A detailed assessment is made of the processing speed of over 100 different computers. This assessment is based upon over 20 benchmark programs which have been run to give more than 700 results. The large number of results have to be analysed using a simple mathematical model to allow detailed comparisons to be easily made.
From Brian Wichmann: Postscript, January 2004. The report was completed but never published. Due to the popularity of the previous report NAC62, it was decided to produce a commercial version. This included the addition of a very large amount of data collected by CCA. Technically, the work is very similar to NAC 62, but the information was re-worked to make is less demanding of the reader.
Clifford Nott did lots of work on the presentation of this material. At the time, the volume of this was too large for convenient word processing, so the appendices were produced separately, mainly by direct computer output (I suspect via paper tape!).
Just prior to the planned publication date, I produced a press release. This went to the Director of NPL, Dr Paul Dean who decided NPL should not publish it!
NPL eventually disposed of all the 300 or so copies in nice red ring files. A person in the publicity section worked out how to remove the gold lettering on the ring files, so those could at least be re-used. I suspect I have the only complete copy (with the acetate sheet to compare performance graphically).
At this point, in mid 1977, I stopped working on computer performance.
Research into the performance of computers has been undertaken at the National Physical Laboratory since J967. This has concentrated upon the use of high level languages, particularly ALGOL 60. Research reports and papers in journals have shown how compilers and computers can be assessed by a number of techniques. Recently it has proved possible to extend the methods to include a number of programming languages and over 100 different computers. It is therefore appropriate that the material should be presented to a wider audience so that the techniques can be applied more extensively.
Although every attempt has been made to ensure the correctness of the data here, no responsibility can be assumed for decisions based upon it. Hence in all important cases, the benchmark programs should be run and all results verified. Continual changes in software make it impossible to keep the performance information up to date. Any additional results from the benchmarks here or on new ones would be most welcome.
If assistance is required in making use of this data, or analysis of new data is required, then NPL can provide a consultancy service for a suitable charge.
One method of measuring computer power is to measure the times of individual programs on the machine in question. Such programs are called benchmarks. It is unwise to rely upon the performance of one benchmark because the relative power of machines varies significantly with the nature of the benchmark or with the work load undertaken.
One approach used to reduce this problem is to combine in one program a carefully adjusted mixture of the different types of work undertaken; this approach was developed in Curnow and Wichmann [3]. However, it has been noted that users still place undue reliance on the benchmark figure.
This can be overcome by demonstrating the variability of the benchmark data by correlating the results from a number of different benchmarks. This report describes a system to collect and correlate such data, and also presents the large number of results.
Most of the benchmarks are written in high level languages; this facilitates their specification and implementation on a large number of machines. An exception is the instruction mixes which are included because a large amount of information on them is available. They provide some indication of the effectiveness of the computer architecture and power independent of compiler software. By making comparisons with benchmarks written in the high level languages, the performance of the compilers can be measured. This is important with the increasing use of high level languages. The wide availability of FORTRAN makes it worthwhile to include the source listings in this guide; for the other languages, the references should be consulted.
It is hoped that the provision of this benchmark data will reduce the need to run further tests since this is an expensive activity.
The usefulness of this data base depends upon keeping it up-to-date and providing data on a wide variety of machines. We should welcome further data and would ask for a copy of any benchmark runs to be sent to NPL.
These benchmarks are restricted to processor limited programs. The reason for this is that peripheral programs are dependent on the configuration and operating system. The variations in time that would then arise cannot be handled by the current system.
This chapter introduces the basic concepts of performance measurement leading to Appendix B, which gives a performance range for a large number of computers together with the performance figures obtained by running benchmarks on them. Further details of the benchmarks, instruction mixes and the data base are given in later chapters.
A useful summary of the measurements is given in Appendix A, where the logarithmic scale gives the performance range of all the computers on only two pages.
One of the main difficulties in measuring computer performance is that no single machine can be compared with all other machines. Two different machines may perform equally well on one task, but are likely to perform differently on other tasks using different mixes of instructions. This is not surprising since there are many possible variations in machine architecture. Some operate only on complete words, others on bytes; the word lengths, and hence the numerical precisions, differ; some have floating point hardware; some have pipelining and instruction look-ahead, and so on. Obviously comparisons cannot be based on a single machine.
Other complications arise in the system software. Two machines may be of similar hardware power, but one may have a much more efficient operating system. Or its FORTRAN compiler may be better. These factors affect performance to the user.
A number of benchmarks are designed to measure the performance of a particular feature of machines, eg Gamma Test and Binomial both use floating point arithmetic, and IF Test uses conditional jump instructions. In general, this type of benchmark is not typical of any one practical workload.
Other benchmarks are written to measure the performance of a machine on a type of workload which is not an average mix of instructions, eg Synth FORD is written in FORTRAN and performs calculations using double precision. Hence, although it is not typical of average workloads, it may accurately reflect the requirements of some particular users.
A better idea of overall performance of a machine on a typical workload may be obtained by running a benchmark which has been designed to represent a more general mix of instructions. Examples of this type of benchmark are Synth FORS, giving a mix of FORTRAN statements and using single precision, and Synth ALGL which gives a general mix of ALGOL 60 statements.
A difficulty similar to the lack of commonality in machine performance is encountered in determining a unit of measurement. Clearly one cannot use any single machine instruction, although a unit based on machine instructions would seem to be the most promising choice. A solution is to use a weighted average of instructions in the form of an instruction mix constructed to reflect an average usage of individual instructions.
These ideas were partly formalized by J C Gibson who proposed a weighted average of instruction times based upon statistics gathered from the 650 and 704 computers [5]. Several different versions of the Gibson mix have been used. In this report, a scientific mix similar to Gibson's is used, together with an ADP mix and a process control mix. The weights assigned to the individual instructions for the mixes are given in THE INSTRUCTION MIXES section.
Machine performance, or the performance of other benchmarks, can now be expressed in terms of Gibson mix instructions. These can be regarded as average machine instructions, and the Gibson mix can be regarded as a benchmark.
It is emphasized that other instruction mixes exist and the choice of mixes based on the Gibson mix is an arbitrary one.
Appendix B gives the performances of various benchmarks (in common units of mix instructions per millisecond) for a large number of machines. It will be seen that some benchmarks have only been measured on certain machines and the table gives all of the available results.
A rough overall comparison of the power of machines can be obtained from the performance range, given for each machine. This is an average performance range obtained from the results of the benchmarks run on that machine. In determining this performance range, account is taken of how relevant each benchmark is to the machine. This is discussed further in THE DATA BASE section and THE OVERALL ANALYSIS section.
The figures in the table can give good indications of areas in which a machine performs poorly. For example, the 370/158 and the 1906S are of comparable overall power, both being approximately in the range 800 - 1350. While the 1906S performs better on most of the benchmarks, it is noticeably inferior on the Ackermann and Synth FORD tests.The Ackermann benchmark is a measure of the efficiency of executing procedure calls, and the better performance on the 370/158 is due to the use of non-standard software on that machine. Synth FORD tests the double precision facilities. The poor performance of the 1906S highlights the lack of double precision hardware on the machine used; however, a double precision hardware unit is available for the 1906S.
Information about the machines and the benchmarks is kept in a data base. This provides a convenient means of storage and updating using a computer. It also allows access to the data so that results can easily be recomputed when new information is added.
In the data base it is necessary to store information about (1) the machines used, (2) the programming language compilers and other relevant software used, (3) the application area of each benchmark and finally, (4) the benchmark data itself. Appendix F lists the data base in textual form with many explanatory comments.
A brief list of the benchmarks used is given below. Further details of the Mixes (1, 2 and 3) are given afterwards. The other benchmarks are described in the BENCHMARKS section, where the order differs from the list below to enable the Fortran benchmarks and their program listings to appear before those written in other Languages.
The benchmark, Dbl Fnctns has not been used in the analysis.
1 | Gibson Mix | represents the Gibson instruction mix |
2 | ADP Mix | represents the ADP instruction mix |
3 | Pr Con Mix | represents the Process Control instruction mix |
4 | Ackn ratio | represents the instruction mix used by the Ackermann benchmark |
5 | Synth FORS | uses an average mix of FORTRAN statements with single precision |
6 | Synth FORD | similar to Synth FORS but uses double precision |
7 | Gamma Test | FORTRAN program to test floating point |
8 | Bit Test | FORTRAN program to test core memory and access |
9 | Binomial | FORTRAN program to test floating point |
10 | IF Test | FORTRAN program to test conditional jump instructions |
11 | DOUBLE FUN | FORTRAN program to test double precision with the main intrinsic functions |
12 | Functions | FORTRAN program to test the coding of some standard mathematical functions |
13 | Synth ALGL | uses an average mix of ALGOL statements |
14 | ALGOL Mix | uses an average mix of ALGOL statements |
15 | Ackermann | ALGOL program to test calls of procedures |
16 | GAMM ALGOL | ALGOL program executing numerical loops |
17 | Chess Mate | ALGOL program executing non-numerical procedures |
18 | GAMM Asmbl | Machine code version of GAMM ALGOL |
19 | POWU | represents the Post Office Work Unit |
20 | GAMM F | FORTRAN version of GAMM ALGOL |
21 | GAMM FD | As GAMM F, but using double precision |
22 | Dbl Fnctns | As Functions, but using double precision |
As noted above, the Gibson mix of machine instructions gives a useful weighted average of instruction times. Several different versions of the Gibson mix have been used and hence it is important to ensure consistency when comparing two machines. Additional data from Carnegie-Mellon University (CMU) is available in [8] and is summarized below.
In this report, three mixes are used: a scientific mix very similar to Gibson's, an ADP Mix and a process oontrol mix. The individual weights are assigned according to the following table.
Machine | 650/704 | 3600 | DEC10 | Any | ||
---|---|---|---|---|---|---|
Source | Gibson | UMASS | CMU | This report | ||
Class | Scientific | ADP | Process Control | |||
Load, store | 31.2 | 30.0 | 42.4 | - | - | 44.9 |
FxP add/subtract | 6.1 | 1.2 | 12.4 | 33.0 | 31.0 | 0.0 |
FxP multiply | 0.6 | 0.1 | 1.1 | 0.6 | 1.3 | 1.7 |
FxP divide | 0.2 | 0.1 | 0.5 | 0.2 | 0.6 | 1.7 |
Compares | 3.8 | 1.2 | - | 4.0 | 6.2 | 0.0 |
Branches | 16.6 | 38.3 | 28.2 | 6.5 | 35.0 | 20.7 |
FlP add/subtract | 6.9 | 0.5 | 4.9 | 7.3 | 0.0 | 0.0 |
FlP multiply | 3.8 | 0.5 | 2.6 | 4.0 | 0.0 | 0.0 |
FlP divide | 1.5 | 0.2 | 1.1 | 1.6 | 0.0 | 0.0 |
Shift | 4.4 | 2.2 | 3.9 | 4.6 | 0.0 | 10.3 |
Logical | 1.6 | 0.5 | 1.0 | 1.7 | 5.4 | 6.9 |
Miscellaneous | 5.3 | 0.0 | 1.5 | - | - | - |
Indexing | 18.0 | 13.4 | - | 19.0 | - | 13.8 |
Fullword | - | 6.9 | - | - | - | - |
I/O control | - | 0.0 | 0.1 | - | - | - |
Register transfer | - | 5.0 | - | - | - | - |
User defined inst | - | - | 0.3 | - | - | - |
Transfer 8 chars | - | - | - | 17.5 | 20.5 | 0.0 |
UMASS = University of Massachusetts CMU = Carnegie Mellon University FxP = Fixed point FlP = Floating point
Entries in the table are percentages. A hyphen implies that the class does not appear in the relevant mix. For the mixes used in this report, the definitions of the various classes are given below. The data used, unless otherwise stated, is one binary word (for binary machines) or six decimal digits (for character machines). It is not necessary to check for overflow, save remainders or allow for rounding.
The three mixes used in this report contain a significant modification on the simple weighting. It is clear that a three-address machine allows algorithms to be coded in fewer instructions than an otherwise identical one-address machine. In fact, some hand coding has shown that a three-address machine requires roughly half as many instructions. This is reflected in the instruction mix by introducing an architectural factor k, as follows.
k=1 one-address machine, less than 8 programmable registers k=1.25 one-address machine, more than 8 registers or hardware stack k=1.5 two-address k=2.0 three-address
The final mix is expressed in instructions per millisecond as is given by the formula
Mix = k / (1000 Σ witi) instructions per millisecond
where the weights ( expressed as a fraction) and execution times for instruction i are wi and ti.
The advantage of this mix is that it is possible with a little desk calculation to obtain a reasonable crude estimate of the raw computing power of a processor. Its disadvantage is that with many machines having unusual architecture the method of calculating the mix is of dubious validity. On the more powerful machines with pipelining, slave memories or instruction look-ahead the technique cannot be applied with any confidence. Some machines, eg the IBM 360, have both binary and decimal arithmetic, but unfortunately the ADP mix makes no allowance for this.
The mix values are more significant when comparing machines of a similar architecture - provided the mix has been calculated on a consistent basis as in this report. The difficulties in interpretation mean that the mix should be calculated independently of the manufacturer, which is again the case for the mixes used here. In the rest of the report, reference to the Gibson Mix refers to the scientific mix in the table above with the architectural factor included.
Two programs have been written to produce an analysis of the benchmark data. The first one, described in this section, attempts to provide an overall view of the data. The aim is to produce a concise summary which necessarily omits some data of only secondary importance. The result of this analysis is in Appendix B. The second program is designed for interactive use and consequently is not described in this report.
In order to make the analysis, the following information was ignored: the characteristics relating the machines, the application areas of the benchmarks, the software used and any extra results that may be available for a benchmark on a particular machine. The benchmark data therefore consists of a list of machines and benchmark results. A relevance factor for each benchmark is used, and plays a critical part in the analysis. This factor is a weighting for each benchmark - the actual value is determined by the data itself as explained below. In essence, therefore, the data consists of a matrix of times for the benchmarks on the machines being considered. The matrix is sparse (about 35% occupied) since many benchmarks have only been run on a few machines. As further machines and benchmarks are added, the matrix is likely to become more sparse.
One problem with benchmark data is that of scaling. Since there is no absolute measurement for performance, each benchmark has its own units - instructions per millisecond, statements per second, microseconds in a loop etc. Reasonable inter-comparisons cannot then be made without resort to a desk calculator. The overall analysis program overcomes this by adjusting each scale to give as close a match as possible to the Gibson Mix scale. The Gibson Mix is not in any special position as far as the analysis itself is concerned, since the scaling is performed merely to print the analysis.
The mathematics behind the analysis uses a simple multiplicative model very similar to that introduced in [14]. The assumption is that the performance of any benchmark on a particular machine will be roughly equal to the product of two factors, one depending upon the benchmark and the other depending upon the machine. These two factors can be calculated by a least squares fitting method from the data matrix. Two improvements have been made upon the analysis used for the ALGOL statements. Firstly the sparse nature of the data has meant that the iterative method was inappropriate and hence it has been replaced. Secondly, a weighting scheme has been introduced to make use of the relevance factors, which increases the stability of the analysis when small changes are made to the less significant data.
The method of solving the fitting problem to find both the factors for the machines and benchmarks is as follows. The multiplicative equations are made additive by taking logarithms. The sum of the squares of the residuals which is to be minimized is a quadratic in the unknowns. On differentiation by the unknowns, one obtains linear equations which can be solved with the aid of classical linear algebra routines. No use is made of the sparseness of the matrices since the main store of the computer used was adequate to handle the full matrices. The mathematical equations themselves are developed in Appendix G.
Initially the relevance factors were set by intuition based upon the supposed importance of each benchmark. The natural question to ask was whether the analysis itself presented any indication of what weighting would be most appropriate. Fortunately such an indication is readily available. The analysis gives an overall indication of performance for each machine and the individual results reduced to a common scale (Appendix B). The degree to which each benchmark conforms to this consensus figure can be calculated. For instance, Synth FORS corresponds very closely to the overall indication of performance and is therefore given a high weight. On the other hand, Ackermann rarely gives good agreement with the overall figure and hence has a very low weight. The relevance factors are explained in more detail in Appendix D.
The relationships between the actual performance measurements and the values printed in the Appendices are given by the benchmark factors listed in Appendix D. The factor for the Gibson Mix is 1.0 because this scale is used as the base throughout the analysis. However, the ADP Mix values are typically larger, so that an ADP Mix value of 154 corresponds to a Gibson Mix value of 100.
To compare a number of computers using the data presented here, the procedure is as follows. Initially, one should consider the likely relevance of information on the execution rate of high-level language programs. Hence one's current system should be surveyed to see what proportion of time is spent executing programs written in high-level languages.
If the current system which one wishes to extend is disc bound or spends over half its time in supervisor state then the data here is unlikely to be of any assistance. On the other hand, if the current problem is to obtain more raw computing power for executing programs written in FORTRAN or ALGOL 60, then the data is highly relevant.
For each computer under consideration, the data presented in Appendix B should be carefully examined. The benchmarks thought likely to be most typical of the work should be listed. The performance range for each machine is used to obtain an upper and lower limit to its likely performance.
For the FORTRAN results, the data on different precisions should be assessed according to the importance, in one's own case, of DOUBLE length working (see below).
Overlapping ranges of power are very likely to be obtained when comparing two similar machines. In this case, one should see whether any additional data which may be relevant is listed in Appendix F.
When selecting a new computer, it often happens that one's final choice is between two machines. Where this is so, one can make a more detailed appraisal of their relative merits starting from the overall analysis provided in Appendix B. To do this, it is convenient to represent the primary performance data for both machines on a single sheet using a logarithmic scale. For each machine for which there is more than one item of data, a graph is included, in Appendix C, of the available data on the common scale derived from the overall analysis. The acetate sheet provided can be used to copy the pages relating to the two machines being compared on to one sheet. The resulting sheet will display the data for both machines thus:-
One knows that, on the basis of all the available data, the ratio of the machine speeds is given by the relative position of the two ranges (actually the ratio between the geometric mean of the limits of each range). This ratio is mirrored by Benchmark X, say, in our example (above). If, however, Benchmark Z is thought to be more typical of the workload undertaken, then machine B may be more powerful than the overall analysis would indicate. The converse is true if Benchmark Y is more typical of the application area. It is very unwise to readjust the scales for A and B on the basis of one benchmark because the overall analysis takes into account more information - in particular, benchmarks like U and V that have not been run on both computers.
The following diagram shows the comparisons between the 370/158 and the 1906S. The 1906S used had no double precision hardware, although this is available if required. The slightly greater power of the 1906S is shown by the nearly parallel and upward sloping lines for most of the benchmarks. The lack of double precision hardware is clearly reflected by the poor results for Synth FORD and DOUBLE FUN on the 1906S.
All the other FORTRAN programs give a consistent slope with the exception of Synth FORS. This is a little unexpected since Synth FORS should reflect typical FORTRAN usage. A likely explanation is that the optimizing FORTRAN compiler on the 1900 (XFEV) has more difficulty producing fast code due to lack of registers than the IBM FORTRAN H compiler on the 370.
The different slope of the Ackermann test is due to the large differences in the efficiency ot the software. The Ackermann test rarely agrees with the consensus of the other benchmarks and hence has a low weight in the overall analysis.
At this point one can examine the factors not used by the overall analysis, especially the availability of additional runs of a benchmark on one machine. Appendix F should be examined to see if the compilers used are comparable. Clearly one should not compare an optimizing compiler on one machine with a diagnostic compiler on another.
As an example to illustrate these points, suppose the two machines being compared are the 370/158 and the 370/168. Results for the 370/158 have been obtained from both the FORTRAN G and H compilers for Synth FORS and Synth FORD, but only from the G compiler for the other benchmarks. For the 370/168, results have been obtained from both compilers for all of the FORTRAN benchmarks.
Where more than one result is available, the overall analysis (Appendix B) gives that for the faster compiler, H, but the * symbol indicates that additional information is available - in this case, a result from the G compiler. This additional information is found in Appendix F under the appropriate benchmark; these values must be divided by the benchmark factor from Appendix D to convert to the common scale used in Appendices B, C and E.
If a diagram is drawn on the acetate sheet from the logarithmic scale machine profiles for the two machines (Appendix C), lines can be drawn joining the common benchmarks, as in the above diagram. The new diagram is shown below.
Since Synth FORS and Bit Test have the same value for the 370/168, one can infer that the values for the 370/158 would also have been equal if the H compiler had been used for Bit Test. The slope of the line joining the Bit Test points would change to that of the line joining the Synth FORS points. If a similar change of slope is now applied to the Binomial" line, an estimate, E, can be made of the speed of this benchmark on the 370/ 158 if the H compiler had been used. This estimate is also near the value for Synth FORS on the 370/158. The estimate can be checked from the figures given in Appendix F since the ratio of the two benchmark values should be the same for each machine.
370/158 370/168 Synth FORS (H) 824 2439 Binomial (H) E 18880 E = (18880 × 824)/2439 = 6378
On dividing by 4.33 (the benchmark factor for Binomial), the value of 1473 is obtained on the common scale.
Any compiling options should also be considered. It may then be clear that not enough comparable data is available and suitable computer runs should be made. (Results of such runs would be a welcome addition to the database and would be appreciated by NPL.)
Some computers have options which significantly affect processing speeds. Dramatic improvements can be obtained by adding a floating point processor when performing scientific calculations. Such a major change is best thought of as producing a distinct machine and all (or almost all) benchmarks should be rerun.
The provision of a hardware double length processing unit clearly only affects a specialist workload. The effectiveness of such a unit can be reasonably judged from the speed improvement of Synth FORD. If this benchmark runs 30% faster, and programs using double length calculations are thought to constitute about a third of the processing load, then a unit should cost no more than 10% of the system to be cost-effective.
Small variations in processing speeds can occur for a variety of reasons. In some cases, store access speeds depend slightly on the amount of store connected. Many processors have speed adjustments which, although intended for marginal testing, are sometimes varied to overcome some reliability problems.
If one is comparing two machines for scientific work which have different word lengths, then one must decide how to make a reasonable comparison. Assuming the times for both Synth FORS and Synth FORD are available, then a weighting is required which reflects the likely use of the two lengths on the two machines. Some idea of the likely weights can be gained from the work of the Numerical Algorithms Group in producing high quality scientific software [4]. They do not use 32-bit arithmetic at all, and use double length precision to a significant but minor extent on 48-bit machines.
Any high level language program that has been run on two or more machines could, in principle, provide information for the system. To be useful, however, the program must be capable of being run on a variety of computers and hence the program should be portable. To permit verification of the results it is essential that the program listing should be available, together with a note of any modifications that were necessary to run on any system.
To assist in handling the raw data, a non-computer-based filing system is used. This file also contains some confidential data that cannot be added to the computer data base. In many cases, the source of the data is confidential but this does not itself necessarily prevent the data itself from being inserted.
There is one important exception to the use of high level language data and that is the instruction mixes. The reason for their insertion is that data calculated in a consistent manner independently of the manufacturers was available for a large number of computers. Other performance data of this nature could be added, but the difficulty of independent verification militates against this.
Although in theory benchmark data can be added for any two machines, in order to ensure good utilization of store in the analysis programs four results on one benchmark is a practical minimum. It is reasonable to add a new computer only if there is sufficient information to justify its inclusion.
Several problems arise with inserting information. The first arises from compiling options, different compilers and language dialects. If more than one result is available for a benchmark on one machine owing to reruns with different software, then the overall analysis program takes the first result listed and ignores the rest. (The existence of additional data is shown by a * following the data analysed.) Hence it is important that the first result should be chosen consistently. The algorithm used is that the fastest result available using the strict language is chosen. Hence if an ALGOL W result is available but faster than the ALGOL 60 one, the ALGOL W one will not appear first. If no ALGOL 60 result was available, then the ALGOL W result cannot be added to the system.
Unavoidable problems arise with results using different software on similar machines. If results on one 360 are available with FORTRAN H but on another 360 only with FORTRAN G, then the apparent relative speeds of the machines will not reflect the hardware speeds. They may, however, reflect the processing speeds of FORTRAN on two actual installations if compiling times and other overheads are ignored.
A further problem which affects all the data to a certain extent is the length of the arithmetic operations. Is it fair to compare a minicomputer using 16-bit integers with a main frame machine using 32 bits or more? The problem with floating point arithmetic is more acute since many programmers believe 32 bits to be inadequate to represent a floating point number upon which a large amount of computation will be performed. Double length on 360 FORTRAN is virtually the same as single length on CDC 6000 series. Since there is nearly a continuous range of representations used for floating point, any decision will cause anomalies. The decision taken in this data base is to rely upon the compiler to interpret the requirements of the program. Hence single length on 360 is six hexadecimal places, which is often inadequate, whereas double on the 6600 is 120 bits which would be rarely needed.
Since ALGOL 60 does not have double length, the only way the length of floating point numbers can be varied is by a compiling option or language extension. Some data on the use of double length in ALGOL is available for the Synthetic Benchmark program.
To obtain an accurate assessment of a new computer, tbe timing data should be added to the data base and the analysis repeated. However, a good approximation can be made by reducing the data to a common scale as determined by the benchmark factors listed in Appendix D.
The procedure is as follows:
A diagram similar to the machine profiles in Appendix C can now be produced enabling direct comparison to be made with the other computers listed in this guide. A more detailed assessment can now be made by varying compiling options.
The system provides a method whereby processor-limited performance data may be collected and correlated. Problems which arise, such as those with the accuracy of numerical calculations, are not due to the system itself but inherent to any comparative work.
A major requirement is to broaden the application area of the data. Benchmarks for COBOL, PL/I and real-time (in CORAL 66 and RTL/2) are needed.
One difficulty with this comparative work is attempting to determine its relevance in any situation. Almost all the benchmarks measure raw computing power, which is not necessarily the most important aspect of the system. Even if processor power is important it is not always being absorbed by high level language programs. It has been reported that about 65% of the available processor capacity can be used by the supervisor in some large operating systems running multi-access facilities. Benchmarks can be used to determine the overhead on user requests in a large system, but unfortunately the interaction between several users typically makes the overhead non-linear in the number of tasks (or user jobs, multi-access terminals etc) being handled. In a procurement situation these problems can be overcome by the use of a complete system benchmark. Unfortunately these benchmarks are highly configuration dependent and hence are of little use outside the confines of a single requirement. Needless to mention, such system benchmarks are expensive, costing £10 000 or more to run and the results are invariably confidential.
The authors would like to thank colleagues from the Central Computer Agency, namely Mr G Brownley, Mr H J Curnow and Mr R Longbottom who have helped to collect much of the data making this system possible. Professor W M Gentleman (University of Waterloo, Canada) also assisted with the problem of assigning weights to the benchmarks. Mr P Verstege wrote the first version of the analysis program while studying at NPL.
This chapter describes the benchmarks; the coding of each one written in FORTRAN is given after its description.
This benchmark measures the average time in microseconds to execute a loop of machine-code instructions. The Loop is that involved in the evaluation of Ackermann's function [10]. The instructions are those generated by a compiler and are therefore more likely to be typical of non-numerical calculations. In fact, this benchmark is really another instruction mix. One would expect results from it to correlate well with those of the ADP mix since both involve a mix of non-floating point instructions.
The Ackermann function was designed to study the efficiency of calling procedures and is highly recursive. The Ackermann benchmark measures this efficiency. Ackn ratio measures the execution speed of the instructions generated, but does not take into account the effects of the compiler which may cause the calling of procedures to be implemented in an inefficient way.
Ackn ratio can be implemented in machine-code, or a high level language which supports recursion. In calculating the ratio, preference is given to results from the high level languages.
This program was constructed from statistical information on the use made of ALGOL 60 [13]. Although data collected by Knuth [7] on the use of FORTRAN shows some differences, this program is believed to provide a mix of instructions typical of many scientific programs.
The program has been carefully written so that optimizing compilers cannot take undue advantage of its loop structure. A good compiler will, however, be able to obtain some advantage by the use of fast registers.
The single parameter, I, is set according to the power of the machine; with I = 10, the benchmark represents one million Whetstone ALGOL [9] instructions. These instructions are similar to the machine instructions of the B6700. The mixes and this program are roughly on the same scale; one Whetstone instruction as measured by this program is about 1.38 Gibson Mix instructions.
On a machine where single length variables do not give adequate precision, this benchmark will not be typical of FORTRAN programs which are run on that machine. A double precision version has been written to overcome this deficiency (see Synth FORD).
The program is described in Curnow and Wichmann [3].
C THE CENTRAL COMPUTER AGENCY / NATIONAL PHYSICAL LABORATORY BENCHMARKS C CHARACTER SET COMMA , POINT . SLASH / ASTERISK * C BRACKETS ( ) EQUALS = PLUS + MINUS - C THE PROGRAMS PERFORM NO INPUT AND ARE SEPARATED BY BLANK CARDS C ALL PROGRAM OUTPUT (WHICH IS SMALL) IS TO DEVICE 6 C CENTRAL COMPUTER AGENCY PROGRAM FOPR12 C BENCHMARK NO 5 = SYNTH FORS IN NPL REPORT NAC 62 C THIS PROGRAM HAS A SINGLE PARAMETER I C SET I=257 FOR ONE MINUTE ON MACHINE OF POWER OF 360/65 C SET I=26 FOR MACHINE 1/10 POWER OF 360/65 C SET I=2570 FOR MACHINE TEN TIMES POWER OF 360/65 COMMON T,T1,T2,E1(4),J,K,L T=0.499975 T1=0.50025 T2=2.0 I = 257 N1=0 N2=12*I N3=14*I N4=345*I N5=0 N6=210*I N7=32*I N8=899*I N9=616*I N10=0 N11=93*I N12=0 X1=1.0 X2=-1.0 X3=-1.0 X4=-1.0 IF (N1)19,19,11 11 DO 18 I=1,N1,1 X1=(X1+X2+X3-X4)*T X2=(X1+X2-X3+X4)*T X3=(X1-X2+X3+X4)*T X4=(-X1+X2+X3+X4)*T 18 CONTINUE 19 CONTINUE CALL POUT(N1,N1,N1,X1,X2,X3,X4) E1(1)=1.0 E1(2)=-1.0 E1(3)=-1.0 E1(4)=-1.0 IF(N2)29,29,21 21 DO 28 I=1,N2,1 E1(1)=(E1(1)+E1(2)+E1(3)-E1(4))*T E1(2)=(E1(1)+E1(2)-E1(3)+E1(4))*T E1(3)=(E1(1)-E1(2)+E1(3)+E1(4))*T E1(4)=(-E1(1)+E1(2)+E1(3)+E1(4))*T 28 CONTINUE 29 CONTINUE CALL POUT(N2,N3,N2,E1(1),E1(2),E1(3),E1(4)) IF(N3)39,39,31 31 DO 38 I=1,N3,1 38 CALL PA(E1) 39 CONTINUE CALL POUT(N3,N2,N2,E1(1),E1(2),E1(3),E1(4)) J=1 IF(N4) 49,49,41 41 DO 48 I=1,N4,1 IF(J-1)43,42,43 42 J=2 GOTO 44 43 J=3 44 IF(J-2)46,46,45 45 J=0 GO TO 47 46 J=1 47 IF(J-1)411,412,412 411 J=1 GO TO 48 412 J=0 48 CONTINUE 49 CONTINUE CALL POUT(N4,J,J,X1,X2,X3,X4) J=1 K=2 L=3 IF(N6)69,69,61 61 DO 68 I=1,N6,1 J=J*(K-J)*(L-K) K=L*K-(L-J)*K L=(L-K)*(K+J) E1(L-1)=J+K+L E1(K-1)=J*K*L 68 CONTINUE 69 CONTINUE CALL POUT(N6,J,K,E1(1),E1(2),E1(3),E1(4)) X=0.5 Y=0.5 IF(N7)79,79,71 71 DO 78 1=1,N7,1 X=T*ATAN(T2*SIN(X)*COS(X)/(COS(X+Y)+COS(X-Y)-1.0)) Y=T*ATAN(T2*SIN(Y)*COS(Y)/(COS(X+Y)+COS(X-Y)-1.0)) 78 CONTINUE 79 CONTINUE CALL POUT(N7,J,K,X,X,Y,Y) X=1.0 Y=1.0 Z=1.0 IF(N8)89,89,81 81 DO 88 I=1,N8,1 88 CALL P3(X,Y,Z) 89 CONTINUE CALL POUT(N8,J,K,X,Y,Z,Z) J=1 K=2 L=3 E1(1)=1.0 E1(2)=2.0 E1(3)=3.0 IF(N9) 99,99,91 91 DO 98 I=1,N9,1 98 CALL P0 99 CONTINUE CALL POUT(N9,J,K,E1(1),E1(2),E1(3),E1(4)) J=2 K=3 IF (N10) 109,109,101 101 DO 108 I=1,N10,1 J=J+K K=J+K J=J-K K=K-J-J 108 CONTINUE 109 CONTINUE CALL POUT(N10,J,K,X1,X2,X3,X4) X=0.75 IF (N11) 119,119,111 111 DO 118 I=1,N11,1 118 X=SQRT(EXP(ALOG(X)/T1)) 119 CONTINUE CALL POUT(N11,J,K,X,X,X,X) STOP END SUBROUTINE PA(E) COMMON T,T1,T2 DIMENSION E(4) J=0 1 E(1)=(E(1)+E(2)+E(3)-E(4))*T E(2)=(E(1)+E(2)-E(3)+E(4))*T E(3)=(E(1)-E(2)+E(3)+E(4))*T E(4)=(-E(1)+E(2)+E(3)+E(4))/T2 J=J+1 IF (J-6) 1,2,2 2 CONTINUE RETURN END SUBROUTINE P0 COMMON T,T1,T2,E1(4),J,K,L E1(J)=E1(K) E1(K)=E1(L) E1(L)=E1(J) RETURN END SUBROUTINE P3(X,Y,Z) COMMON T,T1,T2 X1=X Y1=Y X1=T*(X1+Y1) Y1=T*(X1+Y1) Z=(X1+Y1)/T2 RETURN END SUBROUTINE POUT(N,J,K,X1,X2,X3,X4) WRITE(6,1) N,J,K,X1,X2,X3,X4 1 FORMAT(1H ,3I7,4E12.4) RETURN END
This program is similar to Synth FORS, but uses double precision variables. It is, therefore, typical of FORTRAN programs which use double precision. Apart from this, the program is a very poor indicator of overall machine performance because of its deliberate use of double length operations. These may not be provided by hardware, or, if they are, may not be used as extensively as in this benchmark.
One Whetstone instruction as measured by this program is about 3.1 Gibson Mix instructions.
The provision of floating point hardware as on the IBM 360 gives speeds comparable with FORS, whereas on 1900 machines with no double length hardware, the speed is very slow. Direct comparisons are complicated by the varying precision offered on different machines.
The program is described in Curnow and Wichmann [3].
C CENTRAL COMPUTER AGENCY PROGRAM FOPR13 C BENCHMARK NO 6 = SYNTH FORD IN NPL REPORT NAC62 C THIS PROGRAM HAS A SINGLE PARAMETER I C SET I=91 FOR ONE MINUTE ON MACHINE OF POWER OF 360/65 C SET I=9 FOR MACHINE 1/10 POWER OF 360/65 C SET I=914 FOR MACHINE TEN TIMES POWER OF 360/65 DOUBLE PRECISION X1, X2, X3, X4, X,Y,Z,T,T1,T2,E1 COMMON T,T1,T2,E1(4),J,K,L T=0.499975 T1=0.50025 T2=2.0 I = 91 N1=0 N2=12*I N3=14*I N4=345*I N5=0 N6=210*I N7=32*I N8=899*I N9=616*I N10=0 N11=93*I N12=0 X1=1.0 X2=-1.0 X3=-1.0 X4=-1.0 IF(N1)19,19,11 11 DO 18 I=1,N1,1 X1=(X1+X2+X3-X4)*T X2=(X1+X2-X3+X4)*T X3=(X1-X2+X3+X4)*T X4=(-X1+X2+X3+X4)*T 18 CONTINUE 19 CONTINUE CALL POUT(N1,N1,N1,X1,X2,X3,X4) E1(1)=1.0 E1(2)=-1.0 E1(3)=-1.0 E1(4)=-1.0 IF(N2)29,29,21 21 DO 28 I=1,N2,1 E1(1)=(E1(1)+E1(2)+E1(3)-E1(4))*T E1(2)=(E1(1)+E1(2)-E1(3)+E1(4))*T E1(3)=(E1(1)-E1(2)+E1(3)+E1(4))*T E1(4)=(-E1(1)+E1(2)+E1(3)+E1(4))*T 28 CONTINUE 29 CONTINUE CALL POUT(N2,N3,N2,E1(1),E1(2),E1(3),E1(4)) IF(N3)39,39,31 31 DO 38 I=1,N3,1 38 CALL PA(E1) 39 CONTINUE CALL POUT(N3,N2,N2,E1(1),E1(2),E1(3),E1(4)) J=1 IF(N4) 49,49,41 41 DO 48 I=1,N4,1 IF(J-1)43,42,43 42 J=2 GOTO 44 43 J=3 44 IF(J-2)46,46,45 45 J=0 GO TO 47 46 J=1 47 IF(J-1)411,412,412 411 J=1 GO TO 48 412 J=0 48 CONTINUE 49 CONTINUE CALL POUT(N4,J,J,X1,X2,X3,X4) J=1 K=2 L=3 IF(N6)69,69,61 61 DO 68 I=1,N6,1 J=J*(K-J)*(L-K) K=L*K-(L-J)*K L=(L-K)*(K+J) E1(L-1)=J+K+L E1(K-1)=J*K*L 68 CONTINUE 69 CONTINUE CALL POUT(N6,J,K,E1(1),E1(2),E1(3),E1(4)) X=0.5 Y=0.5 IF(N7)79,79,71 71 DO 78 I=1,N7,1 X=T*DATAN(T2*DSIN(X)*DCOS(X)/(DCOS(X+Y)+DCOS(X-Y)-1.0)) Y=T*DATAN(T2*DSIN(Y)*DCOS(Y)/(DCOS(X+Y)+DCOS(X-Y)-1.0)) 78 CONTINUE 79 CONTINUE CALL POUT(N7,J,K,X,X,Y,Y) X=1.0 Y=1.0 Z=1.0 IF(N8)89,89,81 81 DO 88 I=1,N8,1 88 CALL P3(X,Y,Z) 89 CONTINUE CALL POUT(N8,J,K,X,Y,Z,Z) J=1 K=2 L=3 E1(1)=1.0 E1(2)=2.0 E1(3)=3.0 IF(N9) 99,99,91 91 DO 98 I=1,N9,1 98 CALL P0 99 CONTINUE CALL POUT(N9,J,K,E1(1),E1(2),E1(3),E1(4)) J=2 K=3 IF (N10) 109,109,101 101 DO 108 I=1,N10,1 J=J+K K=J+K J=J-K K=K-J-J 108 CONTINUE 109 CONTINUE CALL POUT(N10,J,K,X1,X2,X3,X4) X=0.75 IF (N11) 119,119,111 111 DO 118 I=1,N11,1 118 X=DSQRT(DEXP(DLOG(X)/T1)) 119 CONTINUE CALL POUT(N11,J,K,X,X,X,X) STOP END SUBROUTINE PA(E) DOUBLE PRECISION T,T1,T2,E COMMON T,T1,T2 DIMENSION E(4) J=0 1 E(1)=(E(1)+E(2)+E(3)-E(4))*T E(2)=(E(1)+E(2)-E(3)+E(4))*T E(3)=(E(1)-E(2)+E(3)+E(4))*T E(4)=(-E(1)+E(2)+E(3)+E(4))/T2 J=J+1 IF (J-6) 1,2,,2 2 CONTINUE RETURN END SUBROUTINE P0 DOUBLE PRECISION T,T1,T2,E1 COMMON T,T1,T2,E1(4),J,K,L E1(J)=E1(K) E1(K)=E1(L) E1(L)=E1(J) RETURN END SUBROUTINE P3(X,Y,Z) DOUBLE PRECISION T,T1,T2,X1,Y1,X,Y,Z COMMON T,T1,T2 X1=X Y1=Y X1=T*(X1+Y1) Y1=T*(X1+Y1) Z=(X1+Y1)/T2 RETURN END SUBROUTINE POUT(N,J,K,X1,X2,X3,X4) DOUBLE PRECISION X1,X2,X3,X4 WRITE(6,1) N,J,K,X1,X2,X3,X4 1 FORMAT(1H ,3I7,4E12.4) RETURN END
This test is designed to give a crude check on the floating point facilities of a machine.
The program calculates the Gamma function for the integers 2 to 11. Two methods are compared:
Gamma (s) = (s - 1)! and Gamma (s) = ∫e-tts-1dt from 0 to ∞ using Simpson's formula.
The program contains two parameters, MM where 2MM is the upper limit of the integration, and MS where 1/MS is the step size for Simpson's rule. The program time in seconds is proportional to MM × MS, and is roughly
(MM × MS) / 0.188 × Gibson
The program will not be a good overall performance indicator because the inner loop consists of calls of the EXP function and use of the ** operator for integer-valued real exponents. The program time therefore measures the speed of these only.
Additional documentation on this program is available from the CCA under the program reference FOPR00.
C CENTRAL COMPUTER AGENCY PROGRAM FOPR00 C BENCHMARK NO 7 = GAMMA IN NPL REPORT NAC62 C THIS PROGRAM HAS TWO PARAMETERS MM AND MS C SET MM=70, MS=100 FOR ONE MINUTE ON MACHINE OF POWER OF 360/65 C SET MM=20, MS=35 FOR MACHINE OF 1/10 POWER OF 360/65 C SET MM=70, MS=1000 FOR MACHINE TEN TIMES POWER OF 360/65 GF=1.0 WRITE(6,50) 50 FORMAT(5H G,10X,14HGAMMA FUNCTION) AC=0.0 AY=0.0 YC=0.0 DO 600 I=1,10 IG=I+1 GS=I A=0.0 Y=0.0 X=0.0 GF=GF*GS WRITE(6,100)IG,GF MM = 70 MS = 100 AN=1.0/FLOAT(MS) DO 500 J=1,MM DO 500 K=1,MS XA=X+AN XB=XA+AN YA=EXP(-XA)*XA**GS YB=EXP(-XB)*XA**GS AY=(AN/3.0)*(Y+4.0*YA+YB) A=A+AY YC=YC+YA-YB X=XB 500 Y=YB WRITE(6,100)IG,A 100 FORMAT(I4,F35.24) 600 CONTINUE WRITE(6,200)AC,YC 200 FORMAT(8H SUMS = ,2E30.24) STOP END
The purpose of this test is to check core memory and access. A single bit is written (by assuming 2's complement arithmetic) in every element of an integer array. This is then rewritten and checked.
The parameters are:
IB size of arrays, typically 64 × 64 IC (number of bits in the word) - 1 NB number of repetitions of the program (at least 2) NA number of rewrites and rechecks (must be 2 for timing)
With NA= 2, the program execution time in seconds is roughly
(NB × IC × IB2) / (4.85 × Gibson)
The program can be altered to interchange rows and columns of the arrays. This change can be made to see whether core access patterns significantly affect the program time.
The program has not been designed as a performance test. An optimizing compiler can remove significant code from the inner loops in a way which is unlikely to be typical.
Additional documentation on this program is available from the CCA under the program reference FOPR01.
C CENTRAL COMPUTER AGENCY PROGRAM FOPR01 C BENCHMARK NO 8 = BIT TEST IN NPL REPORT NAC62 C THIS PROGRAM HAS THREE PARAMETERS, IB, IC AND NB C SET IC=23, IB=64 AND NB=2 FOR ONE MINUTE ON MACHINE OF POWER OF 360/65 C SET IC=15, IB=20 AND NB=3 FOR MACHINE 1/10 POWER OF 360/65 C SET IC=31, IB=64 AND NB=15 FOR MACHINE TEN TIMES 360/65 C THE DIMENSION OF THE ARRAY IA SHOULD BE SET TO IB DIMENSION IA(64,64) IB = 64 ID=1 NA=2 NB=2 DO 200 K=1,NB KK=K/2*2-K WRITE(6,99)KK 99 FORMAT(I6) IC= 23 IE=IC-1 DO 120 N= 1,IC M=N-1 WRITE(6,99)M DO 100 JJ=1,NA DO 100 J=1,IB DO 100 I=1,IB IA(I,J)=KK 100 CONTINUE DO 110 J=1,IB DO 110 I=1,IB MM=2**M+1 MM=MM*ID IA(I,J)=MM+KK M=M+1 IF(M-IE) 110,110,105 105 M=0 110 CONTINUE DO 120 JJ=1,NA M=N-1 DO 120 J=1,IB DO 120 I=1,IB MM=2**M+1 MM=MM*ID+KK IF(IA(I,J)-MM)128,130,128 128 WRITE(6,129)I,J,IA(I,J),MM 129 FORMAT(2H I,I4,2H J,I4,2X,3HWAS,I10,2X,8HEXPECTED,I10) 130 M=M+1 IF(M-IE)120,120,125 125 M=0 120 CONTINUE IF(ID)145,145,140 140 ID=-1 GO TO 200 145 ID=1 200 CONTINUE STOP END
This benchmark tests the floating point facilities of a machine. A binomial expansion is evaluated repeatedly with an argument of two fractions whose sum is unity. Rounding error is accumulated to test the accuracy of floating point.
The parameters are:
IA number of program repetitions (≥2) IE maximum exponent used (≤77) IB and ID, used to determine the two fractions
For reasonable comparisons, one requires that IB > 4 × ID.
The program execution time in seconds is roughly
(IA × IB × IE2) / (4.33 × ID × Gibson)
The program does not contain any loops which can be easily optimized by a compiler. The FORTRAN H compiler executes the program 48% faster than the FORTRAN G compiler.
Additional documentation on this program is available from the CCA under the program reference FOPR02.
C CENTRAL COMPUTER AGENCY PROGRAM FOPR02 C BENCHMARK NO 9 = BINOMIAL IN NPL REPORT NAC62 C THIS PROGRAM HAS FOUR PARAMETERS BUT IA CAN BE FIXED AT 2 C AND ID CAN BE FIXED AT 4 C SET IE=77 AND IB=48 FOR MACHINE OF THE POWER OF 360/65 C SET IE=35 AND IB=23 FOR MACHINE OF 1/10 OF POWER OF 360/65 C SET IE=77 AND IB=485 FOR MACHINE TEN TIMES THE POWER OF 360/65 DIMENSION COMB(501) WRITE(6,250) 250 FORMAT(8H1TSURL02) IA = 2 IB = 48 IC=9*IB ID = 4 IE = 77 AA=10*IB DO 600 M=1,IA BC=0.0 BB=1.0 DO 500 N=1,IE DO 400 K=IB,IC,ID AK=K L=N+1 MM=L/2 COMB(1)=1.0 COMB(L)=1.0 DO 120 I=2,MM AI=I-1 L=L-1 AN=L COMB(I)=AN/AI*COMB(I-1) 120 COMB(L)=COMB(I) IF(N+1-MM*2)150,151,150 150 COMB(L-1)=(AN-1.0)/(AI+1.0)*COMB(L) 151 L=N+1 Q=AK/AA P=1.0-Q BI=Q**N BII=BI DO 160 I=2,L BI=BI*COMB(I)/COMB(I-1)*P/Q 160 BII=BII+BI 400 BB=BB*BII 500 BC=BC+BII WRITE(6,200)M,BB,BC 200 FORMAT(I6,2F30.24) 600 CONTINUE STOP END
This program performs a large number of conditional jumps. An important use is to see how fast it is with computers using pipelining or instruction-look-ahead.
The program consists of 24 IF statements with a minimal number of driving statements.
The single parameter, N, determines the execution time, this being proportional to N2.
The program execution time in seconds is roughly
N2 / (23.7 × Gibson)
The test was first used by Dr Bryant in 1968 for a comparative study of FORTRAN compilers [2]. This version has a few statements added to check for erroneous jumps. In theory, an optimizing compiler could reduce the program to nothing, but so far, no compiler has done so.
It is thought that much commercial data processing involves many conditional jumps, and hence this test could be relevant in that context. It shows that almost all the faster machines perform relatively poorly in executing conditional jumps.
Additional documentation on this program is available from the CCA under the program reference FOPR03.
C CENTRAL COMPUTER AGENCY PROGRAM FOPR03 C BENCHMARK NO 10 = IF TEST IN NPL REPORT NAC62 C THIS PROGRAM HAS ONE PARAMETER N C SET N=897 FOR ONE MINUTE ON A MACHINE OF THE POWER OF 360/65 C SET N=284 FOR ONE MINUTE ON MACHINE 1/10 POWER OF 360/65 C SET N=2840 FOR MACHINE TEN TIMES THE POWER OF 360/65 N = 897 1 FORMAT(I10) WRITE(6,1)N I=-1 J=0 K=1 DO 300 L=1,N DO 300 M=1,N GO TO 200 100 IF(I)101,999,999 101 IF(J)999,102,999 l02 IF(K)999,999,103 103 IF(I)104,999,999 104 IF(J)999,105,999 105 IF(K)999,999,106 106 IF(I)107,999,999 107 IF(J)999,108,999 108 IF(K)999,999,109 109 IF(I)110,999,999 110 IF(J)999,111,999 111 IF(K)999,999,112 112 CONTINUE GO TO 300 200 IF(I)201,999,999 202 IF(K)999,999,203 201 IF(J)999,202,999 203 IF(I)204,999,999 205 IF(K)999,999,206 204 IF(J)999,205,999 206 IF(I)207,999,999 208 IF(K)999,999,209 207 IF(J)999,208,999 209 IF(I)210,999,999 211 IF(K)999,999,212 210 IF(J)999,211,999 212 CONTINUE GO TO 100 999 WRITE(6,2)L,M 2 FORMAT(8H RUBBISH,2I10) WRITE(6,3)I,J,K 300 CONTINUE WRITE(6,3)I,J,K 3 FORMAT(3I10) STOP END
This program checks double precision working with the main intrinsic functions.
A number of identities such as sin2 + cos2 = 1 are used to show that the functions are approximately correct. A sum check is accumulated of the residual from each identity.
There are two parameters, IA and IC, which control the outer and inner loops. For timing purposes, IC = 2 is taken, which gives an execution time in seconds, proportional to IA, of roughly
IA / (0.0504 × Gibson)
Code can be moved from the inner loop by an optimizing compiler to make the program run about twice as fast. Both the FORTRAN H compiler for the 360 and XFEW for the 1900 execute the program about 2.5 times faster than the corresponding non-optimizing compilers.
Additional documentation on this program is available from the CCA under the program reference FOPR04.
C CENTRAL COMPUTER AGENCY PROGRAM FOPR04 C BENCHMARK NO 11 = DOUBLE FUN IN NPL REPORT NAC62 C THIS PROGRAM HAS A SINGLE PARAMETER IA C SET IA=1400 FOR ONE MINUTE ON MACHINE OF THE POWER OF 360/65 C SET IA=140 FOR MACHINE 1/10 POWER OF 360/65 C SET IA=14000 FOR ONE MINUTE ON MACHINE TEN TIMES POWER OF 360/65 DOUBLE PRECISION RM,PII,A,R,B,D(10),E(10) COMPLEX CA COMMON RM PI=3.14159265 PII=DBLE(PI/100000.0) IA = 1400 IC=2 IB=50000/IA WRITE(6,1) 1 FORMAT(10H1FUNCTIONS,18X,6HANSWER,15X,8HSUMCHECK) RM= 1.0 DO 100 1=1,50001,IB A=FLOAT(I-1)*PII DO 100 J=1,IC R=DSIN(A)**2+DCOS(A)**2 100 RM=RM*R WRITE(6,11) 11 FORMAT(8H SIN COS) CALL SUMCK RM=1.0 DO 200 1=1,50001,IB A=FLOAT(I)/300.0 DO 200 J=1,IC SA=SNGL(A) R=(DEXP(A)-DEXP(-A))/(DEXP(A)+DEXP(-A))/TANH(SA) 200 RM=RM*R WRITE(6,12) 12 FORMAT(9H EXP TANH) CALL SUMCK RM=1.0 DO 300 1=1,50001,IB A=FLOAT(I)/300.0 DO 300 J=1,IC B=DEXP(A) R=DLOG(B)/A 300 RM=RM*R WRITE(6,13) 13 FORMAT(8H LOG EXP) CALL SUMCK RM=1.0 DO 400 I= 1,50001,IB DO 400 J=1,IC A=FLOAT(I)*FLOAT(J) B=2.0*DLOG10(A) B=10.0**B R=DSQRT(B)/A 400 RM=RM*R WRITE(6,14) 14 FORMAT(11H LOG10 SQRT) CALL SUMCK RM=1.0 DO 500 I=1,50001,IB C=(FLOAT(I)-0.8)*SNGL(PII) DO 500 J=1,IC CA=CEXP(CMPLX(C,C)) CA=CSQRT(CA*CONJG(CA)*CMPLX(COS(C)**2-SIN(C)**2, 12.0*SIN(C)*COS(C))) CA=CLOG(CA) R=(DABS(DBLE(REAL(CA)))+ABS(AIMAG(CA)))/(2.0*C) 500 RM=RM*R WRITE(6,15) 15 FORMAT(11H CEXP CMPLX,5H CLOG/ 112H CONJG CSQRT) CALL SUMCK RM=1.0 DO 600 I=1,50001,IB CA=CMPLX(FLOAT(I),FLOAT(50001-I)) DO 600 J=1,IC R=CABS(CA)/SQRT(REAL(CA)**2+AIMAG(CA)**2) 600 RM=RM*R WRITE(6,16) 16 FORMAT(5H CABS,5H REAL,6H AIMAG) CALL SUMCK RM=1.0 DO 700 I=1,50001,IB DO 688 M=1,10 D(M)=AMIN0(I,M)*AMAX0(I,M) 688 E(M)=1.0/FLOAT(MIN0(I,M)*MAX0(I,M)) DO 700 J=1,IC DO 699 N=1,10 A=DMIN1(D(1),D(2),D(3),D(4),D(5),D(6),D(7),D(8), 1D(9),D(10)) D(N)=1000000.0 R=A*DMAX1(E(1),E(2),E(3),E(4),E(5),E(6),E(7),E(8), 1E(9),E(10)) RM=RM*R 699 E(N)=0.000001 700 CONTINUE WRITE(6,17) 17 FORMAT(12H DMIN1 DMAX1) CALL SUMCK RM=1.0 PII=3.142/1000000.0 DO 800 I=1,50001,IB C=FLOAT(I)*SNGL(PII*2.0) DO 800 J=1,IC R=DBLE(ATAN(SIN(C)/COS(C))/C) 1*DATAN2(DSIN(DBLE(C)),DCOS(DBLE(C)))/C R=-R*DSIGN(R,DBLE(-1.0))*FLOAT(MOD(I+5000,I+4999)) 1*AINT(SNGL(R)/2.0+FLOAT(I))/FLOAT(I) 800 RM=RM*R WRITE(6,18) 18 FORMAT(18H ATAN DATAN2 DSIGN/9H MOD AINT) CALL SUMCK STOP END SUBROUTINE SUMCK DOUBLE PRECISION RM,V,VW COMMON RM V=RM VW=0.0 IV=0 DO 111 I=1,100 V=(V-VW)*10.0 VV=SNGL(V) IW=IFIX(VV) IV=IV+IW*I 111 VW=FLOAT(IW) WRITE(6,9)RM,IV 9 FORMAT(1H+,15X,D28.20,I10) RETURN END
This program gives a measure of performance of some hand-coded numerical routines which are available with FORTRAN and ALGOL. Although the original data is from the ALGOL statement mix [12], the timings should be independent of the compiler since the routines are hand-coded.
Routines are called by the program loop as follows:
SIN 3 times COS 5 times EXP 3 times SQRT 6 times ARCTAN 2 times
Weights have been assigned to the five routines using small integers so that the total execution time of the program gives the required performance figure.
The 19 calls of the routines are equivalent to about 1400 Gibson Mix instructions.
C BENCHMARK NO 12 = FUNCTIONS IN NPL REPORT NAC62 C THIS PROGRAM HAS ONE PARAMETER N C SET N=26400 FOR ONE MINUTE ON MACHINE OF THE POWER OF A 360/65 C SET N=2640 FOR MACHINE 1/10 POWER OF 360/65 C SET N=264000 FOR MACHINE TEN TIMES THE POWER OF 360/65 N = 26400 C = 1.0 DO 1 I=1,N D = C E = C A = SQRT(C) B = SIN(D) C = COS(E) D = SQRT(A) E = EXP(B) A = SIN(C) B = SQRT(D) C = COS(E) D = ATAN(A) E = SQRT(B) A = COS(C) B = EXP(D) C = SQRT(E) D = SIN(A) E = COS(B) C = SQRT(C) B = COS(D) A = EXP(E) D = ATAN(A) 1 CONTINUE A = A + B + D + E WRITE(6, 100) N, C, A 100 FORMAT(I10, 2F20.15) STOP END
This program is similar to Functions but uses double precision routines and variables in place of single precision.
C NATIONAL PHYSICAL LABORATORY BENCHMARKS, SECOND SET C CHARACTER SET COMMA , POINT . SLASH / ASTERISK * C BRACKETS ( ) EQUALS = PLUS + MINUS - C THE PROGRAMS PERFORM NO INPUT AND ARE SEPARATED BY BLANK CARDS C ALL PROGRAM OUTPUT (WHICH IS SMALL) IS TO DEVICE 6 C THIS PROGRAM IS A DOUBLE LENGTH VERSION OF FUNCTIONS GIVEN IN C NPL REPORT NAC 62. C NATIONAL PHYSICAL LABORATORY BENCHMARK DFUNCTNS C THIS PROGRAM HAS ONE PARAMETER N C SET N=1000 FOR ONE MINUTE ON MACHINE OF THE POWER OF A 360/65 C SET N=10000 FOR MACHINE TEN TIMES THE POWER OF 360/65 DOUBLE PRECISION A, B, C, D, E N = 1000 C = 1.0 DO 1 I=1,N D = C E = C A = DSQRT(C) B = DSIN(D) C = DCOS(E) D = DSQRT(A) E = DEXP(B) A = DSIN(C) B = DSQRT(D) C = DCOS(E) D = DATAN(A) E = DSQRT(B) A = DCOS(C) B = DEXP(D) C = DSQRT(E) D = DSIN(A) E = DCOS(B) C = DSQRT(C) B = DCOS(D) A = DEXP(E) D = DATAN(A) 1 CONTINUE A= A + B + D + E WRITE(6, 100) N, C, A 100 FORMAT( I10, 2D40.30) STOP END
The GAMM figure has been used for many years as a performance measurement [6]. The figure, determined by this program, is a weighted average of the time taken to execute five simple numerical loops.
The single parameter, N, is set according to the power of the machine. The GAMM figure corresponds to about 5.0 Gibson Mix instructions.
If the loops are written in machine code, the figure mainly measures the speed of the floating point facilities. In this program, the loops are written in FORTRAN, and the figure now depends on one-dimensional array access and the implementation of simple loops. The calculations are highly specialized, even for scientific computing, hence the GAMM figure may not be a good measure of general performance. The loops are capable of being optimized by a good compiler so that the code produced should be very similar to that of an assembler program.
Additional documentation is available from NPL on the numerical properties of the computation which allows the program to be used for checking the floating point operations of a computer.
C GAMM IN FORTRAN - SINGLE PRECISION VERSION, VERSION 2 C NATIONAL PHYSICAL LABORATORY BENCHMARK GAMM F C PROGRAM HAS A SINGLE PARAMETER N C SET N = 10000 FOR ABOUT ONE MINUTE ON-COMPUTER THE POWER OF 360/65 C EXTRA DOCUMENTATION IS AVAILABLE ON THESE PROGRAMS FROM NPL INTEGER I, J, REP, THIRTY, TEN, FIVE, N REAL ACC, ACC1, DIVN, Y, ROOT, X DIMENSION A(30), B(30), C(30) N = 10000 FIVE= 5 TEN= 10 THIRTY= 30 DIVN = 1.0 / FLOAT(N) X = .1 ACC = 0.0 C INITIALISE A AND B Y = 1.0 DO 1 I= 1, 30 A(I) = I B(I) = - Y Y = - Y 1 CONTINUE DO 15 REP= 1, N C ONE PASS OF THIS LOOP CORRESPONDS TO 300 GAMM UNITS I= 30 DO 2 J = 1, THIRTY C(I) = A(I) + B(I) I = I - 1 2 CONTINUE Y = 0.0 DO 3 I= 1, TEN Y = (Y + C(I)) * X 3 CONTINUE ACC1 = Y * DIVN Y = C(11) DO 4 I= 12, 20 IF (C(I) .GT. Y) Y = C(I) 4 CONTINUE ROOT = 1.0 DO 5 I=1,5 ROOT= 0.5 *(ROOT + Y/ROOT) 5 CONTINUE ACC1 = ACC1 + ROOT * DIVN DO 6 I= 1, THIRTY A(I) = C(I) - B(I) 6 CONTINUE Y = 0.0 DO 7 I= 1, TEN Y = (Y + A(I)) * X 7 CONTINUE ROOT = 1.0 DO 8 I = 1, FIVE ROOT= 0.5 * (ROOT+ Y/ROOT) 8 CONTINUE ACC1 = ACC1 + ROOT * DIVN DO 9 I= 1, THIRTY C(I) = C(I) * B(I) 9 CONTINUE Y = C(20) DO 10 I= 21, THIRTY IF (C(I) .GT. Y) Y = C(I) 10 CONTINUE ROOT= 1.0 DO 11 I= 1, 5 ROOT= 0.5 * (ROOT+ Y/ROOT) 11 CONTINUE ACC1 = ACC1 + ROOT * DIVN Y = 0.0 DO 12 I= 1, TEN Y = (Y + C(I)) * X 12 CONTINUE ACC1 = ACC1 + Y * DIVN Y = C(1) DO 13 I= 2, TEN IF (C(I) .GT. Y) Y = C(I) 13 CONTINUE ROOT= 1.0 DO 14 I= 1, FIVE ROOT= 0.5 * (ROOT+ Y/ROOT) 14 CONTINUE ACC1 = ACC1 + ROOT * DIVN ACC = ACC + ACC1 15 CONTINUE WRITE( 6, 100) N, ACC, ACC1 C SHOULD PRINT N THEN 16.73343 22410 90064 71684 80142 C 13037 73134 63994 100 FORMAT( I10, 2E30.22 ) C FORMAT SHOULD BE ADJUSTED TO PRINT TO MAXIMUM PRECISION STOP END
The following program is similar to the GAMM F program, but uses double precision variables in place of single precision. The GAMM figures measured by this benchmark is about ( 1 /.120 = 8.33 ) Gibson Mix instructions.
C GAMM IN FORTRAN - DOUBLE PRECISION VERSION, VERSION 2 C NATIONAL PHYSICAL LABORATORY BENCHMARK GAMM FD C THIS PROGRAM HAS A SINGLE PARAMETER N C SET N = 3500 FOR ABOUT ONE MINUTE ON MACHINE THE POWER OF 360/65 INTEGER I, J, REP, THIRTY, TEN, FIVE, N DOUBLE PRECISION ACC, ACC1, DIVN, RN, Y, ROOT, X, A, B, C DIMENSION A(30), B(30), C(30) N = 3500 FIVE = 5 TEN = 10 THIRTY = 30 RN = N DIVN = 1.0 / RN X = .1D0 ACC = 0.0 C INITIALISE A AND B Y = 1.0 DO 1 I= 1, 30 A(I) = I B(I) = - Y Y = - Y 1 CONTINUE DO 15 REP= 1, N C ONE PASS OF THIS LOOP CORRESPONDS TO 300 GAMM UNITS I= 30 DO 2 J = 1, THIRTY C(I) = A(I) + B(I) I = I - 1 2 CONTINUE Y = 0.0 DO 3 I= 1, TEN Y = (Y + C(I)) * X 3 CONTINUE ACC1 = Y * DIVN Y = C(11) DO 4 I= 12, 20 IF (C(I) .GT. Y) Y = C(I) 4 CONTINUE ROOT= 1.0 DO 5 I = 1, 5 ROOT = 0.5D0 * (ROOT+ Y/ROOT) 5 CONTINUE ACC1 = ACC1 + ROOT * DIVN DO 6 I= 1, THIRTY A(I) = C(I) - B(I) 6 CONTINUE Y = 0.0 DO 7 I = 1, TEN Y = (Y + A(I)) * X 7 CONTINUE ROOT= 1.0 DO 8 I= 1, FIVE ROOT= 0.5D0 * (ROOT + Y/ROOT) 8 CONTINUE ACC1 = ACC1 + ROOT * DIVN DO 9 I= 1, THIRTY C(I) = C(I) * B(I) 9 CONTINUE Y = C(20) DO 10 I = 21, THIRTY IF (C(I) .GT. Y) Y = C(I) 10 CONTINUE ROOT= 1.0 DO 11 I = 1, 5 ROOT= 0.5D0 * (ROOT+ Y/ROOT) 11 CONTINUE ACC1 = ACC1 + ROOT * DIVN Y = 0.0 DO 12 I = 1, TEN Y = (Y + C(I)) * X 12 CONTINUE ACC1 = ACC1 + Y * DIVN Y = C(1) DO 13 I = 2, TEN IF (C(I) .GT.- Y) Y = C(I) 13 CONTINUE ROOT= 1.0 DO 14 I = 1, FIVE ROOT= 0.5D0 * (ROOT+ Y/ROOT) 14 CONTINUE ACC1 = ACC1 + ROOT * DIVN ACC = ACC + ACC1 15 CONTINUE WRITE(6, 100) N, ACC, ACC1 C SHOULD PRINT N THEN 16.73343 22410 90064 71684 80142 C 13037 73134 63994 . 100 FORMAT( I10, 2D40.30) C FORMAT SHOULD BE ADJUSTED TO PRINT TO MAXIMUM PRECISION STOP END
This program is the synthetic benchmark referenced in [4]. It should represent a good indication of performance since the program is based upon statistics from nearly 1000 ALGOL 60 programs. Since ALGOL 60 contains no facilities for extended precision, the program may not be directly comparable with FORS or FORD. For instance on 1900, FORS and this program use the same precision but on System 4 FORS used 32 bits whereas this program uses 64 bit reals. On IBM 360 both the Level F and Delft compilers have an option which determines the precision used throughout.
The program should be suitable for measuring the performance of computers with optimizing compilers but not necessarily with certain forms of hardware. The data used by the program is very small and could fit easily into the cache memories of the 370 machines. The program performance gives one Whetstone instruction as being equivalent to ( 1 /.389 = 2.57 ) Gibson Mix instructions. Hence the performance of the program lies about half way between FORS and FORD.
This program is the ALGOL 60 basic statement mix given in [6], but with additions of more recent data. The program is not suitable for measuring the performance of machines with optimizing compilers or with slave memories and instruction look-ahead etc. The ALGOL mix figure is normalized relative to Atlas which corresponds to about 14000 basic statements per second. On this reckoning a basic statement in ALGOL 60 corresponds to about ( 1/(14 x .00246) = 29.0) Gibson Mix instructions.
This program is written in ALGOL 60 and consists of the evaluation of a highly recursive procedure called Ackermann's function. The algorithm can be recoded easily into any language which supports recursion although only ALGOL 60 is considered in the first instance. Results from other languages are only included when the ALGOL 60 figures are available for the same machine. Hence the results included here are not necessarily the same ones listed under Ackn ratio. All of the available data on Ackermann's function is listed in [5]. The times for one call of Ackermann's function (which is taken as the performance measure) is almost entirely dependent upon the software. Figures for the 360 architecture vary from 14.5 to nearly 1000 instructions per call. It is not surprising therefore that this benchmark has a very low relevance. The number of microseconds per call is listed in Appendix F, this figure corresponding to about ( 1/.0147 = 68.0) Gibson Mix instructions.
This benchmark is similar to GAMM F, but is written in ALGOL 60. It gives the GAMM figure, which is a weighted average of the time taken to execute five simple numerical loops. Three arrays are used by the loops which could be either formal parameters or actual arrays. The ALGOL 60 program calculates the figure twice, once for each case. If no accurate interval timer is available then the total execution time of the program can be taken instead. The GAMM figure is not a good indication of performance because, even for scientific computing, the calculations are highly specialized. When written in ALGOL 60, the execution speed depends largely upon the degree of optimization of very simple loops and one-dimensional array access. In contrast, when coded in machine code, the figure measures the speed of the floating point unit and little else. The GAMM figure itself listed in Appendix F is in microseconds and corresponds to about ( 1/.0566 = 17.7 ) Gibson Mix instructions.
This benchmark is a non-numerical algorithm written in ALGOL 60. It has been proposed as a benchmark in [2] where several results are given. Only the results from ALGOL 60 have been used here in order to permit fair comparisons. This program is typical of the type of program and results one ought to be able to add to this system.
The program uses a valid move algorithm to discover a check mate position. The coding is typical of non-numerical applications, involving a large number of procedure calls, equality tests and integer array accessing. The ALGOL 60 version reflects the absence of data structures which forces a higher use of array accessing than would otherwise be necessary.
The performance measure listed in Appendix F is the program execution time in seconds. The execution time corresponds to 49.7 million Gibson Mix instructions.
This benchmark is similar to GAMM F, but is written in Assembly code. The GAMM figure corresponds to about ( 1/.319 = 3.13 ) Gibson Mix instructions.
The Post Office Work Unit was devised by the UK Post Office to reflect the usage of machine instructions for commercial data processing. Hence it is very similar to the ADP Mix. The Unit, which is measured in milliseconds, corresponds to about (1000/1.47 = 680) Gibson Mix instructions.
[1] BELL, A. G., HALLOWELL, P. J. and LONG D. H. A universal benchmark? Software Pract. Exper., 1973, 3, 355-357.
[2] BRYANT, P. FORTRAN - A comparative study. Science Research Council, Atlas Computer Laboratory, 1968.
[3] CURNOW, H. J. and WICHMANN B. A. A synthetic benchmark. Comput. J., 1976, 19, 43-49.
[4] FORD, B. The Evolving NAG Approach to Software Portability. Software Portability, 1977, 249-267, Cambridge University Press.
[5] GIBSON, J. C. The Gibson Mix. Report TR 00.2043, 1970, IBM Systems Development Division, Poughkeepsie, N.Y.
[6] HEINHOLD, J. and BAUER, F. , (Eds) Fachbegriffe der Programmierungstechnik, Ausgearbeitet vom Fachausschutz Programmieren, Munchen, der Gessellschaft fur Angewandte Mathematik und Mechanik (GAMM), 1962.
[7] KNUTH, D. E. An empirical study of FORTRAN programs. Software Pract. Exper., 1971, 1(2), 105-133.
[8] LUNDE, A. Evaluation of Instruction Set Processor Architecture by Program Tracing. Ph D Thesis, 1974, Carnegie-Mellon University.
[9] RANDELL, B. and RUSSELL, L. J. ALGOL 60 Implementation. London, Academic Press, 1964.
[10] SUNDBLAD, Y. The Ackermann function, a theoretical, computational and formula manipulative study. BIT, 1971, ll, 107-119.
[11] WICHMANN, B. A. Ackermann's Function: A study in the efficiency of calling procedures. BIT, 1976, 16, 103-110.
[12] WICHMANN, B. A. Basic statement times for ALGOL 60. NPL NAC Report No. 42, 1973.
[13] WICHMANN, B. A. ALGOL 60 Compilation and Assessment, London, Academic Press, 1973.
(14] WICHMANN, B. A. A comparison of ALGOL 60 execution speeds. NPL Report No. CCU3, 1969.
This Appendix gives the main result of the analysis made from the benchmark data. Each speed is given in instructions per millisecond. It is important to note that if a benchmark took one second, and the speed is given as 1000 instructions per millisecond, this does not imp1y that the benchmark consists or one million instructions. The actual measurement was of time, not machine instructions, but the analysis program has made an estimate of one million instructions. The overall tolerance that should be placed on the speed of a machine is given by the range figures at the top of each column.
In several cases, there is more than one measurement, eg for a different compiler or compiling option. The fastest one is chosen for the analysis, provided it is for the standard language. A star is placed against the results for which other data is available so that this material can be easily located in Appendix F.
Number of machines = 101 Number of benchmarks = 21 Number of results included = 756 Number of results excluded = 181
All performance data is in instructions per millisecond
* before results where additional figures are available
Name | 360/65 | 360/65B | 360/67 | 360/75 | 360/85 |
---|---|---|---|---|---|
Number | 1 | 2 | 3 | 4 | 5 |
Performance Range | |||||
Lower | 422 | 395 | 711 | 365 | 2370 |
Upper | 699 | 819 | 1180 | 1050 | 4110 |
Benchmarks | |||||
Gibson Mix | 543 | 563 | - | 940 | 3240 |
ADP Mix | 351 | 368 | - | 564 | 2220 |
Pr Con Mix | 375 | 401 | - | 640 | 2130 |
Ackn ratio | - | - | - | - | - |
Synth FORS | - | *722 | *835 | - | - |
Synth FORD | - | *1320 | *1430 | - | - |
Gamma Test | 720 | - | 781 | - | 4150 |
Bit Test | 517 | - | 670 | - | 3240 |
Binomial | 766 | - | 1080 | - | 4630 |
IF Test | 306 | - | 555 | - | 1750 |
DOUBLE FUN | - | 936 | 2190 | - | 5830 |
Functions | 861 | - | *927 | - | - |
Synth ALGL | - | *445 | - | - | - |
Algol Mix | - | - | - | - | - |
Ackermann | - | 194 | 563 | *78.2 | - |
GAMM ALGOL | - | - | 1160 | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | 946 | - | - |
GAMM FD | - | - | 1190 | - | - |
Name | 360/195 | 360/50 | 360/30 MI | 370/135 | 370/145 |
---|---|---|---|---|---|
Number | 6 | 7 | 8 | 9 | 10 |
Performance Range | |||||
Lower | 4400 | 112 | 18.1 | 108 | 175 |
Upper | 11700 | 200 | 25.9 | 161 | 261 |
Benchmarks | |||||
Gibson Mix | - | 133 | - | 113 | 178 |
ADP Mix | - | 110 | - | 111 | 214 |
Pr Con Mix | - | 109 | - | 110 | 176 |
Ackn ratio | - | - | - | - | - |
Synth FORS | *6920 | *201 | 21.5 | - | 237 |
Synth FORD | *14900 | *294 | 24.1 | - | 322 |
Gamma Test | 6130 | - | 22.0 | 147 | - |
Bit Test | *4520 | - | 19.9 | 156 | - |
Binomial | *14900 | - | 22.2 | 142 | - |
IF Test | 1750 | - | 17.2 | 125 | - |
DOUBLE FUN | *26200 | 142 | 29.7 | 236 | - |
Functions | - | 225 | - | - | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | *66.2 | - | - | - |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | 370/155 | 370/158 | 370/165 | 370/168 | 370/168 M |
---|---|---|---|---|---|
Number | 11 | 12 | 13 | 14 | 15 |
Performance Range | |||||
Lower | 478 | 840 | 2070 | 3220 | 3500 |
Upper | 655 | 1400 | 3600 | 4890 | 7150 |
Benchmarks | |||||
Gibson Mix | 470 | - | 3070 | - | - |
ADP Mix | 440 | - | 2010 | - | - |
Pr Con Mix | 487 | - | 2140 | - | - |
Ackn ratio | - | 798 | *2320 | - | - |
Synth FORS | *644 | *1140 | - | *3380 | 4200 |
Synth FORD | - | 1750 | - | *5120 | - |
Gamma Test | 608 | 1030 | 3980 | *3530 | *5280 |
Bit Test | - | 1010 | 2600 | *3460 | *3490 |
Binomial | - | 942 | 4570 | *4360 | *6670 |
IF Test | 548 | 728 | 1630 | *2680 | *3190 |
DOUBLE FUN | - | 1440 | 5160 | *8920 | *20700 |
Functions | 683 | - | 3300 | *4740 | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | - | *1650 | 3430 | - |
Ackermann | - | 1750 | *1550 | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | 3940 | - |
GAMM FD | - | - | - | 5390 | - |
Name | Amdahl 470 | ICL 4/50 | ICL 4/70 | ICL 4/72 | ICL 4/75 P |
---|---|---|---|---|---|
Number | 16 | 17 | 18 | 19 | 20 |
Performance Range | |||||
Lower | 6010 | 60.7 | 306 | 380 | 274 |
Upper | 9590 | 88.8 | 518 | 564 | 535 |
Benchmarks | |||||
Gibson Mix | - | 55.0 | 368 | 426 | 333 |
ADP Mix | - | 73.9 | 266 | 311 | 235 |
Pr Con Mix | - | 50.9 | 300 | 371 | 275 |
Ackn ratio | - | - | 226 | - | 273 |
Synth FORS | 6430 | *76.3 | *388 | *415 | *370 |
Synth FORD | 10400 | *85.6 | *653 | *703 | *625 |
Gamma Test | 6930 | 105 | *598 | *661 | - |
Bit Test | 5660 | *89.9 | *381 | *425 | - |
Binomial | 9340 | 86.2 | *549 | *640 | - |
IF Test | 4780 | *74.9 | *304 | *348 | - |
DOUBLE FUN | 16900 | 106 | 769 | 833 | - |
Functions | 7110 | 106 | 579 | 630 | 560 |
Synth ALGL | - | - | - | - | *252 |
Algol Mix | - | 38.6 | 262 | - | 601 |
Ackermann | - | - | 1480 | - | 1480 |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | 532 | - |
POWU | - | - | 350 | 392 | - |
GAMM F | 8850 | 68.9 | 324 | *372 | - |
GAMM FD | 9340 | 64.1 | 446 | *516 | - |
Name | S 4004/55 | 1901A 10SC | 1902A 20SC | 1902S 25S | 1903 EMU |
---|---|---|---|---|---|
Number | 21 | 22 | 23 | 24 | 25 |
Performance Range | |||||
Lower | 25.0 | 17.4 | 33.7 | 43.4 | 45.4 |
Upper | 63.6 | 26.0 | 54.3 | 62.7 | 65.8 |
Benchmarks | |||||
Gibson Mix | - | 26.9 | 56.1 | 59.8 | 59.0 |
ADP Mix | - | 18.9 | 61.7 | 61.6 | 68.9 |
Pr Con Mix | - | 18.7 | 44.5 | 46.5 | - |
Ackn ratio | - | - | - | - | - |
Synth FORS | - | - | - | - | - |
Synth FORD | - | - | - | - | - |
Gamma Test | - | - | 37.0 | - | - |
Bit Test | - | - | 36.1 | - | - |
Binomial | - | - | 41.6 | - | - |
IF Test | - | - | 29.2 | - | - |
DOUBLE FUN | - | - | 14.2 | - | - |
Functions | 24.4 | - | - | - | 38.8 |
Synth ALGL | - | - | - | - | 59.7 |
Algol Mix | 77.5 | - | - | - | 55.2 |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | 28.9 | 36.0 | 36.0 | - |
POWU | - | 15.1 | 59.0 | 59.0 | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | 1903A SC | 1903S 31S | 1903T | 1904A FP | 1904S FP I |
---|---|---|---|---|---|
Number | 26 | 27 | 28 | 29 | 30 |
Performance Range | |||||
Lower | 69.6 | 85.9 | 137.0 | 180.0 | 227.0 |
Upper | 110.0 | 124.0 | 211.0 | 289.0 | 385.0 |
Benchmarks | |||||
Gibson Mix | 114.0 | 118.0 | 220.0 | 261.0 | 302.0 |
ADP Mix | 123.0 | 126.0 | 143.0 | 179.0 | - |
Pr Con Mix | 87.7 | 93.9 | - | 162.0 | - |
Ackn ratio | - | - | - | - | - |
Synth FORS | - | - | - | *266.0 | *307.0 |
Synth FORD | - | - | - | *65.6 | *87.5 |
Gamma Test | 69.3 | - | *175.0 | 236.0 | - |
Bit Test | 73.0 | - | 140.0 | 198.0 | - |
Binomial | 80.6 | - | 195.0 | 240.0 | - |
IF Test | 137.0 | - | 188.0 | 239.0 | - |
DOUBLE FUN | 37.5 | - | 54.4 | 71.4 | - |
Functions | 56.0 | - | - | 297.0 | 337.0 |
Synth ALGL | - | - | - | *322.0 | 396.0 |
Algol Mix | 119.0 | - | - | 369.0 | 367.0 |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | 70.7 | 70.7 | 209.0 | 285.0 | - |
POWU | 115.0 | 115.0 | 170.0 | 226.0 | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | 1905E Acc | 1905F | 1906A | 1906S | ICL 2970 |
---|---|---|---|---|---|
Number | 31 | 32 | 33 | 34 | 35 |
Performance Range | |||||
Lower | 55.8 | 108 | 680 | 822 | 555 |
Upper | 116.0 | 166 | 1020 | 1440 | 896 |
Benchmarks | |||||
Gibson Mix | 144.0 | 196 | 866 | 1150 | - |
ADP Mix | 117.0 | 146 | 588 | 730 | - |
Pr Con Mix | - | - | 494 | 915 | - |
Ackn ratio | - | - | *825 | 1220 | - |
Synth FORS | *92.4 | 76.2 | *810 | *1110 | *604 |
Synth FORD | *33.7 | - | *859 | *381 | *1190 |
Gamma Test | - | 135 | *912 | *1780 | *419 |
Bit Test | *68.8 | - | 929 | 1770 | *555 |
Binomial | *92.9 | 154 | *1290 | 1920 | 834 |
IF Test | *126.0 | 165 | 615 | 1080 | *1010 |
DOUBLE FUN | *33.7 | - | *605 | 195 | *992 |
Functions | *106.0 | - | 1050 | - | *594 |
Synth ALGL | - | 183 | *976 | 1380 | - |
Algol Mix | - | - | *1080 | - | 1020 |
Ackermann | - | - | *2330 | *960 | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | 1660 | - | - |
GAMM Asmbl | - | - | 896 | 1310 | - |
POWU | - | - | 722 | 1040 | - |
GAMM F | *71.2 | - | *647 | - | *691 |
GAMM FD | *13.7 | - | *833 | - | *1050 |
Name | ICL 2980 | MU5 | ICL ATLAS1 | ICL4120/2 | ICL 4130/2 |
---|---|---|---|---|---|
Number | 36 | 37 | 38 | 39 | 40 |
Performance Range | |||||
Lower | 2070 | 5150 | 202 | 13.10 | 65.30 |
Upper | 3600 | 9680 | 393 | 34.10 | 122.00 |
Benchmarks | |||||
Gibson Mix | - | - | - | 25.00 | 112.00 |
ADP Mix | - | - | - | 61.60 | 106.00 |
Pr Con Mix | - | - | - | 42.10 | 85.00 |
Ackn ratio | 1080 | - | - | - | - |
Synth FORS | *2500 | - | - | 14.40 | 104.00 |
Synth FORD | *4330 | - | - | 3.44 | 6.87 |
Gamma Test | *2050 | - | - | - | 79.10 |
Bit Test | *2100 | - | 224 | - | - |
Binomial | *3390 | - | 384 | - | - |
IF Test | *1720 | - | 176 | - | 111.00 |
DOUBLE FUN | 3170 | - | 122 | - | - |
Functions | *2750 | - | 314 | - | 86.10 |
Synth ALGL | - | - | - | 21.40 | 137.00 |
Algol Mix | 3870 | 7060 | 406 | - | 135.00 |
Ackermann | 5670 | - | - | - | - |
GAMM ALGOL | - | - | 313 | - | - |
Chess Mate | - | - | 497 | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | *3220 | - | - | - | - |
GAMM FD | 4230 | - | - | - | - |
Name | Bur 5500 | Bur 6714 F | Bur 6715 S | CDC 3300 | CDC 3600 |
---|---|---|---|---|---|
Number | 41 | 42 | 43 | 44 | 45 |
Performance Range | |||||
Lower | 81.1 | 309 | 252 | 91.8 | 235 |
Upper | 169.0 | 425 | 425 | 178.0 | 413 |
Benchmarks | |||||
Gibson Mix | 144.0 | 298 | 348 | 152.0 | 337 |
ADP Mix | 117.0 | - | - | 102.0 | 211 |
Pr Con Mix | - | - | - | 166.0 | 169 |
Ackn ratio | 104.0 | - | - | - | - |
Synth FORS | 88.6 | *349 | *371 | - | - |
Synth FORD | 65.6 | *341 | *425 | - | - |
Gamma Test | 147.0 | 387 | 366 | - | - |
Bit Test | 53.0 | - | 139 | - | - |
Binomial | 92.4 | 382 | 287 | - | - |
IF Test | 67.0 | *328 | 214 | - | - |
DOUBLE FUN | 45.2 | - | 270 | - | - |
Functions | 109.0 | 350 | - | - | 574 |
Synth ALGL | 242.0 | *605 | *664 | - | - |
Algol Mix | 214.0 | - | - | *125.0 | 326 |
Ackermann | 504.0 | - | - | 47.1 | - |
GAMM ALGOL | 278.0 | - | - | - | - |
Chess Mate | 226.0 | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | Cyber 72 | Cyber 73 | Cyber 173 | CDC 6600 | CDC 7600 |
---|---|---|---|---|---|
Number | 46 | 47 | 48 | 49 | 50 |
Performance Range | |||||
Lower | 902 | 1090 | 6710 | ||
Upper | 2930 | 13900 | |||
Benchmarks | |||||
Gibson Mix | 600 | 800 | - | 2190 | 7000 |
ADP Mix | - | - | - | - | - |
Pr Con Mix | - | - | - | - | - |
Ackn ratio | - | 816 | - | - | - |
Synth FORS | 773 | - | 1410 | 2890 | *12900 |
Synth FORD | 522 | - | 953 | - | 11700 |
Gamma Test | 720 | 548 | 1470 | *3510 | *17100 |
Bit Test | 563 | *624 | 770 | 1350 | *6390 |
Binomial | 809 | 790 | 1890 | *3970 | *27900 |
IF Test | 406 | 417 | 590 | *604 | *3890 |
DOUBLE FUN | 567 | 595 | 1240 | *2540 | *16000 |
Functions | - | 455 | - | 1330 | 6520 |
Synth ALGL | *363 | - | - | - | *5780 |
Algol Mix | - | *491 | - | *1090 | *5480 |
Ackermann | - | 2000 | - | *166 | - |
GAMM ALGOL | - | 310 | - | 1460 | 3190 |
Chess Mate | - | - | - | 497 | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | *16600 |
GAMM FD | - | - | - | - | *10700 |
Name | Hon GE 635 | Hon 6025 | Hon 6030 | Hon 6040 | Hon 6050 |
---|---|---|---|---|---|
Number | 51 | 52 | 53 | 54 | 55 |
Performance Range | |||||
Lower | 189 | 136 | 91.0 | 230 | 201 |
Upper | 327 | 228 | 216.0 | 369 | 428 |
Benchmarks | |||||
Gibson Mix | 379 | 180 | - | 240 | - |
ADP Mix | 281 | 121 | - | - | - |
Pr Con Mix | 210 | - | - | - | - |
Ackn ratio | - | - | - | - | - |
Synth FORS | - | - | - | 317 | - |
Synth FORD | - | - | - | 441 | - |
Gamma Test | - | 206 | - | 315 | - |
Bit Test | - | 108 | - | 152 | - |
Binomial | - | 199 | - | 293 | - |
IF Test | - | 208 | - | 285 | - |
DOUBLE FUN | - | 347 | - | 551 | - |
Functions | 123 | - | 90.4 | - | 207 |
Synth ALGL | - | - | - | 260 | - |
Algol Mix | 317 | - | 253.0 | - | 471 |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | Hon 6060 | Hon 66/40 | Hon 66/60 | DEC 20 | DEC KI10 |
---|---|---|---|---|---|
Number | 56 | 57 | 58 | 59 | 60 |
Performance Range | |||||
Lower | 399 | 487 | 554 | 550 | 513 |
Upper | 682 | 1070 | 1220 | 917 | 959 |
Benchmarks | |||||
Gibson Mix | 479 | 820 | - | - | 590 |
ADP Mix | 311 | - | - | - | - |
Pr Con Mix | - | - | - | - | - |
Ackn ratio | - | - | - | - | 359 |
Synth FORS | 551 | 1100 | - | 613 | 692 |
Synth FORD | 825 | 1850 | - | 789 | 781 |
Gamma Test | 656 | 858 | 1180 | 962 | 1130 |
Bit Test | - | 472 | 385 | 518 | 622 |
Binomial | 654 | 894 | 877 | 825 | 1030 |
IF Test | 301 | 554 | 562 | 508 | 586 |
DOUBLE FUN | 1000 | 2310 | 1870 | 1210 | 1210 |
Functions | - | 319 | - | - | - |
Synth ALGL | 414 | - | - | - | - |
Algol Mix | - | 224 | - | - | - |
Ackermann | - | - | - | - | 215 |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | DEC KL10 | ICL KDF9 | TR4 | TR440 | EL-X8 |
---|---|---|---|---|---|
Number | 61 | 62 | 63 | 64 | 65 |
Performance Range | |||||
Lower | 1150 | 88.2 | 107 | 578 | 110 |
Upper | 1850 | 155.0 | 165 | 843 | 206 |
Benchmarks | |||||
Gibson Mix | 1030 | 170.0 | - | - | - |
ADP Mix | 720 | 128.0 | - | - | - |
Pr Con Mix | - | 90.8 | - | - | - |
Ackn ratio | - | *92.1 | - | - | - |
Synth FORS | 1680 | 111.0 | - | - | - |
Synth FORD | 2010 | 24.1 | - | - | - |
Gamma Test | 2110 | - | - | - | - |
Bit Test | 1280 | - | - | - | - |
Binomial | *1540 | - | - | - | - |
IF Test | 2060 | - | - | - | - |
DOUBLE FUN | *3310 | - | - | - | - |
Functions | - | 109.0 | 142 | 703 | - |
Synth ALGL | *1040 | *160.0 | - | - | - |
Algol Mix | - | *165.0 | 121 | 690 | 150 |
Ackermann | - | 128.0 | - | - | - |
GAMM ALGOL | - | *181.0 | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | *1470 | - | - | - | - |
GAMM FD | *1430 | - | - | - | - |
Name | Univ1106 U | Univ1108 | Univ 1110 | Univ 11 21 | INTER 5 |
---|---|---|---|---|---|
Number | 66 | 67 | 68 | 69 | 70 |
Performance Range | |||||
Lower | 413 | 763 | 1190 | 788 | 36.4 |
Upper | 712 | 1430 | 2320 | 1250 | 92.9 |
Benchmarks | |||||
Gibson Mix | 517 | 932 | 1610 | 709 | 73.9 |
ADP Mix | 335 | 562 | 877 | - | 141.0 |
Pr Con Mix | 316 | 547 | 1030 | - | 101.0 |
Ackn ratio | - | - | - | - | - |
Synth FORS | 741 | 1140 | 2220 | 1120 | - |
Synth FORD | 978 | 1480 | 3500 | 1470 | - |
Gamma Test | 616 | 1570 | - | 1310 | 16.2 |
Bit Test | 470 | 869 | - | 936 | - |
Binomial | 557 | 1240 | - | 1100 | 32.9 |
IF Test | 308 | 847 | - | 607 | 94.8 |
DOUBLE FUN | 1070 | 2780 | - | - | - |
Functions | - | 2040 | - | - | - |
Synth ALGL | - | *726 | - | - | - |
Algol Mix | - | *950 | - | - | - |
Ackermann | - | *389 | - | - | - |
GAMM ALGOL | - | 1450 | - | - | - |
Chess Mate | - | 553 | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | Univ418III | XDS SIG5 | XDS SIG6 | XDS SIG9 | CII IRIS80 |
---|---|---|---|---|---|
Number | 71 | 72 | 73 | 74 | 75 |
Performance Range | |||||
Lower | 31.60 | 285 | 254.0 | 581 | 243 |
Upper | 183.00 | 448 | 553.0 | 964 | 802 |
Benchmarks | |||||
Gibson Mix | - | 272 | - | 972 | - |
ADP Mix | 285.00 | 259 | - | 811 | - |
Pr Con Mix | 287.00 | - | - | 435 | - |
Ackn ratio | - | - | - | - | 442 |
Synth FORS | - | 413 | *458.0 | 669 | - |
Synth FORD | - | 703 | *812.0 | 1130 | - |
Gamma Test | - | 401 | 661.0 | - | - |
Bit Test | 95.90 | 293 | 481.0 | - | - |
Binomial | 2.61 | 326 | 461.0 | - | - |
IF Test | 75.00 | 329 | 371.0 | - | - |
DOUBLE FUN | - | 522 | - | - | - |
Functions | - | - | - | - | - |
Synth ALGL | - | - | *92.7 | - | - |
Algol Mix | - | - | 147.0 | - | - |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | PDP 11/10 | PDP 11/20 | PDP 11/34F | PDP 11/34S | PDP 11/40E |
---|---|---|---|---|---|
Number | 76 | 77 | 78 | 79 | 80 |
Performance Range | |||||
Lower | 12.70 | 34.0 | 228 | 24.0 | 25.9 |
Upper | 32.80 | 144.0 | 443 | 52.7 | 60.2 |
Benchmarks | |||||
Gibson Mix | - | 39.9 | - | - | - |
ADP Mix | - | 186.0 | - | - | - |
Pr Con Mix | - | 152.0 | - | - | - |
Ackn ratio | - | 154.0 | - | - | - |
Synth FORS | 17.90 | - | 283 | 32.5 | - |
Synth FORD | 9.91 | - | 491 | 19.1 | - |
Gamma Test | 11.60 | - | - | 22.5 | 24.3 |
Bit Test | 86.40 | 20.3 | - | 115.0 | 100.0 |
Binomial | 13.30 | 30.1 | - | - | 25.6 |
IF Test | 69.30 | 150.0 | - | 53.5 | 62.2 |
DOUBLE FUN | - | 21.4 | - | 25.9 | 28.0 |
Functions | 15.80 | - | - | - | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | - | - | - | - |
Ackermann | - | *636.0 | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | PDP 11/40S | PDP 11/55 | PDP 11/60 | PDP 11/70 | Hon DDP516 |
---|---|---|---|---|---|
Number | 81 | 82 | 83 | 84 | 85 |
Performance Range | |||||
Lower | 25.6 | 798 | 661 | 751 | 20.4 |
Upper | 56.2 | 1550 | 1260 | 1440 | 67.1 |
Benchmarks | |||||
Gibson Mix | - | - | - | - | 18.0 |
ADP Mix | - | - | - | - | 77.2 |
Pr Con Mix | - | - | - | - | 50.9 |
Ackn ratio | - | - | - | - | 315.0 |
Synth FORS | 24.2 | 989 | 818 | 929 | - |
Synth FORD | 16.2 | 1720 | 1360 | 1570 | - |
Gamma Test | 33.5 | - | - | - | 21.2 |
Bit Test | 144.0 | - | - | - | 86.9 |
Binomial | 34.2 | - | - | - | 20.5 |
IF Test | 85.3 | - | - | - | - |
DOUBLE FUN | 38.2 | - | - | - | 13.2 |
Functions | - | - | - | - | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | - | - | - | - |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | Hon H316 F | CTL MOD1 S | CTL MOD1 F | NOR SM4 | HP 2100A |
---|---|---|---|---|---|
Number | 86 | 87 | 88 | 89 | 90 |
Performance Range | |||||
Lower | 19.2 | 47.4 | 116 | 115 | 72.1 |
Upper | 62.5 | 84.1 | 199 | 379 | 147.0 |
Benchmarks | |||||
Gibson Mix | 16.2 | 45.5 | 113 | - | 99.0 |
ADP Mix | 119.0 | 64.2 | 158 | - | 130.0 |
Pr Con Mix | 114.0 | 111.0 | 247 | - | 151.0 |
Ackn ratio | - | - | - | 209 | - |
Synth FORS | 23.8 | - | - | - | - |
Synth FORD | 17.5 | - | - | - | - |
Gamma Test | 21.3 | - | - | - | 90.6 |
Bit Test | 65.4 | - | - | - | 132.0 |
Binomial | 19.4 | - | - | - | 110.0 |
IF Test | 207.0 | - | - | - | 201.0 |
DOUBLE FUN | 12.4 | - | - | - | 12.4 |
Functions | - | - | - | - | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | - | - | - | - |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | HP 2100S | HP 3000 I | MODCOMP IV | PRIME 300H | GEC 4080 |
---|---|---|---|---|---|
Number | 91 | 92 | 93 | 94 | 95 |
Performance Range | |||||
Lower | 85.3 | 98.3 | 448 | 184 | 289 |
Upper | 150.0 | 256.0 | 777 | 286 | 410 |
Benchmarks | |||||
Gibson Mix | - | 286.0 | - | - | 402 |
ADP Mix | - | - | - | - | 315 |
Pr Con Mix | - | - | - | - | 345 |
Ackn ratio | - | - | - | - | - |
Synth FORS | 105.0 | 192.0 | - | 220 | 367 |
Synth FORD | 147.0 | 26.6 | - | 297 | 450 |
Gamma Test | - | 132.0 | 612 | - | 234 |
Bit Test | - | 250.0 | 728 | 229 | 292 |
Binomial | - | 191.0 | 471 | 163 | 380 |
IF Test | - | 230.0 | 410 | 265 | 310 |
DOUBLE FUN | - | 27.4 | 886 | 278 | 473 |
Functions | - | - | - | - | - |
Synth ALGL | - | - | - | - | - |
Algol Mix | - | - | - | - | - |
Ackermann | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - |
Chess Mate | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - |
POWU | - | - | - | - | - |
GAMM F | - | - | - | - | - |
GAMM FD | - | - | - | - | - |
Name | EAL 1830 | IBM 1130 | VAR620/100 | NOVA 840 S | Ferr 1600B | MINIC 1 |
---|---|---|---|---|---|---|
Number | 96 | 97 | 98 | 99 | 100 | 101 |
Performance Range | ||||||
Lower | 38.7 | 15.40 | 10.40 | 18.6 | 127 | 33.8 |
Upper | 122.0 | 42.70 | 75.70 | 75.2 | 214 | 62.3 |
Benchmarks | ||||||
Gibson Mix | - | 14.00 | - | - | 216 | 32.0 |
ADP Mix | - | 55.80 | - | - | 126 | 45.4 |
Pr Con Mix | - | 64.60 | - | - | 135 | 87.7 |
Ackn ratio | - | - | - | - | - | - |
Synth FORS | - | - | - | 22.6 | - | - |
Synth FORD | - | - | - | - | - | - |
Gamma Test | 47.4 | 16.70 | 17.00 | 17.9 | - | - |
Bit Test | 107.0 | 23.50 | 132.00 | 63.9 | - | - |
Binomial | 24.0 | 8.43 | 17.80 | - | - | - |
IF Test | 251.0 | 92.30 | 137.00 | 357.0 | - | - |
DOUBLE FUN | - | - | 1.57 | - | - | - |
Functions | - | - | - | - | - | - |
Synth ALGL | - | - | - | - | - | - |
Algol Mix | - | - | - | - | - | - |
Ackermann | - | - | - | - | - | - |
GAMM ALGOL | - | - | - | - | - | - |
Chess Mate | - | - | - | - | - | - |
GAMM Asmbl | - | - | - | - | - | - |
POWU | - | - | - | - | - | - |
GAMM F | - | - | - | - | - | - |
GAMM FD | - | - | - | - | - | - |
Profiles do not appear for the E1-X8, MU5, CII IRIS80 and NOR SM4 because there is only one result for each of these machines.
Speed is instructions/millisecond on a logarithmic scale
(* follows results where additional figures are available)
1 630 + + + + + + + + + + + 1 000 + + + + + Functions ; + + + Binomial ; + Gamma Test; upper range + + + + + median value + Gibson Mix; + Bit Test ; 500 + + lower range + + + + Pr Con Mix; + ADP Mix ; + + + IF Test ; + + + + + + + + + + 200 + + + 181 +
Speed is instructions/millisecond on a logarithmic scale
1 710 + + + + + + + Synth FORD* + + + + + 1 000 + + DOUBLE FUN; + + + upper range + + + Synth FORS* + + + + + median value + Gibson Mix; + + 500 + + + Synth ALGL* + lower range + Pr Con Mix; + + ADP Mix ; + + + + + + + + + + + + + 200 + Ackermann ; + 190 +
Speed is instructions/millisecond on a logarithmic scale
2 750 + + + + + + DOUBLE FUN; + + 2 000 + + + + + + + + Synth FORD* + + + GAMM FD ; upper range + GAMM ALGOL; + Binomial ; + + 1 000 + GAMM F ; median value + Functions * + + Synth FORS* + Gamma Test; + Lower range + + + + Bit Test ; + + + Ackermann ; + IF Test ; + 500 + + + + + + + + + + + 306 +
Speed is instructions/millisecond on a logarithmic scale
1 860 + + + + + + + + + + + + + upper range + 1 000 + + Gibson Mix; + + upper range + + + + + + Pr Con Mix; median value + + + ADP Mix ; + + 500 + + + + + + lower range + + + + + + + + + + + + + less than + Ackermann * 206 +
Speed is instructions/millisecond on a logarithmic scale
9 370 + + + + + + + + + + + + DOUBLE FUN; + + 5 000 + + + Binomial ; + + Gamma Test; upper range + + + + + Gibson Mix; Bit Test ; median value + + + + + lower range + + + ADP Mix ; + Pr Con Mix; + 2 000 + + IF Test ; + + + + + + + + + + + + + 1 040 +
Speed is instructions/millisecond on a logarithmic scale
21 500 + DOUBLE FUN; + + 20 000 + + + + + + Synth FORD* Binomial * + + + + + upper range + + + 10 000 + + + + + + + median value + + Synth FORS* + + Gamma Test; + + + + 5 000 + + lower range + Bit Test * + + + + + + + + + + + + + less than + IF Test ; 2 390 +
Speed is instructions/millisecond on a logarithmic scale
449 + + + + + + + + + + + Synth FORD* + + + + + Functions; + + upper range + Synth FORS* + + + + + median value + + DOUBLE FUN; + + Gibson Mix; + + lower range + + ADP Mix ; Pr Con Mix; + 100 + + + + + + + + + ALGOL Mix * + + + + + + + 49.9 +
Speed is instructions/millisecond on a logarithmic scale
65.0 + + + + + + + 50.0 + + + + + + + + + + + DOUBLE FUN; + + + upper range + + Synth FORD; + Binomial ; median value + Synth FORS; Gamma Test; + 20.0 + Bit Test ; lower range + + + IF Test ; + + + + + + + + + + + 10.0 + + + + + + + + 7.22 +
Speed is instructions/millisecond on a logarithmic scale
397 + + + + + + + + + + + + + DOUBLE FUN; + + + 200 + + + + upper range + Bit Test ; + + Gamma Test; Binomial ; + median value + + IF Test ; + + Gibson Mix; lower range + ADP Mix ; Pr Con Mix; + 100 + + + + + + + + + + + + + + + 50.0 + + + + 44.1 +
Speed is instructions/millisecond on a logarithmic scale
640 + + + + + + 500 + + + + + + + + + + Synth FORD; + + + + upper range + + + Synth FORS; + median value + ADP Mix ; + 200 + + lower range + Gibson Mix; Pr Con Mix; + + + + + + + + + + + + 100 + + + + + + + + 71.2 +
Speed is instructions/millisecond on a logarithmic scale
1 680 + + + + + + + + + + + + 1 000 + + + + + + + + + Functions ; upper range + Synth FORS* + Gamma Test; + median value + IF Test ; + + lower range + Pr Con Mix; + Gibson Mix; + ADP Mix ; + + + + + + + + + + + + + + + + + 200 + + 187 +
Speed is instructions/millisecond on a logarithmic scale
3 260 + + + + + + + + + + + 2 000 + + + Synth FORD; Ackermann ; + + + + + DOUBLE FUN; upper range + + + + + Synth FORS* median value + + Gamma Test; 1 000 + Bit Test ; + Binomial ; + lower range + + + Ackn ratio; + + IF Test ; + + + + + + + 500 + + + + + + + + 362 +
Speed is instructions/millisecond on a logarithmic scale
8 190 + + + + + + + + + + + DOUBLE FUN; + 5 000 + + Binomial ; + + Gamma Test; + + upper range + + Functions ; + Gibson Mix; + + + median value + + Bit Test ; + + + Ackn ratio* lower range + Pr Con Mix; + 2 000 + ADP Mix ; + + + IF Test ; ALGOL Mix* + Ackermann * + + + + + + + + + + 1 000 + + + 910 +
Speed is instructions/millisecond on a logarithmic scale
11 900 + + + + + 10 000 + + + DOUBLE FUN; + + + + + + + + + + GAMM FD ; + + Synth FORD* upper range + Functions * + + Binomial * + median value + GAMM F ; + + Gamma Test* + Synth FORS* Bit Test * ALGOL Mix ; lower range + + + + + IF Test * + + + + + + 2 000 + + + + + + + + + 1 320 +
Speed is instructions/millisecond on a logarithmic scale
15 000 + DOUBLE FUN* + + + + + + + + + 10 000 + + + + + + + upper range + + Binomial * + + + + + Gamma Test* median value + 5 000 + + + + Synth FORS; + + lower range + + Bit Test * + + IF Test * + + + + + + + + + 2 000 + + + + + 1 670 +
Speed is instructions/millisecond on a logarithmic scale
22 800 + + + + 20 000 + + + + DOUBLE FUN; + + + + + + + + + + Synth FORD; + upper range + + Binomial ; GAMM FD ; + GAMM F ; + + median value + + Functions ; + Gamma Test; + + Synth FORS; lower range + + Bit Test ; + + 5 000 + + IF Test ; + + + + + + + + + + + + + + 2 530 +
Speed is instructions/millisecond on a logarithmic scale
220 + + + 200 + + + + + + + + + + + + + + Gamma Test; DOUBLE FUN; Functions ; + 100 + + upper range + Bit Test ; + Synth FORD* Binomial ; + + Synth FORS* median value + ADP Mix ; IF Test ; + GAMM F ; + + GAMM FD ; lower range + + + Gibson Mix; + + Pr Con Mix; 50.0 + + + + + + ALGOL Mix; + + + + + + + + + + 24.5 +
Speed is instructions/millisecond on a logarithmic scale
more than 1 200 + Ackermann ; + + + + 1 000 + + + + + + DOUBLE FUN; + + + Synth FORD* + + Gamma Test* + Functions ; + Binomial * + upper range + 500 + + + GAMM FD ; + median value + + Synth FORS* Bit Test * + Gibson Mix; + POWU; + GAMM F; lower range + + Pr Con Mix; IF Test * + + + ADP Mix ; ALGOL Mix ; + + + Ackn ratio; + + 200 + + + + + + + + + + 133 +
Speed is instructions/millisecond on a logarithmic scale
1 300 + + + + + + + + 1 000 + + + + DOUBLE FUN; + + + + Synth FORD* + Gamma Test* + Binomial * Functions ; + + upper range + + GAMM Asmbl; + GAMM FD * 500 + median value + + + Gibson Mix; Synth FORS* Bit Test * + lower range + POWU ; + Pr Con Mix; GAMM F * + IF Test * + + + ADP Mix ; + + + + + + + + 200 + + + + + + + 154 +
Speed is instructions/millisecond on a logarithmic scale
more than 1 150 + Ackermann ; + + + 1 000 + + + + + + + + + + + Synth FORD* ALGOL Mix ; + + Functions ; upper range + + 500 + + + + + median value + + Synth FORS* + + Gibson Mix; + + + lower range + Pr Con Mix; Ackn ratio; + + Synth ALGL* + ADP Mix ; + + + 200 + + + + + + + + + + + 128 +
Speed is instructions/millisecond on a logarithmic scale
120 + + + + + 100 + + + + + + ALGOL Mix ; + + + + upper range + + + + + 50.0 + + + + median value + + + + + + + + + lower range + + + Functions ; + + + 20.0 + + + + + + + + + + 13.3 +
Speed is instructions/millisecond on a logarithmic scale
63.7 + + + + + + 50.0 + + + + + + + + + + + + GAMM Asmbl; + + Gibson Mix; upper range + + + + median value + + 20.0 + ADP Mix ; Pr Con Mix; + lower range + + + + POWU ; + + + + + + + + + 10.0 + + + + + + + + 7.08 +
Speed is instructions/millisecond on a logarithmic scale
128 + + + + + + 100 + + + + + + + + + + + ADP Mix ; + POWU ; + Gibson Mix; upper range + + 50.0 + + + Pr Con Mix; median value + + Binomial ; + + Gamma Test; + Bit Test ; GAMM Asmbl; lower range + + + + IF Test ; + + + + + + + + 20.0 + + + + + + + less than + DOUBLE FUN; 14.3 +
Speed is instructions/millisecond on a logarithmic scale
156 + + + + + + + + + + + 100 + + + + + + + + + + ADP Mix ; upper range + Gibson Mix; POWU ; + + median value + 50.0 + + Pr Con Mix; lower range + + + + + + GAMM Asmbl; + + + + + + + + + + + + 20.0 + + + + 17.4 +
Speed is instructions/millisecond on a logarithmic scale
164 + + + + + + + + + + + + 100 + + + + + + + + ADP Mix ; + upper range + + Gibson Mix; Synth ALGL; + median value + ALGOL Mix ; + 50.0 + lower range + + + + + Functions ; + + + + + + + + + + + + + + 20.0 + + + 18.2 +
Speed is instructions/millisecond on a logarithmic scale
262 + + + + + + + 200 + + + + + + + + IF Test ; + + + ADP Mix ; + Gibson Mix; POWU ; + ALGOL Mix ; upper range + + 100 + + median value + Pr Con Mix; + + Binomial ; + lower range + Bit Test ; + Gamma Test; GAMM Asmbl; + + + + +Functions ; + 50.0 + + + + + + + DOUBLE FUN; + + + + + + 29.1 +
Speed is instructions/millisecond on a logarithmic scale
310 + + + + + + + + + + 200 + + + + + + + + + + + ADP Mix ; upper range + Gibson Mix; + POWU ; + median value + 100 + + Pr Con Mix; lower range + + + + + + GAMM Asmbl; + + + + + + + 50.0 + + + + + + + + + 34.5 +
Speed is instructions/millisecond on a logarithmic scale
510 + + + + + + + + + + + + + + + + + + + Gibson Mix; + upper range + GAMM Asmbl; + Binomial ; + IF Test ; + Gamma Test* median value + POWU ; + + + lower range + ADP Mix ; Bit Test ; + + + + + + + 100 + + + + + + + + + + + + less than + DOUBLE FUN; 56.7 +
Speed is instructions/millisecond on a logarithmic scale
685 + + + + + + + + 500 + + + + + + + ALGOL Mix ; + + + Synth ALGL* + Functions ; upper range + GAMM Asmbl; + + Gibson Mix; Synth FORS* + + Gamma Test; Binomial ; IF Test ; median value + POWU ; + + 200 + Bit Test ; + lower range + ADP Mix ; + + Pr Con Mix; + + + + + + + + + + 100 + + + + + + less than + Synth FORD* DOUBLE FUN; 76.1 +
Speed is instructions/millisecond on a logarithmic scale
886 + + + + + + + + + + + + + 500 + + + + + + Synth ALGL; upper range + ALGOL Mix ; + + Functions ; + + Synth FORS* median value + Gibson Mix; + + + + lower range + + + + 200 + + + + + + + + + + + + + + + less than + Synth FORD* 98.5 +
Speed is instructions/millisecond on a logarithmic scale
241 + + + + + 200 + + + + + + + + Gibson Mix; + + IF Test * + + ADP Mix ; upper range + + Functions * + 100 + + Synth FORS* Binomial * + + median value + + + + Bit Test * GAMM F * + + + lower range + + + + 50.0 + + + + + + + + + Synth FORD* DOUBLE FUN* + + + + less than + GAMM FD * 26.8 +
Speed is instructions/millisecond on a logarithmic scale
401 + + + + + + + + + + + + + + + + 200 + Gibson Mix; + Synth ALGL; + + upper range + IF Test ; + Binomial ; + ADP Mix ; + median value + Gamma Test; + + + lower range + + + 100 + + + + + + Synth FORS; + + + + + + + + + 50.0 + + + 44.6 +
Speed is instructions/millisecond on a logarithmic scale
2 500 + + + Ackermann * + + + 2 000 + + + + Chess Mate; + + + + + + Binomial * + + + ALGOL Mix * + Functions ; upper range + 1 000 + Synth ALGL* + Gamma Test* Bit Test; GAMM Asmbl; + Gibson Mix; Synth FORD* median value + Ackn ratio* GAMM FD* + Synth FORS* + + POWU ; lower range + + GAMM F * + IF Test ; + ADP Mix ; DOUBLE FUN* + + + 500 + Pr Con Mix; + + + + + + + + + + + + + 278 +
Speed is instructions/millisecond on a logarithmic scale
3 260 + + + + + + + + + + + 2 000 + + Binomial ; + Gamma Test* + Bit Test ; + + + + upper range + Synth ALGL; + GAMM Asmbl; + + Ackn ratio; + Gibson Mix; median value + Synth FORS* IF Test ; + POWU ; 1 000 + + Ackermann * + Pr Con Mix; lower range + + + + + ADP Mix ; + + + + + + + 500 + + + + + + Synth FORD* + less than + DOUBLE FUN; 362 +
Speed is instructions/millisecond on a logarithmic scale
2 120 + NOTE: These figures are correct as of April 1977, but + may change significantly 2 000 + + + + + + + + + + + + Synth FORD* + + + IF Test * ALGOL Mi x; GAMM FD * 1 000 + DOUBLE FUN* + upper range + + Binomial ; + + + median value + GAMM F * + + + Synth FORS* + Functions * lower range + Bit Test * + + 500 + + + + Gamma Test* + + + + + + + + + + + + + 235 +
Speed is instructions/millisecond on a logarithmic scale
8 200 + NOTE: These figures are correct as of April 1977, but + may change significantly + + + + + + + Ackermann ; + + + 5 000 + + + Synth FORD* + Gamm FD ; + + ALGOL Mix ; + upper range + Binomial * + + DOUBLE FUN; GAMM F * + + median value + Functions * + + Synth FORS* + + lower range + + Gamma Test* Bit Test * 2 000 + + + + IF Test * + + + + + + + + + + Ackn ratio; + + 1 000 + + 911 +
Speed is instructions/millisecond on a logarithmic scale
846 + + + + + + + + + + + + 500 + Chess Mate; + + + + ALGOL Mix ; upper range + Binomial ; + + + + + Functions ; GAMM ALGOL; + median value + + + + + + Bit Test ; + lower range + 200 + + + IF Test ; + + + + + + + + DOUBLE FUN; + + + + 100 + + 94.0 +
Speed is instructions/millisecond on a logarithmic scale
63.4 + + ADP Mix ; + + + + 50.0 + + + + Pr Con Mix; + + + + upper range + + + + + + + Gibson Mix; + + + median value + Synth ALGOL; + 20.0 + + + + + + + Synth FORS; + lower range + + + + + + + 10.0 + + + + + + + less than + Synth FORD; 7.05 +
Speed is instructions/millisecond on a logarithmic scale
267 + + + + + + + 200 + + + + + + + + + Synth ALGL; ALGOL Mix ; + + upper range + + Gibson Mix; IF Test ; + ADP Mix ; + Synth FORS; 100 + + median value + + Pr Con Mix; Functions ; + + Gamma Test; + + lower range + + + + + + + 50.0 + + + + + + + + + + + less than + Synth FORD; 29.7 +
Speed is instructions/millisecond on a logarithmic scale
more than 352 + Ackermann ; + + + + + + GAMM ALGOL; + + Synth ALGL; + + Chess Mate; + ALGOL Mix ; + 200 + + + + upper range + + + Gamma Test; + Gibson Mix; + + + median value + ADP Mix ; + + Functions ; + Ackn ratio; 100 + + Binomial ; + Synth FORS; lower range + + + + + + Synth FORD; IF Test ; + + + + + Bit Test ; + 50.0 + + DOUBLE FUN; + + + + 39.1 +
Speed is instructions/millisecond on a logarithmic scale
1 090 + + + 1 000 + + + + + + + + + + + Synth ALGL* + + + + 500 + + + upper range + + + Gamma Test; Binomial ; median value + + Synth FORS* Synth FORD* Functions ; + IF Test * lower range + + Gibson Mix; + + + + + + + 200 + + + + + + + + + + + + 121 +
Speed is instructions/millisecond on a logarithmic scale
962 + + + + + + + + + + Synth ALGL* + + + + + 500 + + + + Synth FORD* upper range + + + Synth FORS* + Gamma Test; + Gibson Mix; median value + + + + Binomial ; + DOUBLE FUN; lower range + + + + + IF Test ; + 200 + + + + + + + + Bit test ; + + + + + + 109 +
Speed is instructions/millisecond on a logarithmic scale
383 + + + + + + + + + + + + + + + 200 + + upper range + + Pr Con Mix; + + Gibson Mix; + + + median value + ALGOL Mix * + + + + + ADP Mix ; 100 + lower range + + + + + + + + + + + + + + 50.0 + Ackermann ; + + + 42.5 +
Speed is instructions/millisecond on a logarithmic scale
934 + + + + + + + + + + + + Functions ; + + + + + upper range + + + + + + Gibson Mix; + ALGOL Mix ; median value + + + + + + + lower range + + + ADP Mix ; + 200 + + + + Pr Con Mix; + + + + + + + + + + 104 +
Speed is instructions/millisecond on a logarithmic scale
1 830 + + + + + + + + + + + + + + 1 000 + + + + Binomial ; + Synth FORS; upper range + + Gamma Test; + + median value + Gibson Mix; + + Bit Test ; DOUBLE FUN; + Synth FORD; lower range + 500 + + + + + IF Test ; + + Synth ALGL* + + + + + + + + + + + + + 204 +
Speed is instructions/millisecond on a logarithmic scale
more than 1 800 + Ackermann ; + + + + + + + + + + + + + 1 000 + + + + Ackn ratio; upper range + Gibson Mix; Binomial ; + + + + + Bit Test * median value + DOUBLE FUN; + + Gamma Test; + 500 + ALGOL Mix * + lower range + Functions ; + + IF Test ; + + + + + + GAMM ALGOL; + + + + + + + + + + 200 +
Speed is instructions/millisecond on a logarithmic scale
3 580 + + + + + + + + + + + + + 2 000 + + Binomial ; + + + upper range + + + Synth FORS; Gamma Test; + + + DOUBLE FUN; median value + + + + 1 000 + + Synth FORD; lower range + + + + Bit Test ; + + + + + + IF Test ; + + + 500 + + + + + + 398 +
Speed is instructions/millisecond on a logarithmic scale
5 350 + + 5 000 + + + + + + Binomial ; + + + Gamma Test* + + upper range + Synth FORS; + + DOUBLE FUN* + + + + Gibson Mix; + 2 000 + + median value + + + + + GAMM ALGOL; + + Bit Test ; Functions ; + + + lower range + + ALGOL Mix * + 1 000 + + + + + + + + + + + IF Test * less than + Ackermann * Chess Mate; 595 +
Speed is instructions/millisecond on a logarithmic scale
29 000 + + Binomial * + + + + + + + 20 000 + + + + Gamma Test* GAMM F * + DOUBLE FUN* + + + upper range + + Synth FORS* + + Synth FORD; + + GAMM FD * + 10 000 + + + + + + + lower range + Gibson Mix; + Functions ; + Bit Test * + + Synth ALGL* + ALGOL Mix * + + 5 000 + + + + IF Test * + + + + less than + GAMM ALGOL; 3 220 +
Speed is instructions/millisecond on a logarithmic scale
746 + + + + + + + + + + 500 + + + + + + Gibson Mix; + + + upper range + ALGOL Mix ; + + ADP Mix ; + + median value + + + + + Pr Con Mix; lower range + + + + + + + + + + + Functions ; + + + + 100 + + + + + 82.9 +
Speed is instructions/millisecond on a logarithmic scale
529 + + 500 + + + + + + + + + DOUBLE FUN; + + + + + + + + upper range + + IF Test ; + Gamma Test; Binomial ; 200 + + median value + Gibson Mix; + + + + lower range + + + + ADP Mix ; + + Bit Test ; + + 100 + + + + + + + + + + + 58.8 +
Speed is instructions/millisecond on a logarithmic scale
421 + + + + + + + + + + + + ALGOL Mix ; + + + upper range + + 200 + + + + + + + median value + + + + + + + 100 + lower range + Functions ; + + + + + + + + + + + + + 50.0 + + 46.7 +
Speed is instructions/millisecond on a logarithmic scale
862 + + + + + + + + + + + DOUBLE FUN; + + 500 + + + Synth FORD; + + + + upper range + + + Synth FORS; Gamma Test; + median value + Binomial ; IF Test ; + + Synth ALGL; + lower range + Gibson Mix; + + + 200 + + + + + + + Bit test ; + + + + + + + + 100 + + 95.7 +
Speed is instructions/millisecond on a logarithmic scale
880 + + + + + + + + + + + + + 500 + + ALGOL Mix ; + upper range + + + + + + + + median value + + + + + + + + Functions ; lower range + 200 + + + + + + + + + + + + + + + + 97.8 +
Speed is instructions/millisecond on a logarithmic scale
1 560 + + + + + + + + + + + DOUBLE FUN; 1 000 + + + + Synth FORD; + + + + upper range + Gamma Test; Binomial ; + + + + Synth FORS; median value + 500 + + Gibson Mix; + + lower range + Synth ALGL; + + + + + + ADP Test ; + IF Test ; + + + + + + + + 200 + + + + 174 +
Speed is instructions/millisecond on a logarithmic scale
more than 2 170 + DOUBLE FUN; + + 2 000 + + Synth FORD; + + + + + + + + + + + Synth FORS; upper range + + 1 000 + + Binomial ; + Gamma Test; + Gibson Mix; + + median value + + + + + + + IF Test ; + lower range + + Bit Test ; + + + + + + + + Functions ; + + + + + less than + ALGOL Mix ; 241 +
Speed is instructions/millisecond on a logarithmic scale
2 460 + + + + + 2 000 + + DOUBLE FUN; + + + + + + + + + upper range + Gamma Test; + + + 1 000 + + + + Binomial ; median value + + + + + + + + lower range + IF Test ; + + 500 + + + + + + Bit Test ; + + + + + + + + 274 +
Speed is instructions/millisecond on a logarithmic scale
2 130 + + 2 000 + + + + + + + + + + + + DOUBLE FUN; + + + 1 000 + + Gamma Test; upper range + + + Binomial ; + Synth FORD; + median value + + + + Synth FORS; + lower range + + + Bit Test ; IF Test ; 500 + + + + + + + + + + + + + + + + + 237 +
Speed is instructions/millisecond on a logarithmic scale
2 100 + + 2 000 + + + + + + + + + + + DOUBLE FUN; + + Gamma Test; + + Binomial ; 1 000 + upper range + + + + + Synth FORD; + median value + Synth FORS; + + + Bit Test ; + Gibson Mix; IF Test ; + lower range + + 500 + + + + + + + Ackn ratio; + + + + + + + + + less than + Ackermann ; 234 +
Speed is instructions/millisecond on a logarithmic scale
4 390 + + + + + + + + DOUBLE FUN* + + + + + + + + + Gamma Test; + Synth FORD; IF Test ; 2 000 + upper range + + + Synth FORS; + + Binomial * median value + GAMM F * + GAMM FD * + + Bit Test ; + lower range + + + Gibson Mix; Synth ALGL* + 1 000 + + + + + + + ADP Mix ; + + + + + + + 500 + + 488 +
Speed is instructions/millisecond on a logarithmic scale
351 + + + + + + + + + + + + + 200 + + + GAMM ALGOL* + Gibson Mix; + Synth ALGL* ALGOL Mix * upper range + + + + + ADP Mix ; Ackermann ; + median value + + Synth FORS; + Functions ; + 100 + + Pr Con Mix; Ackn ratio* lower range + + + + + + + + + + + + + 50.0 + + + + + less than + Synth FORD; 39.0 +
Speed is instructions/millisecond on a logarithmic scale
398 + + + + + + + + + + + + + + + + 200 + + + + upper range + + + + Functions ; median value + + + ALGOL Mix ; + lower range + + + 100 + + + + + + + + + + + + + + 50.0 + + + + 44.2 +
Speed is instructions/millisecond on a logarithmic scale
2 090 + + 2 000 + + + + + + + + + + + + + + + 1 000 + + + upper range + + + + median value + Functions ; ALGOL Mix ; + + + lower ramge + + + + 500 + + + + + + + + + + + + + + + + + 233 +
Speed is instructions/millisecond on a logarithmic scale
1 630 + + + + + + + + + + DOUBLE FUN; + 1 000 + Synth FORD; + + + + + + Synth FORS; + upper range + + + Gamma Test; + + Binomial ; median value + + Gibson Mix; 500 + + Bit test ; + lower range + + + + + + ADP Mix ; + + Pr Con Mix; IF Test ; + + + + + + + + + 200 + + + 181 +
Speed is instructions/millisecond on a logarithmic scale
3 130 + + + + DOUBLE FUN; + + + + + + + Functions ; 2 000 + + + + + Gamma Test; + Synth FORD; + GAMM ALGOL; upper range + + + Binomial ; + + Synth FORS; + median value + 1 000 + + Gibson Mix; ALGOL Mix * + + Bit Test ; IF Test ; + lower range + + + Synth ALGL* + + + + + ADP Mix ; + Pr Con Mix; Chess Mate; + 500 + + + + + + Ackermann * + + + 348 +
Speed is instructions/millisecond on a logarithmic scale
4 970 + + + + + + + + + Synth FORD; + + + + + + + + upper range + + Synth FORS; + 2 000 + + + + median value + + Gibson Mix; + + + + + lower range + + + + Pr Con Mix; 1 000 + + + + ADP Mix ; + + + + + + + + + + 553 +
Speed is instructions/millisecond on a logarithmic scale
2 970 + + + + + + + + + 2 000 + + + + + + + + Synth FORD; + + Gamma test; + upper range + + Synth FORS; + Binomial ; + 1 000 + + Bit test ; + + lower range + + + + Gibson Mix; + + + + IF Test ; + + + 500 + + + + + + + + + + 331 +
Speed is instructions/millisecond on a logarithmic scale
175 + + + + + + ADP Mix ; + + + + + + + Pr Con Mix; 100 + upper range + IF Test ; + + + + + Gibson Mix; + + + + median value + + + + 50.0 + + + + + + lower range + + + Binomial ; + + + + + + + + + + 20.0 + less than + Gamma Test; 19.4 +
Speed is instructions/millisecond on a logarithmic scale
228 + ADP Mix ; Pr Con Mix; + + + 200 + + upper range + + + + + + + + + + + + + 100 + Bit Test ; + + + + median value + IF Test ; + + + + + + + + + 50.0 + + + + + + + + lower range + + + + + + less than + Binomial ; 25.3 +
Speed is instructions/millisecond on a logarithmic scale
1 070 + + + 1 000 + + + + + + + + Synth FORD; + + + + + + DOUBLE FUN; 500 + + + upper range + + Synth FORS; + Gamma Test; + median value + + + Binomial ; IF Test ; + lower range + Bit Test ; + + Gibson Mix; + ADP Mix ; + + + + + 200 + + + + + + + + + + + + 119 +
Speed is instructions/millisecond on a logarithmic scale
1 120 + + + + 1 000 + + + + Synth FORD* + + + + + Gamma Test; + + + upper range + + 500 + + Bit Test ; + Synth FORS* Binomial ; + + + median value + IF Test ; + + + + + + + lower range + + + + + + 200 + + + + + + + ALGOL Mix ; + + + less than + Synth ALGL* 125 +
Speed is instructions/millisecond on a logarithmic scale
2 250 + + + 2 000 + + + + + + + + + + + + + Synth FORD; + + 1 000 + Gibson Mix; upper range + + + + ADP Mix ; + median value + + + Synth FORS; + + lower range + + + + 500 + + + + Pr Con Mix; + + + + + + + + + + + + 249 +
Speed is instructions/millisecond on a logarithmic scale
more than 61.3 + Bit Test ; IF Test ; + + + + 50.0 + + + + + + + + + upper range + + + + + + + + + + median value + 20.0 + + + Synth FORS; + + + Functions ; + + + Binomial ; lower range + + + Gamma Test; + + + Synth FORD; 10.0 + + + + + + + + + 6.81 +
Speed is instructions/millisecond on a logarithmic scale
more than 210 + Ackermann * + 200 + + ADP Mix ; + + + + Pr Con Mix; Ackn ratio; + IF Test ; upper range + + + + + + + + 100 + + + + + + + median value + + + + + + + + 50.0 + + + + + Gibson Mix; + + Lower range + + + + Binomial ; + + + + + less than + Bit Test ; DOUBLE FUN; 23.3 +
Speed is instructions/millisecond on a logarithmic scale
954 + + + + + + + + + + + + + + + 500 + Synth FORD; + upper range + + + + + + + median value + + + + Synth FORS; + + + lower range + + + 200 + + + + + + + + + + + + + + + 106 +
Speed is instructions/millisecond on a logarithmic scale
more than 107 + Bit test ; + 100 + + + + + + + + + + + + + + IF Test ; upper range + 50.0 + + + + + + + median value + + + Synth FORS; + + + + + DOUBLE FUN; lower range + + + Gamma Test; + + 20.0 + Synth FORD; + + + + + + + + + + + 11.9 +
Speed is instructions/millisecond on a logarithmic scale
118 + + + + + Bit Test ; 100 + + + + + + + + + + IF Test ; upper range + + + + 50.0 + + + + + median value + + + + + + + + DOUBLE FUN; + lower range + Binomial ; + Gamma Test; + + + + 20.0 + + + + + + + + + + 13.2 +
Speed is instructions/millisecond on a logarithmic scale
more than 114 + Bit Test ; + + + 100 + + + + IF Test ; + + + + + + + + upper range + + + 50.0 + + + + + median value + DOUBLE FUN; + + Binomial ; + Gamma Test; + + + + lower range + + + Synth FORS; + + + 20.0 + + + + + Synth FORD; + + + + + + 12.7 +
Speed is instructions/millisecond on a logarithmic scale
3 340 + + + + + + + + + + + + 2 000 + + + + Synth FORD; + upper range + + + + + + + median value + + + 1 000 + Synth FORS; + + + lower range + + + + + + + + + + + 500 + + + + + + + 371 +
Speed is instructions/millisecond on a logarithmic scale
2 740 + + + + + + + + 2 000 + + + + + + + + + Synth FORD; + upper range + + + + + 1 000 + median value + + + Synth FORS; + + + lower range + + + + + + + 500 + + + + + + + + + + + + 304 +
Speed is instructions/millisecond on a logarithmic scale
3 120 + + + + + + + + + + 2 000 + + + + + + Synth FORD; + + upper range + + + + + + median value + 1 000 + + Synth FORS; + + + lower range + + + + + + + + + + 500 + + + + + + + + + 346 +
Speed is instructions/millisecond on a logarithmic scale
more than 111 + Akcn ratio; + + 100 + + + + Bit Test ; + + ADP Mix ; + + + upper range + + + + + + Pr Con Mix; 50.0 + + + + + + median value + + + + + + + + + + + + lower range + Gamma Test; + Binomial ; 20.0 + + Gibson Mix; + + + + + + + DOUBLE FUN; + + 12.3 +
Speed is instructions/millisecond on a logarithmic scale
more than 104 + ADP Mix ; Pr Con Mix; IF Test ; + 100 + + + + + + + + + Bit test ; + upper range + + + + + 50.0 + + + + + + + median value + + + + + + + + + Synth FORS; + + Gamma Test; + lower range + Binomial ; + + + Synth FORD; + Gibson Mix; + + + + + + DOUBLE FUN; + + 11.6 +
Speed is instructions/millisecond on a logarithmic scale
189 + + + + + + + + + + + + + Pr Con Mix; + + 100 + + + upper range + + + + + + median value + ADP Mix ; + + + + 50.0 + lower range + + Gibson Mix; + + + + + + + + + + + + + + + + + 21.0 +
Speed is instructions/millisecond on a logarithmic scale
456 + + + + + + + + + + + + + + + Pr Con Mix; + + + + upper range + + + + + ADP Mix ; median value + + + + + lower range + + Gibson Mix; + + 100 + + + + + + + + + + + + + + + + 50.7 +
Speed is instructions/millisecond on a logarithmic scale
309 + + + + + + + + + + 200 + IF Test ; + + + + + + Pr Con Mix; upper range + + + ADP Mix ; Bit Test ; + + + + Binomial ; median value + 100 + Gibson Mix; + + Gamma Test; + + + lower range + + + + + + + + + 50.0 + + + + + + + + less than + DOUBLE FUN; 34.4 +
Speed is instructions/millisecond on a logarithmic scale
339 + + + + + + + + + + + + 200 + + + + + + upper range + Synth FORD; + + + + + median value + + + Synth FORS; 100 + + + lower range + + + + + + + + + + + + 50.0 + + + + + + + 37.7 +
Speed is instructions/millisecond on a logarithmic scale
476 + + + + + + + + + + + + Gibson Mix; + + upper range + Bit test ; + + IF Test ; + + + 200 + Synth FORS; Binomial ; + + + median value + + + + + Gamma Test; + + + + + lower range + + + + + + + + + + + + + + less than + Synth FORD; DOUBLE FUN; 52.9 +
Speed is instructions/millisecond on a logarithmic scale
1 770 + + + + + + + + + + + + 1 000 + + + DOUBLE FUN; + + + upper range + + Bit Test ; + + + Gamma Test; median value + + + + 500 + lower range + Binomial ; + + IF Test ; + + + + + + + + + + + + + + + + + 197 +
Speed is instructions/millisecond on a logarithmic scale
687 + + + + + + + + 500 + + + + + + + + + + + Synth FORD; + upper range + DOUBLE FUN; + IF Test ; + + median value + Bit Test ; + Synth FORS; + 200 + lower range + + + + Binomial ; + + + + + + + + + + 100 + + + + + + + 76.4 +
Speed is instructions/millisecond on a logarithmic scale
1 030 + + 1 000 + + + + + + + + + + + + + + + 500 + DOUBLE FUN; + Synth FORD; + + upper range + Gibson Mix; + Binomial ; + Synth FORS; median value + Pr Con Mix; + + ADP Mix ; IF Test ; lower range + Bit Test ; + + + + + Gamma Test; + + + 200 + + + + + + + + + + + + + 115 +
Speed is instructions/millisecond on a logarithmic scale
more than 206 + IF Test ; + 200 + + + + + + + + + + upper range + + + Bit Test ; + 100 + + + + + + + + median value + + + + + + + 50.0 + + Gamma Test; + + + lower range + + + + + + + + + + + Binomial ; + + 22.9 +
Speed is instructions/millisecond on a logarithmic scale
more than 76.9 + IF Test ; + + + + Pr Con Mix; + + + ADP Mix ; + + 50.0 + + + + upper range + + + + + + + + + + median value + + + Bit Test ; + + + 20.0 + + + + Gamma Test; lower range + + + + Gibson Mix; + + + + + + + 10.0 + + + less than + Binomial ; 8.55 +
Speed is instructions/millisecond on a logarithmic scale
more than 84.3 + Bit Test ; IF Test ; + + upper range + + + + + + + + + 50.0 + + + + + + + + + + + + median value + + + + + + + + 20.0 + + + Binomial ; + Gamma Test; + + + + + + + + + lower range + + 10.0 + less than + DOUBLE FUN; 9.37 +
Speed is instructions/millisecond on a logarithmic scale
more than 112 + IF Test ; + + 100 + + + + + + + upper range + + + + Bit Test ; + + + + 50.0 + + + + + + median value + + + + + + + + + + + Synth FORS; + + lower range + + + Gamma Test; + + + + + + + + + 12.5 +
Speed is instructions/millisecond on a logarithmic scale
495 + + + + + + + + + + + + + + + + + + + Gibson Mix; upper range + 200 + + + + median value + + + + + Pr Con Mix; lower range + + ADP Mix ; + + + + 100 + + + + + + + + + + + + + + 55.0 +
Speed is instructions/millisecond on a logarithmic scale
138 + + + + + + + + 100 + + + Pr Con Mix; + + + + + + + upper range + + + + + 50.0 + median value + ADP Mix ; + + + + + lower range + + + Gibson Mix; + + + + + + + + + 20.0 + + + + + + + 15.3 +
This Appendix gives the parameters calculated by the analysis. The performance of each benchmark is assumed to be given approximately by the product of two factors - one from the benchmark and the other from the machine. A least squares process is used to provide the best approximation, taking into account the relevance or weighting factors from the benchmarks.
Some benchmarks give performance ratings very close to the machine factors which are calculated from all the available data. The fits between the machine factors and the available data are expressed as average ratios; these variability factors are listed below under Var. Benchmark. One can see from these that Synth FORS conforms to the consensus view closely (between 1/1.233 = .811 and 1.233). On the other hand, the Synthetic FORD program and Ackermann deviate by factors of 2.1 and 3.5 respectively.
The required weighting (Relevance) for each benchmark is obtained from the inverse of the logarithm of the corresponding variability factor. In fact, rather than take this inverse itself, an estimate is used which allows for the number of results obtained. This is because a small variance on a large sample is more indicative than the same variance on a smaller sample. Use of these weights gave much greater consistency to the consensus data by more of the benchmarks than was achieved with the intuitive weights. The resulting relevance factors do not represent the importance that the benchmarks may have to particular applications but rather give emphasis to those benchmarks which are likely to indicate good overall performance.
The benchmark factors are listed in this Appendix and give the relationships between the various scales. For instance, the ADP Mix is roughly 50% larger than the Gibson Mix. When program times are involved, the factors can be used to predict these times from the Gibson Mix, G. For example we have
Benchmark Program Time in Seconds (Gibson Mix = G) Gamma (MM × MS) / (0.188 × G) Bit Test (NB × IC × IB2) / (4.85 × G) Binomial (IA × IB × IE2) / (4.33 × ID × G) IF Test (N2) / (23.7 × G) DOUBLE FUN (IA) / (0.0504 × G)
No. | Benchmark | Relevance | Benchmarkfactor | Var. Benchmark |
---|---|---|---|---|
1 | Gibson Mix | 3.192 | 1.00000 | 1.328 |
2 | ADP Mix | 1.849 | 1.54000 | 1.616 |
3 | Pr Con Mix | 1.808 | 2.26000 | 1.617 |
4 | Ackn ratio | 1.034 | 1.39000 | 2.005 |
5 | Synth FORS | 4.295 | 0.72200 | 1.233 |
6 | Synth FORD | 1.168 | 0.32000 | 2.144 |
7 | Gamma Test | 2.486 | 0.18800 | 1.437 |
8 | Bit Test | 1.702 | 4.85000 | 1.697 |
9 | Binomial | 1.550 | 4.33000 | 1.791 |
10 | IF Test | 1.374 | 23.70000 | 1.935 |
11 | DOUBLE FUN | 0.988 | 0.05040 | 2.468 |
12 | Functions | 2.435 | 0.00070 | 1.419 |
13 | Synth ALGL | 1.588 | 0.38900 | 1.629 |
14 | ALGOL Mix | 1.804 | 0.00246 | 1.591 |
15 | Ackermann | 0.593 | 0.01470 | 3.546 |
16 | GAMM ALGOL | 1.016 | 0.05660 | 1.810 |
17 | Chess Mate | 0.635 | 0.02010 | 2.243 |
18 | GAM Assmbl | 2.632 | 0.31900 | 1.278 |
19 | POWU | 3.524 | 1.47000 | 1.207 |
20 | GAMM F | 3.269 | 0.20100 | 1.231 |
21 | GAMM FD | 1.217 | 0.12000 | 1.748 |
Relevance factor for the benchmark is 3.192
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + Hon H316 F; 0.5 + Hon DDP516; + + IBM 1130 ; + PDP 11/20 ; + + + + MINIC 1 ; + CDC 7600 ; DEC KL10 ; Univ11 21 ; CTL MOD1 S; + ICL 4/50 ; XDS SIG5 ; CTL MOD1 F; + + 370/145 ; 370/155 ; Bur 6714 F; Hon 6040 ; + 360/50 ; 360/135 ; ICL 4/75 P; DEC KI10 ; + ICL 4/70 ; ICL 4/72 ; Hon 6060 ; Univ1108 ; + Univ1106 U; Univ1110 ; HP 2100A ; 1.0 + 360/65G ; 360/65H ; 1904S FP I; Cyber 72 ; + 360/85 ; 1906A ; 1906S ; Bur 6715 S; + 1903 EMU ; CDC 3600 ; Hon 6025 ; + 370/165 ; 1902S 25S ; 1903S 31S ; 1904A FP ; + ICL 4120/2; Bur 5500 ; CDC 3300 ; CDC 6600 ; + 1901A 10SC; ICL 4130/2; Hon 66/40 ; INTER5 ; + 1902A 20SC; 1903A SC ; 1903T ; Cyber 73 ; + XDS SIG9 ; GEC 4080 ; Ferr 1600B; + 1905F ; ICL KDF9 ; + 360/75 ; Hon GE 635; + + + 1905E Acc ; + HP 3000 I ; + 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.849
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + DEC KL10 ; + Univ1110 ; + Univ1108 ; + + Hon 6060 ; + ICL 4/75 P; Univ1106 U; + 360/65G ; 360/65H ; ICL 4/70 ; ICL 4/72 ; + 1906S ; CDC 3600 ; Hon 6025 ; + 360/85 ; 360/50 ; 370/165 ; 1906A ; + XDS SIG5 ; Ferr 1600B; + 370/155 ; 1904A FP ; CDC 3300 ; + 370/135 ; 1903T ; + + 360/75 ; 1901A 10SC; GEC 4080 ; + 1.0 + 370/145 ; ICL 4/50 ; Bur 5500 ; CTL MOD1 S; + CTL MOD1 F; MINIC 1 ; + 1905F ; ICL KDF9 ; XDS SIG9 ; + Hon GE 635; + 1902S 25S ; 1903S 31S ; ICL 4130/2; + 1903 EMU ; HP 2100A ; + + 1903A SC ; + 1902A 20SC; 1905E Acc ; + + + + + + 2.0 + + Hon DDP516; + IBM 1130 ; + + INTER5 ; + + PDP 11/20 ; + + ICL 4120/2; more than 3.0 + Univ418III; Hon H316 F;
Relevance factor for the benchmark is 1.808
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + + Univ1108 ; + CDC 3600 ; + Univ1106 U; XDS SIG9 ; + 1906A ; + Univ1110 ; + + 360/65G ; 360/85 ; ICL 4/50 ; + 360/65H ; 360/50 ; ICL 4/75 P; 1904A FP ; + ICL 4/70 ; + 370/165 ; ICL 4/72 ; ICL KDF9 ; + 370/135 ; 370/145 ; 1906S ; Hon GE 635; + 370/155 ; Ferr 1600B; + 1901A 10SC; 1902S 25S ; 1903S 31S ; + ICL 4130/2; 1.0 + 1903A SC ; GEC 4080 ; + 360/75 ; 1902A 20SC; + + + + + CDC 3300 ; + Hon DDP516; + HP 2100A ; + + CTL MOD1 F; + + INTER5 ; CTL MOD1 S; + + MINIC 1 ; 2.0 + ICL 4120/2; + + PDP 11/20 ; + + + IBM 1130 ; + + + more than 3.0 + Univ418III; Hon H316 F;
Relevance factor for the benchmark is 1.034
Ratio on a logarithmic scale
less than 1/3 + + + + + ICL 2980 ; + + + + 0.5 + + DEC KI10 ; + + ICL 4/70 ; + + + + + 370/158 ; ICL 4/75 P; + + ICL KDF9 * + 370/165 * + Bur 5500 ; + + 1.0 + 1906A * CII IRIS80; NOR SM4 ; + + 1906S ; + + + + + Cyber 73 ; + + + + + + + 2.0 + + + PDP 11/20 ; + + + + + + more than 3.0 + Hon DDP516;
Relevance factor for the benchmark is 4.295
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + + + + 1905F ; + NOVA 840 S; + PDP 11/40S; + + ICL 4120/2; Hon H316 F; + + Bur 5500 ; + + 360/168 M ; Amdahl 470; + 370/168 * ICL 2970 * DEC 20 ; PDP 11/10 ; + 360/67 * ICL 4/72 * ICL 2980 * XDS SIG9 ; + 360/195 * ICL 4/70 * ICL 4/75 P* 1906A * 1.0 + 360/30 MI ; 1906S * Bur 6714 F* DEC KI10 ; + 370/158 * ICL 4/50 * 1904S FP I* Hon 6060 ; + 370/145 ; Hon 6040 ; ICL KDF9 ; Univ1108 ; + 370/155 * 1904A FP * 1905E Acc * ICL 4130/2; + Bur 6715 S* Cyber 173 ; DEC KL10 ; Univ11 21 ; + 360/65H * Cyber 72 ; XDS SIG5 ; XDS SIG6 * + 360/50 * CDC 7600 * Univ1110 ; PDP 11/34F; + Univ1106 U; PDP 11/34S; PDP 11/55 ; PDP 11/60 ; + PDP 11/70 ; HP 2100 S ; HP 3000 I ; PRIME 300H; + Hon 66/40 ; GEC 4080 ; + CDC 6600 ; + + + + 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.168
Ratio on a logarithmic scale
less than 1/3 + 1904A FP * 1904S FP I* ICL 4120/2; ICL 4130/2; + ICL KDF9 ; HP 3000 I ; + 1906S * + + + 1905E Acc * + PDP 11/40S; + + 0.5 + PDP 11/10 ; Hon H316 F; + + PDP 11/34S; + Bur 5500 ; + + + + + + + Cyber 173 ; + + Cyber 72 ; + + Bur 6714 F* 1.0 + + 1906A * + 360/30 MI ; DEC 20 ; DEC KI10 ; + ICL 4/50 * + CDC 7600 ; + 370/168 * + Bur 6715 S* HP 2100 S ; PRIME 300H; GEC 4080 ; + Amdahl 470; DEC KL10 ; + Univ1108 ; Univ11 21 ; PDP 11/60 ; + 360/67 * 370/145 ; ICL 4/72 * Hon 6040 ; + 370/158 ; ICL 4/75 P* ICL 2980 * Hon 6060 ; + ICL 4/70 * ICL 2970 * XDS SIG9 ; PDP 11/34F; + PDP 11/55 ; PDP 11/70 ; + Univ1106 U; + 360/50 * XDS SIG5 ; 2.0 + + 360/195 * Univ1110 ; + XDS SIG6 * + 360/65H * + + Hon 66/40 ; + + + more than 3.0 +
Relevance factor for the benchmark is 2.486
Ratio on a logarithmic scale
less than 1/3 + INTER5 ; + + + + + + + + NOVA 840 S; 0.5 + + + + PDP 11/10 ; Hon DDP516; + ICL 2970 * VAR620/100; + PDP 11/34S; PDP 11/40E; Hon H316 F; + IBM 1130 ; + GEC 4080 ; EAL 1830 ; + + ICL 2980 * + 1903A SC ; + HP 3000 I ; + 360/67 ; 360/195 ; 1902A 20SC; ICL 4130/2; + 370/168 * Amdahl 470; Cyber 73 ; PDP 11/40S; + 370/158 ; HP 2100A ; 1.0 + 360/30 MI ; 1905F ; + 360/168 M ; 1903T * 1904A FP ; Bur 6714 F; + 370/135 ; 370/155 ; 1906A * Bur 6715 S; + Cyber 72 ; Hon 6025 ; Hon 6040 ; Univ1106 U; + Cyber 173 ; Hon 66/40 ; XDS SIG5 ; MODCOMP IV; + Bur 5500 ; Hon 6060 ; + 360/65G ; 360/85 ; DEC 20 ; Univ11 21 ; + + 370/165 ; ICL 4/50 ; ICL 4/72 ; Hon 66/60 ; + ICL 4/70 * DEC KI10 ; Univ1108 ; + DEC KI10 ; + 1906S * + CDC 7600 * XDS SIG6 ; + + CDC 6600 * 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.702
Ratio on a logarithmic scale
less than 1/3 + PDP 11/20 ; + + + + + + Bur 6715 S; + Bur 5500 ; + Hon 66/60 ; 0.5 + + Hon 6040 ; + + + + 360/195 * Hon 6025 ; + Cyber 173 ; CDC 7600 * Hon 66/40 ; + 360/168 M * + 360/67 ; DEC 20 ; + Amdahl 470; ICL 2980 * CDC 6600 ; + ICL 2970 * ICL ATLAS1; + 1902A 20SC; 1903A SC ; 1903T ; Univ1108 ; + 370/168 * 1904A FP ; 1905E Acc * DEC KI10 ; + 360/30 MI ; 370/158 ; ICL 4/72 * Cyber 72 ; + 360/65G ; 370/165 ; ICL 4/70 * DEC KL10 ; 1.0 + Univ1106 U; Univ11 21 ; XDS SIG5 ; PRIME 300H; + 360/85 ; Cyber 73 * GEC 4080 ; IBM 1130 ; + 1906A ; + + 370/135 ; ICL 4/50 * MODCOMP IV; + Univ418III; XDS SIG6 ; HP 2100A ; + + + + + 1906S ; HP 3000 I ; EAL 1830 ; + NOVA 840 S; + + + Hon H316 F; 2.0 + + + + Hon DDP516; + + PDP 11/40E; + + + more than 3.0 + PDP 11/10 ; PDP 11/34S; PDP 11/40S; VAR620/100;
Relevance factor for the benchmark is 1.550
Ratio on a logarithmic scale
less than 1/3 + Univ418III; IBM 1130 ; + + EAL 1830 ; + + + + PDP 11/20 ; + + 0.5 + + + Hon DDP516; + INTER5 ; Hon H316 F; + + VAR620/100; + PDP 11/10 ; PDP 11/40E; + + PRIME 300H; + + Bur 5500 ; MODCOMP IV; + + 370/158 ; Bur 6715 S; + 1903A SC ; XDS SIG5 ; PDP 11/40S; + 1902A 20SC; 1.0 + Hon 6040 ; + 360/30 MI ; 1904A FP ; Bur 6714 F; Hon 66/60 ; + 370/135 ; 370/168 * DEC KL10 * Univ1106 U; + ICL 4/50 ; 1903T ; 1905E Acc * 1905F ; + 360/67 ; Amdahl 470; ICL 2970 ; Hon 6025 ; + ICL 2980 * Hon 6060 ; Hon 66/40 ; DEC 20 ; + 370/168 M * Cyber 72 ; Cyber 73 ; Univ1108 ; + 360/65G ; ICL 4/70 * ICL 4/72 * ICL ATLAS1; + 360/85 ; DEC KI10 ; Univ11 21 ; XDS SIG6 ; + 1906A * HP 2100A ; HP 3000 I ; GEC 4080 ; + Cyber 173 ; + 370/165 ; + 1906S ; + + 2.0 + + 360/195 * + CDC 6600 * + + + + + + CDC 7600 * more than 3.0 +
Relevance factor for the benchmark is 1.374
Ratio on a logarithmic scale
less than 1/3 + 360/195 ; + CDC 6600 * + + + + CDC 7600 * + + + 0.5 + Cyber 173 ; + + + 360/65G ; 360/85 ; Bur 5500 ; Hon 6060 ; + 360/67 ; 370/165 ; Univ1106 U; Univ11 21 ; + 370/168 M * Amdahl 470; ICL 2980 * ICL ATLAS1; + 370/158 ; Bur 6715 S; Cyber 72 ; + 370/168 * 1902A 20SC; Cyber 73 ; Hon 66/60 ; + DEC 20 ; MODCOMP IV; + ICL 4/70 * ICL 4/72 * 1906A ; Hon 66/40 ; + 360/30 MI ; + DEC KI10 ; Univ1108 ; + + Bur 6714 F* XDS SIG5 ; GEC 4080 ; + 370/135 ; 1.0 + 370/155 ; ICL 4/50 * 1906S ; Hon 6040 ; + 1904A FP ; Univ418III; XDS SIG6 ; + 1903T ; + PRIME 300H; + 1905F ; Hon 6025 ; + ICL 4130/2; + + DEC KL10 ; + ICL 2970 * HP 3000 I ; + PDP 11/34S; + 1903A SC ; 1905E Acc * INTER5 ; PDP 11/40E; + + + + HP 2100A ; 2.0 + + PDP 11/20 ; + PDP 11/40S; + + + + + + more than 3.0 + PDP 11/10 ; Hon H316 F; EAL 1830 ; IBM 1130 ; more than 3.0 + VAR620/100; NOVA 840 S;
Relevance factor for the benchmark is 0.988
Ratio on a logarithmic scale
less than 1/3 + 1902A 20SC; 1903T ; 1904A FP ; 1906S ; + PDP 11/20 ; HP 2100A ; HP 3000 I ; VAR620/100; + Hon DDP516; Hon H316 F; + + Bur 5500 ; + 1905E Acc * + 1903A SC ; ICL ATLAS1; + + 0.5 + + + + + + + + + 1906A * PDP 11/34S; PDP 11/40E; + + + Bur 6715 S; + + Cyber 72 ; + 360/50 ; 1.0 + Cyber 73 ; PDP 11/40S; + Cyber 173 ; + + ICL 2980 ; + PRIME 300H; + + 370/158 ; + 360/30 MI ; ICL 2970 * GEC 4080 ; + ICL 4/50 ; CDC 6600 * XDS SIG5 ; + MODCOMP IV; + + 360/65H ; CDC 7600 * DEC 20 ; + 370/135 ; DEC KI10 ; + 360/85 ; ICL 4/72 ; + 370/165 ; ICL 4/70 ; Hon 6025 ; Hon 6040 ; 2.0 + Hon 6060 ; Univ1106 U; + + 370/168 * Amdahl 470; DEC KL10 * + Hon 66/60 ; + 360/67 ; + + Univ1108 ; + + more than 3.0 + 360/195 * 370/168 M * Hon 66/40 ;
Relevance factor for the benchmark is 2.435
Ratio on a logarithmic scale
less than 1/3 + + + + + + + Hon 66/40 ; + + 0.5 + Hon GE 635; + + + + S 4004/55 ; + 1903A SC ; + Hon 6030 ; + CDC 7600 ; + 1903 EMU ; Hon 6050 ; + Cyber 73 ; CDC 6600 ; PDP 11/10 ; + + ICL 2970 * + + Bur 5500 ; ICL KDF9 ; + Amdahl 470; ICL 4130/2; Bur 6714 F; 1.0 + 360/67 * ICL 2980 * TR440 ; + TR4 ; + ICL ATLAS1; + 1904S FP I; + 370/155 ; 370/165 ; 370/168 * + 1906A ; + 1904A FP ; 1905E Acc * + ICL 4/72 ; + ICL 4/50 ; ICL 4/70 ; ICL 4/75 P; + 360/50 ; + 360/65G ; + + + CDC 3600 ; + Univ1108 ; 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.588
Ratio on a logarithmic scale
less than 1/3 + XDS SIG6 * + + + + + + + + 0.5 + + + + + Cyber 72 * CDC 7600 * + + + ICL 4/75 P* + Univ1108 * + DEC KL10 * + 360/65H * Hon 6060 ; + + + Hon 6040 ; + 1.0 + ICL 4120/2; + + 1903 EMU ; + 1906A * + + 1906S ; + 1904S FP I; + 1904A FP * 1905F ; ICL KDF9 * + + ICL 4130/2; + + Bur 6714 F* + + + 2.0 + Bur 6715 S* + Bur 5500 ; + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.804
Ratio on a logarithmic scale
less than 1/3 + Hon 66/40 ; + + + + XDS SIG6 ; + + + 360/50 * + 0.5 + + ICL 4/50 ; + + CDC 7600 * + 370/165 * CDC 6600 * + + ICL 4/70 ; + + + + + Cyber 73 * + 370/168 ; + TR4 ; Univ1108 * + 1.0 + 1903 EMU ; MU5 ; CDC 3300 * TR440 ; + CDC 3600 ; EL-X8 ; + + + + 1903A SC ; 1904S FP I; Hon GE 635; + 1906A * + ICL 2980 ; ICL KDF9 * + ICL 2970 ; ICL ATLAS1; + ICL 4130/2; + ICL 4/75 P; 1904A FP ; Hon 6050 ; + + + Bur 5500 ; Hon 6030 ; + S 4004/55 ; 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 0.593
Ratio on a logarithmic scale
less than 1/3 + 360/75 * CDC 6600 ; DEC KI10 ; + 360/65H ; + + CDC 3300 ; Univ1108 ; + + + + + 0.5 + + + + 370/165 * + + 360/67 ; + + + + + + + 1906S * + + 1.0 + + + ICL KDF9 ; + + + + + + + + 370/158 ; + + + + 2.0 + + ICL 2980 ; + + + + + + 1906A * + more than 3.0 + ICL 4/70 ; ICL 4/75 P; Bur 5500 ; Cyber 73 ; PDP 11/20 *
Relevance factor for the benchmark is 1.016
Ratio on a logarithmic scale
less than 1/3 + CDC 7600 ; + + + + + + + + 0.5 + + Cyber 73 ; + + + + + + + + + + CDC 6600 ; + + + 1.0 + + + ICL ATLAS1; + + + 360/67 ; + + Univ1108 ; + + ICL KDF9 * + + + + + 2.0 + + + + Bur 5500 ; + + + + + more than 3.0 +
Relevance factor for the benchmark is 0.635
Ratio on a logarithmic scale
less than 1/3 + CDC 6600 ; + + + + + + + + 0.5 + + Univ1108 ; + + + + + + + + + + + + + 1.0 + + + + + + + + + + + + + ICL ATLAS1; + + Bur 5500 ; 2.0 + 1906A ; + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 2.632
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + + + + + + + + 1902S 25S ; 1903S 31S ; + + + 1903A SC ; + 1902A 20SC; + + + 1.0 + + + 1906A ; + ICL 4/72 ; + 1903T ; 1906S ; + 1904A FP ; + + 1901A 10SC; + + + + + + + 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 3.524
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + + + + + + + + + 1901A 10SC; + + + ICL 4/72 ; + ICL 4/70 ; 1906A ; + + 1906S ; 1.0 + 1903T ; 1904A FP ; + + 1903S 31S ; + 1902S 25S ; + + + 1903A SC ; + 1902A 20SC; + + + + + + + 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 3.269
Ratio on a logarithmic scale
less than 1/3 + + + + + + + + + 0.5 + + + + + + + + + + + ICL 4/72 * 1906A * + ICL 4/70 ; + 1905E Acc * + + ICL 4/50 ; 1.0 + 370/168 ; ICL 2970 ; DEC KL10 * + 360/67 ; + + Amdahl 470; ICL 2980 * + + + + + + + + + CDC 7600 * + + 2.0 + + + + + + + + + more than 3.0 +
Relevance factor for the benchmark is 1.217
Ratio on a logarithmic scale
less than 1/3 + 1905E Acc * + + + + + + + + 0.5 + + + + + + + + + + + + + ICL 4/50 ; + + DEC KL10 * 1.0 + 1906A * + + ICL 4/70 ; ICL 4/72 * CDC 7600 * + + Amdahl 470; + + 360/67 ; + 370/168 ; + ICL 2970 * + ICL 2980 ; + + + + + 2.0 + + + + + + + + + more than 3.0 +
The first part gives the structural relationship between the computers considered.
Note the comment against some computers giving further details of model type.
( start of tree structure of machines' ( start of machines with the 360 architecture ( start of the IBM 360 machines themselves (;'360/65G ') (;'360/65H ' also included are I and J processors) (;'360/67 ') (;'360/75 ') (;'360/85 ') (;'360/195 ') (;'360/50 ') (;'360/30 MI') end of IBM 360 machines) ( start of IBM 370 machines (;'370/135 ') (;'370/145 ') (;'370/155 ') (;'370/158 ') (;'370/165 ') (;'370/168 ') (;'370/168 M' high speed multiply) end of IBM 370 machines) ( start of Amdahl machines (;'Amdahl 470' version 6) end of Amdahl machines) ( start of ICL System 4 machines (;'ICL 4/50 ') (;'ICL 4/70 ') (;'ICL 4/72 ') (;'ICL 4/75 P' Paging machine) end of ICL System 4 machines) ( start of Siemens machines (;'S 4004/55 ') end of Siemens machines) end of the 360 architecture machines) ( ( start of ICL 1900 range of machines (;'1901A 10SC' Scientific and Commercial features) (;'1902A 20SC' Scientific and Commercial features) (;'1902S 25S ' Scientific feature) (;'1903 EMU ' floating point) (;'1903A SC ' Scientific and Commercial features) (;'1903S 31S ' Scientific feature) (;'1903T ') (;'1904A FP 'Hardware floating point) (;'1904S FP I' Hardware floating point, Interim store) (;'1905E Acc ' Hardware accumulators) (;'1905F ') (;'1906A ' 4-way interleaved store) (;'1906S ') end of ICL 1900 series) end of 1900 architecture machines) ( ( start of 2900 Series and MU5 (; 'ICL 2970 ') (; 'ICL 2980 ') (; 'MU5 ') end of 2900 Series and MU5) ) ( ( isolated machine (;'ICL ATLAS1') end of ATLAS) ) ( ( start of ICL 4100 Series (;'ICL 4120/2' 2 microsecond store) (;'ICL 4130/2' 2 microsecond store) end of ICL 4100 Series) ) ) ( ( start of Burroughs descriptor machines. Although these machines are not completely compatible, they are sufficiently similar to be regarded as in the same range. (;'Bur 5500 ') (;'Bur 6714 F' fast store) (;'Bur 6715 S' slow store) end of Burroughs descriptor machines) end of descriptor machines) ( ( start of the CDC 3000 Series of machines (;'CDC 3300 ') (;'CDC 3600 ') end of the CDC 3000 Series) end of the 3000 architecture) ( ( start of the CDC 6000 Series of machines (;'Cyber 72 ') (;'Cyber 73 ') (;'Cyber 173 ') (;'CDC 6600 ') (;'CDC 7600 ') end of 6000 Series) end of 6000 architecture) ( ( start of Honeywell 6000 or level 66 machines (;'Hon GE 635') (;'Hon 6025 ') (;'Hon 6030 ') (;'Hon 6040 ') (;'Hon 6050 ') (;'Hon 6060 ') (;'Hon 66/40 ') (;'Hon 66/60 ') end of Honeywell 6000 machines) end of Honeywell architecture) ( ( start of DEC System 10 machines Timing sometimes inconsistent (;'DEC 20 'without cache) (; 'DEC KI10 ') (; 'DEC KL10 ') end of System 10) end of range) ( (isolated machines, not in a range (;'ICL KDF9 ') end of isolated machine type) end of isolated machine range) ( (Telefunken machines (;' TR4 ') (;' TR440 ') end of Telefunken machines) ) ) ( (isolated machine produced by Electrologica (;' EL-X8 ') end of isolated machine) ) ( ( start of Univac 1100 Series of machines (;'Univ1106 U') (;'UniV1108 ' ) (;'UniV1110 ' ) (;'Univ1121 ' The 1100-21 ) end of Univac 1100 Series) end of 1100 architecture) ( ( start of Interdata Series (;'INTER 5 ') end of Interdata Series) ) ( ( start of Univac 400 Series (;'Univ418III') end of Univac 400 Series) ) ( ( Sigma machines (;'XDS SIG5 ') (;'XDS SIG6 ') (;'XDS SIG9 ') end of Sigma machines) ( CII Sigma machines (;'CII IRIS80') ) ) ( start 'range' of all 16-bit mini-computers ( start of DEC PDP11 Series (;'PDP 11/10 ') (;'PDP 11/20 ') (;'PDP 11/34F' floating point) (;'PDP 11/34S' software floating point with RSX11) (;'PDP 11/40E' software floating point and hardware arithmetic) (;'PDP 11/40S' software floating point with RT11) (;'PDP 11/55 ' bipolar memory and floating point) (;'PDP 11/60 ' cache and floating point) (;'PDP 11/70 ' cache and floating point) end of compatible minis) ( start of Honeywell Series 16 machines (;'Hon DDP516' without HS arithmetic) (;'Hon H316 F' with high speed option) end of Honeywell Series 16) ( start of Modular 1 Series (;'CTL MOD1 S' slow store) (;'CTL MOD1 F' fast store ) end of Modular 1 Series) ( (;' NOR SM4 ' Norwegian mini) ) ( Hewlett-Packard 2000 Series (;'HP 2100A ') (;'HP 2100S ') ) ( Hewlett-Packard 3000 series (;'HP 3000 I' Model I) ) ( ( ; 'MODCOMP IV' ) ) ( (;'PRIME 300H' hardware floating point) ) ( (;'GEC 4080 ') ) ( (;'EAL 1830 ') ) ( (;'IBM 1130 ') ) ( (;'VAR620/100') ) ( (;'NOVA 840 S') ) end of 16-bit mini-computers) ( ( start of Ferranti 1600 Series (;'Ferr 1600B') end of 1600 Series) ) ( start or small minicomputers ( start of MINIC Series (;'MINIC 1 ') end of MINIC Series) end of small minis) )' end of tree structure of machines softwarefile: ( ( ;'Assembler ' ;'ALGOL 60 ' ;'ALGOL 60+ ' with extensive optimisation ;'al double ' option with double length ;'ALGOL W ' ;'ALGOL 68 ' ;'Pascal ' ;'CORAL 66 ' ;'RTL/2 ' ;'IMP ' dialect of ALGOL 60 ;'SIMULA ' ;'BCPL ' ;'FORTRAN ' ;'FORTRAN+ ' with extensive optimisation ;'PL/I ' ;'PL/I + ' with extensive optimisation ) )' Application areas described as a tree structure. ( Contents: application areas divided in groups. Main groups are Scientific and Commercial. Each group is divided in language groups. (;'Scientific' (;'FORTRAN ' (;'FOR Numer.' FORTRAN numeric (;'FOR Num fp') FORTRAN numeric fixed point (;'FOR Numflp') FORTRAN numeric floating point end of FORTRAN Numeric) (;'FOR Nonnum' FORTRAN non-numeric (;'FOR Ld St') FORTRAN Load and Store (;'FOR Jumps') FORTRAN Jumps and Subroutine calls end of FORTRAN non-numeric) end of FORTRAN) (;'ALGOL 60 ' (;'ALG numer.' ALGOL numeric (;'ALG num fp') ALGOL fixed point (;'ALG numflp') ALGOL floating point end of ALGOL numeric) (;'ALG nonnum' ALGOL nonnumeric (;'ALG Ld St') ALGOL Load and Store (;'ALG Jumps') ALGOL Jumps and Subroutine calls end of ALGOL non-numeric) (;'Pascal ') (;' IMP ') (;'ALGOL W ') (;'BCPL ') end of other ALGOL dialects) (;'Real Time ' (;'CORAL 66 ' (;'CORAL Num ') CORAL 66 numeric (;'CORAL Int ') CORAL 66 non-numeric end of CORAL 66) (;' RTL/2 ' (;'RTL/2 Num ') RTL/2 numeric (;'RTL/2 Int ') RTL/2 non-numeric end of RTL/2) end of Real Time) end of Scientific languages) (;'Commercial' (;'COBOL ') (;'PL/I ') (;'RPG 11 ') end of Commercial languages) )' end of tree structure for applications benchmark-results E<n> denotes entry in loose leaf file ( (;'Gibson Mix' is the name of the Benchmark ;'Scientific' is the application area relevance=; 3.192, This is a weighted instruction mix which reflects scientific usage machine software result ;'360/65G '; 'Assembler '; 543, ;'360/65H '; 'Assembler '; 563, ;'360/75 '; 'Assembler '; 940, ;'360/85 '; 'Assembler '; 3245, ;'360/50 '; 'Assembler '; 133, ;'370/135 '; 'Assembler '; 113, ;'370/145 '; 'Assembler '; 178, ;'370/155 '; 'Assembler '; 470, ;'370/165 '; 'Assembler '; 3068, ;'ICL 4/50 '; 'Assembler '; 55, ;'ICL 4/70 '; 'Assembler '; 368, ;'ICL 4/72 '; 'Assembler '; 426, ;'ICL 4/75 P'; 'Assembler '; 333, ;'1901A 10SC'; 'Assembler '; 26.9, ;'1902A 20SC'; 'Assembler '; 56.1, ;'1902S 25S '; 'Assembler '; 59.8, ;'1903 EMU '; 'Assembler '; 59, ;'1903A SC '; 'Assembler '; 113.7, ;'1903S 31S '; 'Assembler '; 118, ;'1903T '; 'Assembler '; 220, ;'1904A FP '; 'Assembler '; 261, ;'1904S FP I'; 'Assembler '; 302, ;'1905E Acc '; 'Assembler '; 144, ;'1905F '; 'Assembler '; 196, ;'1906A '; 'Assembler '; 866, ;'1906S '; 'Assembler '; 1150, ;'ICL 4120/2'; 'Assembler '; 25, ;'ICL 4130/2'; 'Assembler '; 112, ;'Bur 5500 '; 'Assembler '; 144, ;'Bur 6714 F'; 'Assembler '; 298, ;'Bur 6715 s'; 'Assembler '; 348, ;'CDC 3300 '; 'Assembler '; 152, ;'CDC 3600 '; 'Assembler '; 337, ;'Cyber 72 '; 'Assembler '; 600, ;'Cyber 73 '; 'Assembler '; 800, ;'CDC 6600 '; 'Assembler '; 2190, ;'CDC 7600 '; 'Assembler '; 7000, ;'Hon GE 635'; 'Assembler '; 379, ;'Hon 6025 '; 'Assembler '; 180, ;'Hon 6040 '; 'Assembler '; 240, ;'Hon 6060 '; 'Assembler '; 479, ;'Hon 66/40 '; 'Assembler '; 820, ;'DEC KI 10 '; 'Assembler '; 590, ;'DEC KL 10 '; 'Assembler '; 1034, ;'ICL KDF9 '; 'Assembler '; 170, ;'Univ1106 U'; 'Assembler '; 517, ;'Univ1108 '; 'Assembler '; 932, ;'Univ1110 '; 'Assembler '; 1606, ;'UniV1121 '; 'Assembler '; 709, ;'INTER 5 '; 'Assembler '; 73.9, ;'XDS SIG5 '; 'Assembler '; 272, ;'XDS SIG9 '; 'Assembler '; 972, ;'PDP 11 /20'; 'Assembler '; 39.9, ;'Hon DDP516'; 'Assembler '; 18, ;'Hon H316 F'; 'Assembler '; 16.2, ;'CTL MOD1 S'; 'Assembler '; 45.5, ;'CTL MOD1 F'; 'Assembler '; 113, ;'HP 2100A '; 'Assembler '; 99, ;'HP 3000 I '; 'Assembler '; 286, ;'GEC 4080 '; 'Assembler '; 402, ;'IBM 1130 '; 'Assembler '; 14, ;'Ferr 1600B'; 'Assembler '; 216, ;'MINIC 1 '; 'Assembler '; 32, end of Gibson Mix) (;'ADP Mix ' is the name of the Benchmark ;'Commerical' is the application area relevance=; 1.849, This is a weighted instruction mix which reflects commercial usage machine software result ;'360/65G '; 'Assembler '; 542, ;'360/65H '; 'Assembler '; 567, ;'360/75 '; 'Assembler '; 870, ;'360/85 '; 'Assembler '; 3418, ;'360/50 '; 'Assembler '; 169, ;'370/135 '; 'Assembler '; 171, ;'370/145 '; 'Assembler '; 330, ;'370/155 '; 'Assembler '; 678, ;'370/165 '; 'Assembler '; 3102, ;'ICL 4/50 '; 'Assembler '; 114, ;'ICL 4/70 '; 'Assembler '; 410, ;'ICL 4/72 '; 'Assembler '; 479, ;'ICL 4/75 P'; 'Assembler '; 363, ;'1901A 10SC'; 'Assembler '; 29.2, ;'1902A 20SC'; 'Assembler '; 95.2, ;'1902S 25S '; 'Assembler '; 95, ;'1903 EMU '; 'Assembler '; 106.3, ;'1903A SC '; 'Assembler '; 189.6, ;'1903S 31S '; 'Assembler '; 195, ;'1903T '; 'Assembler '; 220, ;'1904A FP '; 'Assembler '; 276, ;'1905E Acc '; 'Assembler '; 181, ;'1905F '; 'Assembler '; 225, ;'1906A '; 'Assembler '; 907, ;'1906S '; 'Assembler '; 1125, ;'ICL 4120/2'; 'Assembler '; 95, ;'ICL 4130/2'; 'Assembler '; 164, ;'Bur 5500 '; 'Assembler '; 181, ;'CDC 3300 '; 'Assembler '; 158, ;'CDC 3600 '; 'Assembler '; 326, ;'Hon GE 635'; 'Assembler '; 433, ;'Hon 6025 '; 'Assembler '; 186, ;'Hon 6060 '; 'Assembler '; 480, ;'DEC KL 10 '; 'Assembler '; 1110, ;'ICL KDF9 '; 'Assembler '; 198, ;'Univ1106 U'; 'Assembler '; 517, ;'Univ1108 '; 'Assembler '; 866, ;'Univ1110 '; 'Assembler '; 1353, ;'INTER 5 '; 'Assembler '; 217.5, ;'UNIV418III'; 'Assembler '; 440, ;'XDS SIG5 '; 'Assembler '; 400, ;'XDS SIG9 '; 'Assembler '; 1251, ;'PDP 11 /20'; 'Assembler '; 287.1, ;'Hon DDP516'; 'Assembler '; 119, ;'Hon H316 F'; 'Assembler '; 184, ;'CTL MOD1 S'; 'Assembler '; 99, ;'CTL MOD1 F'; 'Assembler '; 243, ;'HP 2100A '; 'Assembler '; 200, ;'GEC 4080 '; 'Assembler '; 485, ;'IBM 1130 '; 'Assembler '; 86, ;'Ferr 1600B'; 'Assembler '; 195, ;'MINIC 1 '; 'Assembler '; 70, end of ADP Mix) (;'Pr Con Mix' is the name of the Benchmark This is the Process Control Mix. ;'Real Time' is the application area relevance=; 1.808, This is a weighted instruction mix without multiplication and division machine software result ;'360/65G '; 'Assembler '; 848, ;'360/65H '; 'Assembler '; 906, ;'360/75 '; 'Assembler '; 1445, ;'360/85 '; 'Assembler '; 4808, ;'360/50 '; 'Assembler '; 246, ;'370/135 '; 'Assembler '; 249, ;'370/145 '; 'Assembler '; 397, ;'370/155 '; 'Assembler '; 1100, ;'370/165 '; 'Assembler '; 4840, ;'ICL 4/50 '; 'Assembler '; 115, ;'ICL 4/70 '; 'Assembler '; 677, ;'ICL 4/72 '; 'Assembler '; 837, ;'ICL 4/75 P'; 'Assembler '; 622, ;'1901A 10SC'; 'Assembler '; 44.4, ;'1902A 20SC'; 'Assembler '; 100.5, ;'1902S 25S '; 'Assembler '; 105, ;'1903A SC '; 'Assembler '; 198.1, ;'1903S 31S '; 'Assembler '; 212, ;'1904A FP '; 'Assembler '; 365, ;'1906A '; 'Assembler '; 1115, ;'1906S '; 'Assembler '; 2067, ;'ICL 4120/2'; 'Assembler '; 95, ;'ICL 4130/2'; 'Assembler '; 192, ;'CDC 3300 '; 'Assembler '; 374, ;'CDC 3600 '; 'Assembler '; 381, ;'Hon GE 635'; 'Assembler '; 475, ;'ICL KDF9 '; 'Assembler '; 205, ;'Univ1106 U'; 'Assembler '; 713, ;'Univ1108 '; 'Assembler '; 1235, ;'Univ1110 '; 'Assembler '; 2334, ;'INTER 5 '; 'Assembler '; 228.1, ;'UNIV418III'; 'Assembler '; 648, ;'XDS SIG9 '; 'Assembler '; 983, ;'PDP 11 /20'; 'Assembler '; 343.4, ;'Hon DDP516'; 'Assembler '; 115, ;'Hon H316 F'; 'Assembler '; 258, ;'CTL MOD1 S'; 'Assembler '; 250, ;'CTL MOD1 F'; 'Assembler '; 558, ;'HP 2100A '; 'Assembler '; 340, ;'GEC 4080 '; 'Assembler '; 779, ;'IBM 1130 '; 'Assembler '; 146, ;'Ferr 1600B'; 'Assembler '; 304, ;'MINIC 1 '; 'Assembler '; 198, end of Pr Con Mix) (;'Ackn ratio' is the name of the Benchmark ;'ALGOL 60' is the application area relevance=; 1.034, This is the average time to execute an instruction in calculating Ackermanns function, not necessarily in high level language. This is an example of the notation of results as time/instruction [where smallest absolute value is fastest]; machine software result ;'370/158 '; 'Pascal '; -.90, E457 compiler from Canada ;'370/165 '; 'BCPL '; -.31, E456 compiler Cambridge ;'370/165 '; 'Assembler '; -.22, E449 compiler BAL ;'ICL 4/70 '; 'RTL/2 '; -3.17, E452 compiler ICI ;'ICL 4/75 P'; ' IMP '; -2.63, E451 compiler EMAS nocheck ;'1906A '; 'ALGOL 60 '; -.87, E455 compiler Manchester ;'1906A '; 'Pascal '; -.96, E454 compiler XPAC 1B ;'1906S '; 'Pascal '; -.59, E453 compiler XPAC 1B ;'ICL 2980 '; 'ALGOL 60 '; -.667, E1033 compiler Edinburgh ;'Bur 5500 '; 'ALGOL 60 '; -6.92, E446 compiler MK xv.1.01 ;'Cyber 73 '; 'Pascal '; -.88, E450 from Zurich ;'DEC KI 10 '; 'Assembler '; -2, E447 ;'ICL KDF9 '; 'ALGOL 60 '; -7.8, E458 compiler Kidsgrove ;'ICL KDF9 '; 'Assembler '; -7.35, E459 machine coding ;' ;'CII IRIS80'; 'Assembler '; -1.625, E1015 compiler LIS ;'PDP 11 /20'; 'Assembler '; -4.67, E623 compiler Bliss ;'Hon DDP516'; 'Assembler '; -2.28, E461 compiler PL516 ;'NOR SM4 '; 'Assembler '; -3.44, E1016 compiler MARY end of Ackn ratio) (;'Synth FORS' is the name of the Benchmark ;'FORTRAN' is the application area relevance=; 4.295, This is the Curnow Synthetic Benchmark in FORTRAN, FOPR12 machine software result ;'360/65H '; 'FORTRAN + '; 521, E270 compiler H OPT=2 ;'360/65H '; 'FORTRAN '; 530, E269 compiler G ;'360/67 '; 'FORTRAN + '; 602.8, compiler H OPT=2 ;'360/67 '; 'FORTRAN '; 320, E271 compiler G ;'360/195 '; 'FORTRAN + '; 5000, E273 compiler H+ ;'360/195 '; 'FORTRAN + '; 4420, E274 compiler H OPT=2 ;'360/195 '; 'FORTRAN '; 3030, E272 compiler G ;'360/50 '; 'FORTRAN + '; 145, E268 compiler H OPT=2 ;'360/50 '; 'FORTRAN '; 95, E267 compiler CALL/360 ;'360/30 MI '; 'FORTRAN '; 15.5, E736 compiler DOS ;'370/145 '; 'FORTRAN '; 171, E275 compiler G ;'370/155 '; 'FORTRAN + '; 465, E277 compiler H OPT=2 ;'370/155 '; 'FORTRAN '; 395, E276 compiler G ;'370/158 '; 'FORTRAN + '; 824, E279 compiler H OPT=2 ;'370/158 '; 'FORTRAN '; 611, E278 compiler G ;'370/168 '; 'FORTRAN + '; 2439, E281 compiler H OPT=2 ;'370/168 '; 'FORTRAN + '; 1973, compiler Hnot PP OPT=2 ;'370/168 '; 'FORTRAN '; 1887, E280 compiler G ;'370/168 M '; 'FORTRAN + '; 3035, compiler H OPT=2 ;'Amdahl 470'; 'FORTRAN + '; 4640, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 55.1, E679 compiler VOO ;'ICL 4/50 '; 'FORTRAN '; 50, E678 compiler with debug ;'ICL 4/50 '; 'FORTRAN '; 13.5, E680 compiler WATFOR ;'ICL 4/70 '; 'FORTRAN '; 280, E682 ;'ICL 4/70 '; 'FORTRAN '; 224, E681 compiler with debug ;'ICL 4/72 '; 'FORTRAN '; 300, E684 ;'ICL 4/72 '; 'FORTRAN '; 240, E683 with debug ;'ICL 4/75 P'; 'FORTRAN '; 367, E686 ;'ICL 4/75 P'; 'FORTRAN '; 217, E685 compiler with debug ;'ICL 4/75 P'; 'FORTRAN '; 68, E687 compiler WATFOR ;'1904A FP '; 'FORTRAN + '; 192, E285 compiler XFEW TR0 ;'1904A FP '; 'FORTRAN '; 91, E284 compiler XFAT ;'1904S FP I'; 'FORTRAN + '; 222, E288 compiler XFEW TR0 ;'1904S FP I'; 'FORTRAN '; 121, E287 compiler XFAT TR1 ;'1904S FP I'; 'FORTRAN '; 65.5, E289 compiler SOFOR ;'1905E Acc '; 'FORTRAN '; 66.7, compiler XFAT TR0 ;'1905E Acc '; 'FORTRAN '; 43.3, compiler XFAT TR1 ;'1905E Acc '; 'FORTRAN '; 9.03, compiler XFAT TR2 ;'1905F '; 'FORTRAN '; 55, E290 compiler XFAT TR1 ;'1906A '; 'FORTRAN + '; 585, E293 compiler XFEW ;'1906A '; 'FORTRAN + '; 555, E292 compiler XFEH TR0 ;'1906A '; 'FORTRAN '; 282, E291 compiler XFIH TR1 ;'1906S '; 'FORTRAN + '; 800, E296 compiler XFEV TR0 ;'1906S '; 'FORTRAN '; 740, E294 compiler XFIV TR0 ;'1906S '; 'FORTRAN '; 357, E295 compiler XFIV TR1 ;'ICL 2970 '; 'FORTRAN '; 436, compiler F1 5x10 ;'ICL 2970 '; 'FORTRAN '; 412.7, compiler ERCC 4/77 ;'ICL 2980 '; 'FORTRAN '; 1802, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 1181, compiler ERCC 4/77 ;'ICL 4120/2'; 'FORTRAN '; 10.4, E299 ;'ICL 4130/2'; 'FORTRAN '; 75.2, E300 750 ns store ;'Bur 5500 '; 'FORTRAN '; 64, E230 compiler MK XV.1 ;'Bur 6714 F'; 'FORTRAN '; 252, E232 compiler 2.6 OPT=1 ;'Bur 6714 F'; 'FORTRAN '; 193, E231 compiler 2.6 OPT=0 ;'Bur 6715 S'; 'FORTRAN '; 268, compiler 2.7 OPT=1 ;'Bur 6715 S'; 'FORTRAN '; 206, compiler 2.7 OPT=0 ;'Cyber 72 '; 'FORTRAN '; 558, E241 compiler FTN4,3 OPT=2 ;'Cyber 73 '; 'FORTRAN '; 1020, compiler FTN OPT=2 ;'CDC 6600 '; 'FORTRAN '; 2086, E668 compiler FTN4.5 OPT=2 ;'CDC 7600 '; 'FORTRAN '; 9345, compiler FTN OPT=2 UMRCC ;'CDC 7600 '; 'FORTRAN '; 8333, E239 compiler FTN OPT=2 ;'CDC 7600 '; 'FORTRAN '; 7620, E240 compiler FTN4.2 OPT=2 ;'CDC 7600 '; 'FORTRAN '; 3174, compiler FTN OPT=0 UMRCC ;'Hon 6040 '; 'FORTRAN '; 229, ;'Hon 6060 '; 'FORTRAN '; 398, ;'Hon 66/40 '; 'FORTRAN '; 796, ;'DEC 20 '; 'FORTRAN '; 442.5, E1027 ;'DEC KI 10 '; 'FORTRAN '; 500, E250 ;'DEC KL 10 '; 'FORTRAN '; 1210, compiler F10 4A/317 ;'ICL KDF9 '; 'FORTRAN '; 80, E298 compiler EGTRAN ;'Univ1106 U'; 'FORTRAN '; 535, E312 compiler V ;'Univ1108 '; 'FORTRAN '; 820, E313 compiler V ;'Univ1110 '; 'FORTRAN '; 1605, E665 ;'UniV1121 '; 'FORTRAN '; 811, ;'XDS SIG5 '; 'FORTRAN '; 298, E302 compiler EXT FORT ;'XDS SIG6 '; 'FORTRAN '; 331, E303 compiler EXT FORT ;'XDS SIG6 '; 'FORTRAN '; 228, E304 compiler FLAG ;'XDS SIG6 '; 'FORTRAN '; 24, E305 compiler IV H ;'XDS SIG9 '; 'FORTRAN '; 483, E08 compiler EXT FORT ;'PDP 11/10 '; 'FORTRAN '; 12.9, E1005 compiler V01B-080 ;'PDP 11/34F'; 'FORTRAN '; 204, compiler F4+ ;'PDP 11/34S'; 'FORTRAN '; 23.5, ;'PDP 11/40S'; 'FORTRAN '; 17.5, ;'PDP 11/55 '; 'FORTRAN '; 714, compiler F4+ ;'PDP 11/60 '; 'FORTRAN '; 591, compiler F4+ ;'PDP 11/70 '; 'FORTRAN '; 671, compiler F4+ ;'Hon H316 F'; 'FORTRAN '; 17.2, E258 ;'HP 2100S '; 'FORTRAN '; 76, E256 ;'HP 3000 I '; 'FORTRAN '; 139, ;'PRIME 300H'; 'FORTRAN '; 159, ;'GEC 4080 '; 'FORTRAN '; 265, E610 ;'NOVA 840 S'; 'FORTRAN '; 16.3, E1025 end of Synth FORS) (;'Synth FORD' is the name of the Benchmark ;'FORTRAN' is the application area relevance=; 1.168, This is the Curnow Synthetic Benchmark in double length, FOPR13 machine software result ;'360/65H '; 'FORTRAN + '; 421, E357 compiler H OPT=2 ;'360/65H '; 'FORTRAN '; 321, E356 compiler G ;'360/67 '; 'FORTRAN + '; 456.6, compiler H OPT=2 ;'360/67 '; 'FORTRAN '; 278, E358 compiler G ;'360/195 '; 'FORTRAN + '; 4760, E360 compiler H+ ;'360/195 '; 'FORTRAN + '; 4300, E361 compiler H OPT=2 ;'360/195 '; 'FORTRAN '; 3030, E359 compiler G ;'360/50 '; 'FORTRAN + '; 94, E355 compiler H OPT=2 ;'360/50 '; 'FORTRAN '; 67, E354 compiler CALL/360 ;'360/30 MI '; 'FORTRAN '; 7.7, E737 compiler DOS ;'370/145 '; 'FORTRAN '; 103, E362 compiler G ;'370/158 '; 'FORTRAN + '; 560, E363 compiler H OPT=2 ;'370/168 '; 'FORTRAN '; 1639, E364 compiler G ;'370/168 '; 'FORTRAN + '; 1629, compiler Hnot PP OPT=2 ;'Amdahl 470'; 'FORTRAN + '; 3328, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 27.4, E707 compiler ;'ICL 4/50 '; 'FORTRAN '; 26, E376 and E706 compiler VOO debug ;'ICL 4/70 '; 'FORTRAN '; 209, E709 ;'ICL 4/70 '; 'FORTRAN '; 175, E708 compiler with debug ;'ICL 4/72 '; 'FORTRAN '; 225, E711 ;'ICL 4/72 '; 'FORTRAN '; 213, E710 with debug ;'ICL 4/75 P'; 'FORTRAN '; 200, E713 ;'ICL 4/75 P'; 'FORTRAN '; 169, E712 compiler with debug ;'ICL 4/75 P'; 'FORTRAN '; 68, E714 compiler WATFOR ;'1904A FP '; 'FORTRAN + '; 21, E368 compiler XFEW TR0 ;'1904A FP '; 'FORTRAN '; 19, E367 compiler XFAT TR1 ;'1904S FP I'; 'FORTRAN + '; 28, E370 compiler XFEW TR0 ;'1904S FP I'; 'FORTRAN '; 25, E287 compiler XFAT TR1 ;'1905E Acc '; 'FORTRAN '; 10.8, compiler XFAT TR0 ;'1905E Acc '; 'FORTRAN '; 9.89, compiler XFAT TR1 ;'1905E Acc '; 'FORTRAN '; 5.08, compiler XFAT TR2 ;'1906A '; 'FORTRAN + '; 275, E372 compiler XFEH TR0 ;'1906A '; 'FORTRAN '; 187, E371 compiler XFIH TR1 ;'1906S '; 'FORTRAN + '; 122, E375 compiler XFEV TR0 ;'1906S '; 'FORTRAN '; 120, E373 compiler XFIV TR0 ;'1906S '; 'FORTRAN '; 100, E374 compiler XFIV TR1 ;'ICL 2970 '; 'FORTRAN '; 380.1, compiler ERCC 4/77 ;'ICL 2970 '; 'FORTRAN '; 314, compiler F1 5x10 ;'ICL 2980 '; 'FORTRAN '; 1385, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 959.2, compiler ERCC 4/77 ;'ICL 4120/2'; 'FORTRAN '; 1.1, E378 ;'ICL 4130/2'; 'FORTRAN '; 2.2, E370 750 ns store ;'Bur 5500 '; 'FORTRAN '; 21, E319 compiler MK XV.1 ;'Bur 6714 F'; 'FORTRAN '; 109, E321 compiler 2.6 OPT=1 ;'Bur 6714 F'; 'FORTRAN '; 96.4, E320 compiler 2.6 OPT=0 ;'Bur 6715 S'; 'FORTRAN '; 136, compiler 2.7 OPT=1 ;'Bur 6715 s'; 'FORTRAN '; 113, compiler 2.7 OPT=0 ;'Cyber 72 '; 'FORTRAN + '; 167, E329 compiler FTN4,3 OPT=2 ;'Cyber 173 '; 'FORTRAN + '; 305, compiler FTN OPT=2 ;'CDC 7600 '; 'FORTRAN + '; 3745, E328 compiler FTN4.2 OPT=2 ;'Hon 6040 '; 'FORTRAN '; 141, ;'Hon 6060 '; 'FORTRAN '; 264, ;'Hon 66/40 '; 'FORTRAN '; 593, ;'DEC 20 '; 'FORTRAN '; 252.5, E1027 ;'DEC KI 10 '; 'FORTRAN '; 250, E337 ;'DEC KL 10 '; 'FORTRAN '; 644, compiler F10 4A/317 ;'ICL KDF9 '; 'FORTRAN '; 7.7, E377 compiler EGTRAN ;'Univ1106 U'; 'FORTRAN + '; 313, E389 compiler V ;'Univ1108 '; 'FORTRAN + '; 475, E390 compiler V ;'Univ1110 '; 'FORTRAN + '; 1119, E666 ;'UniV1121 '; 'FORTRAN '; 470, ;'XDS SIG5 '; 'FORTRAN '; 225, E381 compiler EXT FORT ;'XDS SIG6 '; 'FORTRAN '; 260, E382 compiler EXT FORT ;'XDS SIG6 '; 'FORTRAN '; 192, E383 compiler FLAG ;'XDS SIG9 '; 'FORTRAN '; 361, E386 compiler EXT FORT ;'PDP 11/10 '; 'FORTRAN '; 3.17, E1006 compiler V01B-080 ;'PDP 11/34F'; 'FORTRAN '; 157, compiler F4+ ;'PDP 11/34S'; 'FORTRAN '; 6.1, ;'PDP 11/40S'; 'FORTRAN '; 5.2, ;'PDP 11/55 '; 'FORTRAN '; 552, compiler F4+ ;'PDP 11/60 '; 'FORTRAN '; 434, compiler F4+ ;'PDP 11/70 '; 'FORTRAN '; 502, compiler F4+ ;'Hon H316 F'; 'FORTRAN '; 5.6, E345 ;'HP 2100S '; 'FORTRAN '; 47, E343 ;'HP 3000 I '; 'FORTRAN '; 8.5, ;'PRIME 300H'; 'FORTRAN '; 95, ;'GEC 4080 '; 'FORTRAN '; 144, E613 end of Synth FORD) (;'Gamma Test' is the name of the Benchmark ;'FOR Numflp' is the application area relevance=; 2.486, This program, FOPROO tests floating point by calculating the Gamma function machine machine software result ;'360/65G '; 'FORTRAN '; 135, E021 compiler G ;'360/67 '; 'FORTRAN + '; 146.5, compiler H OPT=2 ;'360/85 '; 'FORTRAN '; 778, E022 compiler G. ;'360/195 '; 'FORTRAN '; 1150, E061 and E062 compiler G ;'360/30 MI '; 'FORTRAN '; 4.12, E1000 compiler DOS ;'370/135 '; 'FORTRAN '; 27.5, E496 and E497 compiler G ;'370/155 '; 'FORTRAN '; 114, E063 compiler G ;'370/158 '; 'FORTRAN '; 194, E495 compiler G ;'370/165 '; 'FORTRAN '; 746, E064 compiler G ;'370/168 '; 'FORTRAN '; 663, E498 compiler G ;'370/168 '; 'FORTRAN + '; 607.4, compiler H not PP OPT=2 ;'370/168 M '; 'FORTRAN + '; 990, compiler H+ ;'370/168 M '; 'FORTRAN '; 673, compiler G ;'Amdahl470 '; 'FORTRAN + '; 1300, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 19.7, E020 compiler M27 ;'ICL 4/70 '; 'FORTRAN '; 112.1, ;'ICL 4/70 '; 'FORTRAN '; 104, with debug ;'ICL 4/72 '; 'FORTRAN '; 123.9, ;'ICL 4/72 '; 'FORTRAN '; 112, with debug ;'1902A 20SC'; 'FORTRAN '; 6.94, E014 compiler XFAT 2E ;'1903A SC '; 'FORTRAN '; 13, E015 compiler XFAT 2E ;'1903T '; 'FORTRAN '; 32.8, E620 compiler XFIV MK2B ;'1903T '; 'FORTRAN '; 25, E502 TRACE 1 ;'1904A FP '; 'FORTRAN '; 44.3, E016 compiler XFAT 4C ;'1905F '; 'FORTRAN '; 25.4, E013 compiler XFAT 2E ;'1906A '; 'FORTRAN '; 171, E493 compiler XFIH 1C ;'1906A '; 'FORTRAN '; 157, E019 compiler XFAT 2E ;'1906S '; 'FORTRAN '; 333, E474 compiler XFIH 1C ;'1906S '; 'FORTRAN '; 263.2, E773 compiler XFIH ;'1906S '; 'FORTRAN '; 233, E475 compiler XFIH 1C TRACE 1 ;'ICL 2970 '; 'FORTRAN '; 78.6, compiler F1 5x10 ;'ICL 2970 '; 'FORTRAN '; 71.3, compiler ERCC 4/77 ;'ICL 2980 '; 'FORTRAN '; 383.6, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 217, compiler ERCC 4/77 ;'ICL 4130/2'; 'FORTRAN '; 14.84, E1012 ;'Bur 5500 '; 'FORTRAN '; 27.5, E480 compiler 15.1 ;'Bur 6714 F'; 'FORTRAN '; 72.6, E077 compiler I10 ;'Bur 6715 s'; 'FORTRAN '; 68.7, compiler 2.7 ;'Cyber 72 '; 'FORTRAN '; 135, E476 compiler FTN OPT ;'Cyber 73 '; 'FORTRAN '; 102.7, compiler RUN ;'Cyber 173 '; 'FORTRAN '; 276, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 658, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 467, E068 compiler RUN ;'CDC 7600 '; 'FORTRAN + '; 3199, E472 compiler FTN 2.0 OPT=2 ;'CDC 7600 '; 'FORTRAN '; 2621, E473 compiler FTN 2.0 OPT=1 ;'CDC 7600 '; 'FORTRAN '; 1554, E069 and E471 compiler RUN ;'Hon 6025 '; 'FORTRAN '; 38.6, E464 and E479 ;'Hon 6040 '; 'FORTRAN '; 59, ;'Hon 6060 '; 'FORTRAN '; 123, E478 ;'Hon 66/40 '; 'FORTRAN '; 161, ;'Hon 66/60 '; 'FORTRAN '; 221.2, ;'DEC 20 '; 'FORTRAN '; 180.5, E1032 ;'DEC KI 10 '; 'FORTRAN '; 212.8, E492 ;'DEC KL 10 '; 'FORTRAN '; 396.5, E728 ;'Univ1106 U'; 'FORTRAN '; 115.6, E494 ;'Univ1108 '; 'FORTRAN '; 294, E066 ;'UniV1121 '; 'FORTRAN '; 245.7, ;'INTER 5 '; 'FORTRAN '; 3.03, E076 ;'XDS SIG5 '; 'FORTRAN '; 75.3, E070 compiler EXT IVH ;'XDS SIG6 '; 'FORTRAN '; 124, E477 compiler EXT VE00 ;'PDP 11/10 '; 'FORTRAN '; 2.17, E1007 compiler V01B-080 ;'PDP 11/34S'; 'FORTRAN '; 4.214, ;'PDP 11/40E'; 'FORTRAN '; 4.56, E657 ;'PDP 11/40S'; 'FORTRAN '; 6.276, ;'Hon DDP516'; 'FORTRAN '; 3.97, E484 no mult div ;'Hon H316 F'; 'FORTRAN '; 4, E483 ;'HP 2100A '; 'FORTRAN '; 17, E482 ;'HP 3000 I '; 'FORTRAN '; 24.74, ;'MODCOMP IV'; 'FORTRAN '; 114.8, ;'GEC 4080 '; 'FORTRAN '; 43.8, E491 ;'EAL 1830 '; 'FORTRAN '; 8.9, E485 ;'IBM 1130 '; 'FORTRAN '; 3.13, E486 ;'VAR620/100'; 'FORTRAN '; 3.19, E490 ;'NOVA 840 S'; 'FORTRAN '; 3.357, E1022 end of Gamma Test) (;'Bit Test' is the name of the Benchmark ;'FOR Ld St' is the application area relevance=; 1.702, This is the program FOPR01, a core store test in FORTRAN machine software result ;'360/65G '; 'FORTRAN '; 2505, E023 and E024 compiler G ;'360/67 '; 'FORTRAN + '; 3249, compiler H OPT=2 ;'360/85 '; 'FORTRAN '; 15700, E025 and E026 compiler G ;'360/195 '; 'FORTRAN + '; 21900, E078 compiler H OPT=2 ;'360/195 '; 'FORTRAN '; 17740, E027 E79-E82 compiler G ;'360/30 MI '; 'FORTRAN '; 96.6, E1001 compiler DOS ;'370/135 '; 'FORTRAN '; 756, E525 compiler G ;'370/158 '; 'FORTRAN '; 4895, E526. compiler G ;'370/165 '; 'FORTRAN '; 12600, E084 and E085 compiler G ;'370/168 '; 'FORTRAN + '; 16760, compiler H not PP OPT=2 ;'370/168 '; 'FORTRAN '; 13352, E524 compiler G ;'370/168 M '; 'FORTRAN + '; 16931, compiler H+ ;'370/168 M '; 'FORTRAN '; 13653, compiler G ;'Amdahl 470'; 'FORTRAN + '; 27450, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 435.6, ;'ICL 4/50 '; 'FORTRAN '; 391, E028 compiler M27 ;'ICL 4/70 '; 'FORTRAN '; 1846, ;'ICL 4/70 '; 'FORTRAN '; 1459, with debug ;'ICL 4/72 '; 'FORTRAN '; 2059, ;'ICL 4/72 '; 'FORTRAN '; 1568, with debug ;'1902A 20SC'; 'FORTRAN '; 175, compiler XFAT ;'1903A SC '; 'FORTRAN '; 354, compiler XFAT ;'1903T '; 'FORTRAN '; 681, compiler XFAT ;'1904A FP '; 'FORTRAN '; 958, compiler XFAT ;'1905E Acc '; 'FORTRAN '; 333.3, compiler XFAT TR0,TR1 ;'1905E Acc '; 'FORTRAN '; 64.7, compiler XFAT TR2 ;'1906A '; 'FORTRAN '; 4503, compiler XFAT ;'1906S '; 'FORTRAN '; 8597, E772 compiler XFIH 1C ;'ICL 2970 '; 'FORTRAN '; 2690, compiler F1 5x10 ;'ICL 2970 '; 'FORTRAN '; 2135, compiler ERCC 4/77 ;'ICL 2980 '; 'FORTRAN '; 10185, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 5954, compiler ERCC 4/77 ;'ICL ATLAS1'; 'FORTRAN '; 1085, E527 ;'Bur 5500 '; 'FORTRAN '; 257, E511 compiler 15.1 ;'Bur 6715 S'; 'FORTRAN '; 674, compiler 2.7 ;'Cyber 72 '; 'FORTRAN '; 2731, compiler FTN ;'Cyber 73 '; 'FORTRAN '; 3023, compiler FTN ;'Cyber 73 '; 'FORTRAN '; 1693, compiler RUN ;'Cyber 173 '; 'FORTRAN '; 3734, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 6530, E086 compiler RUN ;'CDC 7600 '; 'FORTRAN + '; 30969, compiler FTN ;'CDC 7600 '; 'FORTRAN '; 25798, compiler FTN ;'Hon 6025 '; 'FORTRAN '; 525, ;'Hon 6040 '; 'FORTRAN '; 738, ;'Hon 66/40 '; 'FORTRAN '; 2288, ;'Hon 66/60 '; 'FORTRAN '; 1867, ;'DEC 20 '; 'FORTRAN '; 2511, E1031 ;'DEC KI 10 '; 'FORTRAN '; 3017, E507 ;'DEC KL 10 '; 'FORTRAN '; 6183, E729 ;'Univ1106 U'; 'FORTRAN '; 2278, E517 and E518 ;'Univ1108 '; 'FORTRAN '; 4212, E089 and E090 ;'Univ1121 '; 'FORTRAN '; 4535, ;'UNIV418III'; 'FORTRAN '; 465, E091 ;'XDS SIG5 '; 'FORTRAN '; 1420, E092 and E093 compiler EXT IVH ;'XDS SIG6 '; 'FORTRAN '; 2334, E652 ;'PDP 11 /10'; 'FORTRAN '; 419, E1008 compiler V01B-080 ;'PDP 11 /20'; 'FORTRAN '; 98.5, E506 ;'PDP 11/34S'; 'FORTRAN '; 559, ;'PDP 11/40E'; 'FORTRAN '; 487.1, E658 ;'PDP 11/40S'; 'FORTRAN '; 700, ;'Hon DDP516'; 'FORTRAN '; 421, E514 ;'Hon H316 F'; 'FORTRAN '; 317, E515 ;'HP 2100A '; 'FORTRAN '; 640, E510 ;'HP 3000 I '; 'FORTRAN '; 1212, ;'MODCOMP IV'; 'FORTRAN '; 3527, ;'PRIME 300H'; 'FORTRAN '; 1109, ;'GEC 4080 '; 'FORTRAN '; 1417, E520 ;'EAL 1830 '; 'FORTRAN '; 521, E522 ;'IBM 1130 '; 'FORTRAN '; 114, E523 ;'VAR620/100'; 'FORTRAN '; 640, E519 ;'NOVA 840 S'; 'FORTRAN '; 309.9, E1023 end of Bit Test) (;'Binomial' is the name of the Benchmark ;'FOR Numflp' is the application area relevance=; 1.550, This is the program FOPR02, a floating point test program machine software result ;'360/65G '; 'FORTRAN '; 3317, E103 and E104 compiler G ;'360/67 '; 'FORTRAN + '; 4689, compiler H OPT=2 ;'360/85 '; 'FORTRAN '; 20070, E105 compiler G ;'360/195 '; 'FORTRAN + '; 64700, E107 compiler H OPT=2 ;'360/195 '; 'FORTRAN '; 43630, E106 E108 E109 compiler G ;'360/30 MI '; 'FORTRAN '; 96.1, E1002 compiler DOS ;'370/135 '; 'FORTRAN '; 617, E556 compiler G ;'370/158 '; 'FORTRAN '; 4081, E553 compiler G ;'370/165 '; 'FORTRAN '; 19800, E111 compiler G ;'370/168 '; 'FORTRAN + '; 18880, compiler H not PP OPT=2 ;'370/168 '; 'FORTRAN '; 16494, E557 compiler G ;'370/168 M '; 'FORTRAN + '; 28868, compiler H+ ;'370/168 M '; 'FORTRAN '; 18732, compiler G ;'Amdahl 470'; 'FORTRAN + '; 40450, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 373.2, ;'ICL 4/70 '; 'FORTRAN '; 2377, ;'ICL 4/70 '; 'FORTRAN '; 1844, with debug ;'ICL 4/72 '; 'FORTRAN '; 2772, ;'ICL 4/72 '; 'FORTRAN '; 1990, with debug ;'1902A 20SC'; 'FORTRAN '; 180, E097 compiler XFAT 2E ;'1903A SC '; 'FORTRAN '; 349, E098 compiler XFAT 2E ;'1903T '; 'FORTRAN '; 843, E622 compiler XFIV MK2B ;'1904A FP '; 'FORTRAN '; 1040, E099 compiler XFAT 4C ;'1905E Acc '; 'FORTRAN '; 402.5, compiler XFAT TR0,TR1 ;'1905E Acc '; 'FORTRAN '; 59.7, compiler XFAT TR2 ;'1905F '; 'FORTRAN '; 665, E095 and E096 compiler XFAT 2E ;'1906A '; 'FORTRAN '; 5593, E562 compiler XFIH 1C TR1 ;'1906A '; 'FORTRAN '; 3750, E102 compiler XFAT 3E ;'1906S '; 'FORTRAN '; 8296, E558 compiler XFIH 1C ;'ICL 2970 '; 'FORTRAN '; 3611, compiler ERCC 4/77 ;'ICL 2980 '; 'FORTRAN '; 14671, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 14005, compiler ERCC 4/77 ;'ICL ATLAS1'; 'FORTRAN '; 1662, E561 ;'Bur 5500 '; 'FORTRAN '; 400, E542 compiler 15.1 ;'Bur 6714 F'; 'FORTRAN '; 1653, E543 compiler 2.10 ;'Bur 6715 S'; 'FORTRAN '; 1243, compiler 2.7 ;'Cyber 72 '; 'FORTRAN '; 3505, E551 compiler FTN ;'Cyber 73 '; 'FORTRAN '; 3420, E117 compiler RUN ;'Cyber 173 '; 'FORTRAN '; 8173, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 17204, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 9000, E118 compiler RUN ;'CDC 7600 '; 'FORTRAN + '; 120833, E528 compiler FTN2.0 OPT=2 ;'CDC 7600 '; 'FORTRAN '; 88900, E119 and E120 compiler RUN ;'CDC 7600 '; 'FORTRAN '; 77726, E529 compiler FTN2.0 OPT=1 ;'Hon 6025 '; 'FORTRAN '; 860, E465 and E548 ;'Hon 6040 '; 'FORTRAN '; 1267, ;'Hon 6060 '; 'FORTRAN '; 2832, E549 ;'Hon 66/40 '; 'FORTRAN '; 3873, ;'Hon 66/60 '; 'FORTRAN '; 3797, ;'DEC 20 '; 'FORTRAN '; 3574, E1030 ;'DEC KI 10 '; 'FORTRAN '; 4447, E536 ;'DEC KL 10 '; 'FORTRAN '; 6672, E730 ;'DEC KL 10 '; 'FORTRAN '; 4266, compiler F40 ;'Univ1106 U'; 'FORTRAN '; 2411, E552 ;'Univ1108 '; 'FORTRAN '; 5375, E114 and E115 ;'UniV1121 '; 'FORTRAN '; 4758, ;'INTER 5 '; 'FORTRAN '; 142.5, E124 and E125 ;'UNIV418III'; 'FORTRAN '; 11.3, E116 software fp ;'XDS SIG5 '; 'FORTRAN '; 1410, E121 compiler EXT IVH ;'XDS SIG6 '; 'FORTRAN '; 1996, E550 ;'PDP 11/10 '; 'FORTRAN '; 57.7, E1009 compiler V01B-080 ;'PDP 11/20 '; 'FORTRAN '; 130.4, E537 ;'PDP 11/40E'; 'FORTRAN '; 111, E659 ;'PDP 11/40S'; 'FORTRAN '; 148, ;'Hon DDP516'; 'FORTRAN '; 88.9, E516 ;'Hon H316 F'; 'FORTRAN '; 84.2, E547 ;'HP 2100A '; 'FORTRAN '; 476, E540 ;'HP 3000 I '; 'FORTRAN '; 827, ;'MODCOMP IV'; 'FORTRAN '; 2039, ;'PRIME 300H'; 'FORTRAN '; 708, ;'GEC 4080 '; 'FORTRAN '; 1645, E531 ;'EAL 1830 '; 'FORTRAN '; 104, E539 ;'IBM 1130 '; 'FORTRAN '; 36.5, E538 ;'VAR620/100'; 'FORTRAN '; 77.2, E535 end of Binomial) (;'IF Test' is the name of the Benchmark ;'FOR Jumps' is the application area relevance=; 1.374, This is the program FOPR03 machine software result ;'360/65G '; 'FORTRAN '; 7256, E135 and E136 ;'360/67 '; 'FORTRAN + '; 13165, compiler H OPT=2 ;'360/85 '; 'FORTRAN '; 41600, E138 compiler G ;'360/195 '; 'FORTRAN '; 41550, E139 E140 compiler G ;'360/30 MI '; 'FORTRAN '; 407.5, E1003 compiler DOS ;'370/135 '; 'FORTRAN '; 2976, E587 compiler G ;'370/155 '; 'FORTRAN '; 13000, E141 compiler G ;'370/158 '; 'FORTRAN '; 17278, E590 compiler G ;'370/165 '; 'FORTRAN '; 38700, E142 compiler G ;'370/168 '; 'FORTRAN + '; 63510, compiler H not PP OPT=2 ;'370/168 '; 'FORTRAN '; 41116, E589 compiler G ;'370/168 M '; 'FORTRAN + '; 75798, compiler H+ ;'370/168 M '; 'FORTRAN '; 41379, compiler G ;'Amdahl 470'; 'FORTRAN + '; 13400, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 1778, ;'ICL 4/50 '; 'FORTRAN '; 1572, E134 compiler M27 ;'ICL 4/70 '; 'FORTRAN '; 7218, ;'ICL 4/70 '; 'FORTRAN '; 3077, with debug ;'ICL 4/72 '; 'FORTRAN '; 8266, ;'ICL 4/72 '; 'FORTRAN '; 2995, with debug ;'1902A 20SC'; 'FORTRAN '; 694, E129 compiler XFAT 2E ;'1903A SC '; 'FORTRAN '; 3260, E130 compiler XFAT 2E ;'1903T '; 'FORTRAN '; 4464, E621 compiler XFIV MK2B ;'1904A FP '; 'FORTRAN '; 5680, E131 compiler XFAT 4C ;'1905E Acc '; 'FORTRAN '; 2987, compiler XFAT TR0,TR1 ;'1905E Acc '; 'FORTRAN '; 72.1, compiler XFAT TR2 ;'1905F '; 'FORTRAN '; 3910, E128 compiler XFAT 2E ;'1906A '; 'FORTRAN '; 14600, E132 and E133 compiler XFAT 3E ;'1906S '; 'FORTRAN '; 25633, E591 compiler XFIF 1C ;'ICL 2970 '; 'FORTRAN '; 24040, compiler ERCC 4/77 ;'ICL 2970 '; 'FORTRAN '; 23700, compiler F1 5x10 ;'ICL 2980 '; 'FORTRAN '; 40942, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; 38043, compiler ERCC 4/77 ;'ICL ATLAS1'; 'FORTRAN '; 4167, E127 ;'ICL 4130/2'; 'FORTRAN '; 2632, E1013 ;'Bur 5500 '; 'FORTRAN '; 1591, E581 compiler 15.1 ;'Bur 6714 F'; 'FORTRAN '; 7782, E582 compiler 2.6 ;'Bur 6714 F'; 'FORTRAN '; 7550, E155 compiler I10 ;'Bur 6715 S'; 'FORTRAN '; 5085, compiler 2.7 ;'Cyber 72 '; 'FORTRAN '; 9626, E584 compiler FTN ;'Cyber 73 '; 'FORTRAN '; 9900, E146 compiler RUN ;'Cyber 173 '; 'FORTRAN '; 14008, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 14343, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 11900, E147 and E148 compiler RUN ;'CDC 7600 '; 'FORTRAN + '; 92360, E565 compiler FTN2.0 OPT=2 ;'CDC 7600 '; 'FORTRAN '; 87100, E564 compiler FTN2.0 OPT=1 ;'CDC 7600 '; 'FORTRAN '; 77600, E149 and E150 compiler RUN ;'Hon 6025 '; 'FORTRAN '; 4933, E466 and E569 ;'Hon 6040 '; 'FORTRAN '; 6767, ;'Hon 6060 '; 'FORTRAN '; 7143, E573 ;'Hon 66/40 '; 'FORTRAN '; 13139, ;'Hon 66/60 '; 'FORTRAN '; 13333, ;'DEC 20 '; 'FORTRAN '; 12050, E1029 ;'DEC KI 10 '; 'FORTRAN '; 13900, E570 ;'DEC KL 10 '; 'FORTRAN '; 48880, E731 ;'Univ1106 U'; 'FORTRAN '; 7301, E588 ;'Univ1108 '; 'FORTRAN '; 20100, E143 and E144 ;'UniV1121 '; 'FORTRAN '; 14400, ;'INTER 5 '; 'FORTRAN '; 2250, E154 ;'UNIV418III'; 'FORTRAN '; 1780, E145 ;'XDS SIG5 '; 'FORTRAN '; 7800, E152 compiler EXT IVH ;'XDS SIG6 '; 'FORTRAN '; 8815, E583 ;'PDP 11/10 '; 'FORTRAN '; 1646, E1010 compiler V01B-080 ;'PDP 11/20 '; 'FORTRAN '; 3570, E566 ;'PDP 11/34S'; 'FORTRAN '; 1269, ;'PDP 11/40E'; 'FORTRAN '; 1477, E660 ;'PDP 11/40S'; 'FORTRAN '; 2025, ;'Hon H316 F'; 'FORTRAN '; 4902, E576 ;'HP 2100A '; 'FORTRAN '; 4762, E578 ;'HP 3000 I '; 'FORTRAN '; 5471, ;'MODCOMP IV'; 'FORTRAN '; 9732, ;'PRIME 300H'; 'FORTRAN '; 6294, ;'GEC 4080 '; 'FORTRAN '; 7353, E586 ;'EAL 1830 '; 'FORTRAN '; 5950, E567 ;'IBM 1130 '; 'FORTRAN '; 2190, E568 ;'VAR620/100'; 'FORTRAN '; 3250, E571 ;'NOVA 840 S'; 'FORTRAN '; 8474, E1024 end of IF Test) (;'DOUBLE FUN' is the name of the Benchmark ;'FOR Numflp' is the application area relevance=; .988, This is the program FOPR04 which test the double length functions of FORTRAN machine software result ;'360/65H '; 'FORTRAN '; 47.2, E174 ;'360/67 '; 'FORTRAN + '; 110.6, compiler H OPT=2 ;'360/85 '; 'FORTRAN '; 294.1, compiler G ;'360/195 '; 'FORTRAN + '; 1320, E176 compiler H OPT=2 ;'360/195 '; 'FORTRAN '; 459, E175 compiler G ;'360/50 '; 'FORTRAN '; 7.14, E171 and E172 ;'360/30 MI '; 'FORTRAN '; 1.497, E1004 compiler DOS ;'370/135 '; 'FORTRAN '; 11.9, E600 compiler G ;'370/158 '; 'FORTRAN '; 72.57, compiler G ;'370/165 '; 'FORTRAN '; 260, E177 compiler G ;'370/168 '; 'FORTRAN + '; 449.7, compiler H not PP OPT=2 ;'370/168 '; 'FORTRAN '; 268.8, compiler G ;'370/168 M '; 'FORTRAN + '; 1042, compiler H+ ;'370/168 M '; 'FORTRAN '; 319.5, compiler G ;'Amdahl 470'; 'FORTRAN + '; 849.5, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; 5.33, ;'ICL 4/70 '; 'FORTRAN '; 38.74, ;'ICL 4/72 '; 'FORTRAN '; 42, ;'1902A 20SC'; 'FORTRAN '; .715, E163 and E164 compiler XFAT 2E ;'1903A SC '; 'FORTRAN '; 1.89, E166 compiler XFAT 4E ;'1903T '; 'FORTRAN '; 2.74, ;'1904A FP '; 'FORTRAN '; 3.6, E167 compiler XFAT 4C ;'1905E Acc '; 'FORTRAN '; 1.70, compiler TR0 ;'1905E Acc '; 'FORTRAN '; 1.65, compiler TR1 ;'1905E Acc '; 'FORTRAN '; 1.20, compiler TR2 ;'1906A '; 'FORTRAN + '; 30.5, E170 compiler XFEW 2A ;'1906A '; 'FORTRAN '; 12.4, E168 and E169 compiler XFAT 2B ;'1906S '; 'FORTRAN '; 9.804, compiler XFIH ;'ICL 2970 '; 'FORTRAN '; 50, compiler ERCC 4/77 ;'ICL 2970 '; 'FORTRAN '; 41.2, compiler F1 5x10 ;'ICL 2980 '; 'FORTRAN '; 160, compiler ERCC 4/77 ;'ICL ATLAS1'; 'FORTRAN '; 6.17, E601 ;'Bur 5500 '; 'FORTRAN '; 2.28, E597 compiler 15.1 ;'Bur 6715 S'; 'FORTRAN '; 13.62, compiler 2.7 ;'Cyber 72 '; 'FORTRAN '; 28.6, E598 compiler FTN ;'Cyber 73 '; 'FORTRAN '; 30, E181 compiler RUN ;'Cyber 173 '; 'FORTRAN '; 62.5, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 128.2, compiler FTN ;'CDC 6600 '; 'FORTRAN '; 102, E180 compiler RUN ;'CDC 7600 '; 'FORTRAN '; 806, compiler FTN ;'CDC 7600 '; 'FORTRAN '; 689, E182 and E183 compiler RUN ;'Hon 6025 '; 'FORTRAN '; 17.5, E467 and E603 ;'Hon 6040 '; 'FORTRAN '; 27.78, ;'Hon 6060 '; 'FORTRAN '; 50.5, E604 ;'Hon 66/40 '; 'FORTRAN '; 116.3, ;'Hon 66/60 '; 'FORTRAN '; 94.34, ;'DEC 20 '; 'FORTRAN '; 60.9, E1028 ;'DEC KI 10 '; 'FORTRAN '; 61, E607 ;'DEC KL 10 '; 'FORTRAN + '; 166.7, ;'DEC KL 10 '; 'FORTRAN '; 113.8, E732 ;'Univ1106 U'; 'FORTRAN '; 53.9, E599 ;'Univ1108 '; 'FORTRAN '; 140, E179 ;'XDS SIG5 '; 'FORTRAN '; 26.3, E185 compiler EXT IVH ;'PDP 11/20 '; 'FORTRAN '; 1.08, E608 ;'PDP 11/34S'; 'FORTRAN '; 1.307, ;'PDP 11/40E'; 'FORTRAN '; 1.412, E661 ;'PDP 11/40S'; 'FORTRAN '; 1.923, ;'Hon DDP516'; 'FORTRAN '; .667, E609 ;'Hon H316 F'; 'FORTRAN '; .625, E594 ;'HP 2100A '; 'FORTRAN '; 0.623, E595 ;'HP 3000 I '; 'FORTRAN '; 1.38, ;'MODCOMP IV'; 'FORTRAN '; 44.64, ;'PRIME 300H'; 'FORTRAN '; 14.01, ;'GEC 4080 '; 'FORTRAN '; 23.82, ;'VAR620/100'; 'FORTRAN '; 0.07936, E602 end of DOUBLE FUN) (;'Functions' is the name of the Benchmark ;'Scientific' is the application area relevance=; 2.435, This is a weighted average of the time to call some standard numerical functions. The data comes from the ALGOL statement mix or FORTRAN program machine software result ;'360/65G '; 'Assembler '; -1667, E635 compiler ALGOL F single ;'360/67 '; 'Assembler '; -1549, compiler FORTRAN H2 single ;'360/67 '; 'Assembler '; -1870, E633 compiler ALGOL W single ;'360/50 '; 'Assembler '; -6381, E641 compiler ALGOL F single ;'370/155 '; 'Assembler '; -2103, E642 compiler ALGOL F single ;'370/165 '; 'Assembler '; -434.7, E634 compiler ALGOL F single ;'370/168 '; 'Assembler '; -303, compiler SIMULA ;'370/168 '; 'Assembler '; -460.3, compiler FORTRAN H single ;'Amdahl 470'; 'Assembler '; -201.8, compiler FORTRAN H single ;'ICL 4/50 '; 'Assembler '; -13503, compiler FORTRAN ;'ICL 4/70 '; 'Assembler '; -2480, compiler FORTRAN ;'ICL 4/72 '; 'Assembler '; -2280, compiler FORTRAN ;'ICL 4/75 P'; 'Assembler '; -2566, E632 compiler ALGOL W single ;'S4004/55 '; 'Assembler '; -58700, compiler ALOL double ;'1903 EMU '; 'Assembler '; -36450, compiler XALE Mk5C ;'1903A SC '; 'Assembler '; -25876, E639 compiler ALGOL XALT ;'1904A FP '; 'Assembler '; -4838, E638 compiler ALGOL XALT ;'1904S FP I'; 'Assembler '; -4252, E637 compiler ALGOL XALT ;'1905E Acc '; 'Assembler '; -13600, compiler XFAT TR0 ;'1905E Acc '; 'Assembler '; -20800, compiler XFAT TR1 ;'1905E Acc '; 'Assembler '; -40500, compiler XFAT TR2 ;'1906A '; 'Assembler '; -1366, E625 compiler ALGOL XALV ;'ICL 2970 '; 'Assembler '; -2418, E1035 compiler Edinburgh FORTRAN ;'ICL 2970 '; 'Assembler '; -3636, compiler FORTRAN F1 5x10 ;'ICL 2980 '; 'Assembler '; -522, compiler F1 4/77 ;'ICL 2980 '; 'Assembler '; -813.5, compiler Edinburgh FORTRAN ;'ICL ATLAS1'; 'Assembler '; -4570, E624 compiler ALGOL ;'ICL 4130/2'; 'Assembler '; -16690, E1014 ;'Bur 5500 '; 'Assembler '; -13116, E626 compiler ALGOL ;'Bur 6714 F'; 'Assembler '; -4098, E627 compiler ALGOL ;'CDC 3600 '; 'Assembler '; -2497, compiler ALGOL ;'Cyber 73 '; 'Assembler '; -3158, E640 compiler ALGOL ;'CDC 6600 '; 'Assembler '; -1081, E629 compiler ALGOL ;'CDC 7600 '; 'Assembler '; -220.3, E630 compiler ALGOL ;'Hon GE 635'; 'Assembler '; -11580, compiler ALGOL ;'Hon 6030 '; 'Assembler '; -15850, compiler ALGOL ;'Hon 6050 '; 'Assembler '; -6932, compiler ALGOL ;'Hon 66/40 '; 'Assembler '; -4495, compiler Dartmouth ALGOL ;'ICL KDF9 '; 'Assembler '; -13152, E628 compiler BABEL ;' TR4 '; 'Assembler '; -10140, compiler ALGOL ;' TR440 '; 'Assembler '; -2042, compiler ALGOL ;'Univ1108 '; 'Assembler '; -704.7, E636 compiler NU ALGOL ;'PDP 11/10 '; 'Assembler '; -92400, E1011 compiler V01B-080 FORTRAN end of Functions) (;'Synth ALGL' is the name of the Benchmark ;'ALGOL 60' is the application area relevance=; 1.588, This is the program ALPR12, the ALGOL 60 version of the Curnow Benchmark machine software result ;'360/65H '; 'ALGOL 60 '; 173, E460 compiler Delft LONG Notest ;'360/65H '; 'ALGOL 60 '; 72, E203 compiler F single ;'360/65H '; 'al double '; 65, E224 compiler F double ;'360/65H '; 'PL/I + '; 443, compiler OPT single ;'360/65H '; 'PL/I '; 372, compiler F single ;'ICL 4/75 P'; 'ALGOL 60 '; 98, E724 ;'ICL 4/75 P'; 'ALGOL 60 '; 250, E725 compiler 68-C ;'ICL 4/75 P'; 'ALGOL 60 '; 235, E726 compiler ALGOL W ;'ICL 4/75 P'; 'al double '; 211, E727 compiler ALGOL W LONG ;'1903 EMU '; 'ALGOL 60 '; 23.2, compiler XALE Mk5C ;'1904A FP '; 'ALGOL 60 '; 125, E204 compiler XALT TR0 ;'1904A FP '; 'ALGOL 68 '; 167, E205 compiler 68-R nocheck ;'1904S FP I'; 'ALGOL 60 '; 154, E206 compiler XALT TR0 ;'1905F '; 'ALGOL 60 '; 71, E207 compiler XALT TR0 ;'1906A '; 'ALGOL 60 '; 379, E208 compiler XALV ;'1906A '; 'Pascal '; 661, E210 compiler XPAC ;'1906A '; 'ALGOL 68 '; 578, E209 compiler 68-R ;'1906S '; 'ALGOL 60 '; 536, E211 compiler XALT TR0 ;'ICL 4120/2'; 'ALGOL 60 '; 8.3, E215 ;'ICL 4130/2'; 'ALGOL 60 '; 53.2, E217 750ns store ;'Bur 5500 '; 'ALGOL 60 '; 94, E003 compiler MK XV.1 ;'Bur 6714 F'; 'ALGOL 60 '; 235, E002 compiler 2.6 ;'Bur 6714 F'; 'al double '; 111, E220 compiler 2.6 Double ;'Bur 6715 S'; 'ALGOL 60 '; 258, compiler 2.7 ;'Bur 6715 S'; 'PL/I '; 197, compiler 2.7 ;'Bur 6715 S'; 'al double '; 135, compiler 2.7 ;'Cyber 72 '; 'ALGOL 60 '; 141, E193 compiler 4.0 O=5 ;'Cyber 72 '; 'ALGOL 60 '; 135, E194 compiler 4.0 O=1 ;'CDC 7600 '; 'ALGOL 60 '; 2245, E191 compiler 4.0 O=5 ;'CDC 7600 '; 'ALGOL 60 '; 2105, E192 compiler 4.0 O:1 ;'CDC 7600 '; 'ALGOL 60 '; 1280, E190 compiler 3.0 ;'CDC 7600 '; 'Pascal '; 6850, no bound checking ;'CDC 7600 '; 'Pascal '; 6240, bound checking ;'Hon 6040 '; 'ALGOL 60 '; 101, ;'Hon 6060 '; 'ALGOL 60 '; 161, ;'DEC KL 10 '; 'ALGOL 60 '; 403.6, compiler 6A/634 ;'DEC KL 10 '; 'al double '; 360, compiler 6A/634 ;'ICL KDF9 '; 'ALGOL 60+ '; 62, E214 compiler Kidsgrove OPT ;'ICL KDF9 '; 'ALGOL 60 '; 43, E213 compiler Kidsgrove noopt ;'Univ1108 '; 'ALGOL 60 '; 282, E219 compiler NU ;'Univ1108 '; 'al double '; 221, E226 compiler NU double ;'XDS SIG6 '; 'al double '; 36, E225 double ;'XDS SIG6 '; 'al double '; 38, E218 end of Synth ALGL) (;'ALGOL Mix' is the name of the Benchmark ;'ALGOL 60' is the application area relevance=; 1.804, This is the ALGOL 60 statement mix described in NPL report NAC42 machine software result ;'360/50 '; 'ALGOL 60 '; .163, E437 compiler F single ;'360/50 '; 'al double '; .148, E436 compiler F double ;'370/165 '; 'ALGOL 60 '; 4.06, E402 compiler F ;'370/165 '; 'ALGOL W '; 6.09, E424 compiler with debug 0 ;'370/165 '; 'PL/I + '; 6.01, E428 compiler 2.0 NO.OPT OPT[T] ;'370/165 '; 'PL/I '; 5.48, E427 compiler 5.4 OPT=2 ;'370/168 '; 'SIMULA '; 8.44, compiler SIMULA nosbchk ;'ICL 4/50 '; 'ALGOL 60 '; 0.95, E438 compiler nocheck ;'ICL 4/70 '; 'al double '; .645, E426 compiler ICL double ;'ICL 4/75 P'; 'ALGOL 60 '; 1.48, E400 compiler EMAS nocheck ;'S 4004/55 '; 'ALGOL 60 '; .191, E1020 see NAC42 ;'1903 EMU '; 'ALGOL 60 '; .136, compiler XALE Mk5C ;'1903A SC '; 'ALGOL 60 '; .272, E442 compiler XALT 5 TR0 noopt ;'1904A FP '; 'ALGOL 60 '; .91, E443 compiler XALT 5 TR0 OPT=3 ;'1904S FP I'; 'ALGOL 60 '; .905, E444 compiler XALT ;'1906A '; 'ALGOL 60 '; 2.66, E463 compiler XALV noopt TR0 ;'1906A '; 'ALGOL 60 '; 3.33, E462 compiler 68-R nocheck ;'1906A '; 'Pascal '; 4.68, E429 compiler XPAC 1B ;'ICL 2970 '; 'al double '; 2.506, E1037 compiler Edinburgh ;'ICL 2980 '; 'al double '; 9.53, E1038 compiler Edinburgh ;'ICL MU5 '; 'ALGOL 60 '; 17.4, E1018 see NAC42 ;'ICL ATLAS1'; 'ALGOL 60 '; 1.0, E393 see NPL report NAC42 ;'ICL 4130/2'; 'ALGOL 60 '; .333, E432 ;'Bur 5500 '; 'ALGOL 60 '; .527, E394 see.NPL report NAC42 ;'CDC 3300 '; 'ALGOL 60 '; .309, E418 see NPL report NAC42 ;'CDC 3300 '; 'SIMULA '; .411, E419 SIMULA see NAC42 ;'CDC 3600 '; 'ALGOL 60 '; .804, E1017 Norwegian compiler ;'Cyber 73 '; 'ALGOL 60 '; 1.21, E441 compiler 2.0 see NAC42 ;'Cyber 73 '; 'Pascal '; 2.88, E440 Theoretical times ;'CDC 6600 '; 'ALGOL 60 '; 2.69, E399. compiler 2.0 see NAC42 ;'CDC 6600 '; 'Pascal '; 8.54, E430 Theoretical times see NAC42 ;'CDC 7600 '; 'ALGOL 60 '; 13.5, E401 compiler 3.0 no check ;'CDC 7600 '; 'ALGOL 60 '; 8.62, E439 compiler 3.0 checks ;'Hon GE 635'; 'ALGOL 60 '; .781, E433 ;'Hon 6030 '; 'ALGOL 60 '; .624, see NAC42 ;'Hon 6050 '; 'ALGOL 60 '; 1.16, see NAC42 ;'Hon 66/40 '; 'ALGOL 60 '; .553, compiler Dartmouth ALGOL ;'ICL KDF9 '; 'ALGOL 60 '; .407, E398 compiler Egdon ;'ICL KDF9 '; 'ALGOL 60 '; .393, E397 compiler Kidsgrove noopt ;' TR4 '; 'ALGOL 60 '; .298, E1021 see NAC42 ;' TR440 '; 'ALGOL 60 '; 1.70, ;' EL-X8 '; 'ALGOL 60 '; .37, E1019 compiler Petten ;'Univ1108 '; 'ALGOL 60 '; 2.34, E421 compiler NU ;'UniV1108 '; 'SIMULA '; 3.75, E422 ;'XDS SIG6 '; 'ALGOL 60 '; .362, E435 end of ALGOL Mix) (;'Ackermann' is the name of the Benchmark ;'ALGL Jumps' is the application area relevance=; .593, This program tests the speeds of procedure calls machine software result ;'360/65H '; 'PL/I '; -351, E034 compiler 5.4 ;'360/67 '; 'ALGOL W '; -121, E006 compiler Mk2 ;'360/75 '; 'ALGOL 60 '; -870, E004 compiler F ;'360/75 '; 'PL/I '; -101, E035 compiler OPT 1.2.2 ;'360/75 '; 'ALGOL W '; -103, E007 ;'370/158 '; 'Pascal '; -39, E047 ;'370/165 '; 'ALGOL 60 '; -43.8, E005 compiler Delft ;'370/165 '; 'BCPL '; -5.9, E060 compiler Cambridge ;'ICL 4/70 '; ' RTL/2 '; -46, E051 compiler ICI ;'ICL 4/75 P'; ' IMP '; -46, E008 compiler EMAS nocheck ;'1906A '; 'ALGOL 60 '; -29.2, E040 compiler Manchester ;'1906A '; 'Pascal '; -31.5, E049 compiler XPAC ;'1906S '; 'ALGOL 60 '; -70.9, E039 compiler XALV ;'1906S '; 'Pascal '; -19.1, E048 compiler XPAC ;'ICL 2980 '; 'ALGOL 60 '; -12, E1034 compiler Edinburgh ;'Bur 5500 '; 'ALGOL 60 '; -135, E036 compiler MK XV.1.01 ;'CDC 3300 '; 'SIMULA '; -1445, E030 ;'Cyber 73 '; 'Pascal '; -34, E045 ;'CDC 6600 '; 'ALGOL 60 '; -410, E010 compiler 2.0 ;'CDC 6600 '; 'Pascal '; -18, E046 dated June 1975 ;'CDC 6600 '; 'ALGOL 68 '; -35.8, E011 dated June 1975 ;'CDC 6600 '; 'SIMULA '; -366, E031 and E032 ;'DEC KI 10 '; 'SIMULA '; -317, E033 ;'ICL KDF9 '; 'ALGOL 60 '; -532, E038 compiler Kidsgrove ;'Univ1108 '; 'ALGOL 60 '; -175, E012 compiler NU ;'Univ1108 '; 'SIMULA '; -120, E029 ;'PDP 11/20 '; ' RTL/2 '; -107, E052 compiler ICI ;'PDP 11/20 '; 'Assembler '; -27.5, E057 end of Ackermann) (;'GAMM ALGOL' is the name of the Benchmark ;'ALG numflp' is the application area relevance=; 1.016, This program calculates the GAMM figure for ALGOL 60 reflects scientific usage machine software result ;'360/67 '; 'ALGOL W '; -15.2, E410 no check ;'ICL ATLAS1'; 'ALGOL 60 '; -56.4, E415 compiler Chilton ;'Bur 5500 '; 'ALGOL 60 '; -63.7, E416 compiler level 0 27/2/69 ;'Cyber 73 '; 'ALGOL 60 '; -57, E408 compiler 3.0 2/11/73 ;'CDC 6600 '; 'ALGOL 60 '; -12.12, E413 compiler 2.0 17/5/71 ;'CDC 7600 '; 'ALGOL 60 '; -5.55, E409 compiler 3.0.5 no check ;'ICL KDF9 '; 'ALGOL 60 '; -97.6, E412 compiler Egdon nocheck ;'ICL KDF9 '; 'ALGOL 60 '; -134, E411 compiler Kidsgrove noopt ;'Univ1108 '; 'ALGOL 60 '; -12.2, E414 compiler NU end of GAMM ALGOL) (;'Chess Mate' is the name of the Benchmark ;'ALG nonnum' is the application area relevance=; .635, The program is A G Bells chess algorithm machine software result ;'1906A '; 'ALGOL 60 '; -30, E406 compiler XALT from AG Bell ;'ICL ATLAS1'; 'ALGOL 60 '; -100, E403 from AG Bell ;'Bur 5500 '; 'ALGOL 60 '; -220, E404 from AG Bell ;'CDC 6600 '; 'ALGOL 60 '; -100, E407 compiler 1.0 from AG Bell ;'Univ1108 '; 'ALGOL 60 '; -90, E405 compiler Old from AG Bell end of Chess Mate) (;'GAMM Asmbl' is the name of the Benchmark ;'Scientific' is the application area relevance=; 2.632, This is the GAMM loops programmed in machine code. Performance measure is the loop time in microseconds machine software result ;'ICL 4/72 '; 'Assembler '; -5.9, ;'1901A 10SC'; 'Assembler '; -108.7, E745 ;'1902A 20SC'; 'Assembler '; -87.1, E754 ;'1902S 25S '; 'Assembler '; -87.1, E753 ;'1903A SC '; 'Assembler '; -44.4, E751 ;'1903S 31S '; 'Assembler '; -44.4, E750 ;'1903T '; 'Assembler '; -15, E749 ;'1904A FP '; 'Assembler '; -11, E748 ;'1906A '; 'Assembler '; -3.5, E740 ;'1906S '; 'Assembler '; -2.4, E746 end of GAMM Asmbl) (;'POWU' is the name of the Benchmark ;'Commercial' is the application area relevance=; 3.524, This is the Post Offuce Work Unit devised to measure the commercial ability of a computer in a similar manner to the ADP Mix The performance measure is in millliseconds machine software result ;'ICL 4/70 '; 'Assembler '; -1.94, ;'ICL 4/72 '; 'Assembler '; -1.73, ;'1901A 10SC'; 'Assembler '; -45, E769 ;'1902A 20SC'; 'Assembler '; -11.5, E764 ;'1902S 25S '; 'Assembler '; -11.5, E763 ;'1903A SC '; 'Assembler '; -5.9, E761 ;'1903S 31S '; 'Assembler '; -5.9, E760 ;'1903T '; 'Assembler '; -4, E759 ;'1904A FP '; 'Assembler '; -3. E758 ;'1906A '; 'Assembler '; -94, ;'1906S '; 'Assembler '; -65, E755 end of POWU) (;'GAMM F' is the name of the Benchmark ;'FOR numflp' is the application area relevance=; 3.269, This is the GAMM loops as timed by a program written in FORTRAN using single length arithmetic. The program is available from the National Physical Laboratory on ISO code paper tape. Performance measure is the loop in microseconds. machine software result ;'360/67 '; 'FORTRAN + '; -5.268, compiler H OPT=2 ;'370/168 '; 'FORTRAN + '; -1.264, compiler H not PP OPT=2 ;'Amdahl 470'; 'FORTRAN + '; -.563, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; -72.38, ;'ICL 4/70 '; 'FORTRAN '; -15.4, ;'ICL 4/72 '; 'FORTRAN '; -13.41, ;'ICL 4/72 '; 'FORTRAN '; -21.28, with debug ;'1905E Acc '; 'FORTRAN '; -70, compiler XFAT TR0,TR1 ;'1905E Acc '; 'FORTRAN '; -1043, compiler XFAT TR2 ;'1906A '; 'FORTRAN + '; -7.7, compiler XFEV ;'1906A '; 'FORTRAN '; -8.7, compiler XFIV ;'ICL 2970 '; 'FORTRAN '; -7.21, compiler ERCC 4/77 ;'ICL 2970 '; 'FORTRAN '; -7.67, compiler F1 5x10 ;'ICL 2980 '; 'FORTRAN '; -1.55, compiler F1 4/77 ;'ICL 2980 '; 'FORTRAN '; -1.407, compiler ERCC 4/77 ;'CDC 7600 '; 'FORTRAN + '; -.3, compiler FTN OPT=2 ;'CDC 7600 '; 'FORTRAN '; -1.5, compiler FTN OPT=0 ;'DEC KL 10 '; 'FORTRAN + '; -3.4, compiler F10 4A/317 ;'DEC KL 10 '; 'FORTRAN '; -3.4, compiler F10 4A/317 end of GAMM F) (;'GAMM FD' is the name of the Benchmark ;'FOR numflp' is the application area relevance=; 1.217, This is the GAMM loops as timed by a program written in FORTRAN using single length arithmetic. The program is available from the National Physical Laboratory on ISO code paper tape. Performance measure is the loop in microseconds. machine software result ;'360/67 '; 'FORTRAN + '; -6.977, compiler H OPT=2 ;'370/168 '; 'FORTRAN + '; -1.544, compiler H not PP OPT=2 ;'Amdahl 470'; 'FORTRAN + '; -.892, compiler H not PP OPT=2 ;'ICL 4/50 '; 'FORTRAN '; -129.9, ;'ICL 4/70 '; 'FORTRAN '; -18.67, ;'ICL 4/72 '; 'FORTRAN '; -16.13, ;'ICL 4/72 '; 'FORTRAN '; -24, with debug ;'1905E Acc '; 'FORTRAN '; -609, compiler XFAT TR0,TR1 ;'1905E Acc '; 'FORTRAN '; -1650, compiler XFAT TR2 ;'1906A '; 'FORTRAN + '; -10, compiler XFEV ;'1906A '; 'FORTRAN '; -11, compiler XFIV ;'ICL 2970 '; 'FORTRAN '; -7.955, compiler ERCC 4/77 ;'ICL 2970 '; 'FORTRAN '; -9.5, compiler F1 5x10 ;'ICL 2980 '; 'FORTRAN '; -1.97, compiler F1 4/77 ;'CDC 7600 '; 'FORTRAN + '; -.78, compiler FTN OPT=2 ;'CDC 7600 '; 'FORTRAN '; -1.5, compiler FTN OPT=0 ;'DEC KL 10 '; 'FORTRAN + '; -5.83, compiler F10 4A/317 ;'DEC KL 10 '; 'FORTRAN '; -6.4, compiler F10 4A/317 end of GAMM FD)
Given n benchmarks and m machines, one has the formula
Tij = Bi × Mj × Rij (1)
where Tij is the performance figure for the ith benchmark on machine j, Bi is the factor giving the complexity of the benchmark, Mj is the performance figure for the machine and the Rij are the residuals (which are as near to 1 as possible). The problem is that given some (but not all) the Tij one requires estimates for the Bi and Mj.
Taking logarithms, and denoting the logarithm by a prefix L one has
LTij = LBi + LMj + LRij (2)
We require that the LRij should be of as small a magnitude as possible taking into account the relevance of each benchmark. The relevance is given by a weighting factor to determine the degree of fit on each residual. So we require to minimize
where the wi are the weights for the benchmarks.
If any individual value of the Tij is not available, then the best estimate that could be used in its place would be Bi × Mj. In this case, the corresponding term in E is zero, so the sum in equation (3) is best expressed as
This equation can now be differentiated with respect to the unknowns LBi and LMj to give linear equations in those variables.
Differentiating with respect to LBi gives the n equations
or
or
Differentiating equation (4) with respect to LMj gives the m equations
The sum of all the equation (5) is the same as the sum of all the equation (7), so there appears to be inadequate information to provide a unique solution. However, since only the ratios of the Mj are required, one can take M1 = 1 or LM1 = 0. Substituting the LBi into equation (7) using equation (6) one obtains linear equations for the Mj (j = 2 to m). These equations can be solved by classical linear algebra routines.
Take the following data from three machines and three benchmarks as an example
A B C 1906A 4/50 370/168 1: Gibson Mix 866 55 ? 2: Synth FORS 585 51 2439 3: Bit Test ? 391 13352 _ _ | 6.764 4.007 - | | | LTij = | 6.374 3.932 7.799 | | | | - 5.969 9.499 | |_ _|
Take
w1 = 1.0 w2 = 0.6 w3 = 0.7
The unknowns are LB1, LB2, LB3 and LMB, LMC since LMA = 0
Equation (6) gives the following
LB1 = (6.764 + 4.007 - LMA - LMB)/2 or LB1 = 5.385 - .5 LMB (8) LB2 = (6.374 + 3.932 + 7.799 - LMB - LMC)/3 or LB2 = 6.035 - .3333 LMB - .3333 LMC (9) LB3 = (5.969 + 9.499 - LMB - LMC)/2 or LB3 = 7.734 - .5 LMB - .5 LMC (10)
Equation (7) gives
(6.764 - LB1) + .6 × (6.374 - LB2) = 0 or LB1 + .6 LB2 = 10.59 (11) and (4.007 - LB1 - LMB) + .6 × (3.932 - LB2 - LMB) + .7 × (5.969 - LB3 - LMB) = 0 or LB1 + .6 LB2 + .7 LB3 + 2.3 LMB = 10.54 (12)
Since the last equation from (7) is linearly dependent upon equations 8-12, it need not be considered.
Substituting for the LBi in (11) and (12) , we have
5.385 - .5 LMBB + .6(6035 - .3333 LMC - .3333 LMC) = 10.59 or - .7 LMB - .2 LMC = 1.584 (13) and .7 LB3 + 2.3 LMB = 10.54 - 10.59 or 1.95 LMB - .35 LMC = - 5.464 (14)
By solving (13) and (14) we now have
LMB = - 2.594 LMC = 1.159
Substituting back one gets
LB1 = 6.682 LB2 = 6.513 LB3 = 8.451
Taking exponents of the LMj, the ratio of the machine performances are
1906A : 4/50 : 370/168 1 : 0.075 3.19
Similarly, from the LB1, the benchmark factors are
Gibson Mix 798 Synth FORS 674 Bit Test 4680
In the full analysis, the values obtained for the machine performance ratios and the benchmark factors are slightly different from those above.
The residual Matrix is as follows.
1906A 4/50 370/168 Gibson Mix 1.085 .922 - Synth FORS .868 1.013 1.136 Bit Test - 1.12 .895
It is important to note that the Bi and Mj calculated by this method are comparatively insensitive to the weights wi. the reason for this is that the wi only determine the degree of reliance of the benchmark in the fitting process - it is not a direct weighting method.