ࡱ>   Root Entry`DEST sV>˿PerfectOffice_MAIN& 0PerfectOffice_OBJECTS,@bV>˿@bV>˿ -$Craig T. Dedo$0Craig T. Dedo$J3 / 00-160r1..Interpretation 004" Fortran 2000f17130 W. Burleigh Place Brookfield, WI 53005&/(414) 783-5869   ( )0 ' L-Fortran 2000 Requirement Approved\ NCITS J3 ANSI Fortran Standards Committee*American English U' e37=CIQYag1.a.i.(1)(a)(i)1)a)"Arial Regular fHghQiEjQkElQmEnQo# e37=CIQYag1.a.i.(1)(a)(i)1)a) D IZ<Century Schoolbook Regular\(,Xz`ł3|xo(3D$ V0!   -X    f=)dd=  ****'dxd d(3D$ V0!   -X    f=)dd=  ݀Level 1Level 2Level 3Level 4Level 5,++)UC <<Level 1Level 2Level 3Level 4Level 5($$   1  55' dxdP Pd1, 2, 3,Level 1Level 2Level 3Level 4Level 5B#:.21, 2, 3,NumbersF  .0 Fa, b, c,Level 1Level 2Level 3Level 4Level 5V#:2a, b, c,Lowercase LettersH  .0 Hd V0!   -X    f=)dd=  ݛC  e'#CX ֢ ֣X  r^4` <hDp Lx !(#T$%&-r  %  @! Interpretation004   To:0  0(#(#J3(#(# From:0  JohnReid[>(#(# Date:0 WPC' 0fX[n{o.&, ܉kI1K|3.k^J&SXAHi_:J 'zz.EvI[2Q k-a%tUR t/^R~-t@0[g;9]E8}Jy+lW,:QʴzL|E=uEz~%8{gh۱1"X (`D=O@OyV[/ (I(t0O`*Eh 1I[<d; d3R$bF+tF?4b3Ϸ[w@x 1m%UT-4 0 D3 B 0 AS1 0D D/; Bjjjjjjjjjjjjjjjjj 0sM 0 C  c:\COREL\OFFICE7\TEMPLATE\X3J3DO~1.WPT(V0$   -X    f=)dd=(2F$ V0!   -X    f=)dd=  d:\office\wpwin\template\x3j3-rpt.wpt<6X9`(CourierXx6X@KX@<6X9`(Courierd6X@K@+`4New Century Schoolbooke P'CPA, B,Level 1Level 2Level 3Level 4Level 5 + F'   -X  f  ݛ ,-TT>(# J3/00-160r1##  X NCITSJ3ANSIFortranStandardsCommitteeh h B(#JohnReid Y Interpretation004NN=(#April17,2000  0 0 @(#Page  1  ڀof  1  4#(2Quick A.  .0 c:\Corel\Office7\Template\X3J3 Document.wpt  ,X.F 6C:\ANSI_J3\153\00-160r1.wpdFC:\ANSI_J3\153\00-160r1.wpd/  (3D$ V0!   -X    f=)dd=  H4#:.2Quick 1.F  .0 FInterpretation 004 April17,2000(#(# Subject:0  Interpretation004 (#(#   HerearethedraftsfortheANSWERandEDITSsectionsofInterpretation004.  Malcolmdisagreeswithmeonthisoneandwehavehadto"agreetodisagree".Heobjectsto"shouldbe",butwouldnotobjectto"maybe"(butstillthinksitisthewrongthingtodo).  IhaveaskedafewofmymathematicalfriendswhattheythinkandthosethatarefamiliarwiththeIEEEstandardandthethinkingbehinditagreewithmethataskingforMAXVALforazerosizearrayisjustlikeaskingfor1.0/0.0andshouldreturninf.NUMBER:000004TITLE:ValuereturnedbyMAXVAL/MINVALKEYWORDS:MAXVAL,MINVALDEFECTTYPE:InterpretationSTATUS:J3considerationinprogressQUESTION:TheResultValuesectionoftheMAXVALintrinsicfunctiondescriptionusesthephrasing:0  orhasthevalueofthenegativenumberofthelargestmagnitudesupportedbytheprocessorfornumbersofthetypeandkindtypeparameterofARRAYifARRAYhassizezero (#(# Thisphrasinghasgeneratedatleastthetwofollowingviewsonthereturnvalue:*IfthemachinesupportstheIEEEstandardthentheimplementationshouldreturn-inf.*Forportability,theimplementationshouldreturn-HUGE(ARRAY).Theseviewsleadtothefollowingquestions:1.Istheintentofthestandardtodescribetheresultintermsofmachinevaluesratherthanmodelvalues?2.Iftheanswerto1is"yes",howareprogrammersexpectedtousethisintrinsicfunctionportably?ANSWER:1.  Yes.Processorsmaysupportvaluesthatarenotpresentinthemodelof13.7.1.IEEEinfisanexampleof G*' suchanumberandthisshouldbereturnedonamachinethatsupportstheIEEEstandard.Ifthenegativenumberofthelargestmagnitudeinthemodelhadbeenintended,themodelwouldhavebeenmentionedas,forexample,inthedefinitionofHUGE(13.14.39).2.  AsimpleexampleofitsuseistotestwhetherasetofnumbersSET1hasavaluegreaterthananyvalueinthe v"Y, setSET2.ConsidertheexpressionMAXVAL(SET1)>MAXVAL(SET2).IfSET1isemptyandSET2isnot,thisvalueiscorrectlyfalseevenifallofthevaluesareoutsidethemodelwithvalueslessthanHUGE(SET1).  ItmaybehelpfultoconsiderhowMAXVALmightbecodedforanarrayofrankoneonanIEEEcomputer.Thefollowingcodeissuitable.0  MAXVAL=IEEE_VALUE(1.0,IEEE_NEGATIVE_INF) (#(# 0  DOI=1,SIZE(ARRAY) (#(# 0  0(#(#MAXVAL=MAX(MAXVAL,ARRAY(I)) (#(# 0  ENDDO (#(#   Allofthereductionfunctions(e.g.,MAXVAL,SUM)returnvalueswhentheargumentsarezerosizedarraysgthatgfawhichaaafmaybethesameasthevaluesreturnedforsomenonzerosizedarguments.EDITS:0  None.`+C%:(#(# SUBMITTEDBY:LarryRolisonHISTORY:iJ3/ij0 4 j97-240kkl0 4(#4(#0 (# (#lm143kkmmn0 (# (#nsubmittedI-,'=(#(# h0  0(#(#04(#(#00-160r10 4(#4(#m1530 (# (#hoProposedresponseo-'>(#(# [EndofJ3/00160r1]Root Entry`DESTgͿPerfectOffice_MAIN& 4PerfectOffice_OBJECTS,@bV>˿@bV>˿!"#$%&'()*+,-./012345678MAXVAL_,SUM)returnvalueswhentheargumentsarezerosizedarraysthatmaybethesameasthevaluesreturnedforsomenonzerosizedarguments.EDITS:0  None.*$;(#(# SUBMITTEDBY:Larry_Rolison_HISTORY:iJ3/ij0 4 j97-240kkl0 4(#4(#0 (# (#lm143kkmmn0 (# (#nsubmitted,&>(#(# h0  0(#(#04(#(#00-160r10 4(#4(#m1530 (# (#hoProposedresponseo2-5'?(#(# [EndofJ3/00160r1]WPCH! 4(ﱄuzLO,"B/Uuz֫&k`N %Sױ8sf[j ڛUȮo1(OI( l ¬CFqhMuXcVK[بu6utʷH6W1 :h¡'cuO]vqr҆4B*giαT{눚͘X;F7( NvT"OM(qHg?JG%ԡ^&= N]ylwu_}+Lf/Ps7W:iqsR(;%NO *ÍpU}(#5h J3/00-160r1#5h#  X NCITSJ3ANSIFortranStandardsCommitteed d B(#JohnReid I Interpretation004DD=(#April17,2000  ( ( ( (#Page  1  ڀof  1  4#(2Quick A.  .0 c:\Corel\Office7\Template\X3J3 Document.wpt _ ,X.F 6E:\ANSI_J3\153\00-160r1.wpdFE:\ANSI_J3\153\00-160r1.wpd/  (3$ V0!  5h5h5h -X    f=^~dd=  H4#:.2Quick 1.F  .0 FHP OfficeJet Pro 1170C Series0(,,,,0(0Level 1Level 2Level 3Level 4Level 5Interpretation 004 -$Craig T. Dedo$0Craig T. Dedo$J3 / 00-160r1..Interpretation 004" Fortran 2000f17130 W. Burleigh Place Brookfield, WI 53005&/(414) 783-5869   ( )0 ' L-Fortran 2000 Requirement Approved\ NCITS J3 ANSI Fortran Standards Committee*American English r(U'#137=CIQYag1.a.i.(1)(a)(i)1)a)"Arial Regular hQiEjQkElQmEnQopUqE##37=CIQYag1.a.i.(1)(a)(i)1)a) D IZ<Century Schoolbook RegularDs&,Xz`ł3|xq(3$ V0!  5h5h5h -X    f=^~dd=  ****'dxd d(3$ V0!  5h5h5h -X    f=^~dd=  ݀Level 1Level 2Level 3Level 4Level 5,++)UC <<Level 1Level 2Level 3Level 4Level 5($$   1  55' dxdP Pd1, 2, 3,Level 1Level 2Level 3Level 4Level 5B#:.21, 2, 3,NumbersF  .0 Fa, b, c,Level 1Level 2Level 3Level 4Level 5V#:2a, b, c,Lowercase LettersH  .0 Hd V0!  5h5h5h -X    f=^~dd=  _C  #1'#CX  X  r^4` <hDp Lx !(#T$%&-r  _%  qppq_@! Interpretation004   To:0  0(#(#J3(#(# From:0  JohnReid*-(#(# Date:0  April17,2000(#(# Subject:0 4 Interpretation004f i4(#4(#   HerearethedraftsfortheANSWERandEDITSsectionsofInterpretation004.  Malcolmdisagreeswithmeonthisoneandwehavehadto"agreetodisagree".He_objectsto_Ԁ"shouldbe",butwouldnotobjectto"maybe"(butstillthinksitisthewrongthingtodo).  IhaveaskedafewofmymathematicalfriendswhattheythinkandthosethatarefamiliarwiththeIEEEstandardandthethinkingbehinditagreewithmethataskingfor_MAXVAL_Ԁforazerosizearrayisjustlikeaskingfor1.0/0.0andshouldreturn_inf_.NUMBER:000004TITLE:Valuereturnedby_MAXVAL_/_MINVAL_KEYWORDS:_MAXVAL_,_MINVAL_DEFECTTYPE:InterpretationSTATUS:J3considerationinprogressQUESTION:TheResultValuesectionofthe_MAXVAL_Ԁintrinsicfunctiondescriptionusesthephrasing:0  orhasthevalueofthenegativenumberofthelargestmagnitudesupportedbytheprocessorfornumbersofthetypeandkindtypeparameterofARRAYifARRAYhassizezero (#(# Thisphrasinghasgeneratedatleastthetwofollowingviewsonthereturnvalue:*IfthemachinesupportstheIEEEstandardthentheimplementationshouldreturn-_inf_.*Forportability,theimplementationshouldreturn-HUGE(ARRAY).Theseviewsleadtothefollowingquestions:1.Istheintentofthestandardtodescribetheresultintermsofmachinevaluesratherthanmodelvalues?2.Iftheanswerto1is"yes",howareprogrammersexpectedtousethisintrinsicfunctionportably?ANSWER:1.  Yes.Processorsmaysupportvaluesthatarenotpresentinthemodelof13.7.1.IEEE_inf_Ԁisanexampleof be' suchanumberandthisshouldbereturnedonamachinethatsupportstheIEEEstandard.Ifthenegativenumberofthelargest_magnitude_Ԁinthemodelhadbeenintended,themodelwouldhavebeenmentionedas,forexample,inthedefinitionofHUGE(13.14.39).2.  AsimpleexampleofitsuseistotestwhetherasetofnumbersSET1hasavaluegreaterthananyvaluein x!{, thesetSET2.Considertheexpression_MAXVAL_Ԁ(SET1)>_MAXVAL_Ԁ(SET2).IfSET1isemptyandSET2isnot,thisvalueiscorrectlyfalseevenifallofthevaluesareoutsidethemodelwithvalueslessthanHUGE(SET1).  Itmaybehelpfultoconsiderhow_MAXVAL_ԀmightbecodedforanarrayofrankoneonanIEEEcomputer.Thefollowingcodeissuitable.0  _MAXVAL_Ԁ=_IEEE_VALUE_Ԁ(1.0,_IEEE_NEGATIVE_INF_) (#(# 0  DOI=1,SIZE(ARRAY) (#(# 0  0(#(#_MAXVAL_Ԁ=MAX(_MAXVAL_,ARRAY(I)) (#(# 0  ENDDO (#(#   Allofthereductionfunctions(e.g.,_