{"title":"Performance testing of a FORTRAN library of mathematical function routines - a case study in the application of testing techniques","authors":"D. Lozier, L. Maximon, W. Sadowski","doi":"10.6028/JRES.077B.011","DOIUrl":null,"url":null,"abstract":"Thi s pape r describes a n app lication of tes ting methodology a nd te c hniques de ve loped by th e a uthors to a id in impro vi ng the qu a lity of mathem ati ca l soft wa re. These techni q ues diffe r rad ica ll y in se vera l as pects from techni q ues pre vious ly used . the most import a nt d iffere nce be ing th at th e testin g is not based e xc lus ivel y Oil random a rgume nt s. Ins tead , throughou t the range of eac h fun c tio n use is ma de of s pec ia l a rgum e nt s that a re des ig ned to d etec t progra mming e rrors and to tes t the pe rfnrm a nce of an algorithm in diffe re nt regions. T he fu nction values a re tes ted aga ins t refe re nce valu es whic h are s tored o n re fe re nce tapes ge ne ra ted b y a hi ghl y au the nti ca ted s yste m of s ubro utin es. S ince the eff ec tive ness of s uch a tes tin g sys te m in di scove ring e rrors and pe rform a nce limit ati ons can be fu ll y ascert a ined o nl y th rough ac tu a l use , we report the res ult s of e m ploying our s ys tem to tes t an exi stin g FORTRAN library of ma the matica l fun c tion routines. S pec if, c a s pects of the num e ri ca l accurac y of the libJ·a ry used in thi s te st case are di sc ussed in orde r to ill :~5tra t e the e ffec ti ve ness of a we ll -d es igned tes ting s yste m as a n ana lyti c tool for the e valuation of mathe matica l so ft wa re. S ince docum e nt ati on prov ides information necessary to perform testing and cont ain s s pecifications th at re flect the result s of testin g, our stud y inc ludes com me nts on th e doc ume nt at ion. No information on timing or storage re quire me nts is present ed in thi s case s tud y.","PeriodicalId":166823,"journal":{"name":"Journal of Research of the National Bureau of Standards, Section B: Mathematical Sciences","volume":"21 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1973-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"6","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Research of the National Bureau of Standards, Section B: Mathematical Sciences","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.6028/JRES.077B.011","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 6
Abstract
Thi s pape r describes a n app lication of tes ting methodology a nd te c hniques de ve loped by th e a uthors to a id in impro vi ng the qu a lity of mathem ati ca l soft wa re. These techni q ues diffe r rad ica ll y in se vera l as pects from techni q ues pre vious ly used . the most import a nt d iffere nce be ing th at th e testin g is not based e xc lus ivel y Oil random a rgume nt s. Ins tead , throughou t the range of eac h fun c tio n use is ma de of s pec ia l a rgum e nt s that a re des ig ned to d etec t progra mming e rrors and to tes t the pe rfnrm a nce of an algorithm in diffe re nt regions. T he fu nction values a re tes ted aga ins t refe re nce valu es whic h are s tored o n re fe re nce tapes ge ne ra ted b y a hi ghl y au the nti ca ted s yste m of s ubro utin es. S ince the eff ec tive ness of s uch a tes tin g sys te m in di scove ring e rrors and pe rform a nce limit ati ons can be fu ll y ascert a ined o nl y th rough ac tu a l use , we report the res ult s of e m ploying our s ys tem to tes t an exi stin g FORTRAN library of ma the matica l fun c tion routines. S pec if, c a s pects of the num e ri ca l accurac y of the libJ·a ry used in thi s te st case are di sc ussed in orde r to ill :~5tra t e the e ffec ti ve ness of a we ll -d es igned tes ting s yste m as a n ana lyti c tool for the e valuation of mathe matica l so ft wa re. S ince docum e nt ati on prov ides information necessary to perform testing and cont ain s s pecifications th at re flect the result s of testin g, our stud y inc ludes com me nts on th e doc ume nt at ion. No information on timing or storage re quire me nts is present ed in thi s case s tud y.
Thi s pape r n a描述了测试的app lication婷te methodology a和c hniques de loped了id: th e a uthors to a qu a lity》在impro vi ng mathem洛杉矶ca软肝wa re。这些techni q价值diffe r rad ica ll y在se维拉从美国洛杉矶pects techni q价值pre vious ly过去。《头号进口at a nt d iffere nce成为ing th th g e测试是基于e xc音符家伙ivel y Oil随机a rgume特斯拉s . Ins tead太阳城》throughou t, eac h有趣c tio n用马de l ' s agents of s pec他是rgum e特斯拉s发展到这种a re des ig奈德d etec t progra mming rrors e和t测试《pe rfnrm百万nce在diffe re nt regions of an算法。T他功夫nction价值观a aga re测试泰德ins T refe re nce价值冰whic h是s tored o n re fe re nce的录音带ge ne ra泰德b a嗨ghl y y au之nti ca泰德s yste m s ubro utin冰。S尼斯eff ec基金会》因斯很多a tin g姐姐te测试m在拳台scove e rrors和pe rform a nce盎司肝限额可以成为功夫ll y ascert y a ined o nl th rough屠a l用空调,我们报告e m ploying res奥创S》的S ys tem测试t an exi stin g FORTRAN图书馆《matica我妈的有趣c tion住。S a pec如果c, S pects num e ri ca》洛杉矶《libJ accurac y·瑞过去在thi S te圣凯斯是在r到秩序的sc ussed iii: ~ 5tra t e e ffec ti了尼斯》我们会测试-d冰igned婷美国yste m美国a - n - c ana lyti工具为《妈妈e valuation matica l所以ft wa e re因斯u.s.n ews docum nt上提供肝cybele资讯网测试必要去表演和cont ain ss pecifications th at re flect测试之论点S g,我们的帅哥y公司ludes com送我上nts th e医生ume nt at离子。没有关于时间或存储的信息。