aprof user manual
Contents
1. is an application specific compiled simulator generator The compiled simulators are then executed on a host platform typically x86 32 instrument is a collection of small C and TXL tools that allow for inserting basic block counters in order to obtain basic block execution frequencies e prof is a collection of small C and AWK tools for inserting the necessary code in compiled simulators for generating profiling reports The current NAC specification is detailed in the corresponding reference manual found in the doc subdirectory in HTML and PDF form 2 Obtaining and setting up aprof aprof releases use the aprof src lin win yymmdd tar bz2 naming con vention e Select src for source lin for Linux or win for Windows binaries release e yymmdd is the release date 2 1 Obtaining aprof Download aprof from the ALMA intranet UOP directories Unarchive to a local directory e g C cygwin home user for Windows Cygwin users or home user for a Linux user 2 2 Setting up optional tools For using aprof a Linux or Windows installation is required For Windows Cygwin is suggested optional in order to significantly ease the use of aprof In any case standard Unix Linux tools are expected e bash e make e patch e gawk Boehm s garbage collector is also required but is included both in source and com piled form binary releases only within the thirdparty subdirectory For Windows e Go to htt2. required e app 0 nac which is the NAC representation of the application e app c which is a C implementation that is used in the context of aprof for reference input output data generation As a test vehicle the iterative implementation of a factorial computation will be used namely the fact application Thus the corresponding initial files are fact 0 nac and fact c The contents of fact 0 nac are as follows procedure fact in s32 n out s32 y localvar s32 res localvar s32 x localvar s32 i 10005 x lt mov n res lt ldc 1 i lt lde 1 D_1363 lt jmpun D_1362 res lt mul res i i lt add i 1 D_1363 lt jmpun D_1363 D_1362 D_1364 lt jmple i x D_1364 y lt mov res Since NAC is a relatively low level language a high level language frontend would have to be used for profiling larger applications In this sense fact c would serve as input to a C frontend producing NAC output The reference fact c has the following contents ifdef TEST include lt stdio h gt endif int fact int n int r s xy i x n res 1 for i 1 i lt x itt res res x i return res ifdef TES int main int i int result for i 0 i lt 13 i result fact i printf 08x 08x n i result a return 0 endif To automate the profiling process it is more suitable to use scripting The aprof distri
3. BB _temp_addr _temp_data lt add _temp_data 1 _BB lt store _temp_data _temp_addr x lt mov n res lt lde 1 i lt lde 1 D_1363 lt Jjmpun D_1362 _temp_addr lt ldc 1 _temp_data lt load _BB _temp_addr _temp_data lt add _temp_data 1 _BB lt store _temp_data _temp_addr res lt mul res i i lt add i 1 D_1363 lt Jjmpun D_1363 _temp_addr lt ldc 2 _temp_data lt load _BB _temp_addr _temp_data lt add _temp_data 1 _BB lt store _temp_data _temp_addr D_1362 D_1364 lt jmple i x D_1364 _temp_addr lt ldc 3 _temp_data lt load _BB _temp_addr _temp_data lt add _temp_data 1 _BB lt store _temp_data _temp_addr y lt mov res 6 6 Generation of the backend C files for the given NAC t u The profiling process is based on the generation of a compiled simulator for the NAC program This is accomplished with the use of the nac2c decompiler which is applied on the original form of the application fact 0 nac This is needed in order to extract the static schedule of the initial form of the application Either the sequential or the ASAP scheduler can be used which correspondingly reflect a sequential or intra block parallel abstract machine First a static scheduling extraction run of nac2c must be performed For enabling the sequential scheduler the following should be used SAPROFTOP bin nac2c ex force data types mit ansic mit cdfg sched
4. Internal construction of SSA Static Single Assignment form pseudo ssa Internal construction of local SSA like form use aycockhorspool Enables SSA construction using the Aycock Horspool algorithm keep ssa Does not perform out of SSA conversion and thus keeps PHI statements in the generated CDFGs phi bbs Enable the generation of BB arguments in phi NAC statements no phi bbs Disable the generation of BB arguments in phi statements default emit ansic Emit the equivalent ANSI C program after processing including SSA conversion if enabled emit cdfg Generate the Graphviz representations for all procedure CDFGs emit cfg Generate the Graphviz representations for all procedure CFGs emit cg Generate the Graphviz representation of the application call graph gcc Generate Makefile for GCC compilation default Ilvm_ Generate Makefile for LLVM compilation and or interpretation 4 2 fixnac usage The basic usage of fixnac follows the syntax fixnac exe options i input nac o output nac Additional fixes are applied to the instrumented input nac such as the additon of the declaration of the globalvar BB array for storing BB execution frequencies options is one or more of the following h Print this help decl bb array Declare the _BB globalvar init bb array Initialize the _BB globalvar to zeros Only in the effect if decl bb array has been defined max bbs lt num gt Specify the maximum number of basi
5. aprof user manual Title aprof ALMA profiler Author Nikolaos Kavvadias Contact nkavv uop gr Website http www nkavvadias com Release Date 06 May 2013 Version 0 4 0 Rev history v0 1 0 31 07 2012 Draft preliminary binary release of nac2c the compiled simulator of aprof v0 2 0 31 08 2012 Source release for the Ist increment of nac2c v0 3 0 30 11 2012 Binary release for the Ist draft release of aprof nac2c is now considered a component of aprof v0 4 0 06 05 2013 Added tutorial section in README 1 Introduction aprof ALMA profiler is a performance and resource utilization estimation tool For obtaining these measures aprof implements an abstract machine with unlimited re sources It accepts input specification in either the NAC N Address Code intermedi ate representation or ALMA IR ANSI C form aprof produces two basic outcomes a the number of dynamic abstract machine cycles and b basic block operation sched ule that indicates resource utilization for a given application aprof consists of the following components e libnac is the implementation of an API as a static library that allows for stor ing manipulating and examining NAC IR For instance the scheduler engines are considered as part of libnac As of the 0 3 0 release two schedulers are available for a sequential and an intra block parallel machine model nac2c
6. bution contains reference scripts for profiling Specifically the run aprof app sh can be used The rest of this guide will provide a detailed view of the approach taken by the aforementioned script in the form of a series of steps The SAPROF TOP environmental variable is the path to the top level directory of aprof 6 1 Generation of the reference test data Assuming that gcc is used as the host machine compiler the following prompt gener ates the corresponding executable gcc DTEST DDATAGEN Wall 02 o fact exe fact c Then the reference data can be generated Sfact exe gt amp fact_test_data txt The contents of fact_test_data txt are input and output values for n and y fact n in hexadecimal form 00000000 00000001 00000001 00000001 00000002 00000002 00000003 00000006 00000004 00000018 00000005 00000078 00000006 O000002d0 00000007 000013b0 00000008 00009d80 00000009 00058980 0000000a 00375f00 0000000b 02611500 0000000c 1lc8cfc00 0000000d 7328cc00 6 2 Create a working copy of the NAC representation of fact This can be accomplished by copying fact 0 nac to fact nac cp f fact 0 nac fact nac 6 3 Tracking the number of basic blocks in the unit The following bash script variable 10 num_bbs 0 is used for maintaining the number of basic blocks in the NAC translation unit 6 4 Generate the bbs txt file An AWK script countbbs awk is used for counting the basic blocks in the
7. c blocks in a program De fault 10000 4 3 meascycles usage The basic usage of meascycles follows the syntax meascycles exe input nac Itreads the input nac whichis assumed to be uninstrumented the input_prof txt profiling report file and the corresponding input_sched txt scheduling data file Then it reports the total number of dynamic abstract machine cycles in the following form Number of abstract machine cycles lld as a C based long long int 64 bit signed integer 4 4 TXL passes Executables generated by TXL passes source files share a common invocation style lt trans gt exe input nac q raw gt output nac This scheme applies for executables nacbbinscounters nacparser and nactoglobal 4 5 countbbs awk usage This AWK script generates a textual report named bbs t xt that stores the total num ber of basic blocks in the given NAC translation unit count bbs is invoked as follows gawk f APROFTOP countbbs awk lt S app nac gt bbs txt 5 Running the test suite The basic tests under the tests subdirectory can be exercised by running corresponding test script S cd SAPROFTOP cd tests run aprof sh Alternatively each application can be tested separately using the run aprof app sh script e g as follows for the case of the fibo benchmark run aprof app sh fibo By running a benchmark the following files can be generated if using the appropri ate options assumably for a be
8. e GCPATH in src Makefile linux accordingly Then run the build script from the top level subdirectory cd home user aprof build lin sh e If you want to recompile gc use the build a sh script The script should be changed accordingly comment and uncommented certain lines for selecting ei ther gc6 8 or gc 7 2alpha 6 or for enabling a Windows Cygwin or a Linux build build lin a sh B For Windows users e Windows XP SP2 or older untested on newer systems e Cygwin environment bash make gawk Cygwin can be installed via an auto mated web installer setup exe from http sources redhat com cygwin e TXL installation for Cygwin e Run the build script from the top level subdirectory of aprof cd home user aprof build sh e Similarly to the Windows case for rebuilding gc use the following build a sh 3 File listing The aprof distribution includes the following files Files and or directories denoted by a capital S are available in source releases of aprof Similarly a capital B denotes files directories present solely in binary releases S build lin sh S build lin a sh aprof Top level directory COPYRIGHT aprof binary or source code license S build sh Build script for aprof Windows S build a sh Build script for aprof and gc Windows Build script for aprof Linux Build script for aprof and gc Linux nactoglobal exe cygwin1 dll S clean sh Clea
9. enmacros h Makefile for Windows Cygwin Makefile for Linux Attributed graphs API Control and data flow analyses API includes SSA Call graph API mainly SSA Basic data structures and enums Emitters for graph representations ANSI C code generation routines General purpose C macros graph clh Graph manipulation API item clh CDFG Control Data Flow Graph items API list clh Doubly linked list and iterators API machine clh Machine paramteters for the NAC abstract machine lexer patch Patch for the NAC lexer lex nac c nac clh NAC N Address Code manipulation API nac lly Lexer and parser for the NAC programming language sched clh Scheduling naive ASAP API symtab clh Symbol table API utils clh Various utility functions aprof src nac2c nac2c directory Makefile Makefile for Windows Cygwin Makefile linux Makefile for Linux nac2c c Driver code and option parsing for nac2c aprof src prof prof directory Makefile Makefile for Windows Cygwin and Linux build sh Bash script for building the TXL applications countbbs awk meascycles c Counts the number of BBs in a NAC translation unit Counts the number of abstract machine cycles spent aprof tests Test suite directory O nac The aprof test suite Includes 30 applications each in the corresponding subdirectory binarysearch bitrev bub blesort cordic divider editdist fact factr fibo
10. entire translation unit This is performed by enumerating the labels in the NAC program since all NAC basic blocks have explicit labels gawk f S APROFTOP src prof countbbs awk lt fact nac gt bbs txt Then the bbs txt file is processed to get the number of basic blocks Process the bbs txt file bbsfile bbs txt while read r bbs do num_bbs bbs done lt S bbsfile A while loop is used in order to extract all the basic block counts in bbs txt in case of a multi translation unit application currently unsupported by most features of aprof 6 5 Instrumentation of the NAC file The nacinsbbcount TXL pass inserts profiling code for dynamic basic block count ing in NAC programs SAPROFTOP bin nacinsbbcount exe fact nac txlopts gt fact l nac A usual setup for TXL options is txlcopts q raw Then fixnac is invoked for adding bookkeeping code as for the declaration of the _BB global array its initialization and specifying the maximum number of basic blocks in the program APROFTOP bin fixnac ex decl bb array init bb array max bbs S num_bbs i fact 1l nac o fact 2 nac cp f fact 2 nac fact nac The resulting fact nac representation is as follows globalvar u64 _BB 4 0 0 0 0 procedure fact in s32 n out s32 y 11 localvar s32 res localvar s32 x localvar s32 i localvar u32 _temp_addr localvar u32 _temp_data 10005 _temp_addr lt ldc 0 _temp_data lt load _
11. fibor fir fixsqrt frac gcd knapsack loop1 mandel matmult min imal mips multiply perfect popcount sieve smithwater man sobel tak thornapprox xorshift yuv2rgba ao Reference C implementation for test suite used for gener ating reference data clean tests sh run aprof sh run aprof app sh Clean the debris in all tests subdirectories Run the entire test suite Run a single application from test suite thorn pgm PGM image required for running the thornapprox benchmark aprof thirdparty Third party source binaries directory B gc Garbage collector binaries for Windows Cygwin B gc linux Garbage collector binaries for Linux B gc mingw Garbage collector binaries for Windows MingW src Source code versions of the garbage collector 4 aprof tools usage 4 1 nac2c usage The basic usage of nac2c follows the syntax nac2c exe options input nac The translated C representation of input nac is produced in a series of out put files called input lt i gt _nac c separately for each NAC level procedure where input lt i gt is the name of the corresponding procedure Pre existing files are over written options is one or more of the following d Enable debug output force data types Force predefined data types as given in NAC code Essentially dis ables the effect of both interval analysis and the alternative of using the unknown data type na ssa
12. ining the dynamic basic block counts written in fact_prof txt with the static cycle estimates which are found in fact_sched txt APROFTOP bin meascycles exe fact nac As a result the profiling estimate is produced in the standard output For instance the sequential scheduler produces Number of abstract machine cycles 650 while the ASAP scheduler computes the following Number of abstract machine cycles 551 7 Contact You may contact me for further questions suggestions corrections at Nikolaos Kavvadias lt nkavv uop gr gt lt nikolaos kavvadias gmail com gt http www nkavvadias com Department of Computer Science and Technology University of Peloponnese Tripoli Greece 14
13. naive fact 0 nac 12 The ASAP scheduler is enabled as follows since it mandates at least pseudo SSA Static Single Assignment SAPROFTOP bin nac2c ex force data types ssa pseudo ssa emit ansic mit cdfg sched asap fact 0 nac Then nac2c generates a multitude of files which have been detailed in Section 5 A file named fact_sched txt is expected to be passed to a second run of aprof which is the profiling run cp f fact O0_sched txt fact_sched txt fact sched txt contains the estimated static cycles per basic block 3 4 2 2 aprof proceeds with the second run of nac2c SAPROFTOP bin nac2c ex force data types mit ansic emit cdfg prof fact nac 6 7 Optional step for generating CDFG views Optionally the Graphviz dot representation of each NAC procedure can be visual ized using the following snippet procfile procedure_names txt while read r app2 do echo Creating CDFG view for app2 dot Tpng O app2 dot done lt S procfile 6 8 Building and running the compiled simulator In this step the ansic mk generated Makefile must be run in order to build main exe which is the compiled simulator for the examined application fact make f ansic mk clean S make f ansic mk main This run produces fact_prof txt which contains dynamic basic block counts 14 91 105 14 13 6 9 Calculation of dynamic abstract machine cycles Finally meascycles is used for comb
14. nchmark called app comprising of proc procedures ansic mk Makefile for GCC or LLVM compilation bbs txt Total number of BBs in the NAC translation unit builtin_names txt Name listing of builtin black box functions proc dot proc dot png CDFG representation in Graphviz for procedure proc Visualization of the Graphviz CDFG for procedure proc proc_cfg dot proc_cfg dot png CFG representation in Graphviz for procedure proc Visualization of the Graphviz CFG for procedure proc app_cg dot app_cg dot png Call graph representation in Graphviz for app Visualization of the Graphviz call graph for app app nac Working NAC representation of the application app exe Executable generated by the C implementation of app app_test_data txt Reference test data generated by app exe app_prof txt Basic block profiling report app_sched txt Scheduling report number of static cycles per BB main c Generated C code containing the main function main h Header interface file for the generated files proc_nac c Backend C code generated from the corresponding NAC procedure_names txt Name listing of the procedures used in app 6 Step by step guide to profiling This section provides detailed information on the actual process of profiling First in order to profile an application which is assumed to be contained in a single NAC translation unit two files are
15. ns up the bin and src subdirectories env sh Script to setup the environment B aprof bin Binaries directory fixnac exe fixnac executable for either Windows or Linux meascycles exe meascycles executable for either Windows or Linux nac2c exe nac2c executable for either Windows or Linux nacbbinscount exe nacinsbbcount exec for either Windows or Linux nacparser exe nacparser executable for either Windows or Linux nactoglobal executable for either Windows or Linux Cygwin API DLL not required with a Cygwin setup nac refman txt nac refman html nac refman pdf aprof doc Documentation README This file README html HTML version of README README pdf PDF version of README Reference manual for the NAC programming language HTML version of the above PDF version of the above nactoglobal txl S aprof src Main source directory aprof src instrument instrument directory Makefile Makefile for Windows Cygwin and Linux build sh Bash script for building the TXL applications fixnac c Applies additional fixes to an instrumented NAC file nac Grm TXL grammar for NAC nacinsbbcount txl Inserts basic block counters in NAC programs nacparser txl NAC parser and pretty printer Moves all declarations to the earliest possible site aprof src libnac libnac directory Makefile Makefile linux attrgraph clh cdfa clh cga clh datastructs h emit clh genansic clh g
16. p sources redhat com cygwin e Download the automated web installer setup exe e Copy it to an empty local directory e g C temp cygwin e Click setup exe e Select Install from the Internet Make sure to select make since it might be disabled in the preselection Cygwin will then be setup in the C cygwin directory of your Windows OS For Linux e Any recent Linux distribution should do try using Ubuntu 11 10 2 3 aprof setup There is no actual installation procedure the user should just unzip the aprof lin win yymmdd tar bz2 binary release archive to a local directory Usual choices include C cygwin home user for Windows no Cygwin users and home user for Windows Cygwin Linux users where user is the name of the current user Then change directory to home user aprof On Cygwin for instance type cd home user aprof Set up the APROF TOP environmental variable source env sh The location of the garbage collector is adjusted accordingly in the corresponding makefiles You may add the aprof bin directory to your path export PATH SAPROFTOP bin S PATH 2 4 Building from sources This subsection is relevant only to the source releases of aprof aprof src yymmdd tar bz2 To build aprof from sources the following are required A For Linux users e A typical Linux installation bash make gawk e The TXL compiler from http www txl ca e g version 10 6 e Incase you want to use your system s gc chang
Download Pdf Manuals
Related Search