R.M. Clapp et al. "Toward Real-Time Performance Benchmarks for Ada" Communications of the ACM 29(8):760-7788, August, 1986.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....of methods developed for the deterministic timing prediction of a function. These methods can be grouped into three categories. The first is testing, which is a method of evaluating the timing characteristics of code fragments by actually running them on representative test data (Vestal, 1990)(Clapp et al. 1986)(Plattner, 1984) However, since test data may not fully cover the domain of interest, and measurement may not be possible without setting up an actual environment, this approach may not, on its own, be acceptable in the hard real time domain. Second, the timing properties of the program can be ....

Clapp, R., Duchesneau, L., Volz, R., Mudge, T. and Schultze, T. (1986). Towards real-time performance benchmarks for ada, Communications of the ACM 29(8).

....concentrate on techniques for measuring the performance of individual features of the Ada programming language. The development of the real time performance measurement techniques and the interpretation of the benchmark results are based on the Ada notion of time. An article by the Michigan team [4] begins by reviewing the Ada concept of time and the measurement techniques used in the benchmarks. The specific features measured are then discussed, followed by a summary of the results obtained and an appraisal of these results. A follow up letter about the Michigan benchmarks appears in [3] ....

Clapp, Russell M., et al. Toward Real-Time Performance Benchmarks for Ada. Communications of the ACM 29(8):760-778, August, 1986.

....we run our loops large numbers of times and then repeat each such loop measurement several times. There are residual errors, however, due to clock resolution, external events like interrupts, multiprogramming and I O activity, unreproducible variations in the hit ratio of the cache, and paging [Clap86]. These issues are discussed in more detail in [Saav89] 5 2.4. The Program Analyzer The analysis of programs consists of two phases: the static analysis and the dynamic analysis. In the static phase, we count the number of occurrences of each AbOp in each line of source code. In the dynamic ....

....by the performance of certain sequences of AbOps, and not by a single AbOp in isolation e.g. the IBM RS 6000 multiply add instruction; we discuss this further below. We are not able to account for hardware or software interlocks, non linear interactions between consecutive machine instructions [Clap86], the effectiveness of branch prediction [Lee84] and the effect on timing of branch distance and direction. We have also not accounted for specialized architectural features such as vector operations and vector registers. Another source of errors corresponds to limitations in our measuring tools ....

Clapp, R.M., Duchesneau, L., Volz, R.A., Mudge, T.N., and Schultze, T., " Toward Real-Time Performance Benchmarks for ADA", Comm. of the ACM, Vol.29, No.8, August 1986, pp. 760-778.

....characteristics on the testbed target: initially a DEC MicroVAX II running the VAXELN real time executive, and later a Motorola MC68020 microprocessor. A MIL STD 1750A processor is also a possible future target. This report covers the Ada benchmarks developed at the University of Michigan [3] and by the 1 ACM SIGAda Performance Issues Working Group (PIWG) The PIWG benchmarks include Ada versions of the Whetstone [4] and Dhrystone [7] synthetic benchmarks. The report also discusses briefly the prototype Ada Compiler Evaluation Capability (ACEC) 6] 2. The University of Michigan Ada ....

....In addition, the performance of some run time system features scheduling and storage management is measured. The development of the real time performance measurement techniques and the interpretation of the benchmark results are based on the Ada notion of time. A paper by the Michigan team [3] begins by reviewing the Ada concept of time and the measurement techniques used in the benchmarks. The specific features measured are then discussed, followed by a summary and appraisal of the results obtained. 1 The benchmarks came from the PIWG distribution tape known as TAPE 8 31 86. The ....

[Article contains additional citation context not shown here]

Clapp, Russell M., et al. Toward Real-Time Performance Benchmarks for Ada. Communications of the ACM 29(8):760-778, August, 1986.

....run our loops large numbers of times and then repeat each such loop measurement several times. There are residual errors, however, due to clock resolution, external events like interrupts, multiprogramming and I O activity, and unreproducible variations in the hit ratio of the cache, and paging [Clap86] These issues are discussed in more detail in [Saav89] 2.4. The Program Analyzer The analysis of programs consists of two phases: the static analysis and the dynamic analysis. In the static phase, we count the number of occurrences of each AbOp in each line of source code. In the dynamic ....

....by the performance of certain sequences of AbOps, and not by a single AbOp in isolation e.g. the IBM RS 6000 multiply add instruction; we discuss this further below. We are not able to account for hardware or software interlocks, non linear interactions between consecutive machine instructions [Clap86] the effectiveness of branch prediction [Lee84] and the effect on timing of branch distance and direction. We have also not accounted for specialized architectural features such as vector operations and vector registers. Another source of errors corresponds to limitations in our measuring tools ....

Clapp, R.M., Duchesneau, L., Volz, R.A., Mudge, T.N., and Schultze, T., " Toward Real-Time Performance Benchmarks for ADA", Comm. of the ACM, Vol.29, No.8, August 1986, pp. 760-778.

.... suites, namely the Ada Compiler Evaluation Capability (ACEC) test suite [10] the Performance Issues Working Group (PIWG) test suite developed by a working group of the Association for Computing Machinery s Special Interest Group for Ada (SIGADA) and the University of Michigan test suite [5]. A major weakness of the dual loop benchmark method is that it assumes that textually equivalent code constructs require the same amount of time to execute. In particular, the time required to execute the 1 The Ada program shown here requires that the user initialize the constant LOOP COUNT and ....

Clapp, R. M. et al. 1986. Toward Real-Time Performance Benchmarks for Ada. Communications of the ACM. 29(8): pp. 760-778.

....report. Normally, specification to the lowest level (e.g. board revision level or software fix level for all components) is not available. 38 CMU SEI 87 TR 22 CMU SEI 87 TR 22 39 6. Variation Control Techniques Used by the University of Michigan Benchmarks The University of Michigan tests [3] focus on the execution speed of specific features of the Ada programming language. This approach isolates relatively small items to allow comparisons of critical interest (e.g. rendezvous versus subroutine call) but has led to some very interesting side effects. In particular, these focused ....

Clapp, Russell M., et al. Toward Real-Time Performance Benchmarks for Ada. Communications of the ACM 29(8):760-778, August, 1986.

....This technique allowed us to time single executions of code paths with far greater accuracy than can be obtained using a coarser resolution system clock. It also avoids the problems inherent in the common method of improving system clock resolution by taking averages of repeated invocations [8]. 3.2 TLB Simulation with Tapeworm Many previous TLB studies have used trace driven simulation to explore design trade offs [3, 9, 25] However, there are a number of difficulties with trace driven TLB simulation. First, it is difficult to obtain accurate traces. Code annotation tools like pixie ....

R.M. Clapp, L.J. Duchesneau, R.A. Volz, T.N. Mudge, and T. Schultze, "Toward Real-time Performance Benchmarks for Ada," Communications of the ACM, vol. 29, no. 8, Aug. 1986, pp. 760-778.

No context found.

R.M. Clapp et al. "Toward Real-Time Performance Benchmarks for Ada" Communications of the ACM 29(8):760-7788, August, 1986.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC