTaggedIntTypeOffset+kCFTaggedOffset)
#define MASK (kCFNumberSInt32Type<<kCFTaggedIntTypeOffset)
#define kCFTaggedIntMask (kCFTaggedObjectID_Integer | MASK)

static inline int getInt( NSNumber *o ) {
  long long n=(long long)o;
  if ( n & 1 ) {
    return  n >> kCFTaggedIntValueOffset;
  } else {
    return [o intValue];

static inline NSNumber *makeInt( long long o ) {
  return (NSNumber*)((o << kCFTaggedIntValueOffset) | kCFTaggedIntMask);

int main( int argc , char *argv[] )
  NSNumber* sum = nil;
  for (int k=0;k<1000000; k++ ) {
    sum =makeInt(0);
    for (int i=1;i<=1000;i++) {
      sum =makeInt(getInt(sum)+i);
  NSLog(@"%@/%@ -> '%@'",sum,[sum class],[sum stringValue]);
  return 0;

The reason of course is that Apple has so far hidden this change behind the existing messaging and function call application programming interfaces (APIs) going through CoreFoundation. We are also advised that the representation, including the actual tags, is private and subject to change. What we are leaving on the table is significant: The code in Example 3.2 runs in 1.4 s, compared to 11.4 s for the Foundation/CoreFoundation-based code from Chapter 1.

Hopefully this will change in the future, and the compiler will become aware of these optimizations and be able to generate tagged pointers for integer objects and some of the other tagged types that have been added in the meantime. But as of OS X 10.11 and Xcode 7.3, it hasn’t happened.


A data type that almost qualifies as a primitive in use is the string, even though it is actually variable in length and doesn’t fit in a processor register. In fact, Objective-C strings were the first and for a long time the only object that had compiler support for directly specifying literal objects.

There are actually several distinct major uses for strings:

  1. Human readable text

  2. Bulk storage of serialized data as raw bytes or characters

  3. Tokens or keys for use in programming

While these cases were traditionally all handled uniformly in C using char* pointers, with some NUL terminated and others with a length parameter handled out of band, conflating the separate cases is no longer possible now that text goes beyond 7-bit ASCII.

Cocoa has the NSString class for dealing with human readable text. It handles the subtleties of the Unicode standard, delegating most of the details to iconv library. This sophistication comes at a cost: roughly one order of magnitude slower performance than raw C strings. Table 3.3 shows the cost of comparing 10- and 32-byte C-Strings with 10- and 32-character NSString objects.

Table 3.3 NSString and C-String operations


1 ns




ns append


1 ns




























ns append














Although NSStrings are expensive, this is an expense well spent when the subject matter really is human-readable text. Implementing correct Unicode handling is complex, error prone, and inherently expensive. In addition, the option of having multiple representations with a common interface is valuable, allowing string representations optimized for different usage scenarios to be used interchangeably. For example, literal NSStrings are represented by the NSConstantString class that stores 8-bit characters, whereas the standard NSCFString class (backed by CFString CoreFoundation objects) stores 16-bit unichars internally. Subclasses could also interchangeably provide more sophisticated implementations such as ropes, which store the string as a binary tree of smaller strings and can efficiently insert/delete text into large strings.

Starting with OS X 10.10, string objects on the 64-bit runtime also got the tagged pointer treatment that we previously saw for integers. This may seem odd, as strings are variable-length data structures, arrays of characters. However, 64 is quite a lot of bits, enough to store seven 8-bit characters and some additional identifying information such as the length. In fact, when I myself proposed tagged pointer strings back in 2007, I also had variants with eight 7-bit ASCII strings, or an even tighter packing that ignores most of the control and special characters to use only 6 bits and thus have room for 9 characters. I don’t know if any of those variants are implemented.

Example 3.3 illustrates the different NSString implementation types: a literal is an instance of ___NSCFConstantString, a CF variant of NSConstantString. Creating a mutable copy creates a new string object, whereas creating a copy of that mutable copy creates a tagged pointer string because the string is only 5 characters long. All of this is implementation dependent, but the differences are relevant when looking at NSDictionary lookup performance.

Example 3.3 Show differences between normal, constant, and tagged strings

#import <Foundation/Foundation.h>

void printString( NSString *a ) {
  NSLog(@"string=%@ %p class: %@",a,a,[a class]);

int main()
  NSString *cs=@"Const";
  printString([cs mutableCopy]);
  printString([[cs mutableCopy] copy]);
cc -Wall -o taggedstring taggedstring.m -framework Foundation
string=Const 0x108fe2040 class: __NSCFConstantString
string=Const 0x7fb359c0d630 class: __NSCFString
string=Const 0x74736e6f4355 class: NSTaggedPointerString

While great for human readable text, NSString objects are somewhat heavyweight to be used for serialized data, which is handled more safely and efficiently by the NSData class. Unlike NSString, which requires an encoding to be known for text data and can therefore not be safely used on arbitrary incoming data (it will raise an exception if the data does not conform to the encoding), NSData can be used with arbitrary, potentially binary data read from the network or a disk. For performance, it is possible to get a pointer to the NSData’s contents via the -byte or -mutableBytes methods for processing using straight memory access, whereas NSString (rightfully) protects its internal data representation, with processing only possible by sending high-level messages or by copying the data out of the NSString as 16-bit unichar character data or encoded 8-bit bytes.

When parsing or generating serialized data formats, even textual ones, it is significantly more efficient to treat the serialized representation such as the JSON in Example 3.4 as raw bytes in an NSData, parse any structure delimiters, numbers, and other non-textual entities using C character processing, and create NSString objects exclusively for actual textual content, rather than reading the serialized representation into an NSString and using NSScanner or other high-level string processing routines.

Example 3.4 Textual content of JSON file is shown in bold

[ { "name": "AAPL",   "price": 650.1,     "change": 20.41   },
 {  "name": "MSFT",   "price": 62.79,     "change": -0.9    },
 {  "name": "GOOG",   "price": 340.79,     "change": -5.2    }, ]

Even the strings that appear in such a file tend to be structural rather than actual content, such as the dictionary keys in Example 3.4. These types of structural strings are also represented as NSString objects in Cocoa, just like human-readable text. While convenient due to the literal NSString syntax (@“This is a constant string”), this conflating of human-readable text and functional strings can at times be unfortunate in terms of performance. Fortunately, many types of keys that are more optimized exist—for example, basic C strings, message names, and instance variable names.

InformIT Promotional Mailings & Special Offers

I would like to receive exclusive offers and hear about products from InformIT and its family of brands. I can unsubscribe at any time.


Pearson Education, Inc., 221 River Street, Hoboken, New Jersey 07030, (Pearson) presents this site to provide information about products and services that can be purchased through this site.

This privacy notice provides an overview of our commitment to privacy and describes how we collect, protect, use and share personal information collected through this site. Please note that other Pearson websites and online products and services have their own separate privacy policies.

Collection and Use of Information

To conduct business and deliver products and services, Pearson collects and uses personal information in several ways in connection with this site, including:

Questions and Inquiries

For inquiries and questions, we collect the inquiry or question, together with name, contact details (email address, phone number and mailing address) and any other additional information voluntarily submitted to us through a Contact Us form or an email. We use this information to address the inquiry and respond to the question.

Online Store

For orders and purchases placed through our online store on this site, we collect order details, name, institution name and address (if applicable), email address, phone number, shipping and billing addresses, credit/debit card information, shipping options and any instructions. We use this information to complete transactions, fulfill orders, communicate with individuals placing orders or visiting the online store, and for related purposes.


Pearson may offer opportunities to provide feedback or participate in surveys, including surveys evaluating Pearson products, services or sites. Participation is voluntary. Pearson collects information requested in the survey questions and uses the information to evaluate, support, maintain and improve products, services or sites, develop new products and services, conduct educational research and for other purposes specified in the survey.

Contests and Drawings

Occasionally, we may sponsor a contest or drawing. Participation is optional. Pearson collects name, contact information and other information specified on the entry form for the contest or drawing to conduct the contest or drawing. Pearson may collect additional personal information from the winners of a contest or drawing in order to award the prize and for tax reporting purposes, as required by law.


If you have elected to receive email newsletters or promotional mailings and special offers but want to unsubscribe, simply email information@informit.com.

Service Announcements

On rare occasions it is necessary to send out a strictly service related announcement. For instance, if our service is temporarily suspended for maintenance we might send users an email. Generally, users may not opt-out of these communications, though they can deactivate their account information. However, these communications are not promotional in nature.

Customer Service

We communicate with users on a regular basis to provide requested services and in regard to issues relating to their account we reply via email or phone in accordance with the users' wishes when a user submits their information through our Contact Us form.

Other Collection and Use of Information

Application and System Logs

Pearson automatically collects log data to help ensure the delivery, availability and security of this site. Log data may include technical information about how a user or visitor connected to this site, such as browser type, type of computer/device, operating system, internet service provider and IP address. We use this information for support purposes and to monitor the health of the site, identify problems, improve service, detect unauthorized access and fraudulent activity, prevent and respond to security incidents and appropriately scale computing resources.

Web Analytics

Pearson may use third party web trend analytical services, including Google Analytics, to collect visitor information, such as IP addresses, browser types, referring pages, pages visited and time spent on a particular site. While these analytical services collect and report information on an anonymous basis, they may use cookies to gather web trend information. The information gathered may enable Pearson (but not the third party web trend services) to link information with application and system log data. Pearson uses this information for system administration and to identify problems, improve service, detect unauthorized access and fraudulent activity, prevent and respond to security incidents, appropriately scale computing resources and otherwise support and deliver this site and its services.

Cookies and Related Technologies

This site uses cookies and similar technologies to personalize content, measure traffic patterns, control security, track use and access of information on this site, and provide interest-based messages and advertising. Users can manage and block the use of cookies through their browser. Disabling or blocking certain cookies may limit the functionality of this site.

Do Not Track

This site currently does not respond to Do Not Track signals.


Pearson uses appropriate physical, administrative and technical security measures to protect personal information from unauthorized access, use and disclosure.


This site is not directed to children under the age of 13.


Pearson may send or direct marketing communications to users, provided that

  • Pearson will not use personal information collected or processed as a K-12 school service provider for the purpose of directed or targeted advertising.
  • Such marketing is consistent with applicable law and Pearson's legal obligations.
  • Pearson will not knowingly direct or send marketing communications to an individual who has expressed a preference not to receive marketing.
  • Where required by applicable law, express or implied consent to marketing exists and has not been withdrawn.

Pearson may provide personal information to a third party service provider on a restricted basis to provide marketing solely on behalf of Pearson or an affiliate or customer for whom Pearson is a service provider. Marketing preferences may be changed at any time.

Correcting/Updating Personal Information

If a user's personally identifiable information changes (such as your postal address or email address), we provide a way to correct or update that user's personal data provided to us. This can be done on the Account page. If a user no longer desires our service and desires to delete his or her account, please contact us at customer-service@informit.com and we will process the deletion of a user's account.


Users can always make an informed choice as to whether they should proceed with certain services offered by InformIT. If you choose to remove yourself from our mailing list(s) simply visit the following page and uncheck any communication you no longer want to receive: www.informit.com/u.aspx.

Sale of Personal Information

Pearson does not rent or sell personal information in exchange for any payment of money.

While Pearson does not sell personal information, as defined in Nevada law, Nevada residents may email a request for no sale of their personal information to NevadaDesignatedRequest@pearson.com.

Supplemental Privacy Statement for California Residents

California residents should read our Supplemental privacy statement for California residents in conjunction with this Privacy Notice. The Supplemental privacy statement for California residents explains Pearson's commitment to comply with California law and applies to personal information of California residents collected in connection with this site and the Services.

Sharing and Disclosure

Pearson may disclose personal information, as follows:

  • As required by law.
  • With the consent of the individual (or their parent, if the individual is a minor)
  • In response to a subpoena, court order or legal process, to the extent permitted or required by law
  • To protect the security and safety of individuals, data, assets and systems, consistent with applicable law
  • In connection the sale, joint venture or other transfer of some or all of its company or assets, subject to the provisions of this Privacy Notice
  • To investigate or address actual or suspected fraud or other illegal activities
  • To exercise its legal rights, including enforcement of the Terms of Use for this site or another contract
  • To affiliated Pearson companies and other companies and organizations who perform work for Pearson and are obligated to protect the privacy of personal information consistent with this Privacy Notice
  • To a school, organization, company or government agency, where Pearson collects or processes the personal information in a school setting or on behalf of such organization, company or government agency.


This web site contains links to other sites. Please be aware that we are not responsible for the privacy practices of such other sites. We encourage our users to be aware when they leave our site and to read the privacy statements of each and every web site that collects Personal Information. This privacy statement applies solely to information collected by this web site.

Requests and Contact

Please contact us about this Privacy Notice or if you have any requests or questions relating to the privacy of your personal information.

Changes to this Privacy Notice

We may revise this Privacy Notice through an updated posting. We will identify the effective date of the revision in the posting. Often, updates are made to provide greater clarity or to comply with changes in regulatory requirements. If the updates involve material changes to the collection, protection, use or disclosure of Personal Information, Pearson will provide notice of the change through a conspicuous notice on this site or other appropriate way. Continued use of the site after the effective date of a posted revision evidences acceptance. Please contact us if you have questions or concerns about the Privacy Notice or any objection to any revisions.

Last Update: November 17, 2020