= Dhrystone Benchmark: Rationale for Version 2 and Measurement Rules =
Dhrystone Benchmark: Rationale for Version 2 and Measurement RulesReinhold P. Weicker<br>Siemens AG, E STE 35<br>Postfach 3240<br>D-8520 Erlangen<br>Germany (West)
<br>
Reinhold P. Weicker Siemens AG, E STE 35 Postfach 3240 D-8520 Erlangen Germany (West)== <br>Why a Version 2 of Dhrystone? ==
The Dhrystone benchmark program [1] has become a popular benchmark for CPU/compiler performance measurement, in particular in the area of minicomputers, workstations, PC's and microprocesors. It apparently satisfies a need for an easy-to-use integer benchmark; it gives a first performance indication which is more meaningful than MIPS numbers which, in their literal meaning (million instructions per second), cannot be used across different instruction sets (e.g. RISC vs. CISC). With the increasing use of the benchmark, it seems necessary to reconsider the benchmark and to check whether it can still fulfill this function. Version 2 of Dhrystone is the result of such a re-evaluation, it has been made for two reasons:
a) Dhrystone has been published in Ada [1], and Versions in Ada, Pascal and C have been distributed by Reinhold Weicker via floppy disk. However, the version that was used most often for benchmarking has been the version made by Rick Richardson by another translation from the Ada version into the C programming language, this has been the version distributed via the UNIX network Usenet [2].
There is an obvious need for a common C version of Dhrystone, since C is at present the most popular system programming language for the class of systems (microcomputers, minicomputers, workstations) where Dhrystone is used most. There should be, as far as possible, only one C version of Dhrystone such that results can be compared without restrictions. In the past, the C versions distributed by Rick Richardson (Version 1.1) and by Reinhold Weicker had small (though not significant) differences.
1Together with the new C version, the Ada and Pascal versions have been updated as well. Why a Version 2 of Dhrystone?
The b) As far as it is possible without changes to the Dhrystone benchmark program [1] statistics,optimizing compilers should be prevented from removing significant statements. It has become a popular benchmark forCPU/compiler performance measurement, in particular turned out in the area ofminicomputers, workstations, PC's and microprocesors. It apparently satisfiesa need past that optimizing compilers suppressed code generation for an easy-to-use integer benchmarktoo many statements (by "dead code removal" or "dead variable it gives a first performanceindication which is more meaningful than MIPS numbers which, in their literalmeaning (million instructions per secondelimination"), cannot be used across differentinstruction sets (e.g. RISC vs. CISC). With the increasing use of thebenchmark, it seems necessary This has lead to reconsider the benchmark and to check whetherit can still fulfill this function. Version 2 danger that benchmarking results obtained by a naive application of Dhrystone is - without inspection of the result ofsuch a recode that was generated -evaluation, it has been made for two reasons:could become meaningless.
o Dhrystone The overall policiy for version 2 has been published that the distribution of statements, operand types and operand locality described in Ada [1]should remain unchanged as much as possible. (Very few changes were necessary; their impact should be negligible.) Also, the order of statements should remain unchanged. Although I am aware of some critical remarks on the benchmark - I agree with several of them - and Versions in Adaknow some suggestions for improvement, Pascal and C have been distributed I didn't want to change the benchmark into something different from what has become known as "Dhrystone"; the confusion generated by Reinhold Weicker via floppy disksuch a change would probably outweight the benefits. If I were to write a new benchmark program, I wouldn't give it the name "Dhrystone" since this denotes the program published in [1]. However, I do recognize the version need for a larger number of representative programs that was can be used most often as benchmarks; users should always be encouraged to use more than just one benchmark. The new versions (version 2.1 for benchmarking has been the version made by Rick Richardson by another translation from the Ada version into the C programming language, this has been the Pascal and Ada) will be distributed as widely as possible. (Version 2.1 differs from version 2.0 distributed via the UNIX network Network Usenet [in March 1988 only in a few corrections for minor deficiencies found by users of version 2].0.) Readers who want to use the benchmark for their own measurements can obtain a copy in machine-readable form on floppy disk (MS-DOS or XENIX format) from the author.
There is an obvious need for a common C version == <br>Overall Characteristics of Dhrystone, since C is at present the most popular system programming language for the class of systems (microcomputers, minicomputers, workstations) where Dhrystone is used most. There should be, as far as possible, only one C version of Dhrystone such that results can be compared without restrictions. In the past, Version 2 the C versions distributed by Rick Richardson (Version 1.1) and by Reinhold Weicker had small (though not significant) differences.==
Together with In general, version 2 follows - in the parts that are significant for performance measurement, i.e. within the measurement loop - the published (Ada) version and the new C versions previously distributed. Where the versions distributed by Rick Richardson [2] and Reinhold Weicker have been different, it follows the versiondistributed by Reinhold Weicker. (However, the differences have been so small that their impact on execution time in all likelihood has been negligible.) The initialization and UNIX instrumentation part - which had been omitted in [1] - follows mostly the ideas of Rick Richardson [2]. However, any changes in the Ada initialization part and Pascal versions in the printing of the result have no impact on performance measurement since they are outside the measaurement loop. As a concession to older compilers, names have been updated as wellmade unique within the first 8 characters for the C version.
o As far as The original publication of Dhrystone did not contain any statements for time measurement since they are necessarily system-dependent. However, it turned out that it is possible without changes not enough just to inclose the main procedure of Dhrystone statisticsin a loop and to measure the execution time. If the variables that are computed are not used somehow, optimizing compilers should be prevented from removing significant statements. It has turned out in there is the past danger that optimizing compilers suppressed the compiler considers them as "dead variables" and suppresses code generation for too many a part of the statements (by . Therefore in version 2 all variables of "dead code removalmain" or "dead variable elimination")are printed at the end of the program. This has lead to the danger that benchmarking results obtained by a naive application of Dhrystone - without inspection also permits some plausibility control for correct execution of the code that was generated - could become meaninglessbenchmark.
At several places in the benchmark, code has been added, but only in branches that are not executed. The overall policiy for version 2 intention is that optimizing compilers should be prevented from moving code out of the measurement loop, or from removing code altogether. Statements that are executed have been changed in very few places only. In these cases, only the role of some operands has been changed, and it was made sure that the numbers defining the "Dhrystone distribution" (distribution ofstatements, operand types and operand locality described in [1] should remainunchanged ) still hold as much as possible. (Very few changes were necessary; their impactshould be negligible.) AlsoExcept for sophisticated optimizing compilers, the order of statements should remain unchanged.Although I am aware of some critical remarks on the benchmark - I agree withseveral of them - and know some suggestions execution times for improvement, I didn't want tochange the benchmark into something different from what has become known as"Dhrystone"; the confusion generated by such a change would probably outweightthe benefitsversion 2. If I were to write a new benchmark program, I wouldn't give itthe name "Dhrystone" since this denotes the program published in [1].However, I do recognize should be the need same as for a larger number of representativeprograms that can be used as benchmarks; users should always be encouraged touse more than just one benchmarkprevious versions.
The new versions (version 2.1 for C, Pascal Because of the self-imposed limitation that the order and Ada) will distribution of the executed statements should not be distributed aswidely as possiblechanged, there are still cases where optimizing compilers may not generate code for some statements. (Version 2.1 differs from version 2.0 distributed via theUNIX Network Usenet in March 1988 only in To a few corrections certain degree, this is unavoidable for minordeficiencies found by users small synthetic benchmarks. Users of version 2.0.) Readers who want to use thebenchmark are advised to check code listings whether code is generated for their own measurements can obtain a copy in machine-readableform on floppy disk (MS-DOS or XENIX format) from the authorall statements of Dhrystone.
Contrary to the suggestion in the published paper and its realization in the versions previously distributed, no attempt has been made to subtract the time for the measurement loop overhead. (This calculation has proven difficult to implement in a correct way, and its omission makes the program simpler.) However, since the loop check is now part of the benchmark, this does have an impact - though a very minor one - on the distribution statistics which have been updated for this version.
2. == <br>Discussion of Individual Changes Overall Characteristics of Version 2==
In generalthis section, version 2 follows - in the parts all changes are described that are significant forperformance measurement, i.e. within affect the measurement loop - the published(Ada) version and the C versions previously distributed. Where the versionsdistributed by Rick Richardson [2] and Reinhold Weicker have been different,it follows the version distributed by Reinhold Weicker. (However, thedifferences have been so small that their impact on execution time in alllikelihood has been negligible.) The initialization and UNIX instrumentationpart - which had been omitted in [1] - follows mostly the ideas are not just renamings of RickRichardson [2]variables. However, any changes in All remarks refer to the initialization part and in C version; theprinting of the result have no impact on performance measurement since theyare outside the measaurement loop. As a concession to older compilers, namesother language versions have been made unique within the first 8 characters for the C versionupdated similarly.
The original publication of Dhrystone did not contain any statements for timemeasurement since they are necessarily system-dependent. However, it turnedout that it is not enough just In addition to inclose adding the main procedure of Dhrystone in ameasurement loop and to measure the execution time. If the variables that are computedare not used somehowprintout statements, there is the danger that the compiler considers them as"dead variables" and suppresses code generation for a part of the statements.Therefore in version 2 all variables of "main" are printed changes have been made at the end of theprogram. This also permits some plausibility control for correct execution ofthe benchmark.following places:
At several places in the benchmark*In procedure "main", code has three statements have been added, but only in branchesthat are not executed. The intention is that optimizing compilers should beprevented from moving code out of the measurement loop, or from removing codealtogether. Statements that are non-executed have been changed in very few placesonly. In these cases, only the role "then" part of some operands has been changed, and itwas made sure that the numbers defining the "Dhrystone distribution"statement<pre>if (distribution of statementsEnum_Loc == Func_1 (Ch_Index, operand types and locality'C')) still hold as much aspossible. Except for sophisticated optimizing compilers, execution times forversion 2.1 should be the same as for previous versions.
Because of the self-imposed limitation that the order and distribution of theexecuted statements should not be changed, there they are still cases whereoptimizing compilers may not generate code for some statements. To a certaindegree, this is unavoidable for small synthetic benchmarks. Users of thebenchmark are advised to check code listings whether code is generated for allstatements of Dhrystone.
Contrary to the suggestion in the published paper and its realization in strcpy (Str_2_Loc, "DHRYSTONE PROGRAM, 3'RD STRING");Int_2_Loc = Run_Index;Int_Glob = Run_Index;</pre> The string assignment prevents movement of theversions previously distributed, no attempt has been made preceding assignment to subtract the timefor Str_2_Loc (5'th statement of "main") out of the measurement loop overhead. (This calculation has proven difficult toimplement in a correct wayprobably will not happen for the C version, but it did happen with another language and its omission makes the program simplercompiler.)HoweverThe assignment to Int_2_Loc prevents value propagation for Int_2_Loc, since and the assignment to Int_Glob makes the loop check is now part value of Int_Glob possibly dependent from the benchmark, this does have animpact - though a very minor one - on the distribution statistics which havebeen updated for this versionvalue of Run_Index.
*In the three arithmetic computations at the end of the measurement loop in "main ", the role of some variables has been exchanged, to prevent the division from just cancelling out the multiplication as it was in [1]. A very smart compiler might have recognized this and suppressed code generation for the division.
*For Proc_2, no code has been changed, but the values of the actual parameter have changed due to changes in "main".
*In Proc_4, the second assignment has been changed from <br>
<pre>Bool_Loc = Bool_Loc | Bool_Glob;
3. Discussion of Individual Changesto
In this section, all changes are described that affect the measurement loopBool_Glob = Bool_Loc | Bool_Glob;</pre> and that are not just renamings It now assigns a value to a global variable instead of variables. All remarks refer to the Cversiona local variable (Bool_Loc); the other language versions have been updated similarlyBool_Loc would be a "dead variable" which is not used afterwards.
*In addition to adding the measurement loop and the printout statementsFunc_1,changes have been made at the following places:statement
o In procedure "main", three statements have been added in the non-executed "then" part of the statementCh_1_Glob = Ch_1_Loc;
was added in the non-executed "else" part of the "if (Enum_Loc == Func_1 (Ch_Index" statement, 'C'))to prevent the suppression of code generation for the assignment to Ch_1_Loc.
they are*In Func_2, the second character comparison statement has been changed to
strcpy if (Str_2_Loc, "DHRYSTONE PROGRAM, 3Ch_Loc == 'R'RD STRING"); Int_2_Loc = Run_Index; Int_Glob = Run_Index;
The string assignment prevents movement of the preceding assignment to Str_2_Loc (5'th statement R' instead of "main"'X') out of the measurement loop (This probably will not happen for the C version, but it did happen because a comparison with another language and compiler.) The assignment to Int_2_Loc prevents value propagation for Int_2_Loc, and 'X' is implied in the assignment to Int_Glob makes the value of Int_Glob possibly dependent from the value of Run_Indexpreceding "if" statement.
o In the three arithmetic computations at the end of the measurement loop Also in "main ", the role of some variables has been exchangedFunc_2, to prevent the division from just cancelling out the multiplication as it was in [1]. A very smart compiler might have recognized this and suppressed code generation for the division.statement
o For Proc_2, no code has been changed, but the values of the actual parameter have changed due to changes in "main". Int_Glob = Int_Loc;
o In Proc_4, the second assignment has been changed added in the non-executed part of the last "if" statement, in order to prevent Int_Loc frombecoming a dead variable.
Bool_Loc = Bool_Loc | Bool_Glob;*In Func_3, a non-executed "else" part has been added to the "if" statement
toWhile the program would not be incorrect without this "else" part, it is considered bad programming practice if a function can be left without a return value.
Bool_Glob = Bool_Loc | Bool_GlobTo compensate for this change, the (non-executed) "else" part in the "if" statement of Proc_3 was removed.
It now assigns a value to a global variable instead The distribution statistics have been changed only by the addition of a the measurement loop iteration (1 additional statement, 4 additional local variable integer operands) and by the change in Proc_4 (Bool_Locone operand changed from local to global); Bool_Loc would be a "dead variable" which is not used afterwards. The distribution statistics in the comment headers have been updated accordingly.
o In Func_1, the statement== <br>String Operations ==
Ch_1_Glob = Ch_1_Loc;The string operations (string assignment and string comparison) have not been changed, to keep the program consistent with the original version.
There has been some concern that the string operations are over-represented in the program, and that execution time is dominated by these operations. This was added true in particular when optimizing compilers removed too much code in the non-executed "else" main part of the "if" statementprogram, to prevent the suppression of code generation for the assignment to Ch_1_Locthis should have been mitigated in version 2.
o It should be noted that this is a language-dependent issue: Dhrystone was first published in Ada, and with Ada or Pascal semantics, the time spent in the string operations is, at least in all implementations known to me, considerably smaller. In Func_2Ada and Pascal, assignment and comparison of strings are operators defined in the second character comparison statement language, and the upper bounds of the strings occuring in Dhrystone are part of the type information known at compilation time. The compilers can therefore generate efficient inline code. In C, string assignemt and comparisons are not part of the language, so the string operations must be expressed in terms of the C library functions "strcpy" and "strcmp". (ANSI C allows an implementation to use inline code for these functions.) In addition to the overhead caused by additional function calls, these functions are defined for null-terminated strings where the length of the strings is not known at compilation time; the function has been changed tocheck every byte for the termination condition (the null byte).
if Obviously, a C library which includes efficiently coded "strcpy" and "strcmp" functions helps to obtain good Dhrystone results. However, I don't think that this is unfair since string functions do occur quite frequently in real programs (Ch_Loc == 'R'editors, command interpreters, etc.). If the strings functions are implemented efficiently, this helps real programs as well as benchmark programs.
I admit that the string comparison in Dhrystone terminates later ('R' instead of 'X'after scanning 20 characters) because a comparison than most string comparisons in real programs. For consistency with the original benchmark, I didn'X' is implied in t change the preceding "if" statementprogram despite this weakness.
Also in Func_2, the statement== <br>Intended Use of Dhrystone ==
Int_Glob = Int_Loc;When Dhrystone is used, the following "ground rules" apply:
has been added in the non-executed part of the last "if" statement, in order to prevent Int_Loc from becoming a dead variable.*Separate compilation (Ada and C versions)
o In Func_3As mentioned in [1], a non-executed "else" part has been added Dhrystone was written to reflect actual programming practice in systems programming. The division into several compilation units (5 in the Ada version, 2 in the C version) is intended, as is the "if" statementdistribution of inter-module and intra-module subprogram calls. While Although on many systems there will be no difference in execution time to a Dhrystone version where all compilation units are merged into one file, the program would not rule is that separate compilation should be incorrect without this "else" part, it used. The intention is considered bad that real<br>programming practice if a function can , where programs consist of several independently compiled units, should be reflected. This also has implies that the compiler, while compiling one unit, has no information about the use of variables, register allocation etc. occuring in other compilation units. Although in real life compilation units will probably be left without a return valuelarger, the intention is that these effects of separate compilation are modeled in Dhrystone.
To compensate for this changeA few language systems have post-linkage optimization available (e.g., the (non-executedfinal register allocation is performed after linkage) "else" part in the "if" statement of Proc_3 was removed. <br>
The distribution statistics have been changed only by the addition of themeasurement loop iteration This is a borderline case: Post-linkage optimization involves additional program preparation time (1 additional statement, 4 additional local integeroperands) and by the change although not as much as compilation in Proc_4 (one operand changed from local toglobalunit)which may prevent its<br>general use in practical programming. The distribution statistics in I think that since it defeats the comment headers have been updatedaccordinglyintentions given above, it should not be used for Dhrystone.
Unfortunately, ISO/ANSI Pascal does not contain language features for separate compilation. Although most commercial Pascal compilers provide separate compilation in some way, we cannot use it for Dhrystone since such a version would not be portable. Therefore, no attempt has been made to provide a Pascal version with several compilation units.
4. String Operations*No procedure merging
The string operations Although Dhrystone contains some very short procedures where execution would benefit from procedure merging (string assignment and string comparisoninlining, macro expansion of procedures) have , procedure merging is not beenchanged, to keep be used. The reason is that the percentage of procedure and function calls is part of the program consistent with "Dhrystone distribution" of statements contained in [1]. This restriction does not hold for the original string functions of the C versionsince ANSI C allows an implementation to use inline code for these functions.
There has been some concern that the string operations *Other optimizations are over-represented inthe programallowed, and that execution time is dominated by these operations. Thiswas true in particular when optimizing compilers removed too much code in themain part of the program, this but they should have been mitigated in version 2.be indicated
It should be noted that this is a language-dependent issue: Dhrystone wasfirst published in Ada, often hard to draw an exact line between "normal code generation" and with Ada or Pascal semantics, the time spent "optimization" inthe string compilers: Some compilers perform operations by default that are invoked in other compilers only when optimization isexplicitly requested. Also, at least we cannot avoid that in all implementations known benchmarking people try to me,considerably smallerachieve results that look as good as possible. In Ada and PascalTherefore, assignment and comparison of stringsoptimizations performed by compilers - other than those listed above - are operators defined in the language, and the upper bounds of the stringsoccuring in not forbidden when Dhrystone execution times are part of the type information known at compilationtimemeasured. The compilers can therefore generate efficient inline codeDhrystone is not intended to be non-optimizable but is intended to be similarly optimizable as normal programs. In CFor example,string assignemt and comparisons there are not part of the languageseveral places in Dhrystone where performance benefits from optimizations like common subexpression elimination, so the stringoperations must be expressed in terms of the C library functions "strcpy" and"strcmp"value propagation etc. (ANSI C allows an implementation to use inline code for , but normal programs usually also benefit from thesefunctionsoptimizations.) In addition Therefore, no effort was made to the overhead caused by additional function callsartificially prevent such optimizations. However, measurement reports should indicate which compiler optimization levels have been used,these functions are defined and reporting results with different levels of compiler optimization for null-terminated strings where the length ofthe strings same hardware is not known at compilation time; the function has to check everybyte for the termination condition (the null byte)encouraged.
Obviously, a C library which includes efficiently coded "strcpy" and *Default results are those without "strcmpregister"functions helps to obtain good Dhrystone results. However, I don't think thatthis is unfair since string functions do occur quite frequently in realprograms declarations (editors, command interpreters, etc.C version). If the strings functions areimplemented efficiently, this helps real programs as well as benchmarkprograms.
I admit that When Dhrystone results are quoted without additional qualification, they should be understood as results obtained without use of the string comparison in Dhrystone terminates later "register" attribute. Good compilers should be able to make good use of registers even without explicit register declarations (afterscanning 20 characters[3], p. 193) than most string comparisons in real programs. Forconsistency with the original benchmark, I didn't change the program despitethis weakness.
Of course, for experimental purposes, post-linkage optimization, procedure merging and/or compilation in one unit can be done to determine their effects. However, Dhrystone numbers obtained under these conditions should be explicitly marked as such; "normal" Dhrystone results should be understood as results obtained following the ground rules listed above.
5In any case, for serious performance evaluation, users are advised to ask for code listings and to check them carefully. In this way, when results for different systems are compared, the reader can get a feeling how much performance difference is due to compiler optimization and how much is due to hardware speed. Intended Use of Dhrystone
When Dhrystone is used, the following "ground rules" apply:== <br>Acknowledgements ==
o Separate compilation The C version 2.1 of Dhrystone has been developed in cooperation with Rick Richardson (Tinton Falls, NJ), it incorporates many ideas from the "Version 1.1" distributed previously by him over the UNIX network Usenet. Through his activity with Usenet, Rick Richardson has made a very valuable contribution to the dissemination of the benchmark. I also thank Chaim Benedelac (Ada National Semiconductor), David Ditzel (SUN), Earl Killian and C John Mashey (MIPS), Alan Smith and Rafael Saavedra-Barrera (UC at Berkeley) for their help with comments on earlier versions)of the benchmark.
As mentioned in [1], Dhrystone was written == <br>Bibliography to reflect actual programming practice in systems programming. The division into several compilation units (5 in the Ada version, 2 in the C version) is intended, as is the distribution of inter-module and intra-module subprogram calls. Although on many systems there will be no difference in execution time to a Dhrystone version where all compilation units are merged into one file, the rule is that separate compilation should be used. The intention is that real programming practice, where programs consist of several independently compiled units, should be reflected. This also has implies that the compiler, while compiling one unit, has no information about the use of variables, register allocation etc. occuring in other compilation units. Although in real life compilation units will probably be larger, the intention is that these effects of separate compilation are modeled in Dhrystone.==
[1] Reinhold P. Weicker: Dhrystone: A few language systems have post-linkage optimization available (e.gSynthetic Systems Programming Benchmark.Communications of the ACM 27, final register allocation is performed after linkage)10 (Oct. This is a borderline case: Post-linkage optimization involves additional program preparation time (although not as much as compilation in one unit1984) which may prevent its general use in practical programming. I think that since it defeats the intentions given above, it should not be used for Dhrystone.1013-1030
Unfortunately, ISO/ANSI Pascal does not contain language features for separate compilation. Although most commercial Pascal compilers provide separate compilation in some way, we cannot use it for [2]Rick Richardson: Dhrystone since such a version would not be portable1. Therefore1 Benchmark Summary (and Program Text) Informal Distribution via "Usenet", no attempt has been made Last Version Known to provide a Pascal version with several compilation unitsme: Sept.21, 1987
o No procedure merging[3]Brian W. Kernighan and Dennis M. Ritchie: The C Programming Language. Prentice-Hall, Englewood Cliffs (NJ) 1978
Although Dhrystone contains some very short procedures where execution would benefit from procedure merging (inlining, macro expansion of procedures), procedure merging is not to be used. The reason is that the percentage of procedure and function calls is part of the "Dhrystone distribution" of statements contained in [1]. This restriction does not hold for the string functions of the C version since ANSI C allows an implementation to use inline code for these functions.<br>
o Other optimizations are allowed, but they should be indicated= IGEP Dhrystone 2.1 MIPS Test =
It is often hard to draw an exact line between "normal code generation" == Test Software and "optimization" in compilers: Some compilers perform operations by default that are invoked in other compilers only when optimization is explicitly requested. Also, we cannot avoid that in benchmarking people try to achieve results that look as good as possible. Therefore, optimizations performed by compilers - other than those listed above - are not forbidden when Dhrystone execution times are measured. Dhrystone is not intended to be non-optimizable but is intended to be similarly optimizable as normal programs. For example, there are several places in Dhrystone where performance benefits from optimizations like common subexpression elimination, value propagation etc., but normal programs usually also benefit from these optimizations. Therefore, no effort was made to artificially prevent such optimizations. However, measurement reports should indicate which compiler optimization levels have been used, and reporting results with different levels of compiler optimization for the same hardware is encouraged.==
o Default results are those without "register" declarations (C version)You can donwload the Dhrystone 2.1 MIPS test from [http://downloads.isee.biz/pub/files/dhrystone-2.1.tar.gz here].
When Dhrystone results are quoted without additional qualificationThe software it's compiled for OMAP / DM processors, they should inside be understood as results obtained without use of the "register" attribute. Good compilers should be able to make good use of registers even without explicit register declarations ([3], p. 193).available 2 executables:
Of course, for experimental purposes, post-linkage optimization, proceduremerging and/or compilation in one unit can be done to determine their effects.However, Dhrystone numbers obtained under these conditions should beexplicitly marked as such; "normal" Dhrystone results should be understood asresults obtained following the ground rules listed above.*gcc_dry2reg<br>
In any case, for serious performance evaluation, users are advised to ask forcode listings and to check them carefully. In this way, when results fordifferent systems are compared, the reader can get a feeling how muchperformance difference is due to compiler optimization and how much is due tohardware speed.<u>Tune Parameters:</u>
GCCOPTIM= -O
6Compiler: Linaro & Ubuntu <pre>$ arm-linux-gnueabi-gcc -vUsing built-in specs. AcknowledgementsCOLLECT_GCC=arm-linux-gnueabi-gccCOLLECT_LTO_WRAPPER=/usr/lib/gcc/arm-linux-gnueabi/4.5.2/lto-wrapperTarget: arm-linux-gnueabiConfigured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 4.5.2-8ubuntu3' --with-bugurl=file:///usr/share/doc/gcc-4.5/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.5 --enable-shared --enable-multiarch --with-multiarch-defaults=i386-linux-gnu --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/arm-linux-gnueabi/include/c++/4.5.2 --libdir=/usr/lib --enable-nls --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin --enable-gold --enable-ld=default --with-plugin-ld=ld.gold --enable-objc-gc --disable-sjlj-exceptions --with-arch=armv7-a --with-float=softfp --with-fpu=vfpv3-d16 --with-mode=thumb --disable-werror --enable-checking=release --program-prefix=arm-linux-gnueabi- --includedir=/usr/arm-linux-gnueabi/include --build=i686-linux-gnu --host=i686-linux-gnu --target=arm-linux-gnueabi --with-headers=/usr/arm-linux-gnueabi/include --with-libs=/usr/arm-linux-gnueabi/libThread model: posixgcc version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu3)</pre> *cc_dry2reg<br>
The C version 2.1 of Dhrystone has been developed in cooperation with RickRichardson (Tinton Falls, NJ), it incorporates many ideas from the "Version1.1" distributed previously by him over the UNIX network Usenet. Through hisactivity with Usenet, Rick Richardson has made a very valuable contribution tothe dissemination of the benchmark. I also thank Chaim Benedelac (NationalSemiconductor), David Ditzel (SUN), Earl Killian and John Mashey (MIPS), AlanSmith and Rafael Saavedra-Barrera (UC at Berkeley) for their help withcomments on earlier versions of the benchmark.<u>Tune Parameters:</u><br>
OPTIMIZE= -O4 -march=armv7-a -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp -fno-tree-vectorize
7Compiler: Linaro & Ubuntu<br> <pre>$ arm-linux-gnueabi-gcc -vUsing built-in specs. BibliographyCOLLECT_GCC=arm-linux-gnueabi-gccCOLLECT_LTO_WRAPPER=/usr/lib/gcc/arm-linux-gnueabi/4.5.2/lto-wrapperTarget: arm-linux-gnueabiConfigured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 4.5.2-8ubuntu3' --with-bugurl=file:///usr/share/doc/gcc-4.5/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.5 --enable-shared --enable-multiarch --with-multiarch-defaults=i386-linux-gnu --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/arm-linux-gnueabi/include/c++/4.5.2 --libdir=/usr/lib --enable-nls --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin --enable-gold --enable-ld=default --with-plugin-ld=ld.gold --enable-objc-gc --disable-sjlj-exceptions --with-arch=armv7-a --with-float=softfp --with-fpu=vfpv3-d16 --with-mode=thumb --disable-werror --enable-checking=release --program-prefix=arm-linux-gnueabi- --includedir=/usr/arm-linux-gnueabi/include --build=i686-linux-gnu --host=i686-linux-gnu --target=arm-linux-gnueabi --with-headers=/usr/arm-linux-gnueabi/include --with-libs=/usr/arm-linux-gnueabi/libThread model: posixgcc version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu3)</pre> Calculation References:<br>
*[1] Reinhold Phttp://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.faqs/4160. Weicker: html ARM Dhrystone: A Synthetic Systems Programming Benchmark. Communications of the ACM 27, 10 (Octreference]. 1984), 1013-1030
[2] Rick Richardson== Test Case 1: Dhrystone IGEPv2 Revision C- DM3730 @ 1.1 Benchmark Summary (and Program Text) Informal Distribution via "Usenet", Last Version Known to me: Sept. Ghz 21, 1987==
<u>'''Board'''</u>: IGEPv2 Revision C - RC5 - DM3730 @ 1Ghz - 512 MBytes LPDDR RAM + 512 MBytes OneNand Flash<br> <u>'''Operating System'''</u>: Linux version 2.6.35.13 (mcaro@manel-p) (gcc version 4.5.2 (Ubuntu/Linaro 4.5.2-8ubuntu3) ) #3 Fri Jun 10 19:58:16 CEST 2011 <u>'''Boot Software:'''</u> IGEP X-Loader 2.0.1-2<br> '''<u>a) Test case: Execution 10000000 loops (gcc_dry2)</u>'''<br> <u>Result</u><br> <pre>root@localhost:/tmp# ./gcc_dry2Dhrystone Benchmark, Version 2.1 (Language: C)Program compiled without 'register' attributePlease give the number of runs through the benchmark: 10000000Execution starts, 10000000 runs through DhrystoneExecution endsFinal values of the variables used in the benchmark:Int_Glob: 5 should be: 5Bool_Glob: 1 should be: 1Ch_1_Glob: A should be: ACh_2_Glob: B should be: BArr_1_Glob[8]: 7 should be: 7Arr_2_Glob[38][7]: 10000010 should be: Number_Of_Runs + 10Ptr_Glob-> Ptr_Comp: 13295624 should be: (implementation-dependent) Discr: 0 should be: 0 Enum_Comp: 2 should be: 2 Int_Comp: 17 should be: 17 Str_Comp: DHRYSTONE PROGRAM, SOME STRING should be: DHRYSTONE PROGRAM, SOME STRINGNext_Ptr_Glob-> Ptr_Comp: 13295624 should be: (implementation-dependent), same as above Discr: 0 should be: 0 Enum_Comp: 1 should be: 1 Int_Comp: 18 should be: 18 Str_Comp: DHRYSTONE PROGRAM, SOME STRING should be: DHRYSTONE PROGRAM, SOME STRINGInt_1_Loc: 5 should be: 5Int_2_Loc: 13 should be: 13Int_3_Loc: 7 should be: 7Enum_Loc: 1 should be: 1Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING should be: DHRYSTONE PROGRAM, 1'ST STRINGStr_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING should be: DHRYSTONE PROGRAM, 2'ND STRING Microseconds for one run through Dhrystone: Brian W0.4Dhrystones per Second: 2788671.0</pre> '''''DMIPS: 2788671. Kernighan and Dennis M0 / 1757 = 1587. Ritchie17''''' '''<u>b) Test case: Execution 10000000 loops (cc_dry2)</u>'''<br> <u>Result</u><br> <pre>root@localhost:/tmp# ./cc_dry2reg Dhrystone Benchmark, Version 2.1 (Language: The C Programming Language.) Program compiled with 'register' attribute Please give the number of runs through the benchmark: 10000000 Execution starts, 10000000 runs through DhrystoneExecution ends Final values of the variables used in the benchmark: Int_Glob: 5 should be: 5Bool_Glob: 1 should be: 1Ch_1_Glob: A should be: ACh_2_Glob: B should be: BArr_1_Glob[8]: 7 should be: 7Arr_2_Glob[8][7]: Prentice10000010 should be: Number_Of_Runs + 10Ptr_Glob-Hall> Ptr_Comp: 4169736 should be: (implementation-dependent) Discr: 0 should be: 0 Enum_Comp: 2 should be: 2 Int_Comp: 17 should be: 17 Str_Comp: DHRYSTONE PROGRAM, Englewood Cliffs SOME STRING should be: DHRYSTONE PROGRAM, SOME STRINGNext_Ptr_Glob-> Ptr_Comp: 4169736 should be: (NJimplementation-dependent) 1978, same as above Discr: 0 should be: 0 Enum_Comp: 1 should be: 1 Int_Comp: 18 should be: 18 Str_Comp: DHRYSTONE PROGRAM, SOME STRING should be: DHRYSTONE PROGRAM, SOME STRINGInt_1_Loc: 5 should be: 5Int_2_Loc: 13 should be: 13Int_3_Loc: 7 should be: 7Enum_Loc: 1 should be: 1Str_1_Loc: DHRYSTONE PROGRAM, 1'ST STRING should be: DHRYSTONE PROGRAM, 1'ST STRINGStr_2_Loc: DHRYSTONE PROGRAM, 2'ND STRING should be: DHRYSTONE PROGRAM, 2'ND STRING Microseconds for one run through Dhrystone: 0.3Dhrystones per Second: 3987539.0</pre> '''''DMIPS: 3987539.0 / 1757 = 2269.51'''''<br> [[Category:How_to_forge]][[Category:Software]][[Category:Software applications]]