This is an unusual question, but I do hope there's a definitive answer.
There's a longstanding debate in our office about how efficiently compilers generate code, specifically number of instructions. We write code for low power embedded systems with virtually no loops. Therefore, the number of instructions emitted is directly proportional to power consumed.
Much of our code looks like this (notice, no dynamic memory allocation, no system calls, very few function calls, very few loops).
foo += 3 * (77 + bar);
if (baz > 18 - qux)
bar -= 19 + 7 >> spam;
I can compile the above snippet with -O3 and read the assembly, but I couldn't write it myself.
The claim I would like to prove or disprove is that compilers generate code that is 2-4X "fatter" (and therefore consume 2-4X times as much power) compared with hand written assembly code.
I'm interested in any compiler with which you have experience.
From this answer I know that GCC and clang can emit assembly interleaved with the C code with
gcc -g -c -Wa,-alh foo.cc
These answers provide solid basis:
When is assembly faster?
Why do you program in assembly?
How can I measure the efficiency with which a compiler generates code?
Hand assembly can always at least match if not beat the compiler, because at the very least, you can start with the compiler generated assembly code and tweak it to make it better. To really do a good job, you need to understand the CPU architecture (pipeline, functional units, memory hierarchy, out-of-order dispatch units, etc.) so that you can schedule each instruction for maximum efficiency.
Another thing to consider is that the number of instructions is not necessarily directly proportional to performance, whether it is speed or power (see Hennessey and Patterson's Computer Architecture: A Quantitative Approach). Basically, you have to look at how many clock cycles each instruction takes, in addition to the number of instructions (and clock rate) to know how long it will take. To know how much energy will be consumed, you also need to know how much energy each instruction takes.
How the CPU implements each instruction affects how many cycles it takes to execute. As an example, your code sequence has a >> operator. The compiler might translate that to a single ASR instruction, but without knowing the architecture, there is no telling how many clock cycles it might take -- some architectures can do an arbitrary shift in a single cycle, while others need one cycle for each bit shift.
Memory access contributes to the number of cycles and power consumption, too. When there are too many variables to store in registers some of them will have to be stored in memory. If you are accessing off chip memory and have a fairly high CPU clock rate, the memory bus can be pretty power hungry. A longer sequence of instructions that avoids reading from and writing to memory (e.g., by computing the same result twice) can be less expensive.
As several others have suggested, there is no substitute for benchmarking. Assuming you are using a microcontroller-based system with a constant input voltage, your best bet is to measure the current draw of your system with each alternative set of code and see which does best (one way would be with a current probe and a digital storage oscilloscope).
Even if you can always write better assembler than the compiler, there is a cost in development time and maintainability. In The Mythical Man Month Brooks estimated 3-5x more effort at time when many, if not most, programmers wrote code in assembler. Unless your code is really tiny, you are probably best off only coding the most critical parts in assembly. Even so, the person writing the assembly should be able to prove that their (more expensive) code is worth the cost by comparing running code vs. running code.
If the question is "how can I measure the efficiency with which a compiler generates code" (your actual question), the answer is "that depends". It depends on how you define "efficiency". Mostly, compilers are designed to optimize for speed. As you change the optimization level (-O1, -O2, -O3), the compiler will spend more time looking for "clever things to do to make it just a bit faster". This can involve loop unrolling, order of execution, use of registers, and many other things.
It seems that your "efficiency" criterion is not one that compilers are designed for: you say you want "fewest cycles" because you think that == lowest power. However I would argue that "fastest execution" == "shortest time before processor can go into standby mode again". Unless you believe that the power consumption of the processor in "awake" mode changes significantly with instructions executed, I think that it is safe to say that fastest execution == shortest time awake == lowest power consumption.
In which case "fat code" doesn't matter - it's back to speed only. Note also that not all instructions take the same number of clock cycles (although to be fair, that depends on the processor).
EDIT, okay that was fun...
Folks that make the blanket statement that compilers outperform humans, are the ones that have not actually checked. Anything a compiler can create a human can create. But a compiler cannot always create the code a human can create. It is that simple. For projects anywhere from a few lines to a few dozen lines or larger, it becomes easier and easier to hand fix the optimizations made by a compiler. Compiler and target help close that gap but there will always be the educated someone that will be able to meet or exceed the compilers output.
The claim I would like to prove or disprove is that compilers generate
code that is 2-4X "fatter" (and therefore consume 2-4X times as much
power) compared with hand written assembly code.
Unless you are defining "fatter" to mean uses that much power. Size of a binary and power consumption are not related. If this whole question/project is related to power consumption, the compiler wont take into account the bios settings you have chosen (assuming you are talking about pcs), the video card, hard disk, monitor, mouse, keyboard, etc, etc. In addition to the processor which is only one (relatively small) part of the equation. And even if it did would someone make a compiler that only makes your code efficient, they cant and wont tune the compiler for every system on the planet. Aint gonna happen.
If you are talking a mobile phone which is a very controlled environment the app may get tuned to save power, but the compiler is not the master of that, it is the user, the compiler does part of it the rest is hand tuned by the programmer.
I can compile the above snippet with -O3 and read the assembly, but I couldn't write it myself.
If you go into this with that kind of attitude then you have automatically failed. Yes you can meet or beat the compiler, period. It is a matter of confidence and will power and time/effort. That statement means you have not really studied the problem, which is why you are asking the question you are asking. Take some time, do some more research, ask detailed questions at stackoverflow (not open ended ones like this one), and with time you will understand what compilers do and dont do and why, in particular why they are not perfect (for any one or many rulers by which that opinion is defined). This question is wholly about opinion and will spark flame wars, and such and will get closed and eventually removed from this site. Instead write and compile and publish code segments and ask questions about "why did the compiler produce this output, why didnt it to [this] instead?" Those kinds of questions have a better chance at real answers and of staying here for others to learn from.
I will break down this question in to sub questions. I am confused if I should ask them separately or in one question. So I will just stick to one SO question.
What are generally the steps to analyze and improve performance of C applications?
Do these steps change if I am developing for an embedded system?
What tools are out there which can help me?
Recently I have been given a task to improve the performance of our product on ARM11 platform. I am relatively new to this field of embedded systems and need gurus here on SO to help me out.
simply changing compilers can improve your C performance for the same source code by many times over. GCC has not necessarily gotten better for performance over the years, for some programs gcc 3.x produces much tighter code than 4.x. Back when I had access to the tools, ARMs compiler produced significantly better code than gcc. As much as 3 or 4 times faster. LLVM has caught up to GCC 4.x and I suspect will pass gcc by in terms of performance and overall use for cross compiling embedded code. Try different versions of gcc, 3.x and 4.x if you are using gcc. Metaware's compiler and arms adt ran circles around gcc3.x, gcc3.x will give gcc4.x a run for its money with arm code, for thumb code gcc4.x is better and for thumb2 (which doesnt apply to you) gcc4.x also better. Remember I have not said a word about changing a single line of code (yet).
LLVM is capable of full program optimization in addition to infinitely more tuning knobs than gcc. Despite that the code generated (ver 27) is only just catching up to the current gcc 4.x in terms of performance for the few programs I tried. And I didnt try the n factoral number of optimization combinations (optimize on the compile step, different options for each file, or combine two files or three files or all files and optimize those bundles, my theory is do no optimization on the C to bc steps, link all the bc together then do a single optimization pass on the whole program, the allow the default optimization when llc takes it to the target).
By the same token simply knowing your compiler and the optimizations can greatly improve the performance of the code without having to change any of it. You have an ARM11 arr you compiling for arm11 or generic arm? You can gain a few to a dozen percent by telling the compiler specifically which architecture/family (armv6 for example) over the generic armv4 (ARM7) that is often chosen as the default. Knowing to use -O2 or -O3 if you are brave.
It is often not the case but switching to thumb mode can improve performance for specific platforms. Doesnt apply to you but the gameboy advance is a perfect example, loaded with non-zero wait state 16 bit busses. Thumb has a handful of a percent overhead because it takes more instructions to do the same thing, but by increasing the fetch times, and taking advantage of some of the sequential read features of the gba thumb code can run significantly faster than arm code for the same source code.
having an arm11 you probably have an L1 and maybe L2 cache, are they on? Are they configured? Do you have an mmu and is your heavy use memory cached? or are you running zero wait state memory and dont need a cache and should turn it off? In addition to not realizing that you can take the same source code and make it run many times faster by changing compilers or options, folks often dont realize that when you use a cache simply adding a single up to a few nops in your startup code (as a trick to adjust where code lands in memory by one, two, a few words) you can change your codes execution speed by as much as 10 to 20 percent. Where those cache line reads hit in heavily used functions/loops makes a big difference. Even saving one cache line read by adjusting where the code lands is noticeable (cutting it from 3 to 2 or 2 to 1 for example).
Knowing your architecture, both the processor and your memory environment is where the tuning if any would start. Most C libraries if you are high level enough to use one (I often dont use a C library as I run without an operating system and with very limited resources) both in their C code and sometimes add some assembler to make bottleneck routines like memcpy, much faster. If your programs are operating on aligned 32 or even better 64 bit addresses, and you adjust even if it means using a handful of bytes more memory for every structure/array/memcpy to be an integral multiple of 32 bits or 64 bits you will see noticeable improvements (if your code uses structs or copies data in other ways). In addition to getting your structures (if you use them, I certainly dont with embedded code) size aligned, even if you waste memory, getting elements aligned, consider using 32 bit integers for every element instead of bytes or halfwords. Depending on your memory system this can help (it can hurt too btw). As with the GBA example above looking at specific functions that either by profiling or intuition you know are not being implemented in a manner that takes advantage of your processor or platform or libraries you may want to turn to assembler either from scratch or compiling from C initially then disassembling and hand tuning. Memcpy is a good example you may know your systems memory performance and may chose to create your own memcpy specifically for aligned data, copying 64 or 128 or more bits per instruction.
Likewise mixing global and local variables can make a noticeable performance difference. Traditionally folks are told never to use globals, but in embedded this isnt necessarily true, depends on how deeply embedded and how much tuning and speed and other factors you are interested in. This is a touchy subject and I may get flamed for it, so I will leave it at that.
The compiler has to burn and evict registers in order to make function calls, plus if you use local variables a stack frame may be required, so function calls are expensive, but at the same time, depending on the code within a function that has now grown in size by avoiding functions, you may create the problem you were trying to avoid, evicting registers to re-use them. Even a single line of C code can make the difference between all the variables in a function fits in registers to having to start evicting a bunch of registers. For functions or segments of code where you know you need some performance gain compile and disassemble (and look at register usage, how often it fetches memory or writes to memory). You can and will find places where you need to take a well used loop and make it its own function even though the function call has a penalty because by doing that the compiler can better optimize the loop and not evict/reuse registers and you get an overall net gain. Even a single extra instruction in a loop that goes around hundreds of times is a measurable performance hit.
Hopefully you already know to absolutely not compile for debug, turn all of the compile for debug options off. You may already know that code compile for debug that runs without bugs doesnt mean it is debugged, compiling for debug and using debuggers hide bugs leaving them as time bombs in your code for your final compile for release. Learn to always compile for release and test with the release version both for performance and finding bugs in your code.
Most instruction sets do not have a divide function. Avoid using divides or modulo in your code as much as humanly possible they are performance killers. Naturally this is not the case for powers of two, to save the compiler and to mentally avoid divides and modulos try to use shifts and ands. Multplies are easier and more often found in instruction sets, but are still costly. This is a good case to write assembler to do your multiplies instead of letting the C copiler do it. The arm multiply is a 32bit * 32bit = 32 bit so to do accurate math without overflowing there has to be extra C code wrapped around the multiply, if you already know you wont overflow, burn the registers for a function call and do the multiply in assembler (for the arm).
Likewise most instruction sets do not have a floating point unit, with yours you might, even so avoid float if at all possible. If you have to use float that is a whole other pandora's box of performance issues. Most folks dont see the performance problems with code as simple as this:
float a,b;
...
a = b * 7.0;
The rest of the problem is not understanding floating point accuracy and how good or bad the C libraries are just trying to get your constants into floating point form. Again float is a whole other long discussion on performance problems.
I am a product of Michael Abrash (I actually have a print copy of zen of assembly language) and the bottom line is time your code. Come up with an accurate way to time the code, you may think you know where the bottlenecks are and you may think you know your architecture but trying different things even if you think they are wrong, and timing them you may find and eventually have to figure out the error in your thinking. Adding nops to start.S as a final tuning step is a good example of this, all the other work you have done for performance can be instantly erased by not having a good alignment with the cache, this also means re-arranging functions within your source code so that they land in different places in the binary image. I have seen 10 to 20 percent swings of speed increase and decrease as a result of cache line alignments.
Code Review:
What are good code review techniques ?
Static and dynamic analysis of the code.
Tools for static analysis: Sparrow, Prevent, Klockworks
Tools for dynamic analysis : Valgrind, purify
Gprof allows you to learn where your program spent its time and which functions called which other functions while it was executing.
Steps are same
Apart from what is listed is point 1, there are tools like memcheck etc.
There is a big list here based on platform
Phew!! Quite a big question!
What are generally the steps to
analyze and improve performance of C
applications?
As well as other static code analysers mentioned here there is a fairly cheap version called PC-Lint which has been around for ages. Sometimes throws up lots of errors and warnings for one error but by the end of it you'll be happy and know waaaaay more about C/C++ because of it.
With all code analysers some of the issues may be more structural to the code so best to start analysing it from day 1 of coding; running analysis on old software may swamp you with issues which may take a while to untangle, best to keep it clean from the beginning.
But code analysers will not catch all logical errors, i.e. it doesn't do what you want it to do! These are best done by code reviews first, then testing. Performance is often improved by by trying to keep the algorithms as simple as possible, keeping instructions in loops tight, possibly unrolling loops (your compiler optimisations may do this), use of fast caches when accessing data which is slow to get.
Code reviews can raise a lot of issues from lots of other peoples eyes looking at it. Don't get too many people, try to get 3 other people if possible, sometimes junior developers ask the most insightful questions like, "why are we doing this?".
Testing can be roughly split into two sections, automated and manual. Automated testing requires effort producing test handlers for functions/units but once run can be run again and again very quickly. Manual testing requires planning, self-discipline to perform them all to the required, imagination to think up of scenarios that may impair performance and you have to be observant (you may have passed the test but the 'scope trace has a bit of an anomaly before/after the test).
"Do these steps change if I am
developing for an embedded system?"
Performance ananlysis can be different on embedded systems to applications systems; with the very broad brush that "embedded" now covers it depends how hardware-centric you are. It can be done using profilers, if you want a more cheap and chearful method then use test output pins to measure sections of code, or measure them with breakpoints on simulators that come with the development environment.
Make sure that not just a typical length of task is measured but also a maximum, as that is where one task may start impeding on other tasks and your scheduled tasks are not completed in time.
What tools are out there which can
help me?
Simulators on the IDEs, static analysis tools, dynamic analysis tools, but most of all you and other humans getting the requirements right, decent reviewing (of code and testing) and thorough testing (automated and manual).
Good luck!
My experiences.
Function calls are slow, eliminate with macros or inlined methods. Look at the disassembler listing to see.
If using GCC, mark optimized sections with #pragma GCC optimize("O3") or compile them separately.
Play with different combinations of applying the inline attribute (basically find a balance between size and speed).
It is a difficult question to be answered shortly since various techniques have been proposed such as flowchart and state diagram,so you can take a look at some titles:
ARM System-on-Chip Architecture, 2nd Edition -- Steve Furber
ARM System Developer's Guide - Designing and Optimizing System Software -- Andrew N. Sloss, Dominic Symes, Chris Wright & John Rayfield
The Definitive Guide to the ARM Cortex-M3 --Joseph Yiu
C Programming for Embedded Systems --Kirk Zurell
Embedded C -- Michael J. Pont
Programming Embedded Systems in C and C++ --Michael Barr
An Embedded Software Primer --David E, Simon
Embedded Microprocessor Systems 3rd Edition --Stuart Ball
Global Specification and Validation of Embedded Systems - Integrating Heterogeneous Components --G. Nicolescu & A.A Jerraya
Embedded Systems: Modeling, Technology and Applications --Gunter Hommel & Sheng Huanye
Embedded Systems and Computer Architecture --Graham Wilson
Designing Embedded Hardware --John Catsoulis
You have to use a profiler. It will help you identify your application's bottleneck(s). Then focus on improving the functions you spend the most time in and the ones you call the most. Repeat this procedure until you're satisfied with your application performance.
No they don't.
Depending on the platform you're developing onto :
Windows : AMD Code Analyst, VTune, Sleepy
Linux : valgrind / callgrind / cachegrind
Mac : the Xcode profiler is quite good.
Try to find a profiler for the architecture you actually work on.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 2 years ago.
Improve this question
In listening to the Stack Overflow podcast, the jab keeps coming up that "real programmers" write in C, and that C is so much faster because it's "close to the machine." Leaving the former assertion for another post, what is special about C that allows it to be faster than other languages?
Or put another way: what's to stop other languages from being able to compile down to binary that runs every bit as fast as C?
There isn't much that's special about C. That's one of the reasons why it's fast.
Newer languages which have support for garbage collection, dynamic typing and other facilities which make it easier for the programmer to write programs.
The catch is, there is additional processing overhead which will degrade the performance of the application. C doesn't have any of that, which means that there is no overhead, but that means that the programmer needs to be able to allocate memory and free them to prevent memory leaks, and must deal with static typing of variables.
That said, many languages and platforms, such as Java (with its Java Virtual Machine) and .NET (with its Common Language Runtime) have improved performance over the years with advents such as just-in-time compilation which produces native machine code from bytecode to achieve higher performance.
There is a trade-off the C designers have made. That's to say, they made the decision to put speed above safety. C won't
Check array index bounds
Check for uninitialized variable values
Check for memory leaks
Check for null pointer dereference
When you index into an array, in Java it takes some method call in the virtual machine, bound checking and other sanity checks. That is valid and absolutely fine, because it adds safety where it's due. But in C, even pretty trivial things are not put in safety. For example, C doesn't require memcpy to check whether the regions to copy overlap. It's not designed as a language to program a big business application.
But these design decisions are not bugs in the C language. They are by design, as it allows compilers and library writers to get every bit of performance out of the computer. Here is the spirit of C how the C Rationale document explains it:
C code can be non-portable. Although it strove to give programmers the opportunity to write truly portable programs, the Committee did not want to force programmers into writing portably, to preclude the use of C as a ``high-level assembler'': the ability to write machine-specific code is one of the strengths of C.
Keep the spirit of C. The Committee kept as a major goal to preserve the traditional spirit of C. There are many facets of the spirit of C, but the essence is a community sentiment of the underlying principles upon which the C language is based. Some of the facets of the spirit of C can be summarized in phrases like
Trust the programmer.
Don't prevent the programmer from doing what needs to be done.
Keep the language small and simple.
Provide only one way to do an operation.
Make it fast, even if it is not guaranteed to be portable.
The last proverb needs a little explanation. The potential for efficient code generation is one of the most important strengths of C. To help ensure that no code explosion occurs for what appears to be a very simple operation, many operations are defined to be how the target machine's hardware does it rather than by a general abstract rule. An example of this willingness to live with what the machine does can be seen in the rules that govern the widening of char objects for use in expressions: whether the values of char objects widen to signed or unsigned quantities typically depends on which byte operation is more efficient on the target machine.
If you spend a month to build something in C that runs in 0.05 seconds, and I spend a day writing the same thing in Java, and it runs in 0.10 seconds, then is C really faster?
But to answer your question, well-written C code will generally run faster than well-written code in other languages because part of writing C code "well" includes doing manual optimizations at a near-machine level.
Although compilers are very clever indeed, they are not yet able to creatively come up with code that competes with hand-massaged algorithms (assuming the "hands" belong to a good C programmer).
Edit:
A lot of comments are along the lines of "I write in C and I don't think about optimizations."
But to take a specific example from this post:
In Delphi I could write this:
function RemoveAllAFromB(a, b: string): string;
var
before, after :string;
begin
Result := b;
if 0 < Pos(a,b) then begin
before := Copy(b,1,Pos(a,b)-Length(a));
after := Copy(b,Pos(a,b)+Length(a),Length(b));
Result := before + after;
Result := RemoveAllAFromB(a,Result); //recursive
end;
end;
and in C I write this:
char *s1, *s2, *result; /* original strings and the result string */
int len1, len2; /* lengths of the strings */
for (i = 0; i < len1; i++) {
for (j = 0; j < len2; j++) {
if (s1[i] == s2[j]) {
break;
}
}
if (j == len2) { /* s1[i] is not found in s2 */
*result = s1[i];
result++; /* assuming your result array is long enough */
}
}
But how many optimizations are there in the C version? We make lots of decisions about implementation that I don't think about in the Delphi version. How is a string implemented? In Delphi I don't see it. In C, I've decided it will be a pointer to an array of ASCII integers, which we call chars. In C, we test for character existence one at a time. In Delphi, I use Pos.
And this is just a small example. In a large program, a C programmer has to make these kinds of low-level decisions with every few lines of code. It adds up to a hand-crafted, hand-optimized executable.
I didn't see it already, so I'll say it: C tends to be faster because almost everything else is written in C.
Java is built on C, Python is built on C (or Java, or .NET, etc.), Perl is, etc. The OS is written in C, the virtual machines are written in C, the compilers are written in C, the interpreters are written in C. Some things are still written in Assembly language, which tends to be even faster. More and more things are being written in something else, which is itself written in C.
Each statement that you write in other languages (not Assembly) is typically implemented underneath as several statements in C, which are compiled down to native machine code. Since those other languages tend to exist in order to obtain a higher level of abstraction than C, those extra statements required in C tend to be focused on adding safety, adding complexity, and providing error handling. Those are often good things, but they have a cost, and its names are speed and size.
Personally, I have written in literally dozens of languages spanning most of the available spectrum, and I personally have sought the magic that you hint at:
How can I have my cake and eat it, too? How can I play with high-level abstractions in my favorite language, then drop down to the nitty gritty of C for speed?
After a couple of years of research, my answer is Python (on C). You might want to give it a look. By the way, you can also drop down to Assembly from Python, too (with some minor help from a special library).
On the other hand, bad code can be written in any language. Therefore, C (or Assembly) code is not automatically faster. Likewise, some optimization tricks can bring portions of higher-level language code close to the performance level of raw C. But, for most applications, your program spends most of its time waiting on people or hardware, so the difference really does not matter.
Enjoy.
There are a lot of questions in there - mostly ones I am not qualified to answer. But for this last one:
what's to stop other languages from being able to compile down to binary that runs every bit as fast as C?
In a word, abstraction.
C is only one or two levels of abstraction away from machine language. Java and the .NET languages are at a minimum three levels of abstraction away from assembler. I'm not sure about Python and Ruby.
Typically, the more programmer toys (complex data types, etc.), the further you are from machine language and the more translation has to be done.
I'm off here and there, but that's the basic gist.
There are some good comments on this post with more details.
It is not so much that C is fast as that C's cost model is transparent. If a C program is slow, it is slow in an obvious way: by executing a lot of statements. Compared with the cost of operations in C, high-level operations on objects (especially reflection) or strings can have costs that are not obvious.
Two languages that generally compile to binaries which are just as fast as C are Standard ML (using the MLton compiler) and Objective Caml. If you check out the benchmarks game you'll find that for some benchmarks, like binary trees, the OCaml version is faster than C. (I didn't find any MLton entries.) But don't take the shootout too seriously; it is, as it says, a game, the the results often reflect how much effort people have put in tuning the code.
C is not always faster.
C is slower than, for example, Modern Fortran.
C is often slower than Java for some things (especially after the JIT compiler has had a go at your code).
C lets pointer aliasing happen, which means some good optimizations are not possible. Particularly when you have multiple execution units, this causes data fetch stalls. Ow.
The assumption that pointer arithmetic works really causes slow bloated performance on some CPU families (PIC particularly!) It used to suck the big one on segmented x86.
Basically, when you get a vector unit, or a parallelizing compiler, C stinks and modern Fortran runs faster.
C programmer tricks, like thunking (modifying the executable on the fly), cause CPU prefetch stalls.
Do you get the drift?
And our good friend, the x86, executes an instruction set that these days bears little relationship to the actual CPU architecture. Shadow registers, load-store optimizers, all in the CPU. So C is then close to the virtual metal. The real metal, Intel don't let you see. (Historically VLIW CPU's were a bit of a bust so, maybe that's no so bad.)
If you program in C on a high-performance DSP (maybe a TI DSP?), the compiler has to do some tricky stuff to unroll the C across the multiple parallel execution units. So in that case, C isn't close to the metal, but it is close to the compiler, which will do whole program optimization. Weird.
And finally, some CPUs (www.ajile.com) run Java bytecodes in hardware. C would a PITA to use on that CPU.
what's to stop other languages from
being able to compile down to binary
that runs every bit as fast as C?
Nothing. Modern languages like Java or .NET languages are oriented more toward programmer productivity rather than performance. Hardware is cheap nowadays. Also compilation to intermediate representation gives a lot of bonuses such as security, portability, etc. The .NET CLR can take advantage of different hardware. For example, you don't need to manually optimize/recompile program to use the SSE instructions set.
I guess you forgot that Assembly language is also a language :)
But seriously, C programs are faster only when the programmer knows what he's doing. You can easily write a C program that runs slower than programs written in other languages that do the same job.
The reason why C is faster is because it is designed in this way. It lets you do a lot of "lower level" stuff that helps the compiler to optimize the code. Or, shall we say, you the programmer are responsible for optimizing the code. But it's often quite tricky and error prone.
Other languages, like others already mentioned, focus more on productivity of the programmer. It is commonly believed that programmer time is much more expensive than machine time (even in the old days). So it makes a lot of sense to minimize the time programmers spend on writing and debugging programs instead of the running time of the programs. To do that, you will sacrifice a bit on what you can do to make the program faster because a lot of things are automated.
The main factors are that it's a statically-typed language and that's compiled to machine code. Also, since it's a low-level language, it generally doesn't do anything you don't tell it to.
These are some other factors that come to mind.
Variables are not automatically initialized
No bounds checking on arrays
Unchecked pointer manipulation
No integer overflow checking
Statically-typed variables
Function calls are static (unless you use function pointers)
Compiler writers have had lots of time to improve the optimizing code. Also, people program in C for the purpose of getting the best performance, so there's pressure to optimize the code.
Parts of the language specification are implementation-defined, so compilers are free to do things in the most optimal way
Most static-typed languages could be compiled just as fast or faster than C though, especially if they can make assumptions that C can't because of pointer aliasing, etc.
C++ is faster on average (as it was initially, largely a superset of C, though there are some differences). However, for specific benchmarks, there is often another language which is faster.
From The Computer Language Benchmarks Game:
fannjuch-redux was fastest in Scala
n-body and fasta were faster in Ada.
spectral-norm was fastest in Fortran.
reverse-complement, mandelbrot and pidigits were fastest in ATS.
regex-dna was fastest in JavaScript.
chameneou-redux was fastest is Java 7.
thread-ring was fastest in Haskell.
The rest of the benchmarks were fastest in C or C++.
For the most part, every C instruction corresponds to a very few assembler instructions. You are essentially writing higher level machine code, so you have control over almost everything the processor does. Many other compiled languages, such as C++, have a lot of simple looking instructions that can turn into much more code than you think it does (virtual functions, copy constructors, etc..) And interpreted languages like Java or Ruby have another layer of instructions that you never see - the Virtual Machine or Interpreter.
I know plenty of people have said it in a long winded way, but:
C is faster because it does less (for you).
Many of these answers give valid reasons for why C is, or is not, faster (either in general or in specific scenarios). It's undeniable that:
Many other languages provide automatic features that we take for granted. Bounds checking, run-time type checking, and automatic memory management, for example, don't come for free. There is at least some cost associated with these features, which we may not think about—or even realize—while writing code that uses these features.
The step from source to machine is often not as direct in other languages as it is in C.
OTOH, to say that compiled C code executes faster than other code written in other languages is a generalization that isn't always true. Counter-examples are easy to find (or contrive).
All of this notwithstanding, there is something else I have noticed that, I think, affects the comparative performance of C vs. many other languages more greatly than any other factor. To wit:
Other languages often make it easier to write code that executes more slowly. Often, it's even encouraged by the design philosophies of the language. Corollary: a C programmer is more likely to write code that doesn't perform unnecessary operations.
As an example, consider a simple Windows program in which a single main window is created. A C version would populate a WNDCLASS[EX] structure which would be passed to RegisterClass[Ex], then call CreateWindow[Ex] and enter a message loop. Highly simplified and abbreviated code follows:
WNDCLASS wc;
MSG msg;
wc.style = 0;
wc.lpfnWndProc = &WndProc;
wc.cbClsExtra = 0;
wc.cbWndExtra = 0;
wc.hInstance = hInstance;
wc.hIcon = NULL;
wc.hCursor = LoadCursor(NULL, IDC_ARROW);
wc.hbrBackground = (HBRUSH)(COLOR_BTNFACE + 1);
wc.lpszMenuName = NULL;
wc.lpszClassName = "MainWndCls";
RegisterClass(&wc);
CreateWindow("MainWndCls", "", WS_OVERLAPPEDWINDOW | WS_VISIBLE,
CW_USEDEFAULT, 0, CW_USEDEFAULT, 0, NULL, NULL, hInstance, NULL);
while(GetMessage(&msg, NULL, 0, 0)){
TranslateMessage(&msg);
DispatchMessage(&msg);
}
An equivalent program in C# could be just one line of code:
Application.Run(new Form());
This one line of code provides all of the functionality that nearly 20 lines of C code did, and adds some things we left out, such as error checking. The richer, fuller library (compared to those used in a typical C project) did a lot of work for us, freeing our time to write many more snippets of code that look short to us but involve many steps behind the scenes.
But a rich library enabling easy and quick code bloat isn't really my point. My point is more apparent when you start examining what actually happens when our little one-liner actually executes. For fun sometime, enable .NET source access in Visual Studio 2008 or higher, and step into the simple one-linef above. One of the fun little gems you'll come across is this comment in the getter for Control.CreateParams:
// In a typical control this is accessed ten times to create and show a control.
// It is a net memory savings, then, to maintain a copy on control.
//
if (createParams == null) {
createParams = new CreateParams();
}
Ten times. The information roughly equivalent to the sum of what's stored in a WNDCLASSEX structure and what's passed to CreateWindowEx is retrieved from the Control class ten times before it's stored in a WNDCLASSEX structure and passed on to RegisterClassEx and CreateWindowEx.
All in all, the number of instructions executed to perform this very basic task is 2–3 orders of magnitude more in C# than in C. Part of this is due to the use of a feature-rich library, which is necessarily generalized, versus our simple C code which does exactly what we need and nothing more. But part of it is due to the fact that the modularized, object-oriented nature of .NET framework, lends itself to a lot of repetition of execution that often is avoided by a procedural approach.
I'm not trying to pick on C# or the .NET framework. Nor am I saying that modularization, generalization, library/language features, OOP, etc. are bad things. I used to do most of my development in C, later in C++, and most lately in C#. Similarly, before C, I used mostly assembly. And with each step "higher" my language goes, I write better, more maintainable, more robust programs in less time. They do, however, tend to execute a little more slowly.
I don't think anyone has mentioned the fact that much more effort has been put into C compilers than any other compiler, with perhaps the exception of Java.
C is extremely optimizable for many of the reasons already stated - more than almost any other language. So if the same amount of effort is put into other language compilers, C will probably still come out on top.
I think there is at least one candidate language that, with effort, could be optimized better than C and thus we could see implementations that produce faster binaries. I'm thinking of Digital Mars' D, because the creator took care to build a language that could potentially be better optimized than C. There may be other languages that have this possibility. However, I cannot imagine that any language will have compilers more than just a few percent faster than the best C compilers. I would love to be wrong.
I think the real "low hanging fruit" will be in languages that are designed to be easy for humans to optimize. A skilled programmer can make any language go faster, but sometimes you have to do ridiculous things or use unnatural constructs to make this happen. Although it will always take effort, a good language should produce relatively fast code without having to obsess over exactly how the program is written.
It's also important (at least to me) that the worst case code tends to be fast. There are numerous "proofs" on the web that Java is as fast or faster than C, but that is based on cherry picking examples.
I'm not big fan of C, but I know that anything I write in C is going to run well. With Java, it will "probably" run within 15% of the speed, usually within 25%, but in some cases it can be far worse. Any cases where it's just as fast or within a couple of percent are usually due to most of the time being spent in the library code which is heavily optimized C code anyway.
This is actually a bit of a perpetuated falsehood. While it is true that C programs are frequently faster, this is not always the case, especially if the C programmer isn't very good at it.
One big glaring hole that people tend to forget about is when the program has to block for some sort of I/O, such as user input in any GUI program. In these cases, it doesn't really matter what language you use since you are limited by the rate at which data can come in rather than how fast you can process it. In this case, it doesn't matter much if you are using C, Java, C# or even Perl; you just cannot go any faster than the data can come in.
The other major thing is that using garbage collection (GC) and not using proper pointers allows the virtual machine to make a number of optimizations not available in other languages. For instance, the JVM is capable of moving objects around on the heap to defragment it. This makes future allocations much faster since the next index can simply be used rather than looking it up in a table. Modern JVMs also don't have to actually deallocate memory; instead, they just move the live objects around when they GC and the spent memory from the dead objects is recovered essentially for free.
This also brings up an interesting point about C and even more so in C++. There is something of a design philosophy of "If you don't need it, you don't pay for it." The problem is that if you do want it, you end up paying through the nose for it. For instance, the vtable implementation in Java tends to be a lot better than C++ implementations, so virtual function calls are a lot faster. On the other hand, you have no choice but to use virtual functions in Java and they still cost something, but in programs that use a lot of virtual functions, the reduced cost adds up.
It's not so much about the language as the tools and libraries. The available libraries and compilers for C are much older than for newer languages. You might think this would make them slower, but au contraire.
These libraries were written at a time when processing power and memory were at a premium. They had to be written very efficiently in order to work at all. Developers of C compilers have also had a long time to work in all sorts of clever optimizations for different processors. C's maturity and wide adoption makes for a signficant advantage over other languages of the same age. It also gives C a speed advantage over newer tools that don't emphasize raw performance as much as C had to.
Amazing to see the old "C/C++ must be faster than Java because Java is interpreted" myth is still alive and kicking. There are articles going back a few years, as well as more recent ones, that explain with concepts or measurements why this simply isn't always the case.
Current virtual machine implementations (and not just the JVM, by the way) can take advantage of information gathered during program execution to dynamically tune the code as it runs, using a variety of techniques:
rendering frequent methods to machine code,
inlining small methods,
adjustment of locking
and a variety of other adjustments based on knowing what the code is actually doing, and on the actual characteristics of the environment in which it's running.
The lack of abstraction is what makes C faster. If you write an output statement you know exactly what is happening. If you write an output statement in Java it is getting compiled to a class file which then gets run on a virtual machine, introducing a layer of abstraction.
The lack of object-oriented features as a part of the language also increases its speed do to less code being generated. If you use C as an object-oriented language, then you are doing all the coding for things such as classes, inheritance, etc. This means rather than make something generalized enough for everyone with the amount of code and the performance penalty that requires you only write what you need to get the job done.
The fastest running code would be carefully handcrafted machine code. Assembler will be almost as good. Both are very low level and it takes a lot of writing code to do things. C is a little above assembler. You still have the ability to control things at a very low level in the actual machine, but there is enough abstraction, make writing it faster and easier then assembler.
Other languages, such as C# and Java, are even more abstract. While Assembler and machine code are called low-level languages, C# and JAVA (and many others) are called high-level languages. C is sometimes called a midlevel language.
Don't take someone’s word for it; look at the disassembly for both C and your language-of-choice in any performance critical part of your code. I think you can just look in the disassembly window at runtime in Visual Studio to see disassembled .NET code. It should be possible, if tricky, for Java using WinDbg, though if you do it with .NET, many of the issues would be the same.
I don't like to write in C if I don't need to, but I think many of the claims made in these answers that tout the speed of languages other than C can be put aside by simply disassembling the same routine in C and in your higher level language of choice, especially if lots of data is involved as is common in performance critical applications. Fortran may be an exception in its area of expertise; I don't know. Is it higher level than C?
The first time I did compare JITed code with native code resolved any and all questions whether .NET code could run comparably to C code. The extra level of abstraction and all the safety checks come with a significant cost. The same costs would probably apply to Java, but don't take my word for it; try it on something where performance is critical. (Does anyone know enough about JITed Java to locate a compiled procedure in memory? It should certainly be possible.)
Setting aside advanced optimization techniques such as hot-spot optimization, pre-compiled meta-algorithms, and various forms of parallelism, the fundamental speed of a language correlates strongly with the implicit behind-the-scenes complexity required to support the operations that would commonly be specified within inner loops.
Perhaps the most obvious is validity checking on indirect memory references—such as checking pointers for null and checking indexes against array boundaries. Most high-level languages perform these checks implicitly, but C does not. However, this is not necessarily a fundamental limitation of these other languages—a sufficiently clever compiler may be capable of removing these checks from the inner loops of an algorithm through some form of loop-invariant code motion.
The more fundamental advantage of C (and to a similar extent the closely related C++) is a heavy reliance on stack-based memory allocation, which is inherently fast for allocation, deallocation, and access. In C (and C++) the primary call stack can be used for allocation of primitives, arrays, and aggregates (struct/class).
While C does offer the capability to dynamically allocate memory of arbitrary size and lifetime (using the so called 'heap'), doing so is avoided by default (the stack is used instead).
Tantalizingly, it is sometimes possible to replicate the C memory allocation strategy within the runtime environments of other programming languages. This has been demonstrated by asm.js, which allows code written in C or C++ to be translated into a subset of JavaScript and run safely in a web browser environment—with near-native speed.
As somewhat of an aside, another area where C and C++ outshine most other languages for speed is the ability to seamlessly integrate with native machine instruction sets. A notable example of this is the (compiler and platform dependent) availability of SIMD intrinsics which support the construction of custom algorithms that take advantage of the now nearly ubiquitous parallel processing hardware—while still utilizing the data allocation abstractions provided by the language (lower-level register allocation is managed by the compiler).
1) As others have said, C does less for you. No initializing variables, no array bounds checking, no memory management, etc. Those features in other languages cost memory and CPU cycles that C doesn't spend.
2) Answers saying that C is less abstracted and therefore faster are only half correct I think. Technically speaking, if you had a "sufficiently advanced compiler" for language X, then language X could approach or equal the speed of C. The difference with C is that since it maps so obviously (if you've taken an architecture course) and directly to assembly language that even a naive compiler can do a decent job. For something like Python, you need a very advanced compiler to predict the probable types of objects and generate machine code on the fly -- C's semantics are simple enough that a simple compiler can do well.
Back in the good ole days, there were just two types of languages: compiled and interpreted.
Compiled languages utilized a "compiler" to read the language syntax and convert it into identical assembly language code, which could than just directly on the CPU. Interpreted languages used a couple of different schemes, but essentially the language syntax was converted into an intermediate form, and then run in a "interpreter", an environment for executing the code.
Thus, in a sense, there was another "layer" -- the interpreter -- between the code and the machine. And, as always the case in a computer, more means more resources get used. Interpreters were slower, because they had to perform more operations.
More recently, we've seen more hybrid languages like Java, that employ both a compiler and an interpreter to make them work. It's complicated, but a JVM is faster, more sophisticated and way more optimized than the old interpreters, so it stands a much better change of performing (over time) closer to just straight compiled code. Of course, the newer compilers also have more fancy optimizing tricks so they tend to generate way better code than they used to as well. But most optimizations, most often (although not always) make some type of trade-off such that they are not always faster in all circumstances. Like everything else, nothing comes for free, so the optimizers must get their boast from somewhere (although often times it using compile-time CPU to save runtime CPU).
Getting back to C, it is a simple language, that can be compiled into fairly optimized assembly and then run directly on the target machine. In C, if you increment an integer, it's more than likely that it is only one assembler step in the CPU, in Java however, it could end up being a lot more than that (and could include a bit of garbage collection as well :-) C offers you an abstraction that is way closer to the machine (assembler is the closest), but you end up having to do way more work to get it going and it is not as protected, easy to use or error friendly. Most other languages give you a higher abstraction and take care of more of the underlying details for you, but in exchange for their advanced functionality they require more resources to run. As you generalize some solutions, you have to handle a broader range of computing, which often requires more resources.
I have found an answer on a link about why some languages are faster and some are slower, I hope this will clear more about why C or C++ is faster than others, There are some other languages also that is faster than C, but we can not use all of them. Some explanation -
One of the big reasons that Fortran remains important is because it's fast: number crunching routines written in Fortran tend to be quicker than equivalent routines written in most other languages. The languages that are competing with Fortran in this space—C and C++—are used because they're competitive with this performance.
This raises the question: why? What is it about C++ and Fortran that make them fast, and why do they outperform other popular languages, such as Java or Python?
Interpreting versus compiling
There are many ways to categorize and define programming languages, according to the style of programming they encourage and features they offer. When looking at performance, the biggest single distinction is between interpreted languages and compiled ones.
The divide is not hard; rather, there's a spectrum. At one end, we have traditional compiled languages, a group that includes Fortran, C, and C++. In these languages, there is a discrete compilation stage that translates the source code of a program into an executable form that the processor can use.
This compilation process has several steps. The source code is analyzed and parsed. Basic coding mistakes such as typos and spelling errors can be detected at this point. The parsed code is used to generate an in-memory representation, which too can be used to detect mistakes—this time, semantic mistakes, such as calling functions that don't exist, or trying to perform arithmetic operations on strings of text.
This in-memory representation is then used to drive a code generator, the part that produces executable code. Code optimization, to improve the performance of the generated code, is performed at various times within this process: high-level optimizations can be performed on the code representation, and lower-level optimizations are used on the output of the code generator.
Actually executing the code happens later. The entire compilation process is simply used to create something that can be executed.
At the opposite end, we have interpreters. The interpreters will include a parsing stage similar to that of the compiler, but this is then used to drive direct execution, with the program being run immediately.
The simplest interpreter has within it executable code corresponding to the various features the language supports—so it will have functions for adding numbers, joining strings, whatever else a given language has. As it parses the code, it will look up the corresponding function and execute it. Variables created in the program will be kept in some kind of lookup table that maps their names to their data.
The most extreme example of the interpreter style is something like a batch file or shell script. In these languages, the executable code is often not even built into the interpreter itself, but rather separate, standalone programs.
So why does this make a difference to performance? In general, each layer of indirection reduces performance. For example, the fastest way to add two numbers is to have both of those numbers in registers in the processor, and to use the processor's add instruction. That's what compiled programs can do; they can put variables into registers and take advantage of processor instructions. But in interpreted programs, that same addition might require two lookups in a table of variables to fetch the values to add, then calling a function to perform the addition. That function may very well use the same processor instruction as the compiled program uses to perform the actual addition, but all the extra work before the instruction can actually be used makes things slower.
If you want to know more please check the source.
Some C++ algorithms are faster than C, and some implementations of algorithms or design patterns in other languages can be faster than C.
When people say that C is fast, and then move on to talking about some other language, they are generally using C's performance as a benchmark.
Just step through the machine code in your IDE, and you'll see why it's faster (if it's faster). It leaves out a lot of hand-holding. Chances are your Cxx can also be told to leave it out too, in which case it should be about the same.
Compiler optimizations are overrated, as are almost all perceptions about language speed.
Optimization of generated code only makes a difference in hotspot code, that is, tight algorithms devoid of function calls (explicit or implicit). Anywhere else, it achieves very little.
With modern optimizing compilers, it's highly unlikely that a pure C program is going to be all that much faster than compiled .NET code, if at all. With the productivity enhancement that frameworks like .NET provide the developer, you can do things in a day that used to take weeks or months in regular C. Coupled with the cheap cost of hardware compared to a developer's salary, it's just way cheaper to write the stuff in a high-level language and throw hardware at any slowness.
The reason Jeff and Joel talk about C being the "real programmer" language is because there isn't any hand-holding in C. You must allocate your own memory, deallocate that memory, do your own bounds-checking, etc. There isn't any such thing as new object(); There isn't any garbage collection, classes, OOP, entity frameworks, LINQ, properties, attributes, fields, or anything like that.
You have to know things like pointer arithmetic and how to dereference a pointer. And, for that matter, know and understand what a pointer is. You have to know what a stack frame is and what the instruction pointer is. You have to know the memory model of the CPU architecture you're working on. There is a lot of implicit understanding of the architecture of a microcomputer (usually the microcomputer you're working on) when programming in C that simply is not present nor necessary when programming in something like C# or Java. All of that information has been off-loaded to the compiler (or VM) programmer.
It's the difference between automatic and manual. Higher-level languages are abstractions, thus automated. C/C++ are manually controlled and handled; even error checking code is sometimes a manual labor.
C and C++ are also compiled languages which means none of that run-everywhere business. These languages have to be fine-tuned for the hardware you work with, thus adding an extra layer of gotcha. Though this is slightly phasing out now as C/C++ compilers are becoming more common across all platforms. You can do cross compilations between platforms. It's still not a run everywhere situation, and you’re basically instructing compiler A to compile against compiler B the same code on a different architecture.
Bottom line, C languages are not meant to be easy to understand or reason. This is also why they’re referred to as systems languages. They came out before all this high-level abstraction nonsense. This is also why they are not used for front end web programming. They’re just not suited to the task; they’re meant to solve complex problems that can't be resolved with conventional language tooling.
This is why you get crazy stuff, like micro-architectures, drivers, quantum physics, AAA games, and operating systems. There are things C and C++ are just well suited for. Speed and number crunching being the chief areas.
C is fast because it is natively compiled, low-level language. But C is not the fastest. The Recursive Fibonacci Benchmark shows that Rust, Crystal, and Nim can be faster.