Search Results

Search found 3512 results on 141 pages for 'premature optimization'.

Page 5/141 | < Previous Page | 1 2 3 4 5 6 7 8 9 10 11 12  | Next Page >

• Objective-C optimization

  - by SpecialK

  Are there standard optimization tricks for Objective-C to make for faster execution along the lines of "inlining" frequent methods as in C++ or the "g++ -fast" tag?

  Read the article

• Performance Optimization for Matrix Rotation

  - by Summer_More_More_Tea

  Hello everyone: I'm now trapped by a performance optimization lab in the book "Computer System from a Programmer's Perspective" described as following: In a N*N matrix M, where N is multiple of 32, the rotate operation can be represented as: Transpose: interchange elements M(i,j) and M(j,i) Exchange rows: Row i is exchanged with row N-1-i A example for matrix rotation(N is 3 instead of 32 for simplicity): ------- ------- |1|2|3| |3|6|9| ------- ------- |4|5|6| after rotate is |2|5|8| ------- ------- |7|8|9| |1|4|7| ------- ------- A naive implementation is: #define RIDX(i,j,n) ((i)*(n)+(j)) void naive_rotate(int dim, pixel *src, pixel *dst) { int i, j; for (i = 0; i < dim; i++) for (j = 0; j < dim; j++) dst[RIDX(dim-1-j, i, dim)] = src[RIDX(i, j, dim)]; } I come up with an idea by inner-loop-unroll. The result is: Code Version Speed Up original 1x unrolled by 2 1.33x unrolled by 4 1.33x unrolled by 8 1.55x unrolled by 16 1.67x unrolled by 32 1.61x I also get a code snippet from pastebin.com that seems can solve this problem: void rotate(int dim, pixel *src, pixel *dst) { int stride = 32; int count = dim >> 5; src += dim - 1; int a1 = count; do { int a2 = dim; do { int a3 = stride; do { *dst++ = *src; src += dim; } while(--a3); src -= dim * stride + 1; dst += dim - stride; } while(--a2); src += dim * (stride + 1); dst -= dim * dim - stride; } while(--a1); } After carefully read the code, I think main idea of this solution is treat 32 rows as a data zone, and perform the rotating operation respectively. Speed up of this version is 1.85x, overwhelming all the loop-unroll version. Here are the questions: In the inner-loop-unroll version, why does increment slow down if the unrolling factor increase, especially change the unrolling factor from 8 to 16, which does not effect the same when switch from 4 to 8? Does the result have some relationship with depth of the CPU pipeline? If the answer is yes, could the degrade of increment reflect pipeline length? What is the probable reason for the optimization of data-zone version? It seems that there is no too much essential difference from the original naive version. EDIT: My test environment is Intel Centrino Duo processor and the verion of gcc is 4.4 Any advice will be highly appreciated! Kind regards!

  Read the article

• Need help with basic optimization problem

  - by ??iu

  I know little of optimization problems, so hopefully this will be didactic for me: rotors = [1, 2, 3, 4...] widgets = ['a', 'b', 'c', 'd' ...] assert len(rotors) == len(widgets) part_values = [ (1, 'a', 34), (1, 'b', 26), (1, 'c', 11), (1, 'd', 8), (2, 'a', 5), (2, 'b', 17), .... ] Given a fixed number of widgets and a fixed number of rotors, how can you get a series of widget-rotor pairs that maximizes the total value where each widget and rotor can only be used once?

  Read the article

• C++ Performance/memory optimization guidelines

  - by ML

  Hi All, Does anyone have a resource for C++ memory optimization guidelines? Best practices, tuning, etc? As an example: Class xxx { public: xxx(); virtual ~xxx(); protected: private: }; Would there be ANY benefit on the compiler or memory allocation to get rid of protected and private since there there are no items that are protected and private in this class?

  Read the article

• Any Javascript optimization benchmarks?

  - by int3

  I watched Nicholas Zakas' talk, Speed up your Javascript, with some interest. I liked how he benchmarked the various performance improvements created by various optimization techniques, e.g. reducing calls to deeply nested objects, changing loops to count down instead of up, etc. I would like to run these benchmarks myself though, to see exactly how our current browsers are faring. I guess it wouldn't be too difficult to cook up some timed loops, but I'd like to know if there are any existing implementations out there.

  Read the article

• Java code optimization leads to numerical inaccuracies and errors

  - by rano

  I'm trying to implement a version of the Fuzzy C-Means algorithm in Java and I'm trying to do some optimization by computing just once everything that can be computed just once. This is an iterative algorithm and regarding the updating of a matrix, the clusters x pixels membership matrix U, this is the update rule I want to optimize: where the x are the element of a matrix X (pixels x features) and v belongs to the matrix V (clusters x features). And m is a parameter that ranges from 1.1 to infinity. The distance used is the euclidean norm. If I had to implement this formula in a banal way I'd do: for(int i = 0; i < X.length; i++) { int count = 0; for(int j = 0; j < V.length; j++) { double num = D[i][j]; double sumTerms = 0; for(int k = 0; k < V.length; k++) { double thisDistance = D[i][k]; sumTerms += Math.pow(num / thisDistance, (1.0 / (m - 1.0))); } U[i][j] = (float) (1f / sumTerms); } } In this way some optimization is already done, I precomputed all the possible squared distances between X and V and stored them in a matrix D but that is not enough, since I'm cycling througn the elements of V two times resulting in two nested loops. Looking at the formula the numerator of the fraction is independent of the sum so I can compute numerator and denominator independently and the denominator can be computed just once for each pixel. So I came to a solution like this: int nClusters = V.length; double exp = (1.0 / (m - 1.0)); for(int i = 0; i < X.length; i++) { int count = 0; for(int j = 0; j < nClusters; j++) { double distance = D[i][j]; double denominator = D[i][nClusters]; double numerator = Math.pow(distance, exp); U[i][j] = (float) (1f / (numerator * denominator)); } } Where I precomputed the denominator into an additional column of the matrix D while I was computing the distances: for (int i = 0; i < X.length; i++) { for (int j = 0; j < V.length; j++) { double sum = 0; for (int k = 0; k < nDims; k++) { final double d = X[i][k] - V[j][k]; sum += d * d; } D[i][j] = sum; D[i][B.length] += Math.pow(1 / D[i][j], exp); } } By doing so I encounter numerical differences between the 'banal' computation and the second one that leads to different numerical value in U (not in the first iterates but soon enough). I guess that the problem is that exponentiate very small numbers to high values (the elements of U can range from 0.0 to 1.0 and exp , for m = 1.1, is 10) leads to ver y small values, whereas by dividing the numerator and the denominator and THEN exponentiating the result seems to be better numerically. The problem is it involves much more operations. Am I doing something wrong? Is there a possible solution to get both the code optimized and numerically stable? Any suggestion or criticism will be appreciated.

  Read the article

• PHP website Optimization

  - by ana

  I have a high traffic website and I need make sure my site is fast enough to display my pages to everyone rapidly. I searched on Google many articles about speed and optimization and here's what I found: Cache the page Save it to the disk Caching the page in memory: This is very fast but if I need to change the content of my page I have to remove it from cache and then re-save the file on the disk. Save it to disk This is very easy to maintain but every time the page is accessed I have to read on the disk. Which method should I go with? Thanks

  Read the article

• Does MATLAB perform tail call optimization?

  - by Shea Levy

  I've recently learned Haskell, and am trying to carry the pure functional style over to my other code when possible. An important aspect of this is treating all variables as immutable, i.e. constants. In order to do so, many computations that would be implemented using loops in an imperative style have to be performed using recursion, which typically incurs a memory penalty due to the allocation a new stack frame for each function call. In the special case of a tail call (where the return value of a called function is immediately returned to the callee's caller), however, this penalty can be bypassed by a process called tail call optimization (in one method, this can be done by essentially replacing a call with a jmp after setting up the stack properly). Does MATLAB perform TCO by default, or is there a way to tell it to?

  Read the article

• Including associations optimization in Rails

  - by Vitaly

  Hey, I'm looking for help with Ruby optimization regarding loading of associations on demand. This is simplified example. I have 3 models: Post, Comment, User. References are: Post has many comments and Comment has reference to User (:author). Now when I go to the post page, I expect to see post body + all comments (and their respective authors names). This requires following 2 queries: select * from Post -- to get post data (1 row) select * from Comment inner join User -- to get comment + usernames (N rows) In the code I have: Post.find(params[:id], :include => { :comments => [:author] } But it doesn't work as expected: as I see in the back end, there're still N+1 hits (some of them are cached though). How can I optimize that?

  Read the article

• why optimization does not happen?

  - by aaa

  hi. I have C/C++ code, that looks like this: static int function(double *I) { int n = 0; // more instructions, loops, for (int i; ...; ++i) n += fabs(I[i] > tolerance); return n; } function(I); // return value is not used. compiler inlines function, however it does not optimize out n manipulations. I would expect compiler is able to recognize that value is never used as rhs only. Is there some side effect, which prevents optimization? Thanks

  Read the article

• Static variable for optimization

  - by keithjgrant

  I'm wondering if I can use a static variable for optimization: public function Bar() { static $i = moderatelyExpensiveFunctionCall(); if ($i) { return something(); } else { return somethingElse(); } } I know that once $i is initialized, it won't be changed by by that line of code on successive calls to Bar(). I assume this means that moderatelyExpensiveFunctionCall() won't be evaluated every time I call, but I'd like to know for certain. Once PHP sees a static variable that has been initialized, does it skip over that line of code? In other words, is this going to optimize my execution time if I make a lot of calls to Bar(), or am I wasting my time?

  Read the article

• Copy method optimization in compilers

  - by Dženan

  Hi All! I have the following code: void Stack::operator =(Stack &rhs) { //do the actual copying } Stack::Stack(Stack &rhs) //copy-constructor { top=NULL; //initialize this as an empty stack (which it is) *this=rhs; //invoke assignment operator } Stack& Stack::CopyStack() { return *this; //this statement will invoke copy contructor } It is being used like this: unsigned Stack::count() { unsigned c=0; Stack copy=CopyStack(); while (!copy.empty()) { copy.pop(); c++; } return c; } Removing reference symbol from declaration of CopyStack (returning a copy instead of reference) makes no difference in visual studio 2008 (with respect to number of times copying is invoked). I guess it gets optimized away - normally it should first make a copy for the return value, then call assignment operator once more to assign it to variable sc. What is your experience with this sort of optimization in different compilers? Regards, Dženan

  Read the article

• Which isometric angles can be mirrored (and otherwise transformed) for optimization?

  - by Tom

  I am working on a basic isometric game, and am struggling to find the correct mirrors. Mirror can be any form of transform. I have managed to get SE out of SW, by scaling the sprite on X axis by -1. Same applies for NE angle. Something is bugging me, that I should be able to also mirror N to S, but I cannot manage to pull this one off. Am I just too sleepy and trying to do the impossible, or a basic -1 scale on Y axis is not enough? What are the common used mirror table for optimizing 8 angle (N, NE, E, SE, S, SW, W, NW) isometric sprites?

  Read the article

• When should an API favour optimization over readability and ease-of-use?

  - by jmlane

  I am in the process of designing a small library, where one of my design goals is to use as much of the native domain language as possible in the API. While doing so, I've noticed that there are some cases in the API outline where a more intuitive, readable attribute/method call requires some functionally unnecessary encapsulation. Since the final product will not necessarily require high performance, I am unconcerned about making the decision to favour ease-of-use in my current project over the most efficient implementation of the code in question. I know not to assume readability and ease-of-use are paramount in all expected use-cases, such as when performance is required. I would like to know if there are more general reasons that argue for an API design preferring (marginally) more efficient implementations?

  Read the article

• When should code favour optimization over readability and ease-of-use?

  - by jmlane

  I am in the process of designing a small library, where one of my design goals is that the API should be as close to the domain language as possible. While working on the design, I've noticed that there are some cases in the code where a more intuitive, readable attribute/method call requires some functionally unnecessary encapsulation. Since the final product will not necessarily require high performance, I am unconcerned about making the decision to favour ease-of-use in my current project over the most efficient implementation of the code in question. I know not to assume readability and ease-of-use are paramount in all expected use-cases, such as when performance is required. I would like to know if there are more general reasons that argue for a design preferring more efficient implementations—even if only marginally so?

  Read the article

• Finding perfect numbers in C# (optimization)

  - by paradox

  I coded up a program in C# to find perfect numbers within a certain range as part of a programming challenge . However, I realized it is very slow when calculating perfect numbers upwards of 10000. Are there any methods of optimization that exist for finding perfect numbers? My code is as follows: using System; using System.Collections.Generic; using System.Linq; namespace ConsoleTest { class Program { public static List<int> FindDivisors(int inputNo) { List<int> Divisors = new List<int>(); for (int i = 1; i<inputNo; i++) { if (inputNo%i==0) Divisors.Add(i); } return Divisors; } public static void Main(string[] args) { const int limit = 100000; List<int> PerfectNumbers = new List<int>(); List<int> Divisors=new List<int>(); for (int i=1; i<limit; i++) { Divisors = FindDivisors(i); if (i==Divisors.Sum()) PerfectNumbers.Add(i); } Console.Write("Output ="); for (int i=0; i<PerfectNumbers.Count; i++) { Console.Write(" {0} ",PerfectNumbers[i]); } Console.Write("\n\n\nPress any key to continue . . . "); Console.ReadKey(true); } } }

  Read the article

• sql-server performance optimization by removing print statements

  - by AG

  We're going through a round of sql-server stored procedure optimizations. The one recommendation we've found that clearly applies for us is 'SET NOCOUNT ON' at the top of each procedure. (Yes, I've seen the posts that point out issues with this depending on what client objects you run the stored procedures from but these are not issues for us.) So now I'm just trying to add in a bit of common sense. If the benefit of SET NOCOUNT ON is simply to reduce network traffic by some small amount every time, wouldn't it also make sense to turn off all the PRINT statements we have in the stored procedures that we only use for debugging? I can't see how it can hurt performance. OTOH, it's a bit of a hassle to implement due to the fact that some of the print statements are the only thing within else clauses, so you can't just always comment out the one line and be done. The change carries some amount of risk so I don't want to do it if it isn't going to actually help. But I don't see eliminating print statements mentioned anywhere in articles on optimization. Is that because it is so obvious no one bothers to mention it?

  Read the article

• Simplification / optimization of GPS track

  - by GreyCat

  I've got a GPS track, produces by gpxlogger(1) (supplied as a client for gpsd). GPS receiver updates its coordinates every 1 second, gpxlogger's logic is very simple, it writes down location (lat, lon, ele) and a timestamp (time) received from GPS every n seconds (n = 3 in my case). After writing down a several hours worth of track, gpxlogger saves several megabyte long GPX file that includes several thousands of points. Afterwards, I try to plot this track on a map and use it with OpenLayers. It works, but several thousands of points make using the map a sloppy and slow experience. I understand that having several thousands of points of suboptimal. There are myriads of points that can be deleted without losing almost anything: when there are several points making up roughly the straight line and we're moving with the same constant speed between them, we can just leave the first and the last point and throw anything else. I thought of using gpsbabel for such track simplification / optimization job, but, alas, it's simplification filter works only with routes, i.e. analyzing only geometrical shape of path, without timestamps (i.e. not checking that the speed was roughly constant). Is there some ready-made utility / library / algorithm available to optimize tracks? Or may be I'm missing some clever option with gpsbabel?

  Read the article

• Optimization of a c++ matrix/bitmap class

  - by Andrew

  I am searching a 2D matrix (or bitmap) class which is flexible but also fast element access. The contents A flexible class should allow you to choose dimensions during runtime, and would look something like this (simplified): class Matrix { public: Matrix(int w, int h) : data(new int[x*y]), width(w) {} void SetElement(int x, int y, int val) { data[x+y*width] = val; } // ... private: // symbols int width; int* data; }; A faster often proposed solution using templates is (simplified): template <int W, int H> class TMatrix { TMatrix() data(new int[W*H]) {} void SetElement(int x, int y, int val) { data[x+y*W] = val; } private: int* data; }; This is faster as the width can be "inlined" in the code. The first solution does not do this. However this is not very flexible anymore, as you can't change the size anymore at runtime. So my question is: Is there a possibility to tell the compiler to generate faster code (like when using the template solution), when the size in the code is fixed and generate flexible code when its runtime dependend? I tried to achieve this by writing "const" where ever possible. I tried it with gcc and VS2005, but no success. This kind of optimization would be useful for many other similar cases.

  Read the article

• Is there a way that I can hard code a const XmlNameTable to be reused by all of my XmlTextReader(s)?

  - by highone

  Before I continue I would just like to say I know that "Premature optimization is the root of all evil." However this program is only a hobby project and I enjoy trying to find ways to optimize it. That being said, I was reading an article on improving xml performance and it recommended sharing "the XmlNameTable class that is used to store element and attribute names across multiple XML documents of the same type to improve performance." I wasn't able to find any information about doing this in my googling, so it is likely that this is either not possible, a no-no, or a stupid question, but what's the harm in asking?

  Read the article

• MySQL table organization and optimization (Rails)

  - by aguynamedloren

  I've been learning Ruby on Rails over the past few months with no prior programming experience. Lately, I've been thinking about database optimization and table organization. I know there are great books on the subject, but I typically learn by example / as I go. Here's a hypothetical situation: Let's say I am building a social network for a niche community with 250,000 members (users). The users have the ability to attend events. Let's say there are 50,000 past/present/future events. Much like Facebook events, a user can attend any number of events and an event can have any number of attendees. In the database, there would be a table for users and a table for events. Somehow I would have to create an association between the users and events. I could create an "events" column in the users table such that each user row would contain a hash of event IDs, or I could create an "attendees" column in the events table such that each event row would contain a hash of user IDs. Neither of these solutions seem ideal, however. On a users profile page, I want to display the list of events they are associated with, which would require scanning the 50,000 event rows for the user ID of said user if I include an "attendees" column in the events table. Likewise, on an event page, I want to display a list of attendees for the event, which would require scanning the 250,000 user rows for the event ID of said event if I include an "events" column in the users table. Option 3 would be to create a third table that contains the attendee information for each and every event - but I don't see how this would solve any problems. Are these non-issues? Rails makes accessing all of this information easy, but I guess I'm worried about scale. It is entirely possible that I am under-estimating the speed and processing power of modern databases / servers / etc. How long would it take to scan 250,000 user rows for specific event IDs - 10ms? 100ms? 1,000ms? I guess that's not that bad. Am I just over-thinking this?

  Read the article

• SQL query performance optimization (TimesTen)

  - by Sergey Mikhanov

  Hi community, I need some help with TimesTen DB query optimization. I made some measures with Java profiler and found the code section that takes most of the time (this code section executes the SQL query). What is strange that this query becomes expensive only for some specific input data. Here’s the example. We have two tables that we are querying, one represents the objects we want to fetch (T_PROFILEGROUP), another represents the many-to-many link from some other table (T_PROFILECONTEXT_PROFILEGROUPS). We are not querying linked table. These are the queries that I executed with DB profiler running (they are the same except for the ID): Command> select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; < 1169655247309537280 > < 1169655249792565248 > < 1464837997699399681 > 3 rows found. Command> select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928; < 1169655247309537280 > 1 row found. This is what I have in the profiler: 12:14:31.147 1 SQL 2L 6C 10825P Preparing: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272 12:14:31.147 2 SQL 4L 6C 10825P sbSqlCmdCompile ()(E): (Found already compiled version: refCount:01, bucket:47) cmdType:100, cmdNum:1146695. 12:14:31.147 3 SQL 4L 6C 10825P Opening: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:31.147 4 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:31.148 5 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:31.148 6 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:31.228 7 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:31.228 8 SQL 4L 6C 10825P Closing: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1464837998949302272; 12:14:35.243 9 SQL 2L 6C 10825P Preparing: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928 12:14:35.243 10 SQL 4L 6C 10825P sbSqlCmdCompile ()(E): (Found already compiled version: refCount:01, bucket:44) cmdType:100, cmdNum:1146697. 12:14:35.243 11 SQL 4L 6C 10825P Opening: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928; 12:14:35.243 12 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928; 12:14:35.243 13 SQL 4L 6C 10825P Fetching: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928; 12:14:35.243 14 SQL 4L 6C 10825P Closing: select G.M_ID from T_PROFILECONTEXT_PROFILEGROUPS CG, T_PROFILEGROUP G where CG.M_ID_EID = G.M_ID and CG.M_ID_OID = 1466585677823868928; It’s clear that the first query took almost 100ms, while the second was executed instantly. It’s not about queries precompilation (the first one is precompiled too, as same queries happened earlier). We have DB indices for all columns used here: T_PROFILEGROUP.M_ID, T_PROFILECONTEXT_PROFILEGROUPS.M_ID_OID and T_PROFILECONTEXT_PROFILEGROUPS.M_ID_EID. My questions are: Why querying the same set of tables yields such a different performance for different parameters? Which indices are involved here? Is there any way to improve this simple query and/or the DB to make it faster? UPDATE: to give the feeling of size: Command> select count(*) from T_PROFILEGROUP; < 183840 > 1 row found. Command> select count(*) from T_PROFILECONTEXT_PROFILEGROUPS; < 2279104 > 1 row found.

  Read the article

• Which libraries use the "We Know Where You Live" optimization for std::make_shared?

  - by KnowItAllWannabe

  Over two years ago, Stephan T. Lavavej described a space-saving optimization he implemented in Microsoft's implementation of std::make_shared, and I know from speaking with him that Microsoft has nothing against other library implementations adopting this optimization. If you know for sure whether other libraries (e.g., for Gnu C++, Clang, Intel C++, plus Boost (for boost::make_shared)) have adopted this implementation, please contribute an answer. I don't have ready access to that many make_shared implementations, nor am I wild about digging into the bowels of the ones I have to see if they've implemented the WKWYL optimization, but I'm hoping that SO readers know the answers for some libraries off-hand. I know from looking at the code that as of Boost 1.52, the WKWYL optimization had not been implemented, but Boost is now up to version 1.55. Note that this optimization is different from std::make_shared's ability to avoid a dedicated heap allocation for the reference count used by std::shared_ptr. For a discussion of the difference between WKWYL and that optimication, consult this question.

  Read the article

• Overhead of calling tiny functions from a tight inner loop? [C++]

  - by John

  Say you see a loop like this one: for(int i=0; i<thing.getParent().getObjectModel().getElements(SOME_TYPE).count(); ++i) { thing.getData().insert( thing.GetData().Count(), thing.getParent().getObjectModel().getElements(SOME_TYPE)[i].getName() ); } if this was Java I'd probably not think twice. But in performance-critical sections of C++, it makes me want to tinker with it... however I don't know if the compiler is smart enough to make it futile. This is a made up example but all it's doing is inserting strings into a container. Please don't assume any of these are STL types, think in general terms about the following: Is having a messy condition in the for loop going to get evaluated each time, or only once? If those get methods are simply returning references to member variables on the objects, will they be inlined away? Would you expect custom [] operators to get optimized at all? In other words is it worth the time (in performance only, not readability) to convert it to something like: ElementContainer &source = thing.getParent().getObjectModel().getElements(SOME_TYPE); int num = source.count(); Store &destination = thing.getData(); for(int i=0;i<num;++i) { destination.insert(thing.GetData().Count(), source[i].getName(); } Remember, this is a tight loop, called millions of times a second. What I wonder is if all this will shave a couple of cycles per loop or something more substantial? Yes I know the quote about "premature optimisation". And I know that profiling is important. But this is a more general question about modern compilers, Visual Studio in particular.

  Read the article

• How John Got 15x Improvement Without Really Trying

  - by rchrd

  The following article was published on a Sun Microsystems website a number of years ago by John Feo. It is still useful and worth preserving. So I'm republishing it here.  How I Got 15x Improvement Without Really Trying John Feo, Sun Microsystems Taking ten "personal" program codes used in scientific and engineering research, the author was able to get from 2 to 15 times performance improvement easily by applying some simple general optimization techniques. Introduction Scientific research based on computer simulation depends on the simulation for advancement. The research can advance only as fast as the computational codes can execute. The codes' efficiency determines both the rate and quality of results. In the same amount of time, a faster program can generate more results and can carry out a more detailed simulation of physical phenomena than a slower program. Highly optimized programs help science advance quickly and insure that monies supporting scientific research are used as effectively as possible. Scientific computer codes divide into three broad categories: ISV, community, and personal. ISV codes are large, mature production codes developed and sold commercially. The codes improve slowly over time both in methods and capabilities, and they are well tuned for most vendor platforms. Since the codes are mature and complex, there are few opportunities to improve their performance solely through code optimization. Improvements of 10% to 15% are typical. Examples of ISV codes are DYNA3D, Gaussian, and Nastran. Community codes are non-commercial production codes used by a particular research field. Generally, they are developed and distributed by a single academic or research institution with assistance from the community. Most users just run the codes, but some develop new methods and extensions that feed back into the general release. The codes are available on most vendor platforms. Since these codes are younger than ISV codes, there are more opportunities to optimize the source code. Improvements of 50% are not unusual. Examples of community codes are AMBER, CHARM, BLAST, and FASTA. Personal codes are those written by single users or small research groups for their own use. These codes are not distributed, but may be passed from professor-to-student or student-to-student over several years. They form the primordial ocean of applications from which community and ISV codes emerge. Government research grants pay for the development of most personal codes. This paper reports on the nature and performance of this class of codes. Over the last year, I have looked at over two dozen personal codes from more than a dozen research institutions. The codes cover a variety of scientific fields, including astronomy, atmospheric sciences, bioinformatics, biology, chemistry, geology, and physics. The sources range from a few hundred lines to more than ten thousand lines, and are written in Fortran, Fortran 90, C, and C++. For the most part, the codes are modular, documented, and written in a clear, straightforward manner. They do not use complex language features, advanced data structures, programming tricks, or libraries. I had little trouble understanding what the codes did or how data structures were used. Most came with a makefile. Surprisingly, only one of the applications is parallel. All developers have access to parallel machines, so availability is not an issue. Several tried to parallelize their applications, but stopped after encountering difficulties. Lack of education and a perception that parallelism is difficult prevented most from trying. I parallelized several of the codes using OpenMP, and did not judge any of the codes as difficult to parallelize. Even more surprising than the lack of parallelism is the inefficiency of the codes. I was able to get large improvements in performance in a matter of a few days applying simple optimization techniques. Table 1 lists ten representative codes [names and affiliation are omitted to preserve anonymity]. Improvements on one processor range from 2x to 15.5x with a simple average of 4.75x. I did not use sophisticated performance tools or drill deep into the program's execution character as one would do when tuning ISV or community codes. Using only a profiler and source line timers, I identified inefficient sections of code and improved their performance by inspection. The changes were at a high level. I am sure there is another factor of 2 or 3 in each code, and more if the codes are parallelized. The study’s results show that personal scientific codes are running many times slower than they should and that the problem is pervasive. Computational scientists are not sloppy programmers; however, few are trained in the art of computer programming or code optimization. I found that most have a working knowledge of some programming language and standard software engineering practices; but they do not know, or think about, how to make their programs run faster. They simply do not know the standard techniques used to make codes run faster. In fact, they do not even perceive that such techniques exist. The case studies described in this paper show that applying simple, well known techniques can significantly increase the performance of personal codes. It is important that the scientific community and the Government agencies that support scientific research find ways to better educate academic scientific programmers. The inefficiency of their codes is so bad that it is retarding both the quality and progress of scientific research. # cacheperformance redundantoperations loopstructures performanceimprovement 1 x x 15.5 2 x 2.8 3 x x 2.5 4 x 2.1 5 x x 2.0 6 x 5.0 7 x 5.8 8 x 6.3 9 2.2 10 x x 3.3 Table 1 — Area of improvement and performance gains of 10 codes The remainder of the paper is organized as follows: sections 2, 3, and 4 discuss the three most common sources of inefficiencies in the codes studied. These are cache performance, redundant operations, and loop structures. Each section includes several examples. The last section summaries the work and suggests a possible solution to the issues raised. Optimizing cache performance Commodity microprocessor systems use caches to increase memory bandwidth and reduce memory latencies. Typical latencies from processor to L1, L2, local, and remote memory are 3, 10, 50, and 200 cycles, respectively. Moreover, bandwidth falls off dramatically as memory distances increase. Programs that do not use cache effectively run many times slower than programs that do. When optimizing for cache, the biggest performance gains are achieved by accessing data in cache order and reusing data to amortize the overhead of cache misses. Secondary considerations are prefetching, associativity, and replacement; however, the understanding and analysis required to optimize for the latter are probably beyond the capabilities of the non-expert. Much can be gained simply by accessing data in the correct order and maximizing data reuse. 6 out of the 10 codes studied here benefited from such high level optimizations. Array Accesses The most important cache optimization is the most basic: accessing Fortran array elements in column order and C array elements in row order. Four of the ten codes—1, 2, 4, and 10—got it wrong. Compilers will restructure nested loops to optimize cache performance, but may not do so if the loop structure is too complex, or the loop body includes conditionals, complex addressing, or function calls. In code 1, the compiler failed to invert a key loop because of complex addressing do I = 0, 1010, delta_x IM = I - delta_x IP = I + delta_x do J = 5, 995, delta_x JM = J - delta_x JP = J + delta_x T1 = CA1(IP, J) + CA1(I, JP) T2 = CA1(IM, J) + CA1(I, JM) S1 = T1 + T2 - 4 * CA1(I, J) CA(I, J) = CA1(I, J) + D * S1 end do end do In code 2, the culprit is conditionals do I = 1, N do J = 1, N If (IFLAG(I,J) .EQ. 0) then T1 = Value(I, J-1) T2 = Value(I-1, J) T3 = Value(I, J) T4 = Value(I+1, J) T5 = Value(I, J+1) Value(I,J) = 0.25 * (T1 + T2 + T5 + T4) Delta = ABS(T3 - Value(I,J)) If (Delta .GT. MaxDelta) MaxDelta = Delta endif enddo enddo I fixed both programs by inverting the loops by hand. Code 10 has three-dimensional arrays and triply nested loops. The structure of the most computationally intensive loops is too complex to invert automatically or by hand. The only practical solution is to transpose the arrays so that the dimension accessed by the innermost loop is in cache order. The arrays can be transposed at construction or prior to entering a computationally intensive section of code. The former requires all array references to be modified, while the latter is cost effective only if the cost of the transpose is amortized over many accesses. I used the second approach to optimize code 10. Code 5 has four-dimensional arrays and loops are nested four deep. For all of the reasons cited above the compiler is not able to restructure three key loops. Assume C arrays and let the four dimensions of the arrays be i, j, k, and l. In the original code, the index structure of the three loops is L1: for i L2: for i L3: for i for l for l for j for k for j for k for j for k for l So only L3 accesses array elements in cache order. L1 is a very complex loop—much too complex to invert. I brought the loop into cache alignment by transposing the second and fourth dimensions of the arrays. Since the code uses a macro to compute all array indexes, I effected the transpose at construction and changed the macro appropriately. The dimensions of the new arrays are now: i, l, k, and j. L3 is a simple loop and easily inverted. L2 has a loop-carried scalar dependence in k. By promoting the scalar name that carries the dependence to an array, I was able to invert the third and fourth subloops aligning the loop with cache. Code 5 is by far the most difficult of the four codes to optimize for array accesses; but the knowledge required to fix the problems is no more than that required for the other codes. I would judge this code at the limits of, but not beyond, the capabilities of appropriately trained computational scientists. Array Strides When a cache miss occurs, a line (64 bytes) rather than just one word is loaded into the cache. If data is accessed stride 1, than the cost of the miss is amortized over 8 words. Any stride other than one reduces the cost savings. Two of the ten codes studied suffered from non-unit strides. The codes represent two important classes of "strided" codes. Code 1 employs a multi-grid algorithm to reduce time to convergence. The grids are every tenth, fifth, second, and unit element. Since time to convergence is inversely proportional to the distance between elements, coarse grids converge quickly providing good starting values for finer grids. The better starting values further reduce the time to convergence. The downside is that grids of every nth element, n > 1, introduce non-unit strides into the computation. In the original code, much of the savings of the multi-grid algorithm were lost due to this problem. I eliminated the problem by compressing (copying) coarse grids into continuous memory, and rewriting the computation as a function of the compressed grid. On convergence, I copied the final values of the compressed grid back to the original grid. The savings gained from unit stride access of the compressed grid more than paid for the cost of copying. Using compressed grids, the loop from code 1 included in the previous section becomes do j = 1, GZ do i = 1, GZ T1 = CA(i+0, j-1) + CA(i-1, j+0) T4 = CA1(i+1, j+0) + CA1(i+0, j+1) S1 = T1 + T4 - 4 * CA1(i+0, j+0) CA(i+0, j+0) = CA1(i+0, j+0) + DD * S1 enddo enddo where CA and CA1 are compressed arrays of size GZ. Code 7 traverses a list of objects selecting objects for later processing. The labels of the selected objects are stored in an array. The selection step has unit stride, but the processing steps have irregular stride. A fix is to save the parameters of the selected objects in temporary arrays as they are selected, and pass the temporary arrays to the processing functions. The fix is practical if the same parameters are used in selection as in processing, or if processing comprises a series of distinct steps which use overlapping subsets of the parameters. Both conditions are true for code 7, so I achieved significant improvement by copying parameters to temporary arrays during selection. Data reuse In the previous sections, we optimized for spatial locality. It is also important to optimize for temporal locality. Once read, a datum should be used as much as possible before it is forced from cache. Loop fusion and loop unrolling are two techniques that increase temporal locality. Unfortunately, both techniques increase register pressure—as loop bodies become larger, the number of registers required to hold temporary values grows. Once register spilling occurs, any gains evaporate quickly. For multiprocessors with small register sets or small caches, the sweet spot can be very small. In the ten codes presented here, I found no opportunities for loop fusion and only two opportunities for loop unrolling (codes 1 and 3). In code 1, unrolling the outer and inner loop one iteration increases the number of result values computed by the loop body from 1 to 4, do J = 1, GZ-2, 2 do I = 1, GZ-2, 2 T1 = CA1(i+0, j-1) + CA1(i-1, j+0) T2 = CA1(i+1, j-1) + CA1(i+0, j+0) T3 = CA1(i+0, j+0) + CA1(i-1, j+1) T4 = CA1(i+1, j+0) + CA1(i+0, j+1) T5 = CA1(i+2, j+0) + CA1(i+1, j+1) T6 = CA1(i+1, j+1) + CA1(i+0, j+2) T7 = CA1(i+2, j+1) + CA1(i+1, j+2) S1 = T1 + T4 - 4 * CA1(i+0, j+0) S2 = T2 + T5 - 4 * CA1(i+1, j+0) S3 = T3 + T6 - 4 * CA1(i+0, j+1) S4 = T4 + T7 - 4 * CA1(i+1, j+1) CA(i+0, j+0) = CA1(i+0, j+0) + DD * S1 CA(i+1, j+0) = CA1(i+1, j+0) + DD * S2 CA(i+0, j+1) = CA1(i+0, j+1) + DD * S3 CA(i+1, j+1) = CA1(i+1, j+1) + DD * S4 enddo enddo The loop body executes 12 reads, whereas as the rolled loop shown in the previous section executes 20 reads to compute the same four values. In code 3, two loops are unrolled 8 times and one loop is unrolled 4 times. Here is the before for (k = 0; k < NK[u]; k++) { sum = 0.0; for (y = 0; y < NY; y++) { sum += W[y][u][k] * delta[y]; } backprop[i++]=sum; } and after code for (k = 0; k < KK - 8; k+=8) { sum0 = 0.0; sum1 = 0.0; sum2 = 0.0; sum3 = 0.0; sum4 = 0.0; sum5 = 0.0; sum6 = 0.0; sum7 = 0.0; for (y = 0; y < NY; y++) { sum0 += W[y][0][k+0] * delta[y]; sum1 += W[y][0][k+1] * delta[y]; sum2 += W[y][0][k+2] * delta[y]; sum3 += W[y][0][k+3] * delta[y]; sum4 += W[y][0][k+4] * delta[y]; sum5 += W[y][0][k+5] * delta[y]; sum6 += W[y][0][k+6] * delta[y]; sum7 += W[y][0][k+7] * delta[y]; } backprop[k+0] = sum0; backprop[k+1] = sum1; backprop[k+2] = sum2; backprop[k+3] = sum3; backprop[k+4] = sum4; backprop[k+5] = sum5; backprop[k+6] = sum6; backprop[k+7] = sum7; } for one of the loops unrolled 8 times. Optimizing for temporal locality is the most difficult optimization considered in this paper. The concepts are not difficult, but the sweet spot is small. Identifying where the program can benefit from loop unrolling or loop fusion is not trivial. Moreover, it takes some effort to get it right. Still, educating scientific programmers about temporal locality and teaching them how to optimize for it will pay dividends. Reducing instruction count Execution time is a function of instruction count. Reduce the count and you usually reduce the time. The best solution is to use a more efficient algorithm; that is, an algorithm whose order of complexity is smaller, that converges quicker, or is more accurate. Optimizing source code without changing the algorithm yields smaller, but still significant, gains. This paper considers only the latter because the intent is to study how much better codes can run if written by programmers schooled in basic code optimization techniques. The ten codes studied benefited from three types of "instruction reducing" optimizations. The two most prevalent were hoisting invariant memory and data operations out of inner loops. The third was eliminating unnecessary data copying. The nature of these inefficiencies is language dependent. Memory operations The semantics of C make it difficult for the compiler to determine all the invariant memory operations in a loop. The problem is particularly acute for loops in functions since the compiler may not know the values of the function's parameters at every call site when compiling the function. Most compilers support pragmas to help resolve ambiguities; however, these pragmas are not comprehensive and there is no standard syntax. To guarantee that invariant memory operations are not executed repetitively, the user has little choice but to hoist the operations by hand. The problem is not as severe in Fortran programs because in the absence of equivalence statements, it is a violation of the language's semantics for two names to share memory. Codes 3 and 5 are C programs. In both cases, the compiler did not hoist all invariant memory operations from inner loops. Consider the following loop from code 3 for (y = 0; y < NY; y++) { i = 0; for (u = 0; u < NU; u++) { for (k = 0; k < NK[u]; k++) { dW[y][u][k] += delta[y] * I1[i++]; } } } Since dW[y][u] can point to the same memory space as delta for one or more values of y and u, assignment to dW[y][u][k] may change the value of delta[y]. In reality, dW and delta do not overlap in memory, so I rewrote the loop as for (y = 0; y < NY; y++) { i = 0; Dy = delta[y]; for (u = 0; u < NU; u++) { for (k = 0; k < NK[u]; k++) { dW[y][u][k] += Dy * I1[i++]; } } } Failure to hoist invariant memory operations may be due to complex address calculations. If the compiler can not determine that the address calculation is invariant, then it can hoist neither the calculation nor the associated memory operations. As noted above, code 5 uses a macro to address four-dimensional arrays #define MAT4D(a,q,i,j,k) (double *)((a)->data + (q)*(a)->strides[0] + (i)*(a)->strides[3] + (j)*(a)->strides[2] + (k)*(a)->strides[1]) The macro is too complex for the compiler to understand and so, it does not identify any subexpressions as loop invariant. The simplest way to eliminate the address calculation from the innermost loop (over i) is to define a0 = MAT4D(a,q,0,j,k) before the loop and then replace all instances of *MAT4D(a,q,i,j,k) in the loop with a0[i] A similar problem appears in code 6, a Fortran program. The key loop in this program is do n1 = 1, nh nx1 = (n1 - 1) / nz + 1 nz1 = n1 - nz * (nx1 - 1) do n2 = 1, nh nx2 = (n2 - 1) / nz + 1 nz2 = n2 - nz * (nx2 - 1) ndx = nx2 - nx1 ndy = nz2 - nz1 gxx = grn(1,ndx,ndy) gyy = grn(2,ndx,ndy) gxy = grn(3,ndx,ndy) balance(n1,1) = balance(n1,1) + (force(n2,1) * gxx + force(n2,2) * gxy) * h1 balance(n1,2) = balance(n1,2) + (force(n2,1) * gxy + force(n2,2) * gyy)*h1 end do end do The programmer has written this loop well—there are no loop invariant operations with respect to n1 and n2. However, the loop resides within an iterative loop over time and the index calculations are independent with respect to time. Trading space for time, I precomputed the index values prior to the entering the time loop and stored the values in two arrays. I then replaced the index calculations with reads of the arrays. Data operations Ways to reduce data operations can appear in many forms. Implementing a more efficient algorithm produces the biggest gains. The closest I came to an algorithm change was in code 4. This code computes the inner product of K-vectors A(i) and B(j), 0 = i < N, 0 = j < M, for most values of i and j. Since the program computes most of the NM possible inner products, it is more efficient to compute all the inner products in one triply-nested loop rather than one at a time when needed. The savings accrue from reading A(i) once for all B(j) vectors and from loop unrolling. for (i = 0; i < N; i+=8) { for (j = 0; j < M; j++) { sum0 = 0.0; sum1 = 0.0; sum2 = 0.0; sum3 = 0.0; sum4 = 0.0; sum5 = 0.0; sum6 = 0.0; sum7 = 0.0; for (k = 0; k < K; k++) { sum0 += A[i+0][k] * B[j][k]; sum1 += A[i+1][k] * B[j][k]; sum2 += A[i+2][k] * B[j][k]; sum3 += A[i+3][k] * B[j][k]; sum4 += A[i+4][k] * B[j][k]; sum5 += A[i+5][k] * B[j][k]; sum6 += A[i+6][k] * B[j][k]; sum7 += A[i+7][k] * B[j][k]; } C[i+0][j] = sum0; C[i+1][j] = sum1; C[i+2][j] = sum2; C[i+3][j] = sum3; C[i+4][j] = sum4; C[i+5][j] = sum5; C[i+6][j] = sum6; C[i+7][j] = sum7; }} This change requires knowledge of a typical run; i.e., that most inner products are computed. The reasons for the change, however, derive from basic optimization concepts. It is the type of change easily made at development time by a knowledgeable programmer. In code 5, we have the data version of the index optimization in code 6. Here a very expensive computation is a function of the loop indices and so cannot be hoisted out of the loop; however, the computation is invariant with respect to an outer iterative loop over time. We can compute its value for each iteration of the computation loop prior to entering the time loop and save the values in an array. The increase in memory required to store the values is small in comparison to the large savings in time. The main loop in Code 8 is doubly nested. The inner loop includes a series of guarded computations; some are a function of the inner loop index but not the outer loop index while others are a function of the outer loop index but not the inner loop index for (j = 0; j < N; j++) { for (i = 0; i < M; i++) { r = i * hrmax; R = A[j]; temp = (PRM[3] == 0.0) ? 1.0 : pow(r, PRM[3]); high = temp * kcoeff * B[j] * PRM[2] * PRM[4]; low = high * PRM[6] * PRM[6] / (1.0 + pow(PRM[4] * PRM[6], 2.0)); kap = (R > PRM[6]) ? high * R * R / (1.0 + pow(PRM[4]*r, 2.0) : low * pow(R/PRM[6], PRM[5]); < rest of loop omitted > }} Note that the value of temp is invariant to j. Thus, we can hoist the computation for temp out of the loop and save its values in an array. for (i = 0; i < M; i++) { r = i * hrmax; TEMP[i] = pow(r, PRM[3]); } [N.B. – the case for PRM[3] = 0 is omitted and will be reintroduced later.] We now hoist out of the inner loop the computations invariant to i. Since the conditional guarding the value of kap is invariant to i, it behooves us to hoist the computation out of the inner loop, thereby executing the guard once rather than M times. The final version of the code is for (j = 0; j < N; j++) { R = rig[j] / 1000.; tmp1 = kcoeff * par[2] * beta[j] * par[4]; tmp2 = 1.0 + (par[4] * par[4] * par[6] * par[6]); tmp3 = 1.0 + (par[4] * par[4] * R * R); tmp4 = par[6] * par[6] / tmp2; tmp5 = R * R / tmp3; tmp6 = pow(R / par[6], par[5]); if ((par[3] == 0.0) && (R > par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * tmp5; } else if ((par[3] == 0.0) && (R <= par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * tmp4 * tmp6; } else if ((par[3] != 0.0) && (R > par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * TEMP[i] * tmp5; } else if ((par[3] != 0.0) && (R <= par[6])) { for (i = 1; i <= imax1; i++) KAP[i] = tmp1 * TEMP[i] * tmp4 * tmp6; } for (i = 0; i < M; i++) { kap = KAP[i]; r = i * hrmax; < rest of loop omitted > } } Maybe not the prettiest piece of code, but certainly much more efficient than the original loop, Copy operations Several programs unnecessarily copy data from one data structure to another. This problem occurs in both Fortran and C programs, although it manifests itself differently in the two languages. Code 1 declares two arrays—one for old values and one for new values. At the end of each iteration, the array of new values is copied to the array of old values to reset the data structures for the next iteration. This problem occurs in Fortran programs not included in this study and in both Fortran 77 and Fortran 90 code. Introducing pointers to the arrays and swapping pointer values is an obvious way to eliminate the copying; but pointers is not a feature that many Fortran programmers know well or are comfortable using. An easy solution not involving pointers is to extend the dimension of the value array by 1 and use the last dimension to differentiate between arrays at different times. For example, if the data space is N x N, declare the array (N, N, 2). Then store the problem’s initial values in (_, _, 2) and define the scalar names new = 2 and old = 1. At the start of each iteration, swap old and new to reset the arrays. The old–new copy problem did not appear in any C program. In programs that had new and old values, the code swapped pointers to reset data structures. Where unnecessary coping did occur is in structure assignment and parameter passing. Structures in C are handled much like scalars. Assignment causes the data space of the right-hand name to be copied to the data space of the left-hand name. Similarly, when a structure is passed to a function, the data space of the actual parameter is copied to the data space of the formal parameter. If the structure is large and the assignment or function call is in an inner loop, then copying costs can grow quite large. While none of the ten programs considered here manifested this problem, it did occur in programs not included in the study. A simple fix is always to refer to structures via pointers. Optimizing loop structures Since scientific programs spend almost all their time in loops, efficient loops are the key to good performance. Conditionals, function calls, little instruction level parallelism, and large numbers of temporary values make it difficult for the compiler to generate tightly packed, highly efficient code. Conditionals and function calls introduce jumps that disrupt code flow. Users should eliminate or isolate conditionls to their own loops as much as possible. Often logical expressions can be substituted for if-then-else statements. For example, code 2 includes the following snippet MaxDelta = 0.0 do J = 1, N do I = 1, M < code omitted > Delta = abs(OldValue ? NewValue) if (Delta > MaxDelta) MaxDelta = Delta enddo enddo if (MaxDelta .gt. 0.001) goto 200 Since the only use of MaxDelta is to control the jump to 200 and all that matters is whether or not it is greater than 0.001, I made MaxDelta a boolean and rewrote the snippet as MaxDelta = .false. do J = 1, N do I = 1, M < code omitted > Delta = abs(OldValue ? NewValue) MaxDelta = MaxDelta .or. (Delta .gt. 0.001) enddo enddo if (MaxDelta) goto 200 thereby, eliminating the conditional expression from the inner loop. A microprocessor can execute many instructions per instruction cycle. Typically, it can execute one or more memory, floating point, integer, and jump operations. To be executed simultaneously, the operations must be independent. Thick loops tend to have more instruction level parallelism than thin loops. Moreover, they reduce memory traffice by maximizing data reuse. Loop unrolling and loop fusion are two techniques to increase the size of loop bodies. Several of the codes studied benefitted from loop unrolling, but none benefitted from loop fusion. This observation is not too surpising since it is the general tendency of programmers to write thick loops. As loops become thicker, the number of temporary values grows, increasing register pressure. If registers spill, then memory traffic increases and code flow is disrupted. A thick loop with many temporary values may execute slower than an equivalent series of thin loops. The biggest gain will be achieved if the thick loop can be split into a series of independent loops eliminating the need to write and read temporary arrays. I found such an occasion in code 10 where I split the loop do i = 1, n do j = 1, m A24(j,i)= S24(j,i) * T24(j,i) + S25(j,i) * U25(j,i) B24(j,i)= S24(j,i) * T25(j,i) + S25(j,i) * U24(j,i) A25(j,i)= S24(j,i) * C24(j,i) + S25(j,i) * V24(j,i) B25(j,i)= S24(j,i) * U25(j,i) + S25(j,i) * V25(j,i) C24(j,i)= S26(j,i) * T26(j,i) + S27(j,i) * U26(j,i) D24(j,i)= S26(j,i) * T27(j,i) + S27(j,i) * V26(j,i) C25(j,i)= S27(j,i) * S28(j,i) + S26(j,i) * U28(j,i) D25(j,i)= S27(j,i) * T28(j,i) + S26(j,i) * V28(j,i) end do end do into two disjoint loops do i = 1, n do j = 1, m A24(j,i)= S24(j,i) * T24(j,i) + S25(j,i) * U25(j,i) B24(j,i)= S24(j,i) * T25(j,i) + S25(j,i) * U24(j,i) A25(j,i)= S24(j,i) * C24(j,i) + S25(j,i) * V24(j,i) B25(j,i)= S24(j,i) * U25(j,i) + S25(j,i) * V25(j,i) end do end do do i = 1, n do j = 1, m C24(j,i)= S26(j,i) * T26(j,i) + S27(j,i) * U26(j,i) D24(j,i)= S26(j,i) * T27(j,i) + S27(j,i) * V26(j,i) C25(j,i)= S27(j,i) * S28(j,i) + S26(j,i) * U28(j,i) D25(j,i)= S27(j,i) * T28(j,i) + S26(j,i) * V28(j,i) end do end do Conclusions Over the course of the last year, I have had the opportunity to work with over two dozen academic scientific programmers at leading research universities. Their research interests span a broad range of scientific fields. Except for two programs that relied almost exclusively on library routines (matrix multiply and fast Fourier transform), I was able to improve significantly the single processor performance of all codes. Improvements range from 2x to 15.5x with a simple average of 4.75x. Changes to the source code were at a very high level. I did not use sophisticated techniques or programming tools to discover inefficiencies or effect the changes. Only one code was parallel despite the availability of parallel systems to all developers. Clearly, we have a problem—personal scientific research codes are highly inefficient and not running parallel. The developers are unaware of simple optimization techniques to make programs run faster. They lack education in the art of code optimization and parallel programming. I do not believe we can fix the problem by publishing additional books or training manuals. To date, the developers in questions have not studied the books or manual available, and are unlikely to do so in the future. Short courses are a possible solution, but I believe they are too concentrated to be much use. The general concepts can be taught in a three or four day course, but that is not enough time for students to practice what they learn and acquire the experience to apply and extend the concepts to their codes. Practice is the key to becoming proficient at optimization. I recommend that graduate students be required to take a semester length course in optimization and parallel programming. We would never give someone access to state-of-the-art scientific equipment costing hundreds of thousands of dollars without first requiring them to demonstrate that they know how to use the equipment. Yet the criterion for time on state-of-the-art supercomputers is at most an interesting project. Requestors are never asked to demonstrate that they know how to use the system, or can use the system effectively. A semester course would teach them the required skills. Government agencies that fund academic scientific research pay for most of the computer systems supporting scientific research as well as the development of most personal scientific codes. These agencies should require graduate schools to offer a course in optimization and parallel programming as a requirement for funding. About the Author John Feo received his Ph.D. in Computer Science from The University of Texas at Austin in 1986. After graduate school, Dr. Feo worked at Lawrence Livermore National Laboratory where he was the Group Leader of the Computer Research Group and principal investigator of the Sisal Language Project. In 1997, Dr. Feo joined Tera Computer Company where he was project manager for the MTA, and oversaw the programming and evaluation of the MTA at the San Diego Supercomputer Center. In 2000, Dr. Feo joined Sun Microsystems as an HPC application specialist. He works with university research groups to optimize and parallelize scientific codes. Dr. Feo has published over two dozen research articles in the areas of parallel parallel programming, parallel programming languages, and application performance.

  Read the article

< Previous Page | 1 2 3 4 5 6 7 8 9 10 11 12  | Next Page >