Profiling

Profiling profiling cost-centre profiling Glasgow Haskell comes with a time and space profiling system. Its purpose is to help you improve your understanding of your program's execution behaviour, so you can improve it. Any comments, suggestions and/or improvements you have are welcome. Recommended “profiling tricks” would be especially cool! Profiling a program is a three-step process: Re-compile your program for profiling with the -prof option, and probably one of the -auto or -auto-all options. These options are described in more detail in -prof -auto -auto-all Run your program with one of the profiling options, eg. +RTS -p -RTS. This generates a file of profiling information. Note that multi-processor execution (e.g. +RTS -N2) is not supported while profiling. RTS option Examine the generated profiling information, using one of GHC's profiling tools. The tool to use will depend on the kind of profiling information generated. Cost centres and cost-centre stacks GHC's profiling system assigns costs to cost centres. A cost is simply the time or space required to evaluate an expression. Cost centres are program annotations around expressions; all costs incurred by the annotated expression are assigned to the enclosing cost centre. Furthermore, GHC will remember the stack of enclosing cost centres for any given expression at run-time and generate a call-graph of cost attributions. Let's take a look at an example: main = print (nfib 25) nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2) Compile and run this program as follows: $ ghc -prof -auto-all -o Main Main.hs $ ./Main +RTS -p 121393 $ When a GHC-compiled program is run with the RTS option, it generates a file called <prog>.prof. In this case, the file will contain something like this: Fri May 12 14:06 2000 Time and Allocation Profiling Report (Final) Main +RTS -p -RTS total time = 0.14 secs (7 ticks @ 20 ms) total alloc = 8,741,204 bytes (excludes profiling overheads) COST CENTRE MODULE %time %alloc nfib Main 100.0 100.0 individual inherited COST CENTRE MODULE entries %time %alloc %time %alloc MAIN MAIN 0 0.0 0.0 100.0 100.0 main Main 0 0.0 0.0 0.0 0.0 CAF PrelHandle 3 0.0 0.0 0.0 0.0 CAF PrelAddr 1 0.0 0.0 0.0 0.0 CAF Main 6 0.0 0.0 100.0 100.0 main Main 1 0.0 0.0 100.0 100.0 nfib Main 242785 100.0 100.0 100.0 100.0 The first part of the file gives the program name and options, and the total time and total memory allocation measured during the run of the program (note that the total memory allocation figure isn't the same as the amount of live memory needed by the program at any one time; the latter can be determined using heap profiling, which we will describe shortly). The second part of the file is a break-down by cost centre of the most costly functions in the program. In this case, there was only one significant function in the program, namely nfib, and it was responsible for 100% of both the time and allocation costs of the program. The third and final section of the file gives a profile break-down by cost-centre stack. This is roughly a call-graph profile of the program. In the example above, it is clear that the costly call to nfib came from main. The time and allocation incurred by a given part of the program is displayed in two ways: “individual”, which are the costs incurred by the code covered by this cost centre stack alone, and “inherited”, which includes the costs incurred by all the children of this node. The usefulness of cost-centre stacks is better demonstrated by modifying the example slightly: main = print (f 25 + g 25) f n = nfib n g n = nfib (n `div` 2) nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2) Compile and run this program as before, and take a look at the new profiling results: COST CENTRE MODULE scc %time %alloc %time %alloc MAIN MAIN 0 0.0 0.0 100.0 100.0 main Main 0 0.0 0.0 0.0 0.0 CAF PrelHandle 3 0.0 0.0 0.0 0.0 CAF PrelAddr 1 0.0 0.0 0.0 0.0 CAF Main 9 0.0 0.0 100.0 100.0 main Main 1 0.0 0.0 100.0 100.0 g Main 1 0.0 0.0 0.0 0.2 nfib Main 465 0.0 0.2 0.0 0.2 f Main 1 0.0 0.0 100.0 99.8 nfib Main 242785 100.0 99.8 100.0 99.8 Now although we had two calls to nfib in the program, it is immediately clear that it was the call from f which took all the time. The actual meaning of the various columns in the output is: entries The number of times this particular point in the call graph was entered. individual %time The percentage of the total run time of the program spent at this point in the call graph. individual %alloc The percentage of the total memory allocations (excluding profiling overheads) of the program made by this call. inherited %time The percentage of the total run time of the program spent below this point in the call graph. inherited %alloc The percentage of the total memory allocations (excluding profiling overheads) of the program made by this call and all of its sub-calls. In addition you can use the RTS option to get the following additional information: ticks The raw number of time “ticks” which were attributed to this cost-centre; from this, we get the %time figure mentioned above. bytes Number of bytes allocated in the heap while in this cost-centre; again, this is the raw number from which we get the %alloc figure mentioned above. What about recursive functions, and mutually recursive groups of functions? Where are the costs attributed? Well, although GHC does keep information about which groups of functions called each other recursively, this information isn't displayed in the basic time and allocation profile, instead the call-graph is flattened into a tree. Inserting cost centres by hand Cost centres are just program annotations. When you say to the compiler, it automatically inserts a cost centre annotation around every top-level function in your program, but you are entirely free to add the cost centre annotations yourself. The syntax of a cost centre annotation is {-# SCC "name" #-} <expression> where "name" is an arbitrary string, that will become the name of your cost centre as it appears in the profiling output, and <expression> is any Haskell expression. An SCC annotation extends as far to the right as possible when parsing. (SCC stands for "Set Cost Centre"). Rules for attributing costs The cost of evaluating any expression in your program is attributed to a cost-centre stack using the following rules: If the expression is part of the one-off costs of evaluating the enclosing top-level definition, then costs are attributed to the stack of lexically enclosing SCC annotations on top of the special CAF cost-centre. Otherwise, costs are attributed to the stack of lexically-enclosing SCC annotations, appended to the cost-centre stack in effect at the call site of the current top-level definition The call-site is just the place in the source code which mentions the particular function or variable.. Notice that this is a recursive definition. Time spent in foreign code (see ) is always attributed to the cost centre in force at the Haskell call-site of the foreign function. What do we mean by one-off costs? Well, Haskell is a lazy language, and certain expressions are only ever evaluated once. For example, if we write: x = nfib 25 then x will only be evaluated once (if at all), and subsequent demands for x will immediately get to see the cached result. The definition x is called a CAF (Constant Applicative Form), because it has no arguments. For the purposes of profiling, we say that the expression nfib 25 belongs to the one-off costs of evaluating x. Since one-off costs aren't strictly speaking part of the call-graph of the program, they are attributed to a special top-level cost centre, CAF. There may be one CAF cost centre for each module (the default), or one for each top-level definition with any one-off costs (this behaviour can be selected by giving GHC the flag). -caf-all If you think you have a weird profile, or the call-graph doesn't look like you expect it to, feel free to send it (and your program) to us at glasgow-haskell-bugs@haskell.org. Compiler options for profiling profilingoptions optionsfor profiling : To make use of the profiling system all modules must be compiled and linked with the option. Any SCC annotations you've put in your source will spring to life. Without a option, your SCCs are ignored; so you can compile SCC-laden code without changing it. There are a few other profiling-related compilation options. Use them in addition to . These do not have to be used consistently for all modules in a program. : cost centresautomatically inserting GHC will automatically add _scc_ constructs for all top-level, exported functions. : All top-level functions, exported or not, will be automatically _scc_'d. : The costs of all CAFs in a module are usually attributed to one “big” CAF cost-centre. With this option, all CAFs get their own cost-centre. An “if all else fails” option… : Ignore any _scc_ constructs, so a module which already has _scc_s can be compiled for profiling with the annotations ignored. Time and allocation profiling To generate a time and allocation profile, give one of the following RTS options to the compiled program when you run it (RTS options should be enclosed between +RTS...-RTS as usual): or : time profile The option produces a standard time profile report. It is written into the file program.prof. The option produces a more detailed report containing the actual time and allocation data as well. (Not used much.) RTS option This option makes use of the extra information maintained by the cost-centre-stack profiler to provide useful information about the location of runtime errors. See . Profiling memory usage In addition to profiling the time and allocation behaviour of your program, you can also generate a graph of its memory usage over time. This is useful for detecting the causes of space leaks, when your program holds on to more memory at run-time that it needs to. Space leaks lead to longer run-times due to heavy garbage collector activity, and may even cause the program to run out of memory altogether. To generate a heap profile from your program: Compile the program for profiling (). Run it with one of the heap profiling options described below (eg. for a basic producer profile). This generates the file prog.hp. Run hp2ps to produce a Postscript file, prog.ps. The hp2ps utility is described in detail in . Display the heap profile using a postscript viewer such as Ghostview, or print it out on a Postscript-capable printer. RTS options for heap profiling There are several different kinds of heap profile that can be generated. All the different profile types yield a graph of live heap against time, but they differ in how the live heap is broken down into bands. The following RTS options select which break-down to use: RTS option Breaks down the graph by the cost-centre stack which produced the data. RTS option Break down the live heap by the module containing the code which produced the data. RTS option Breaks down the graph by closure description. For actual data, the description is just the constructor name, for other closures it is a compiler-generated string identifying the closure. RTS option Breaks down the graph by type. For closures which have function type or unknown/polymorphic type, the string will represent an approximation to the actual type. RTS option Break down the graph by retainer set. Retainer profiling is described in more detail below (). RTS option Break down the graph by biography. Biographical profiling is described in more detail below (). In addition, the profile can be restricted to heap data which satisfies certain criteria - for example, you might want to display a profile by type but only for data produced by a certain module, or a profile by retainer for a certain type of data. Restrictions are specified as follows: name,... RTS option Restrict the profile to closures produced by cost-centre stacks with one of the specified cost centres at the top. name,... RTS option Restrict the profile to closures produced by cost-centre stacks with one of the specified cost centres anywhere in the stack. module,... RTS option Restrict the profile to closures produced by the specified modules. desc,... RTS option Restrict the profile to closures with the specified description strings. type,... RTS option Restrict the profile to closures with the specified types. cc,... RTS option Restrict the profile to closures with retainer sets containing cost-centre stacks with one of the specified cost centres at the top. bio,... RTS option Restrict the profile to closures with one of the specified biographies, where bio is one of lag, drag, void, or use. For example, the following options will generate a retainer profile restricted to Branch and Leaf constructors: prog +RTS -hr -hdBranch,Leaf There can only be one "break-down" option (eg. in the example above), but there is no limit on the number of further restrictions that may be applied. All the options may be combined, with one exception: GHC doesn't currently support mixing the and options. There are three more options which relate to heap profiling: : Set the profiling (sampling) interval to secs seconds (the default is 0.1 second). Fractions are allowed: for example will get 5 samples per second. This only affects heap profiling; time profiles are always sampled on a 1/50 second frequency. RTS option Include the memory occupied by threads in a heap profile. Each thread takes up a small area for its thread state in addition to the space allocated for its stack (stacks normally start small and then grow as necessary). This includes the main thread, so using is a good way to see how much stack space the program is using. Memory occupied by threads and their stacks is labelled as “TSO” when displaying the profile by closure description or type description. RTS option Sets the maximum length of a cost-centre stack name in a heap profile. Defaults to 25. Retainer Profiling Retainer profiling is designed to help answer questions like why is this data being retained?. We start by defining what we mean by a retainer:

A retainer is either the system stack, or an unevaluated closure (thunk).

In particular, constructors are not retainers. An object B retains object A if (i) B is a retainer object and (ii) object A can be reached by recursively following pointers starting from object B, but not meeting any other retainer objects on the way. Each live object is retained by one or more retainer objects, collectively called its retainer set, or its retainer set, or its retainers. When retainer profiling is requested by giving the program the option, a graph is generated which is broken down by retainer set. A retainer set is displayed as a set of cost-centre stacks; because this is usually too large to fit on the profile graph, each retainer set is numbered and shown abbreviated on the graph along with its number, and the full list of retainer sets is dumped into the file prog.prof. Retainer profiling requires multiple passes over the live heap in order to discover the full retainer set for each object, which can be quite slow. So we set a limit on the maximum size of a retainer set, where all retainer sets larger than the maximum retainer set size are replaced by the special set MANY. The maximum set size defaults to 8 and can be altered with the RTS option: size Restrict the number of elements in a retainer set to size (default 8). Hints for using retainer profiling The definition of retainers is designed to reflect a common cause of space leaks: a large structure is retained by an unevaluated computation, and will be released once the computation is forced. A good example is looking up a value in a finite map, where unless the lookup is forced in a timely manner the unevaluated lookup will cause the whole mapping to be retained. These kind of space leaks can often be eliminated by forcing the relevant computations to be performed eagerly, using seq or strictness annotations on data constructor fields. Often a particular data structure is being retained by a chain of unevaluated closures, only the nearest of which will be reported by retainer profiling - for example A retains B, B retains C, and C retains a large structure. There might be a large number of Bs but only a single A, so A is really the one we're interested in eliminating. However, retainer profiling will in this case report B as the retainer of the large structure. To move further up the chain of retainers, we can ask for another retainer profile but this time restrict the profile to B objects, so we get a profile of the retainers of B: prog +RTS -hr -hcB This trick isn't foolproof, because there might be other B closures in the heap which aren't the retainers we are interested in, but we've found this to be a useful technique in most cases. Biographical Profiling A typical heap object may be in one of the following four states at each point in its lifetime: The lag stage, which is the time between creation and the first use of the object, the use stage, which lasts from the first use until the last use of the object, and The drag stage, which lasts from the final use until the last reference to the object is dropped. An object which is never used is said to be in the void state for its whole lifetime. A biographical heap profile displays the portion of the live heap in each of the four states listed above. Usually the most interesting states are the void and drag states: live heap in these states is more likely to be wasted space than heap in the lag or use states. It is also possible to break down the heap in one or more of these states by a different criteria, by restricting a profile by biography. For example, to show the portion of the heap in the drag or void state by producer: prog +RTS -hc -hbdrag,void Once you know the producer or the type of the heap in the drag or void states, the next step is usually to find the retainer(s): prog +RTS -hr -hccc... NOTE: this two stage process is required because GHC cannot currently profile using both biographical and retainer information simultaneously. Actual memory residency How does the heap residency reported by the heap profiler relate to the actual memory residency of your program when you run it? You might see a large discrepancy between the residency reported by the heap profiler, and the residency reported by tools on your system (eg. ps or top on Unix, or the Task Manager on Windows). There are several reasons for this: There is an overhead of profiling itself, which is subtracted from the residency figures by the profiler. This overhead goes away when compiling without profiling support, of course. The space overhead is currently 2 extra words per heap object, which probably results in about a 30% overhead. Garbage collection requires more memory than the actual residency. The factor depends on the kind of garbage collection algorithm in use: a major GC in the standard generation copying collector will usually require 3L bytes of memory, where L is the amount of live data. This is because by default (see the option) we allow the old generation to grow to twice its size (2L) before collecting it, and we require additionally L bytes to copy the live data into. When using compacting collection (see the option), this is reduced to 2L, and can further be reduced by tweaking the option. Also add the size of the allocation area (currently a fixed 512Kb). The stack isn't counted in the heap profile by default. See the option. The program text itself, the C stack, any non-heap data (eg. data allocated by foreign libraries, and data allocated by the RTS), and mmap()'d memory are not counted in the heap profile. <command>hp2ps</command>––heap profile to PostScript hp2ps heap profiles postscript, from heap profiles Usage: hp2ps [flags] [<file>[.hp]] The program hp2pshp2ps program converts a heap profile as produced by the runtime option into a PostScript graph of the heap profile. By convention, the file to be processed by hp2ps has a .hp extension. The PostScript output is written to <file>@.ps. If <file> is omitted entirely, then the program behaves as a filter. hp2ps is distributed in ghc/utils/hp2ps in a GHC source distribution. It was originally developed by Dave Wakeling as part of the HBC/LML heap profiler. The flags are: In order to make graphs more readable, hp2ps sorts the shaded bands for each identifier. The default sort ordering is for the bands with the largest area to be stacked on top of the smaller ones. The option causes rougher bands (those representing series of values with the largest standard deviations) to be stacked on top of smoother ones. Normally, hp2ps puts the title of the graph in a small box at the top of the page. However, if the JOB string is too long to fit in a small box (more than 35 characters), then hp2ps will choose to use a big box instead. The option forces hp2ps to use a big box. Generate encapsulated PostScript suitable for inclusion in LaTeX documents. Usually, the PostScript graph is drawn in landscape mode in an area 9 inches wide by 6 inches high, and hp2ps arranges for this area to be approximately centred on a sheet of a4 paper. This format is convenient of studying the graph in detail, but it is unsuitable for inclusion in LaTeX documents. The option causes the graph to be drawn in portrait mode, with float specifying the width in inches, millimetres or points (the default). The resulting PostScript file conforms to the Encapsulated PostScript (EPS) convention, and it can be included in a LaTeX document using Rokicki's dvi-to-PostScript converter dvips. Create output suitable for the gs PostScript previewer (or similar). In this case the graph is printed in portrait mode without scaling. The output is unsuitable for a laser printer. Normally a profile is limited to 20 bands with additional identifiers being grouped into an OTHER band. The flag removes this 20 band and limit, producing as many bands as necessary. No key is produced as it won't fit!. It is useful for creation time profiles with many bands. Normally a profile is limited to 20 bands with additional identifiers being grouped into an OTHER band. The flag specifies an alternative band limit (the maximum is 20). requests the band limit to be removed. As many bands as necessary are produced. However no key is produced as it won't fit! It is useful for displaying creation time profiles with many bands. Use previous parameters. By default, the PostScript graph is automatically scaled both horizontally and vertically so that it fills the page. However, when preparing a series of graphs for use in a presentation, it is often useful to draw a new graph using the same scale, shading and ordering as a previous one. The flag causes the graph to be drawn using the parameters determined by a previous run of hp2ps on file. These are extracted from file@.aux. Use a small box for the title. Normally trace elements which sum to a total of less than 1% of the profile are removed from the profile. The option allows this percentage to be modified (maximum 5%). requests no trace elements to be removed from the profile, ensuring that all the data will be displayed. Generate colour output. Ignore marks. Print out usage information. Manipulating the hp file (Notes kindly offered by Jan-Willhem Maessen.) The FOO.hp file produced when you ask for the heap profile of a program FOO is a text file with a particularly simple structure. Here's a representative example, with much of the actual data omitted: JOB "FOO -hC" DATE "Thu Dec 26 18:17 2002" SAMPLE_UNIT "seconds" VALUE_UNIT "bytes" BEGIN_SAMPLE 0.00 END_SAMPLE 0.00 BEGIN_SAMPLE 15.07 ... sample data ... END_SAMPLE 15.07 BEGIN_SAMPLE 30.23 ... sample data ... END_SAMPLE 30.23 ... etc. BEGIN_SAMPLE 11695.47 END_SAMPLE 11695.47 The first four lines (JOB, DATE, SAMPLE_UNIT, VALUE_UNIT) form a header. Each block of lines starting with BEGIN_SAMPLE and ending with END_SAMPLE forms a single sample (you can think of this as a vertical slice of your heap profile). The hp2ps utility should accept any input with a properly-formatted header followed by a series of *complete* samples. Zooming in on regions of your profile You can look at particular regions of your profile simply by loading a copy of the .hp file into a text editor and deleting the unwanted samples. The resulting .hp file can be run through hp2ps and viewed or printed. Viewing the heap profile of a running program The .hp file is generated incrementally as your program runs. In principle, running hp2ps on the incomplete file should produce a snapshot of your program's heap usage. However, the last sample in the file may be incomplete, causing hp2ps to fail. If you are using a machine with UNIX utilities installed, it's not too hard to work around this problem (though the resulting command line looks rather Byzantine): head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \ | hp2ps > FOO.ps The command fgrep -n END_SAMPLE FOO.hp finds the end of every complete sample in FOO.hp, and labels each sample with its ending line number. We then select the line number of the last complete sample using tail and cut. This is used as a parameter to head; the result is as if we deleted the final incomplete sample from FOO.hp. This results in a properly-formatted .hp file which we feed directly to hp2ps. Viewing a heap profile in real time The gv and ghostview programs have a "watch file" option can be used to view an up-to-date heap profile of your program as it runs. Simply generate an incremental heap profile as described in the previous section. Run gv on your profile: gv -watch -seascape FOO.ps If you forget the -watch flag you can still select "Watch file" from the "State" menu. Now each time you generate a new profile FOO.ps the view will update automatically. This can all be encapsulated in a little script: #!/bin/sh head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \ | hp2ps > FOO.ps gv -watch -seascape FOO.ps & while [ 1 ] ; do sleep 10 # We generate a new profile every 10 seconds. head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \ | hp2ps > FOO.ps done Occasionally gv will choke as it tries to read an incomplete copy of FOO.ps (because hp2ps is still running as an update occurs). A slightly more complicated script works around this problem, by using the fact that sending a SIGHUP to gv will cause it to re-read its input file: #!/bin/sh head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \ | hp2ps > FOO.ps gv FOO.ps & gvpsnum=$! while [ 1 ] ; do sleep 10 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \ | hp2ps > FOO.ps kill -HUP $gvpsnum done Observing Code Coverage code coverage Haskell Program Coverage hpc Code coverage tools allow a programmer to determine what parts of their code have been actually executed, and which parts have never actually been invoked. GHC has an option for generating instrumented code that records code coverage as part of the Haskell Program Coverage (HPC) toolkit, which is included with GHC. HPC tools can be used to render the generated code coverage information into human understandable format. Correctly instrumented code provides coverage information of two kinds: source coverage and boolean-control coverage. Source coverage is the extent to which every part of the program was used, measured at three different levels: declarations (both top-level and local), alternatives (among several equations or case branches) and expressions (at every level). Boolean coverage is the extent to which each of the values True and False is obtained in every syntactic boolean context (ie. guard, condition, qualifier). HPC displays both kinds of information in two primary ways: textual reports with summary statistics (hpc report) and sources with color mark-up (hpc markup). For boolean coverage, there are four possible outcomes for each guard, condition or qualifier: both True and False values occur; only True; only False; never evaluated. In hpc-markup output, highlighting with a yellow background indicates a part of the program that was never evaluated; a green background indicates an always-True expression and a red background indicates an always-False one. A small example: Reciprocation For an example we have a program, called Recip.hs, which computes exact decimal representations of reciprocals, with recurring parts indicated in brackets. reciprocal :: Int -> (String, Int) reciprocal n | n > 1 = ('0' : '.' : digits, recur) | otherwise = error "attempting to compute reciprocal of number <= 1" where (digits, recur) = divide n 1 [] divide :: Int -> Int -> [Int] -> (String, Int) divide n c cs | c `elem` cs = ([], position c cs) | r == 0 = (show q, 0) | r /= 0 = (show q ++ digits, recur) where (q, r) = (c*10) `quotRem` n (digits, recur) = divide n r (c:cs) position :: Int -> [Int] -> Int position n (x:xs) | n==x = 1 | otherwise = 1 + position n xs showRecip :: Int -> String showRecip n = "1/" ++ show n ++ " = " ++ if r==0 then d else take p d ++ "(" ++ drop p d ++ ")" where p = length d - r (d, r) = reciprocal n main = do number <- readLn putStrLn (showRecip number) main The HPC instrumentation is enabled using the -fhpc flag. $ ghc -fhpc Recip.hs --make HPC index (.mix) files are placed placed in .hpc subdirectory. These can be considered like the .hi files for HPC. $ ./Recip 1/3 = 0.(3) We can generate a textual summary of coverage: $ hpc report Recip 80% expressions used (81/101) 12% boolean coverage (1/8) 14% guards (1/7), 3 always True, 1 always False, 2 unevaluated 0% 'if' conditions (0/1), 1 always False 100% qualifiers (0/0) 55% alternatives used (5/9) 100% local declarations used (9/9) 100% top-level declarations used (5/5) We can also generate a marked-up version of the source. $ hpc markup Recip writing Recip.hs.html This generates one file per Haskell module, and 4 index files, hpc_index.html, hpc_index_alt.html, hpc_index_exp.html, hpc_index_fun.html. Options for instrumenting code for coverage Turning on code coverage is easy, use the -fhpc flag. Instrumented and non-instrumented can be freely mixed. When compiling the Main module GHC automatically detects when there is an hpc compiled file, and adds the correct initialization code. The hpc toolkit The hpc toolkit uses a cvs/svn/darcs-like interface, where a single binary contains many function units. $ hpc Usage: hpc COMMAND ... Commands: help Display help for hpc or a single command Reporting Coverage: report Output textual report about program coverage markup Markup Haskell source with program coverage Processing Coverage files: sum Sum multiple .tix files in a single .tix file combine Combine two .tix files in a single .tix file map Map a function over a single .tix file Coverage Overlays: overlay Generate a .tix file from an overlay file draft Generate draft overlay that provides 100% coverage Others: show Show .tix file in readable, verbose format version Display version for hpc In general, these options act on .tix file after an instrumented binary has generated it, which hpc acting as a conduit between the raw .tix file, and the more detailed reports produced. The hpc tool assumes you are in the top-level directory of the location where you built your application, and the .tix file is in the same top-level directory. You can use the flag --srcdir to use hpc for any other directory, and use --srcdir multiple times to analyse programs compiled from difference locations, as is typical for packages. We now explain in more details the major modes of hpc. hpc report hpc report gives a textual report of coverage. By default, all modules and packages are considered in generating report, unless include or exclude are used. The report is a summary unless the --per-module flag is used. The --xml-output option allows for tools to use hpc to glean coverage. $ hpc help report Usage: hpc report [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]] Options: --per-module show module level detail --decl-list show unused decls --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --srcdir=DIR path to source directory of .hs files multi-use of srcdir possible --hpcdir=DIR sub-directory that contains .mix files default .hpc [rarely used] --xml-output show output in XML hpc markup hpc markup marks up source files into colored html. $ hpc help markup Usage: hpc markup [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]] Options: --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --srcdir=DIR path to source directory of .hs files multi-use of srcdir possible --hpcdir=DIR sub-directory that contains .mix files default .hpc [rarely used] --fun-entry-count show top-level function entry counts --highlight-covered highlight covered code, rather that code gaps --destdir=DIR path to write output to hpc sum hpc sum adds together any number of .tix files into a single .tix file. hpc sum does not change the original .tix file; it generates a new .tix file. $ hpc help sum Usage: hpc sum [OPTION] .. <TIX_FILE> [<TIX_FILE> [<TIX_FILE> ..]] Sum multiple .tix files in a single .tix file Options: --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --output=FILE output FILE --union use the union of the module namespace (default is intersection) hpc combine hpc combine is the swiss army knife of hpc. It can be used to take the difference between .tix files, to subtract one .tix file from another, or to add two .tix files. hpc combine does not change the original .tix file; it generates a new .tix file. $ hpc help combine Usage: hpc combine [OPTION] .. <TIX_FILE> <TIX_FILE> Combine two .tix files in a single .tix file Options: --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --output=FILE output FILE --function=FUNCTION combine .tix files with join function, default = ADD FUNCTION = ADD | DIFF | SUB --union use the union of the module namespace (default is intersection) hpc map hpc map inverts or zeros a .tix file. hpc map does not change the original .tix file; it generates a new .tix file. $ hpc help map Usage: hpc map [OPTION] .. <TIX_FILE> Map a function over a single .tix file Options: --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --output=FILE output FILE --function=FUNCTION apply function to .tix files, default = ID FUNCTION = ID | INV | ZERO --union use the union of the module namespace (default is intersection) hpc overlay and hpc draft Overlays are an experimental feature of HPC, a textual description of coverage. hpc draft is used to generate a draft overlay from a .tix file, and hpc overlay generates a .tix files from an overlay. % hpc help overlay Usage: hpc overlay [OPTION] .. <OVERLAY_FILE> [<OVERLAY_FILE> [...]] Options: --srcdir=DIR path to source directory of .hs files multi-use of srcdir possible --hpcdir=DIR sub-directory that contains .mix files default .hpc [rarely used] --output=FILE output FILE % hpc help draft Usage: hpc draft [OPTION] .. <TIX_FILE> Options: --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE --srcdir=DIR path to source directory of .hs files multi-use of srcdir possible --hpcdir=DIR sub-directory that contains .mix files default .hpc [rarely used] --output=FILE output FILE Caveats and Shortcomings of Haskell Program Coverage HPC does not attempt to lock the .tix file, so multiple concurrently running binaries in the same directory will exhibit a race condition. There is no way to change the name of the .tix file generated, apart from renaming the binary. HPC does not work with GHCi. Using “ticky-ticky” profiling (for implementors) ticky-ticky profiling (ToDo: document properly.) It is possible to compile Glasgow Haskell programs so that they will count lots and lots of interesting things, e.g., number of updates, number of data constructors entered, etc., etc. We call this “ticky-ticky” profiling,ticky-ticky profiling profiling, ticky-ticky because that's the sound a Sun4 makes when it is running up all those counters (slowly). Ticky-ticky profiling is mainly intended for implementors; it is quite separate from the main “cost-centre” profiling system, intended for all users everywhere. To be able to use ticky-ticky profiling, you will need to have built the ticky RTS. (This should be described in the building guide, but amounts to building the RTS with way "t" enabled.) To get your compiled program to spit out the ticky-ticky numbers, use a RTS option-r RTS option. See . Compiling your program with the switch yields an executable that performs these counts. Here is a sample ticky-ticky statistics file, generated by the invocation foo +RTS -rfoo.ticky. foo +RTS -rfoo.ticky ALLOCATIONS: 3964631 (11330900 words total: 3999476 admin, 6098829 goods, 1232595 slop) total words: 2 3 4 5 6+ 69647 ( 1.8%) function values 50.0 50.0 0.0 0.0 0.0 2382937 ( 60.1%) thunks 0.0 83.9 16.1 0.0 0.0 1477218 ( 37.3%) data values 66.8 33.2 0.0 0.0 0.0 0 ( 0.0%) big tuples 2 ( 0.0%) black holes 0.0 100.0 0.0 0.0 0.0 0 ( 0.0%) prim things 34825 ( 0.9%) partial applications 0.0 0.0 0.0 100.0 0.0 2 ( 0.0%) thread state objects 0.0 0.0 0.0 0.0 100.0 Total storage-manager allocations: 3647137 (11882004 words) [551104 words lost to speculative heap-checks] STACK USAGE: ENTERS: 9400092 of which 2005772 (21.3%) direct to the entry code [the rest indirected via Node's info ptr] 1860318 ( 19.8%) thunks 3733184 ( 39.7%) data values 3149544 ( 33.5%) function values [of which 1999880 (63.5%) bypassed arg-satisfaction chk] 348140 ( 3.7%) partial applications 308906 ( 3.3%) normal indirections 0 ( 0.0%) permanent indirections RETURNS: 5870443 2137257 ( 36.4%) from entering a new constructor [the rest from entering an existing constructor] 2349219 ( 40.0%) vectored [the rest unvectored] RET_NEW: 2137257: 32.5% 46.2% 21.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% RET_OLD: 3733184: 2.8% 67.9% 29.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% RET_UNBOXED_TUP: 2: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% RET_VEC_RETURN : 2349219: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% UPDATE FRAMES: 2241725 (0 omitted from thunks) SEQ FRAMES: 1 CATCH FRAMES: 1 UPDATES: 2241725 0 ( 0.0%) data values 34827 ( 1.6%) partial applications [2 in place, 34825 allocated new space] 2206898 ( 98.4%) updates to existing heap objects (46 by squeezing) UPD_CON_IN_NEW: 0: 0 0 0 0 0 0 0 0 0 UPD_PAP_IN_NEW: 34825: 0 0 0 34825 0 0 0 0 0 NEW GEN UPDATES: 2274700 ( 99.9%) OLD GEN UPDATES: 1852 ( 0.1%) Total bytes copied during GC: 190096 ************************************************** 3647137 ALLOC_HEAP_ctr 11882004 ALLOC_HEAP_tot 69647 ALLOC_FUN_ctr 69647 ALLOC_FUN_adm 69644 ALLOC_FUN_gds 34819 ALLOC_FUN_slp 34831 ALLOC_FUN_hst_0 34816 ALLOC_FUN_hst_1 0 ALLOC_FUN_hst_2 0 ALLOC_FUN_hst_3 0 ALLOC_FUN_hst_4 2382937 ALLOC_UP_THK_ctr 0 ALLOC_SE_THK_ctr 308906 ENT_IND_ctr 0 E!NT_PERM_IND_ctr requires +RTS -Z [... lots more info omitted ...] 0 GC_SEL_ABANDONED_ctr 0 GC_SEL_MINOR_ctr 0 GC_SEL_MAJOR_ctr 0 GC_FAILED_PROMOTION_ctr 47524 GC_WORDS_COPIED_ctr The formatting of the information above the row of asterisks is subject to change, but hopefully provides a useful human-readable summary. Below the asterisks all counters maintained by the ticky-ticky system are dumped, in a format intended to be machine-readable: zero or more spaces, an integer, a space, the counter name, and a newline. In fact, not all counters are necessarily dumped; compile- or run-time flags can render certain counters invalid. In this case, either the counter will simply not appear, or it will appear with a modified counter name, possibly along with an explanation for the omission (notice ENT_PERM_IND_ctr appears with an inserted ! above). Software analysing this output should always check that it has the counters it expects. Also, beware: some of the counters can have large values!