C and MPI: function works differently with same data - c

I have successfully wrote a complicate function with PETSc library (it's a MPI-based scientific library for parallel solving huge linear systems). This library provides its own "malloc" version and basic datatypes (i.e. "PetscInt" as standard "int"). For this function, I've always been using PETSc stuff instead of standard stuff such as "malloc" and "int". The function has been extensevely tested and always worked fine. Despite the use of MPI, the function is fully serial, and all processors perform it on the same data (each processor has its copy): no communication involved at all.
Then, I decided to not use PETSc and write a standard MPI version. Basically, I rewrote all code substituting PETSc stuff with classic C stuff, not with brutal force but paying attention for substitutions (no "Replace" tool of any editor, I mean! All done by hands). During substitution, few minor changes have been made, such as declaring two different variables a and b, instead of declaring a[2]. These are the substitutions:
PetscMalloc -> malloc
PetscScalar -> double
PetscInt -> int
PetscBool -> created an enum structure to replicate it, as C doesn't have boolean datatype.
Basically, algorithms have not been changed during the substitution process. The main function is a "for" loop (actually 4 nested loops). At each iteration, it calls another function. Let's call it Disfunction. Well, Disfunction works perfectly outside the 4-cycle (as I tested it separately), but inside the 4-cycle, in some cases works, in some doesn't. Also, I checked the data passed to Disfunction at each iteration: with ECXACTELY the same input, Disfunction performs different computations between one iteration and another.
Also, computed data doesn't seem to be Undefined Behaviour, as Disfunction always gives back the same results with different runs of the program.
I've noticed that changing the number of processors for "mpiexec" gives different computational results.
That's my problem. Few other considerations: the program use extensively "malloc"; computed data is the same for all processes, correct or not; Valgrind doesn't detect errors (apart from detecting error with normal use of printf, which is another problem and an OT); Disfunction calls recursively two other functions (extensively tested in PETSc version as well); algorithms involved are mathematically correct; Disfunction depends on an integer parameter p>0: for p=1,2,3,4,5 it works PERFECTELY, for p>=6 it does not.
If asked, I can post the code but it's long and complicated (scientifically, not informatically) and I think it requires time to be explained.
My idea is that I mess up with memory allocations, but I can't understand where.
Sorry for my english and for bad formattation.

Well, I don't know if anyone is stll interested, but the problem was that PETSc functon PetscMalloc zero-initialize the data, not like standard C malloc. Stupid mistake... – user3029623

The only suggestion I can offer without reference to the code itself is to try to construct progressively simpler test cases that demonstrate your issue.
When you narrow down the iterative process to a single point in your data set or a single step (by eliminating some loops), does the error still occur? If not, that might suggest their bounds are wrong.
Does the erroneous output always occur on particular loop indices, especially the first or last? Perhaps there are some ghost or halo values you're missing or some boundary condition that you're not properly accounting for.

Related

Is there any use for recursions?

I've recently learned about tail-recursions as a way to make a recursion that doesn't crash when you give it too big of a number to work with. I realised that I could easily rewrite a tail-recursion as a while loop and have it do basically exactly the same thing, which lead me wondering - is there any use for recursions when you can do everything with a normal loop?
Yes, recursion code looks smaller and is easier to understand, but it also has a chance of completely crashing, while a simple loop cannot crash doing the same task.
I'll take for example the Haskell language, it is Purely functional:
Every function in Haskell is a function in the mathematical sense
(i.e., "pure"). Even side-effecting IO operations are but a
description of what to do, produced by pure code. There are no
statements or instructions, only expressions which cannot mutate
variables (local or global) nor access state like time or random
numbers.
So, in haskell a recursive function is tail recursive if the final result of the
recursive call is the final result of the function itself. If the
result of the recursive call must be further processed (say, by adding
1 to it, or consing another element onto the beginning of it), it is
not tail recursive. (see here)
On the other hand, in many programming languages, calling a function uses stack space, so a function that is tail recursive can build up a large stack of calls to itself, which wastes memory. Since in a tail call, the containing function is about to return, its environment can actually be discarded and the recursive call can be entered without creating a new stack frame. This trick is called tail call elimination or tail call optimisation and allows tail-recursive functions to recur indefinitely.
It's been a long while since I posted this question and my opinion on the topic has changed. Here's why:
I learned Haskell, and it's a language that fixes everything bad about recursion - recursive definitions and algorithms are turned into normal looping algorithms and most of the time you don't even use recursion directly and instead use map, fold, filter, or a combination of those. And with everything bad removed, the good sides of functional programming start to shine through - everything is closer to its mathematical definition, not obscured by clunky loops and variables.
To someone else who is struggling to understand why recursion is great, go learn Haskell. It has a lot of other very interesting features like being lazy (values are evaluated only when they're requested), static (variables can never be modified), pure (functions cannot do anything other than take input and return output, so no printing to the console), strongly typed with a very expressive type system, filled with mind-blowing abstractions like Functor, Monad, State, and much more. I can almost say it's life-changing.

Nested for loops extremely slow in MATLAB (preallocated)

I am trying to learn MATLAB and one of the first problems I encountered was to guess the background from an image sequence with a static camera and moving objects. For a start I just want to do a mean or median on pixels over time, so it's just a single function I would like to apply to one of the rows of the 4 dimensional array.
I have loaded my RGB images in a 4 dimensional array with the following dimensions:
uint8 [ num_images, width, height, RGB ]
Here is the function I wrote which includes 4 nested loops. I use preallocation but still, it is extremely slow. In C++ I believe this function could run at least 10x-20x faster, and I think on CUDA it could actually run in real time. In MATLAB it takes about 20 seconds with the 4 nested loops. My stack is 100 images with 640x480x3 dimensions.
function background = calc_background(stack)
tic;
si = size(stack,1);
sy = size(stack,2);
sx = size(stack,3);
sc = size(stack,4);
background = zeros(sy,sx,sc);
A = zeros(si,1);
for x = 1:sx
for y = 1:sy
for c = 1:sc
for i = 1:si
A(i) = stack(i,y,x,c);
end
background(y,x,c) = median(A);
end
end
end
background = uint8(background);
disp(toc);
end
Could you tell me how to make this code much faster? I have tried experimenting with somehow getting the data directly from the array using only the indexes and it seems MUCH faster. It completes in 3 seconds vs. 20 seconds, so that’s a 7x performance difference, just by writing a smaller function.
function background = calc_background2(stack)
tic;
% bad code, confusing
% background = uint8(squeeze(median(stack(:, 1:size(stack,2), 1:size(stack,3), 1:3 ))));
% good code (credits: Laurent)
background=uint8((squeeze(median(stack,1)));
disp(toc);
end
So now I don't understand if MATLAB could be this fast then why is the nested loop version so slow? I am not making any dynamic resizing and MATLAB must be running the same 4 nested loops inside.
Why is this happening?
Is there any way to make nested loops run fast, like it would happen naturally in C++?
Or should I get used to the idea of programming MATLAB in this crazy one line statements way to get optimal performance?
Update
Thank you for all the great answers, now I understand a lot more. My original code with stack(:, 1:size(stack,2), 1:size(stack,3), 1:3 )) didn't make any sense, it is exactly the same as stack, I was just lucky with median's default option of using the 1st dimension for its working range.
I think it's better to ask how to write an efficient question in an other question, so I asked it here:
How to write vectorized functions in MATLAB
If I understand your question, you're asking why Matlab is faster for matrix operations than for procedural programming calls. The answer is simply that that's how it's designed. If you really want to know what makes it that way, you can read this newsletter from Matlab's website which discusses some of the underlying technology, but you probably won't get a great answer, as the software is proprietary. I also found some relevant pages by simply googling, and this old SO question
also seems to address your question.
Matlab is an interpreted language, meaning that it must evaluate each line of code of your script.
Evaluating is a lengthy process since it must parse, 'compile' and interpret each line*.
Using for loops with simple operations means that matlab takes far more time parsing/compiling than actually executing your code.
Builtin functions, on the other hand are coded in a compiled language and heavily optimized. They're very fast, hence the speed difference.
Bottom line: we're very used to procedural language and for loops, but there's almost always a nice and fast way to do the same things in a vectorized way.
* To be complete and to pay honour to whom honour is due: recent versions of Matlab actually tries to accelerate loops by analyzing repeated operations to compile chunks of repetitive operations into native executable. This is called Just In Time compilation (JIT) and was pointed out by Jonas in the following comments.
Original answer:
If I understood well (and you want the median of the first dimension) you might try:
background=uint8((squeeze(median(stack,1)));
Well, the difference between both is their method of executing code. To sketch it very roughly: in C you feed your code to a compiler which will try to optimize your code or at any rate convert it to machine code. This takes some time, but when you actually execute your program, it is in machine code already and therefore executes very fast. You compiler can take a lot of time trying to optimize the code for you, in general you don't care whether it takes 1 minute or 10 minutes to compile a distribution-ready program.
MATLAB (and other interpreted languages) don't generally work that way. When you execute your program, an interpreter will interprete each line of code and transform it into a sequence of machine code on the fly. This is a bit slower if you write for-loops as it has to interprete the code over and over again (at least in principle, there are other overheads which might matter more for the newest versions of MATLAB). Here the hurdle is the fact that everything has to be done at runtime: the interpreter can perform some optimizations, but it is not useful to perform time-consuming optimizations that might increase performance by a lot in some cases as they will cause performance to suffer in most other cases.
You might ask what you gain by using MATLAB? You gain flexibility and clear semantics. When you want to do a matrix multiplication, you just write it as such; in C this would yield a double for loop. You have to worry very little about data types, memory management, ...
Behind the scenes, MATLAB uses compiled code (Fortan/C/C++ if I'm not mistaken) to perform large operations: so a matrix multiplication is really performed by a piece of machine code which was compiled from another language. For smaller operations, this is the case as well, but you won't notice the speed of these calculations as most of your time is spent in management code (passing variables, allocating memory, ...).
To sum it all up: yes you should get used to such compact statements. If you see a line of code like Laurent's example, you immediately see that it computes a median of stack. Your code requires 11 lines of code to express the same, so when you are looking at code like yours (which might be embedded in hundreds of lines of other code), you will have a harder time understanding what is happening and pinpointing where a certain operation is performed.
To argue even further: you shouldn't program in MATLAB in the same way as you'd program in C/C++; nor should you do the other way round. Each language has its stronger and weaker points, learn to know them and use each language for what it's made for. E.g. you could write a whole compiler or webserver in MATLAB but in general that will be really slow as MATLAB was not intended to handle or concatenate strings (it can, but it might be very slow).

Is it okay to use functions to stay organized in C?

I'm a relatively new C programmer, and I've noticed that many conventions from other higher-level OOP languages don't exactly hold true on C.
Is it okay to use short functions to have your coding stay organized (even though it will likely be called only once)? An example of this would be 10-15 lines in something like void init_file(void), then calling it first in main().
I would have to say, not only is it OK, but it's generally encouraged. Just don't overly fragment the train of thought by creating myriads of tiny functions. Try to ensure that each function performs a single cohesive, well... function, with a clean interface (too many parameters can be a hint that the function is performing work which is not sufficiently separate from it's caller).
Furthermore, well-named functions can serve to replace comments that would otherwise be needed. As well as providing re-use, functions can also (or instead) provide a means to organize the code and break it down into smaller units which can be more readily understood. Using functions in this way is very much like creating packages and classes/modules, though at a more fine-grained level.
Yes. Please. Don't write long functions. Write short ones that do one thing and do it well. The fact that they may only be called once is fine. One benefit is that if you name your function well, you can avoid writing comments that will get out of sync with the code over time.
If I can take the liberty to do some quoting from Code Complete:
(These reason details have been abbreviated and in spots paraphrased, for the full explanation see the complete text.)
Valid Reasons to Create a Routine
Note the reasons overlap and are not intended to be independent of each other.
Reduce complexity - The single most important reason to create a routine is to reduce a program's complexity (hide away details so you don't need to think about them).
Introduce an intermediate, understandable abstraction - Putting a section of code int o a well-named routine is one of the best ways to document its purpose.
Avoid duplicate code - The most popular reason for creating a routine. Saves space and is easier to maintain (only have to check and/or modify one place).
Hide sequences - It's a good idea to hide the order in which events happen to be processed.
Hide pointer operations - Pointer operations tend to be hard to read and error prone. Isolating them into routines shifts focus to the intent of the operation instead of the mechanics of pointer manipulation.
Improve portability - Use routines to isolate nonportable capabilities.
Simplify complicated boolean tests - Putting complicated boolean tests into a function makes the code more readable because the details of the test are out of the way and a descriptive function name summarizes the purpose of the tests.
Improve performance - You can optimize the code in one place instead of several.
To ensure all routines are small? - No. With so many good reasons for putting code into a routine, this one is unnecessary. (This is the one thrown into the list to make sure you are paying attention!)
And one final quote from the text (Chapter 7: High-Quality Routines)
One of the strongest mental blocks to
creating effective routines is a
reluctance to create a simple routine
for a simple purpose. Constructing a
whole routine to contain two or three
lines of code might seem like
overkill, but experience shows how
helpful a good small routine can be.
If a group of statements can be thought of as a thing - then make them a function
i think it is more than OK, I would recommend it! short easy to prove correct functions with well thought out names lead to code which is more self documenting than long complex functions.
Any compiler worth using will be able to inline these calls to generate efficient code if needed.
Functions are absolutely necessary to stay organized. You need to first design the problem, and then depending on the different functionality you need to split them into functions. Some segment of code which is used multiple times, probably needs to be written in a function.
I think first thinking about what problem you have in hand, break down the components and for each component try writing a function. When writing the function see if there are some code segment doing the same thing, then break it into a sub function, or if there is a sub module then it is also a candidate for another function. But at some time this breaking job should stop, and it depends on you. Generally, do not make many too big functions and not many too small functions.
When construction the function please consider the design to have high cohesion and low coupling.
EDIT1::
you might want to also consider separate modules. For example if you need to use a stack or queue for some application. Make it separate modules whose functions could be called from other functions. This way you can save re-coding commonly used modules by programming them as a group of functions stored separately.
Yes
I follow a few guidelines:
DRY (aka DIE)
Keep Cyclomatic Complexity low
Functions should fit in a Terminal window
Each one of these principles at some point will require that a function be broken up, although I suppose #2 could imply that two functions with straight-line code should be combined. It's somewhat more common to do what is called method extraction than actually splitting a function into a top and bottom half, because the usual reason is to extract common code to be called more than once.
#1 is quite useful as a decision aid. It's the same thing as saying, as I do, "never copy code".
#2 gives you a good reason to break up a function even if there is no repeated code. If the decision logic passes a certain complexity threshold, we break it up into more functions that make fewer decisions.
It is indeed a good practice to refactor code into functions, irrespective of the language being used. Even if your code is short, it will make it more readable.
If your function is quite short, you can consider inlining it.
IBM Publib article on inlining

how do call graphs resolve function pointers?

I am implementing a call graph program for a C using perl script. I wonder how to resolve call graphs for function pointers using output of 'objdump'?
How different call graph applications resolve function pointers?
Are function pointers resolved at run time or they can be done statically?
EDIT
How do call graphs resolve cycles in static evaluation of program?
It is easy to build a call graph of A-calls-B when the call statement explicitly mentions B. It is much harder to handle indirect calls, as you've noticed.
Good static analysis tools form estimates of the contents of pointer variables by propagating pointer assignments/copies/arithmetic across program data flows (inter and intra-procedural ["global"]) using a variety of schemes, often conservative ("you get too much").
Without such an estimate, you cannot have any idea what a pointer contains and therefore simply cannot make a useful prediction (well, you can use the ultimate conservative estimate that it will go anywhere, but I think you've already rejected that solution).
Our DMS Software Reengineering Toolkit has static control/dataflow/points-to/call graph analysis that has been applied to huge systems (~~25 million lines) of C code, and produced such call graphs. The machinery to do this
is pretty complex but you can find it in advanced topics in the compiler literature. I doubt you want to implement this in Perl.
This is easier when you have source code, because you at least reliably know what is code, and what is not. You're trying to do this on object code, which means you can't even eliminate data.
Using function pointers is a way of choosing the actual function to call at runtime, so in general, it wouldn't be possible to know what would actually happen statically.
However, you could look at all functions that are possible to call and perhaps show those in some way. Often the callbacks have a unique enough signature (not always).
If you want to do better, you have to analyze the source code, to see which functions are assigned to pointers to begin with.

Have you written very long functions? If so, why?

I am writing an academic project about extremely long functions in the Linux kernel.
For that purpose, I am looking for examples for real-life functions that are extremely long (few hundreds of lines of code), that you don't consider bad programming (i.e., they won't benefit from decomposition or usage of a dispatch table).
Have you ever written or seen such a code? Can you post or link to it, and give explanation of why is it so long?
I have been getting amazing help from the community here - any idea that will be taken into the project will be properly credited.
Thanks,
Udi
The longest functions that I have ever written all have one thing in common, a very large switch statement. There are times, when you have to switch on a long list of items and it would only make things harder to understand if you tried to refactor some of the options into a separate function. Having large switch statements makes the Cyclomatic complexity go through the roof, but it is often better than the alternative implementations.
It was the last one before I got fired.
A previous job: An extremely long case statement, IIRC 1000+ lines. This was long before objects. Each option was only a few lines long. Breaking it up would have made it less clear. There were actually a pair of such routines doing different things to the same underlying set of data types.
Sorry, I don't have the code anymore and it isn't mine to post, anyway.
The longest function that I didn't see as being horrible would be the key method of a custom CPU VM. As with #epotter, this involved a big switch statement. In fact I'd say a lot of method that I find resist being cleanly broken down or improved in readability involve switch statements.
Unfortunately, you won't often find this type of subroutine checked in or posted somewhere if it's auto-generated during a build step using some sort of code generator.
So look for projects that have C generated from another language.
Beside the performance, I think the size of the call stack in Kernel space is 8K (please verify the size). Also, as far as I know, code in kernel is fairly specific. If some code is unlikely to be re-used in the future why bother make it a function considering function call overhead.
I could imagine that when speed is important (such as when holding some sort of lock in the kernel) you would not want to break up a function because of the overhead due to making a functional call. When compiled, parameters have to be pushed onto the stack and data has to be popped off before returning. Therefor you may have a large function for efficiency reasons.

Resources