I need to optimize code to get room for some new code. I do not have the space for all the changes. I can not use code bank switching (80c31 with 64k).
You haven't really given a lot to go on here, but there are two main levels of optimizations you can consider:
Micro-Optimizations:
eg. XOR A instead of MOV A,0
Adam has covered some of these nicely earlier.
Macro-Optimizations:
Look at the structure of your program, the data structures and algorithms used, the tasks performed, and think VERY hard about how these could be rearranged or even removed. Are there whole chunks of code that actually aren't used? Is your code full of debug output statements that the user never sees? Are there functions specific to a single customer that you could leave out of a general release?
To get a good handle on that, you'll need to work out WHERE your memory is being used up. The Linker map is a good place to start with this. Macro-optimizations are where the BIG wins can be made.
As an aside, you could - seriously- try rewriting parts of your code with a good optimizing C compiler. You may be amazed at how tight the code can be. A true assembler hotshot may be able to improve on it, but it can easily be better than most coders. I used the IAR one about 20 years ago, and it blew my socks off.
With assembly language, you'll have to optimize by hand. Here are a few techniques:
Note: IANA8051P (I am not an 8501 programmer but I have done lots of assembly on other 8 bit chips).
Go through the code looking for any duplicated bits, no matter how small and make them functions.
Learn some of the more unusual instructions and see if you can use them to optimize, eg. A nice trick is to use XOR A to clear the accumulator instead of MOV A,0 - it saves a byte.
Another neat trick is if you call a function before returning, just jump to it eg, instead of:
CALL otherfunc
RET
Just do:
JMP otherfunc
Always make sure you are doing relative jumps and branches wherever possible, they use less memory than absolute jumps.
That's all I can think of off the top of my head for the moment.
Sorry I am coming to this late, but I once had exactly the same problem, and it became a repeated problem that kept coming back to me. In my case the project was a telephone, on an 8051 family processor, and I had totally maxed out the ROM (code) memory. It kept coming back to me because management kept requesting new features, so each new feature became a two step process. 1) Optimize old stuff to make room 2) Implement the new feature, using up the room I just made.
There are two approaches to optimization. Tactical and Strategical. Tactical optimizations save a few bytes at a time with a micro optimization idea. I think you need strategic optimizations which involve a more radical rethinking about how you are doing things.
Something I remember worked for me and could work for you;
Look at the essence of what your code has to do and try to distill out some really strong flexible primitive operations. Then rebuild your top level code so that it does nothing low level at all except call on the primitives. Ideally use a table based approach, your table contains stuff like; Input state, event, output state, primitives.... In other words when an event happens, look up a cell in the table for that event in the current state. That cell tells you what new state to change to (optionally) and what primitive(s) (if any) to execute. You might need multiple sets of states/events/tables/primitives for different layers/subsystems.
One of the many benefits of this approach is that you can think of it as building a custom language for your particular problem, in which you can very efficiently (i.e. with minimal extra code) create new functionality simply by modifying the table.
Sorry I am months late and you probably didn't have time to do something this radical anyway. For all I know you were already using a similar approach! But my answer might help someone else someday who knows.
In the whacked-out department, you could also consider compressing part of your code and only keeping some part that is actively used decompressed at any particular point in time. I have a hard time believing that the code required for the compress/decompress system would be small enough a portion of the tiny memory of the 8051 to make this worthwhile, but has worked wonders on slightly larger systems.
Yet another approach is to turn to a byte-code format or the kind of table-driven code that some state machine tools output -- having a machine understand what your app is doing and generating a completely incomprehensible implementation can be a great way to save room :)
Finally, if the code is indeed compiled in C, I would suggest compiling with a range of different options to see what happens. Also, I wrote a piece on compact C coding for the ESC back in 2001 that is still pretty current. See that text for other tricks for small machines.
1) Where possible save your variables in Idata not in xdata
2) Look at your Jmp statements – make use of SJmp and AJmp
I assume you know it won't fit because you wrote/complied and got the "out of memory" error. :) It appears the answers address your question pretty accurately; short of getting code examples.
I would, however, recommend a few additional thoughts;
Make sure all the code is really
being used -- code coverage test? An
unused sub is a big win -- this is a
tough step -- if you're the original
author, it may be easier -- (well, maybe) :)
Ensure the level of "verification"
and initialization -- sometimes we
have a tendency to be over zealous
in insuring we have initialized
variables/memory and sure enough
rightly so, how many times have we
been bitten by it. Not saying don't
initialize (duh), but if we're doing
a memory move, the destination
doesn't need to be zero'd first --
this dovetails with
1 --
Eval the new features -- can an
existing sub be be enhanced to cover
both functions or perhaps an
existing feature replaced?
Break up big code if a piece of the
big code can save creating a new
little code.
or perhaps there's an argument for hardware version 2.0 on the table now ... :)
regards
Besides the already mentioned (more or less) obvious optimizations, here is a really weird (and almost impossible to achieve) one: Code reuse. And with Code reuse I dont mean the normal reuse, but to a) reuse your code as data or b) to reuse your code as other code. Maybe you can create a lut (or whatever static data) that it can represented by the asm hex opcodes (here you have to look harvard vs von neumann architecture).
The other would reuse code by giving code a different meaning when you address it different. Here an example to make clear what I mean. If the bytes for your code look like this: AABCCCDDEEFFGGHH at address X where each letter stands for one opcode, imagine you would now jump to X+1. Maybe you get a complete different functionality where the now by space seperated bytes form the new opcodes: ABC CCD DE EF GH.
But beware: This is not only tricky to achieve (maybe its impossible), but its a horror to maintain. So if you are not a demo code (or something similiar exotic), I would recommend to use the already other mentioned ways to save mem.
Related
I am in the second year of my bachelor study in information technology. Last year in one of my courses they taught me to write clean code so other programmers have an easier time working with your code. I learned a lot about writing clean code from a video ("clean code") on pluralsight (paid website for learning which my school uses). There was an example in there about assigning if conditions to boolean variables and using them to enhance readability. In my course today my teacher told me it's very bad code because it decreases performance (in bigger programs) due to increased tests being executed. I was wondering now whether I should continue using boolean variables for readability or not use them for performance. I will illustrate in an example (I am using python code for this example):
example boolean variable
Let's say we need to check whether somebody is legal to drink alcohol we get the persons age and we know the legal drinking age is 21.
is_old_enough = persons_age >= legal_drinking_age
if is_old_enough:
do something
My teacher told me today that this would be very bad for performance since 2 tests are performed first persons_age >= legal_drinking_age is tested and secondly in the if another test occurs whether the person is_old_enough.
My teacher told me that I should just put the condition in the if, but in the video they said that code should be read like natural language to make it clear for other programmers. I was wondering now which would be the better coding practice.
example condition in if:
if persons_age >= legal_drinking_age:
do something
In this example only 1 test is tested whether persons_age >= legal_drinking_age. According to my teacher this is better code.
Thank you in advance!
yours faithfully
Jonas
I was wondering now which would be the better coding practice.
The real safe answer is : Depends..
I hate to use this answer, but you won't be asking unless you have faithful doubt. (:
IMHO:
If the code will be used for long-term use, where maintainability is important, then a clearly readable code is preferred.
If the program speed performance crucial, then any code operation that use less resource (smaller dataSize/dataType /less loop needed to achieve the same thing/ optimized task sequencing/maximize cpu task per clock cycle/ reduced data re-loading cycle) is better. (example keyword : space-for-time code)
If the program minimizing memory usage is crucial, then any code operation that use less storage and memory resource to complete its operation (which may take more cpu cycle/loop for the same task) is better. (example: small devices that have limited data storage/RAM)
If you are in a race, then you may what to code as short as possible, (even if it may take a slightly longer cpu time later). example : Hackathon
If you are programming to teach a team of student/friend something.. Then readable code + a lot of comment is definitely preferred .
If it is me.. I'll stick to anything closest to assembly language as possible (as much control on the bit manipulation) for backend development. and anything closest to mathematica-like code (less code, max output, don't really care how much cpu/memory resource is needed) for frontend development. ( :
So.. If it is you.. you may have your own requirement/preference.. from the user/outsiders/customers point of view.. it is just a working/notWorking program. YOur definition of good program may defer from others.. but this shouldn't stop us to be flexible in the coding style/method.
Happy exploring. Hope it helps.. in any way possible.
Performance
Performance is one of the least interesting concerns for this question, and I say this as one working in very performance-critical areas like image processing and raytracing who believes in effective micro-optimizations (but my ideas of effective micro-optimization would be things like improving memory access patterns and memory layouts for cache efficiency, not eliminating temporary variables out of fear that your compiler or interpreter might allocate additional registers and/or utilize additional instructions).
The reason it's not so interesting is, because, as pointed out in the comments, any decent optimizing compiler is going to treat those two you wrote as equivalent by the time it finishes optimizing the intermediate representation and generates the final results of the instruction selection/register allocation to produce the final output (machine code). And if you aren't using a decent optimizing compiler, then this sort of microscopic efficiency is probably the last thing you should be worrying about either way.
Variable Scopes
With performance aside, the only concern I'd have with this convention, and I think it's generally a good one to apply liberally, is for languages that don't have a concept of a named constant to distinguish it from a variable.
In those cases, the more variables you introduce to a meaty function, the more intellectual overhead it can have as the number of variables with a relatively wide scope increases, and that can translate to practical burdens in maintenance and debugging in extreme cases. If you imagine a case like this:
some_variable = ...
...
some_other_variable = ...
...
yet_another_variable = ...
(300 lines more code to the function)
... in some function, and you're trying to debug it, then those variables combined with the monstrous size of the function starts to multiply the difficulty of trying to figure out what went wrong. That's a practical concern I've encountered when debugging codebases spanning millions of lines of code written by all sorts of people (including those no longer on the team) where it's not so fun to look at the locals watch window in a debugger and see two pages worth of variables in some monstrous function that appears to be doing something incorrectly (or in one of the functions it calls).
But that's only an issue when it's combined with questionable programming practices like writing functions that span hundreds or thousands of lines of code. In those cases it will often improve everything just focusing on making reasonable-sized functions that perform one clear logical operation and don't have more than one side effect (or none ideally if the function can be programmed as a pure function). If you design your functions reasonably then I wouldn't worry about this at all and favor whatever is readable and easiest to comprehend at a glance and maybe even what is most writable and "pliable" (to make changes to the function easier if you anticipate a future need).
A Pragmatic View on Variable Scopes
So I think a lot of programming concepts can be understood to some degree by just understanding the need to narrow variable scopes. People say avoid global variables like the plague. We can go into issues with how that shared state can interfere with multithreading and how it makes programs difficult to change and debug, but you can understand a lot of the problems just through the desire to narrow variable scopes. If you have a codebase which spans a hundred thousand lines of code, then a global variable is going to have the scope of a hundred thousands of lines of code for both access and modification, and crudely speaking a hundred thousand ways to go wrong.
At the same time that pragmatic sort of view will find it pointless to make a one-shot program which only spans 100 lines of code with no future need for extension avoid global variables like the plague, since a global here is only going to have 100 lines worth of scope, so to speak. Meanwhile even someone who avoids those like the plague in all contexts might still write a class with member variables (including some superfluous ones for "convenience") whose implementation spans 8,000 lines of code, at which point those variables have a much wider scope than even the global variable in the former example, and this realization could drive someone to design smaller classes and/or reduce the number of superfluous member variables to include as part of the state management for the class (which can also translate to simplified multithreading and all the similar types of benefits of avoiding global variables in some non-trivial codebase).
And finally it'll tend to tempt you to write smaller functions as well, since a variable towards the top of some function spanning 500 lines of code is going to also have a fairly wide scope. So anyway, my only concern when you do this is to not let the scope of those temporary, local variables get too wide. And if they do, then the general answer is not necessarily to avoid those variables but to narrow their scope.
when I get a kernel using too many registers there are basically 3 options I can do:
leave the kernel as it is, which results in low occupancy
set compiler to use lower number of registers, spilling them, causing worse performance
rewrite the kernel
For option 3, I'd like to know which part of the kernel needs the maximum number of registers. Is there any tool or technique allowing me to identify this part? Reading through the PTX code (I develop on NVidia) is not helpful, the registers have various high numbers and to be honest, the best I can do is to identify which part of the assembly code maps to which part of the C code.
Just commenting out some code is not much a way to go - for example, I noticed that if I just put the code into loop, the number of registers raises dramatically, not only by one for the loop control variable. I personally suspect the NVidia compiler from imperfect variable liveness analysis, but of course I cannot do much with that :-)
If you're running on NVidia hardware, you can pass -cl-nv-verbose compile option to clBuildProgram then clGetProgramInfo CL_PROGRAM_BINARIES to get human readable text about the compile. In there it will say the number of registers it uses. Note that NVidia caches compiles and it only produces that register info when the kernel source actually changes, so you may want to inject some superfluous change in the source code to force it to do the full analysis.
If you're running on AMD hardware, just set the environment variable GPU_DUMP_DEVICE_KERNEL=1. It will produce a text file of the IL during the compile. Not sure it explicitly says the number of registers used, but it's what is equivalent to the NVidia technique above.
When looking at that output (at least on nvidia), you'll find that it seems to use an infinite number of registers (if you go by the register numbers). In reality, it does a flow analysis and actually reuses registers in a way that is not at all obvious when looking at the IL.
This is a tough question in any language, and there's probably not one correct answer. But here are some things to think about:
Look for the code in the "deepest" scope you can find, keeping in mind that most functions are probably inlined by your OpenCL compiler. Count the variables used in this scope, and walk up the containing scopes. In each containing scope, count variables that are used both before and after the inner scope. These are potentially live while the inner scope executes. This process could help you account for the live registers at a particular part of the program.
Try to take variables that "span" deep scopes and make them not span the scope if possible. For example, if you have something like this:
int i = func1();
int j = func2(); // perhaps lots of live registers here
int k = func3(i,j);
you could try to reorder the first two lines if func2 has lots of live registers. That would remove i from the set of live registers while func2 is running. This is a trivial pattern, of course, but hopefully it's illustrative.
Think about getting rid of variables that just keep around the results of simple computations. You might be able to recompute these when you need them. For example, if you have something like int i = get_local_id(0) you might be able to just use get_local_id(0) wherever you would have used i.
Think about getting rid of variables that are keeping around values stored in memory.
Without good tools for this kind of thing, it ends up being more art than science. But hopefully some of this is helpful.
The idea is somewhat similar to what Apple has done in the OpenGL stack. I want to have that a bit more general.
Basically, I want to have specialised and optimised variants of some code for some specific cases.
In other words: I have given an algorithm/code for a function (let B = {0,1})
f : B^n -> B^m
Now, I special a specific case by a function (which predefines part of the input of f)
preset : {1..n} -> {0,1,unset}
The amount of predefinitions (∈ {0..n}) is then given by
pn := |preset⁻¹({0,1})|
Canonically, we now get a specialised function
f_preset : B^(n-pn) -> B^m
Also canonically, we get the code/algorithm for this specialised function. Naturally, the code for f_preset will be somewhat more fast than f with pn > 0. Then, you also can optimise this code further (there might be some dead code now, some loops can be unpacked now, some calculations can be precalculated, etc). In some cases, it can have noteable improvements.
Apple does roughly this for their OpenGL stack (from what I have read / know): They try to find a good preset at runtime after everything is setup for variables which will not change anymore, then make an optimised version of the specialised function and only use that one instead of the original function.
Initially, I thought about a way to optimise the physics simulation of some own game. There I have a lot of particle objects and a set of particle types (which is unknown at compile time). A particle type is a set of attributes. The particle types are fixed and constant once they are loaded. Each particle object is of one of theye particle types. The physic simulation for a particle object is some very heavy peace of code with many many branches and very heavily depends on the particle type. My idea was now to have an optimised physics simulation function for each particle type.
After thinking a bit about this, I wanted to go a bit further:
I want to automatically calculate a set of such presets at runtime and maintain the optimised code for each. And I want to automatically add or remove presets when the circumstances change.
There are several questions now:
Is there an easy way to calculate a good preset? How do I know what variables are constant for a given situation?
Is there an easy way to check how good a preset is? 'Good' refers to the performance of the resulting optimised code.
How to compare two algorithms/codes for performance? Via some heuristic? Or by testing with random input?
How many presets (and optimised code variants) should there be for a function? A fixed limit for all functions? Or is this different for every function? Is it maybe even depending on the current computer state?
How to maintain the different optimised code variants? A wrapper function around f which chooses automatically the best optimised variant doesn't seem to be very nice as this maybe not so easy check would be needed for every single call. A solution to this problem might also be deeply related to the question about how to find the set/amount of good presets. (In the particle type case, the optimised code would be attached to / saved together with the particle type. The amount of particle types also define the amount of presets.)
For my initial case, most of these questions are kind of obsolete but am really interested now in how to do this in a more general way. Of course, most/all of these questions are also uncalculateable but I wonder to what degree you may still get good results.
This whole topic is also very important for optimisations in JIT compilers. Are they doing these kind of optimisations already? To what degree?
Are there good recent research works which answers some of my questions? Or maybe also some results which say that it is just too hard to do this in such a general way?
It seems to me you are asking about partial evaluation.
I actually have a bit of a problem with that concept, because it is usually couched in terms that are over-academic and over-difficult.
The way it is usually expressed is that you have some general function F(Islow, Ifast) having arguments that can take different values at different times. The Islow arguments change seldom, and the Ifast arguments can be different every time it is called.
Then the problem is to write some kind of partial-evaluator function G(F, Islow) -> F1(Ifast) that takes function F and the Islow arguments, and generates a new (simpler) function F1 that only takes the Ifast arguments.
The problem with this is 1) somebody has to write the general function F, and 2) somebody has to write the general partial evaluator G.
What makes more sense to me is to write from scratch a function H(Islow) -> F1(Ifast), that is, write a code-generator specifically for F1, rather than writing two functions F and G, especially where G is very difficult to write.
H is usually much easier to write than F, and G need not be written at all! The result function F1 usually is smaller and has much higher performance than F, so it's a win-win situation.
When people write code generators, that is what they are doing, and it is a very effective programming technique.
This is a dangerous question, so let me try to phrase it correctly. Premature optimization is the root of all evil, but if you know you need it, there is a basic set of rules that should be considered. This set is what I'm wondering about.
For instance, imagine you got a list of a few thousand items. How do you look up an item with a specific, unique ID? Of course, you simply use a Dictionary to map the ID to the item.
And if you know that there is a setting stored in a database that is required all the time, you simply cache it instead of issuing a database request hundred times a second.
Or even something as simple as using a release instead of a debug build in prod.
I guess there are a few even more basic ideas.
I am specifically not looking for "don't do it, for experts: don't do it yet" or "use a profiler" answers, but for really simple, general hints. If you feel this is an argumentative question, you probably misunderstood my intention.
I am also not looking for concrete advice in any of my projects nor any sophisticated low level tricks. Think of it as an overview of how to avoid the most important performance mistakes you made as a very beginner.
Edit: This might be a good description of what I am looking for: Create a presentation (not a practical example) of common optimization rules for people who have a basic technical understanding (let's say they got a CS degree) but for some reason never wrote a single line of code. Point out the most important aspects. Pseudocode is fine. Do not assume specific languages or even architectures.
Two rules:
Use the right data structures.
Use the right algorithms.
I think that covers it.
Minimize the number of network roundtrips
Minimize the number of harddisk seeks
These are several orders of magnitude slower than anything else your program is likely to do, so avoiding them can be very important indeed. Typical methods to achieve this are:
Caching
Increasing the granularity of network and HD accesses
For example, B-Trees are absolutely ubiquitous in DB systems because the reduce the granularity of HD access for on-disk index lookups.
I think something extremely important is to be very carefully on all code that is frequently executed. This is normally the code in critical inner loops.
Rule 1: Know this code
For this code avoid all overhead. Small differences in runtime can make a big impact on the overall performance. E.g. if you implement an image filter a difference of 0.001ms per pixel will make a difference in 1s in the filter runtime on a image with size 1000x1000 (which is not big).
Things to avoid/do in inner loops are:
don't go through interfaces (e.g DB queries, RPC calls etc)
don't jump around in the RAM, try to access it linearly
if you have to read from disk then read large chunks outside the inner loop (paging)
avoid virtual function calls
avoid function calls / use inline functions
use float instead of double if possible
avoid numerical casts if possible
use ++a instead of a++ in C++
iterate directly on pointers if possible
The second general advice: Each layer/interface costs, try to avoid large stacks of different technologies, the system will spend more time in data transformation then in doing the actual job, keep things simple.
And as the others said, use the right algorithm, try to optimize the algorithm complexity first before you optimize the algorithm implementation.
I know you're looking for specific coding hints, but those are easy to find: cacheing, loop unrolling, code hoisting, data & code locality, blah, blah...
The biggest hint of all is don't use them.
Would it help to make this point if I said "This is the secret that the almighty Powers That Be don't want you to know!!"? Pick your Powers: Microsoft, Google, Sun, etc. etc.
Don't Use Them
Until you know, with dead certainty, what the problems are, and then the coding hints are obvious.
Here's an example where many coding tricks were used, but the heart and soul of the exercise is not the coding techniques, but the diagnostic technique.
Are your algorithms correct for the situation or are there better ones available?
I've always hated comments that fill half the screen with asterisks just to tell you that the function returns a string, I never read those comments.
However, I do read comments that describe why something is done and how it's done (usually the single line comments in the code); those come in really handy when trying to understand someone else's code.
But when it comes to writing comments, I don't write that, rather, I use comments only when writing algorithms in programming contests, I'd think of how the algorithm will do what it does then I'd write each one in a comment, then write the code that corresponds to that comment.
An example would be:
//loop though all the names from n to j - 1
Other than that I can't imagine why anyone would waste valuable time writing comments when he could be writing code.
Am I right or wrong? Am I missing something? What other good use cases of comments am I not aware of?
Comments should express why you are doing something not what you are doing
It's an old adage, but a good metric to use is:
Comment why you're doing something, not how you're doing it.
Saying "loop through all the names from n to j-1" should be immediately clear to even a novice programmer from the code alone. Giving the reason why you're doing that can help with readability.
If you use something like Doxygen, you can fully document your return types, arguments, etc. and generate a nice "source code manual." I often do this for clients so that the team that inherits my code isn't entirely lost (or forced to review every header).
Documentation blocks are often overdone, especially is strongly typed languages. It makes a lot more sense to be verbose with something like Python or PHP than C++ or Java. That said, it's still nice to do for methods & members that aren't self explanatory (not named update, for instance).
I've been saved many hours of thinking, simply by commenting what I'd want to tell myself if I were reading my code for the first time. More narrative and less observation. Comments should not only help others, but yourself as well... especially if you haven't touched it in five years. I have some ten year old Perl that I wrote and I still don't know what it does anymore.
Something very dirty, that I've done in PHP & Python, is use reflection to retrieve comment blocks and label elements in the user interface. It's a use case, albeit nasty.
If using a bug tracker, I'll also drop the bug ID near my changes, so that I have a reference back to the tracker. This is in addition to a brief description of the change (inline change logs).
I also violate the "only comment why not what" rule when I'm doing something that my colleagues rarely see... or when subtlety is important. For instance:
for (int i = 50; i--; ) cout << i; // looping from 49..0 in reverse
for (int i = 50; --i; ) cout << i; // looping from 49..1 in reverse
I use comments in the following situations:
High-level API documentation comments, i.e. what is this class or function for?
Commenting the "why".
A short, high-level summary of what a much longer block of code does. The key word here is summary. If someone wants more detail, the code should be clear enough that they can get it from the code. The point here is to make it easy for someone browsing the code to figure out where some piece of logic is without having to wade through the details of how it's performed. Ideally these cases should be factored out into separate functions instead, but sometimes it's just not do-able because the function would have 15 parameters and/or not be nameable.
Pointing out subtleties that are visible from reading the code if you're really paying attention, but don't stand out as much as they should given their importance.
When I have a good reason why I need to do something in a hackish way (performance, etc.) and can't write the code more clearly instead of using a comment.
Comment everything that you think is not straightforward and you won't be able to understand the next time you see your code.
It's not a bad idea to record what you think your code should be achieving (especially if the code is non-intuitive, if you want to keep comments down to a minimum) so that someone reading it a later date, has an easier time when debugging/bugfixing. Although one of the most frustrating things to encounter in reading someone else's code is cases where the code has been updated, but not the comments....
I've always hated comments that fill half the screen with asterisks just to tell you that the function returns a string, I never read those comments.
Some comments in that vein, not usually with formatting that extreme, actually exist to help tools like JavaDoc and Doxygen generate documentation for your code. This, I think, is a good form of comment, because it has both a human- and machine-readable format for documentation (so the machine can translate it to other, more useful formats like HTML), puts the documentation close to the code that it documents (so that if the code changes, the documentation is more likely to be updated to reflect these changes), and generally gives a good (and immediate) explanation to someone new to a large codebase of why a particular function exists.
Otherwise, I agree with everything else that's been stated. Comment why, and only comment when it's not obvious. Other than Doxygen comments, my code generally has very few comments.
Another type of comment that is generally useless is:
// Commented out by Lumpy Cheetosian on 1/17/2009
...uh, OK, the source control system would have told me that. What it won't tell me is WHY Lumpy commented out this seemingly necessary piece of code. Since Lumpy is located in Elbonia, I won't be able to find out until Monday when they all return from the Snerkrumph holiday festival.
Consider your audience, and keep the noise level down. If your comments include too much irrelevant crap, developers will just ignore them in practice.
BTW: Javadoc (or Doxygen, or equiv.) is a Good Thing(tm), IMHO.
I also use comments to document where a specific requirement came from. That way the developer later can look at the requirement that caused the code to be like it was and, if the new requirement conflicts with the other requirment get that resolved before breaking an existing process. Where I work requirments can often come from different groups of people who may not be aware of other requirements then system must meet. We also get frequently asked why we are doing a certain thing a certain way for a particular client and it helps to be able to research to know what requests in our tracking system caused the code to be the way it is. This can also be done on saving the code in the source contol system, but I consider those notes to be comments as well.
Reminds me of
Real programmers don't write documentation
I wrote this comment a while ago, and it's saved me hours since:
// NOTE: the close-bracket above is NOT the class Items.
// There are multiple classes in this file.
// I've already wasted lots of time wondering,
// "why does this new method I added at the end of the class not exist?".