From owner-hpff-distribute Tue Jan 2 09:46:33 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id JAA14588 for hpff-distribute-out; Tue, 2 Jan 1996 09:46:33 -0600 (CST) Received: from fontainebleau.ensmp.fr (root@fontainebleau.ensmp.fr [192.54.148.100]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id JAA14582 for ; Tue, 2 Jan 1996 09:46:24 -0600 (CST) Received: from cri.ensmp.fr (chailly.ensmp.fr) by fontainebleau.ensmp.fr with SMTP id AA07957 (5.67a8/IDA-1.5 for ); Tue, 2 Jan 1996 16:46:15 +0100 Received: from provins.caii by cri.ensmp.fr (4.1/SMI-4.0) id AA10263; Tue, 2 Jan 96 16:46:14 +0100 Date: Tue, 2 Jan 96 16:46:14 +0100 Message-Id: <9601021546.AA10263@cri.ensmp.fr> Received: by provins.caii (4.1/SMI-4.1) id AA03312; Tue, 2 Jan 96 16:46:04 +0100 From: Fabien COELHO To: schreiber@hpl.hp.com Subject: hpff-distribute: HPF reduce directive Cc: hpff-distribute@cs.rice.edu Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Hi Rob, I'm not sure you're in charge of the reduce directive, but anyway, I guess someone on the "hpff-distribute" list is: I just had a look the "reduce" directive proposal. I was wondering how the compiler should guess at which level in a loop nest the reduction must be performed. My point is that the way the "reduce" directive is placed in the loop nest does not give this information to the compiler, thus some analysis will be needed. This analysis could be avoided with some other syntax in the "new" and "independent" spirit. EXAMPLE: -------- Here is an example to illustrate this point. (This example is not designed to be meaningful from the computation point of view, but to provide a set of problems in the same loop nest.) The reduction on s1 must be performed at the j level, the ones on s2 and s3 at the i level. The s3 reduction has two contributions from B and A. The reduction on s1 is initialized with B. It is used for the p reduction. real A(n,n), B(n) ! there may be some external loop around this piece of code ... p = 1. s2 = 0. s3 = 0. do i=1, n B(i) = 2.*real(i-j) s1 = B(i) s3 = s3 + n*B(i) do j=1, n A(j,i) = 1./real(i+j) s1 = s1 + A(j,i) s2 = s2 + A(j,i) s3 = s3 + A(j,i) enddo p = p * s1 enddo ... ! p, s2 and s3 might be used... CURRENT "INTERNAL" REDUCE SYNTAX: --------------------------------- With the current syntax, it's quite verbose and the compiler must decide at with level the reduction must be performed. Another issue is whether or not several contributions are allowed: p = 1. s2 = 0. s3 = 0. chpf$ independent(i), new(s1, j) do i=1, n B(i) = 2.*real(i-j) s1 = B(i) ! initialization of s1 chpf$ reduce ! first contribution to s3... s3 = s3 + n*B(i) chpf$ independent(j) do j=1, n A(j,i) = 1./real(i+j) chpf$ reduce ! must guess that it is at the j level ! what are the needed analysis for this? ! should it trust the "new" outside? ! "see" that s1 is "simply" (as opposed ! to s3) assigned and used in the i loop? ! note that both s1 and s3 are read and ! written within the i loop. s1 = s1 + A(j,i) chpf$ reduce ! i level (2 loops outward) s2 = s2 + A(j,i) chpf$ reduce ! i level, second contribution to s3... s3 = s3 + A(j,i) enddo chpf$ reduce ! i level p = p * s1 enddo "EXTERNAL" REDUCE SYNTAX: ------------------------- I would rather suggest the less verbose (and nothing/few to guess): p = 1. s2 = 0. s3 = 0. chpf$ independent(i), new(j,s1), reduce(s2,s3,p) do i=1, n B(i) = 2.*real(i-j) s1 = B(i) s3 = s3 + n*B(i) chpf$ independent(j), reduce(s1[,s2,s3]) ! [] optional? do j=1, n A(j,i) = 1./real(i+j) s1 = s1 + A(j,i) s2 = s2 + A(j,i) s3 = s3 + A(j,i) enddo p = p * s1 enddo Apart from the reduced (!) verbosity, the very fact that the reduce directive is attached to a loop gives the compiler a hint about the level at which the reduction (=> [partial] synchronization) is to be performed. For checking purposes every scalar assigned in an independent loop should appear in some "new" or "reduce" directive... Also the multi-contribution to a reduction does not require multiple annotations from the user (rising the question of the semantics of the operation if some are forgotten...) The compiler has to guess what are the reduction statement(s) for each scalar, that is it has to find all the assignments to that scalar within the loop nest. The spirit of such an "external" reduce directive is quite different from the "internal" one. It asserts at a higher level in the code a general property about a loop nest and a scalar manipulated in this loop nest. This is homogeneous with the existing "independent" and "new" directives, ans thus improves the readability of the code. NOTES: ------ As far as the forall is concerned, since it may not be independent, the "external" reduce should exist without attachment: chpf$ reduce(s) forall(i=2:n-1) s = s + a(i) a(i) = a(i)+a(i-1)+a(i+1) end forall Also it is not very clear to me what to do with reductions into arrays. (I'm not sure from the minutes whether this is allowed or not)... chpf$ independent(i,j), reduce(S(i)) ??? do i=1, n do j=1, n S(i) = S(i) + A(k,j) enddo enddo means that an "internal" directive might be of some help? CONCLUSION: ----------- To conclude, I suggest that another "external" (as opposed to the current "internal") syntax for the reduce directive would (1) leave less work to the compiler about "guessing" things, (2) be more high level and readable (3) be less verbose. Hope this help. Thanks for having read this mail. Comments are welcome. Have a nice day, Fabien. -- Fabien COELHO __ http://www.cri.ensmp.fr/~coelho __ coelho@cri.ensmp.fr CRI, ENSMP, 35, rue Saint Honoré, 77305 Fontainebleau Cedex, France phone: 33 1 64 69 {voice: 48 52, fax: 47 09, standard: 47 08} ________ All opinions expressed here are mine ________ --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Tue Jan 2 10:57:27 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id KAA16230 for hpff-distribute-out; Tue, 2 Jan 1996 10:57:27 -0600 (CST) Received: from fontainebleau.ensmp.fr (root@fontainebleau.ensmp.fr [192.54.148.100]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id KAA16215; Tue, 2 Jan 1996 10:57:03 -0600 (CST) Received: from cri.ensmp.fr (chailly.ensmp.fr) by fontainebleau.ensmp.fr with SMTP id AA08477 (5.67a8/IDA-1.5); Tue, 2 Jan 1996 17:57:00 +0100 Received: from provins.caii by cri.ensmp.fr (4.1/SMI-4.0) id AA12171; Tue, 2 Jan 96 17:56:59 +0100 Date: Tue, 2 Jan 96 17:56:59 +0100 Message-Id: <9601021656.AA12171@cri.ensmp.fr> Received: by provins.caii (4.1/SMI-4.1) id AA03405; Tue, 2 Jan 96 17:56:58 +0100 From: Fabien COELHO To: chk@cs.rice.edu Cc: hpff-distribute@cs.rice.edu In-Reply-To: message from Fabien COELHO on Tue, 2 Jan 96 16:46:14 +0100 Subject: hpff-distribute: not atached "new" directive Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Hi Chuck, > chpf$ reduce(s) > forall(i=2:n-1) > s = s + a(i) > a(i) = a(i)+a(i-1)+a(i+1) > end forall The example I just posted on the net reminds me my (really real) application with "private" scalars that required a "new" clause outside a loop. Forall constructs may also use scalars to avoid redundant computations: chpf$ new(x) ! apply on the next loop, here the forall... forall(i=2:n) x = a(i-1)+2*a(i)+a(i+1) ! x used to store an intermediate value. b(i) = f1(x) ! some pure function. a(i) = f2(x) ! idem. end forall Here the loop is NOT independent and x must be private to each iteration for parallelism. I guess the compiler must expand the scalar into an array for such a loop. I think that this kind of loop should be allowed by HPF, and that it is not with the current definition of forall/new. The other way around is to assume that a scalar in a forall *must* be private (?) or to perform some analysis to prove it, but I understood that one of the goal of the HPF Forum is that no advanced analysis technique is needed by compilers for handling HPF. Hope this help. Have a nice day, Fabien. -- Fabien COELHO __ http://www.cri.ensmp.fr/~coelho __ coelho@cri.ensmp.fr CRI, ENSMP, 35, rue Saint Honoré, 77305 Fontainebleau Cedex, France phone: 33 1 64 69 {voice: 48 52, fax: 47 09, standard: 47 08} ________ All opinions expressed here are mine ________ --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Tue Jan 2 11:18:15 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id LAA16972 for hpff-distribute-out; Tue, 2 Jan 1996 11:18:15 -0600 (CST) Received: from hplms26.hpl.hp.com (hplms26.hpl.hp.com [15.255.168.31]) by cs.rice.edu (8.7.1/8.7.1) with ESMTP id LAA16967 for ; Tue, 2 Jan 1996 11:18:12 -0600 (CST) Received: from hplpp3.hpl.hp.com by hplms26.hpl.hp.com with ESMTP ($Revision: 1.36.108.11 $/15.5+ECS 3.3+HPL1.1S) id AA230183003; Tue, 2 Jan 1996 09:16:43 -0800 Received: by hplpp3.hpl.hp.com (1.37.109.14/15.5+ECS 3.3+HPL1.1) id AA197913000; Tue, 2 Jan 1996 09:16:40 -0800 Date: Tue, 2 Jan 1996 09:16:40 -0800 From: Rob Schreiber Message-Id: <199601021716.AA197913000@hplpp3.hpl.hp.com> To: coelho@chailly.ensmp.fr, schreiber@hplms26.hpl.hp.com Subject: hpff-distribute: Re: HPF reduce directive Cc: hpff-distribute@cs.rice.edu Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Dear Fabien, I know this is a subtle issue. I thought we had unabiguously defined the answer to the "shich loop" question as the outermost DO loop that: i) Is independent ii) Does not have a NEW clause for the reduction variable, and iii) Does not *contain* in its range an independent DO loop with a NEW clause for the reduction variable. Also, the reduction statement must be in the lexical range of the DO loop. Thus, we have addressed the issues of whether more than one reduction statment may occur (there is no restriction on the number or the place) and which loops are the scope of the reduction. Nevertheless, your points about economy of expression are very well taken. We should discuss this alternative. Rob From coelho@chailly.ensmp.fr Tue Jan 2 07:49 PST 1996 Hi Rob, I'm not sure you're in charge of the reduce directive, but anyway, I guess someone on the "hpff-distribute" list is: I just had a look the "reduce" directive proposal. I was wondering how the compiler should guess at which level in a loop nest the reduction must be performed. My point is that the way the "reduce" directive is placed in the loop nest does not give this information to the compiler, thus some analysis will be needed. This analysis could be avoided with some other syntax in the "new" and "independent" spirit. EXAMPLE: -------- Here is an example to illustrate this point. (This example is not designed to be meaningful from the computation point of view, but to provide a set of problems in the same loop nest.) The reduction on s1 must be performed at the j level, the ones on s2 and s3 at the i level. The s3 reduction has two contributions from B and A. The reduction on s1 is initialized with B. It is used for the p reduction. real A(n,n), B(n) ! there may be some external loop around this piece of code ... p = 1. s2 = 0. s3 = 0. do i=1, n B(i) = 2.*real(i-j) s1 = B(i) s3 = s3 + n*B(i) do j=1, n A(j,i) = 1./real(i+j) s1 = s1 + A(j,i) s2 = s2 + A(j,i) s3 = s3 + A(j,i) enddo p = p * s1 enddo ... ! p, s2 and s3 might be used... CURRENT "INTERNAL" REDUCE SYNTAX: --------------------------------- With the current syntax, it's quite verbose and the compiler must decide at with level the reduction must be performed. Another issue is whether or not several contributions are allowed: p = 1. s2 = 0. s3 = 0. chpf$ independent(i), new(s1, j) do i=1, n B(i) = 2.*real(i-j) s1 = B(i) ! initialization of s1 chpf$ reduce ! first contribution to s3... s3 = s3 + n*B(i) chpf$ independent(j) do j=1, n A(j,i) = 1./real(i+j) chpf$ reduce ! must guess that it is at the j level ! what are the needed analysis for this? ! should it trust the "new" outside? ! "see" that s1 is "simply" (as opposed ! to s3) assigned and used in the i loop? ! note that both s1 and s3 are read and ! written within the i loop. s1 = s1 + A(j,i) chpf$ reduce ! i level (2 loops outward) s2 = s2 + A(j,i) chpf$ reduce ! i level, second contribution to s3... s3 = s3 + A(j,i) enddo chpf$ reduce ! i level p = p * s1 enddo "EXTERNAL" REDUCE SYNTAX: ------------------------- I would rather suggest the less verbose (and nothing/few to guess): p = 1. s2 = 0. s3 = 0. chpf$ independent(i), new(j,s1), reduce(s2,s3,p) do i=1, n B(i) = 2.*real(i-j) s1 = B(i) s3 = s3 + n*B(i) chpf$ independent(j), reduce(s1[,s2,s3]) ! [] optional? do j=1, n A(j,i) = 1./real(i+j) s1 = s1 + A(j,i) s2 = s2 + A(j,i) s3 = s3 + A(j,i) enddo p = p * s1 enddo Apart from the reduced (!) verbosity, the very fact that the reduce directive is attached to a loop gives the compiler a hint about the level at which the reduction (=> [partial] synchronization) is to be performed. For checking purposes every scalar assigned in an independent loop should appear in some "new" or "reduce" directive... Also the multi-contribution to a reduction does not require multiple annotations from the user (rising the question of the semantics of the operation if some are forgotten...) The compiler has to guess what are the reduction statement(s) for each scalar, that is it has to find all the assignments to that scalar within the loop nest. The spirit of such an "external" reduce directive is quite different from the "internal" one. It asserts at a higher level in the code a general property about a loop nest and a scalar manipulated in this loop nest. This is homogeneous with the existing "independent" and "new" directives, ans thus improves the readability of the code. NOTES: ------ As far as the forall is concerned, since it may not be independent, the "external" reduce should exist without attachment: chpf$ reduce(s) forall(i=2:n-1) s = s + a(i) a(i) = a(i)+a(i-1)+a(i+1) end forall Also it is not very clear to me what to do with reductions into arrays. (I'm not sure from the minutes whether this is allowed or not)... chpf$ independent(i,j), reduce(S(i)) ??? do i=1, n do j=1, n S(i) = S(i) + A(k,j) enddo enddo means that an "internal" directive might be of some help? No, I think one could require that ALL of S be treated as a redction variable; so the list in reduce(something-list) would be a list of names, not subscripted instances of names (whatever the F90 syntactic class is). CONCLUSION: ----------- To conclude, I suggest that another "external" (as opposed to the current "internal") syntax for the reduce directive would (1) leave less work to the compiler about "guessing" things, (2) be more high level and readable (3) be less verbose. Hope this help. Thanks for having read this mail. Comments are welcome. Have a nice day, Fabien. --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Tue Jan 2 12:44:03 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id MAA20317 for hpff-distribute-out; Tue, 2 Jan 1996 12:44:03 -0600 (CST) Received: from fontainebleau.ensmp.fr (root@fontainebleau.ensmp.fr [192.54.148.100]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id MAA20300; Tue, 2 Jan 1996 12:43:54 -0600 (CST) Received: from cri.ensmp.fr (chailly.ensmp.fr) by fontainebleau.ensmp.fr with SMTP id AA09225 (5.67a8/IDA-1.5); Tue, 2 Jan 1996 19:43:43 +0100 Received: from provins.caii by cri.ensmp.fr (4.1/SMI-4.0) id AA13899; Tue, 2 Jan 96 19:43:42 +0100 Date: Tue, 2 Jan 96 19:43:42 +0100 Message-Id: <9601021843.AA13899@cri.ensmp.fr> Received: by provins.caii (4.1/SMI-4.1) id AA03502; Tue, 2 Jan 96 19:43:41 +0100 From: Fabien COELHO To: chk@cs.rice.edu In-Reply-To: (chk@cs.rice.edu) Subject: hpff-distribute: Re: HPF reduce directive Cc: hpff-distribute@cs.rice.edu Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Hi Chuck, > Actually, the mailing list that is discussing REDUCE is hpff-task. Sorry for the mis-posting, and thanks for the pointer. About independent: > I'll comment that I generally like the "external" syntax, modulo a > couple syntax nits (for example, INDEPEDENT does not name the loop index, Oops... Your're right (for sure:-). I noticed the syntax from early versions of the HPF document (0.2 for instance) and never notice ever since that it was dropped out! I parse it with my compiler (as well as the implicite one). I found it quite convenient for loop nests (the inner loop indexes being implicitely private...) chpf$ independent(i,j) do i = ... do j = ... ... enddo enddo instead of chpf$ independent, new(j) do i = ... chpf$ independent do j = ... ... enddo enddo But this is just a matter of syntactic sugar. And verbosity. > and doesn't apply to nested loops.) Do you mean that my inner loop cannot be tagged as independent? Because it is not perfectly nested? I have not seen such a restriction in HPF 1.0 (or it is not listed in the constraint list?). I cannot see any rational for it. The compiler is not required to execute the loop in parallel, it's just a matter of asserting some semantical property of a piece of code, in my understanding. BTW, Happy new year! Fabien. -- Fabien COELHO __ http://www.cri.ensmp.fr/~coelho __ coelho@cri.ensmp.fr CRI, ENSMP, 35, rue Saint Honoré, 77305 Fontainebleau Cedex, France phone: 33 1 64 69 {voice: 48 52, fax: 47 09, standard: 47 08} ________ All opinions expressed here are mine ________ --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Wed Jan 3 10:12:59 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id KAA12195 for hpff-distribute-out; Wed, 3 Jan 1996 10:12:59 -0600 (CST) Date: Wed, 3 Jan 1996 10:12:59 -0600 (CST) Message-Id: <199601031612.KAA12195@cs.rice.edu> From: Rob Schreiber Subject: hpff-distribute: New inquiry procedures Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Draft of mapping inquiry procedures revised to allow new (general block, indirect, and ONTO sections of processors arrangements) mapping possibilities. Rob %!PS-Adobe-2.0 %%Creator: dvipsk 5.58a Copyright 1986, 1994 Radical Eye Software %%Title: temp.dvi %%Pages: 12 %%PageOrder: Ascend %%BoundingBox: 0 0 612 792 %%EndComments %DVIPSCommandLine: dvips temp -o temp.ps %DVIPSParameters: dpi=600, compressed, comments removed %DVIPSSource: TeX output 1996.01.02:1513 %%BeginProcSet: texc.pro /TeXDict 250 dict def TeXDict begin /N{def}def /B{bind def}N /S{exch}N /X{S N}B /TR{translate}N /isls false N /vsize 11 72 mul N /hsize 8.5 72 mul N /landplus90{false}def /@rigin{isls{[0 landplus90{1 -1}{-1 1} ifelse 0 0 0]concat}if 72 Resolution div 72 VResolution div neg scale isls{landplus90{VResolution 72 div vsize mul 0 exch}{Resolution -72 div hsize mul 0}ifelse TR}if Resolution VResolution vsize -72 div 1 add mul TR[matrix currentmatrix{dup dup round sub abs 0.00001 lt{round}if} forall round exch round exch]setmatrix}N /@landscape{/isls true N}B /@manualfeed{statusdict /manualfeed true put}B /@copies{/#copies X}B /FMat[1 0 0 -1 0 0]N /FBB[0 0 0 0]N /nn 0 N /IE 0 N /ctr 0 N /df-tail{ /nn 8 dict N nn begin /FontType 3 N /FontMatrix fntrx N /FontBBox FBB N string /base X array /BitMaps X /BuildChar{CharBuilder}N /Encoding IE N end dup{/foo setfont}2 array copy cvx N load 0 nn put /ctr 0 N[}B /df{ /sf 1 N /fntrx FMat N df-tail}B /dfs{div /sf X /fntrx[sf 0 0 sf neg 0 0] N df-tail}B /E{pop nn dup definefont setfont}B /ch-width{ch-data dup length 5 sub get}B /ch-height{ch-data dup length 4 sub get}B /ch-xoff{ 128 ch-data dup length 3 sub get sub}B /ch-yoff{ch-data dup length 2 sub get 127 sub}B /ch-dx{ch-data dup length 1 sub get}B /ch-image{ch-data dup type /stringtype ne{ctr get /ctr ctr 1 add N}if}B /id 0 N /rw 0 N /rc 0 N /gp 0 N /cp 0 N /G 0 N /sf 0 N /CharBuilder{save 3 1 roll S dup /base get 2 index get S /BitMaps get S get /ch-data X pop /ctr 0 N ch-dx 0 ch-xoff ch-yoff ch-height sub ch-xoff ch-width add ch-yoff setcachedevice ch-width ch-height true[1 0 0 -1 -.1 ch-xoff sub ch-yoff .1 sub]/id ch-image N /rw ch-width 7 add 8 idiv string N /rc 0 N /gp 0 N /cp 0 N{rc 0 ne{rc 1 sub /rc X rw}{G}ifelse}imagemask restore}B /G{{id gp get /gp gp 1 add N dup 18 mod S 18 idiv pl S get exec}loop}B /adv{cp add /cp X}B /chg{rw cp id gp 4 index getinterval putinterval dup gp add /gp X adv}B /nd{/cp 0 N rw exit}B /lsh{rw cp 2 copy get dup 0 eq{pop 1}{ dup 255 eq{pop 254}{dup dup add 255 and S 1 and or}ifelse}ifelse put 1 adv}B /rsh{rw cp 2 copy get dup 0 eq{pop 128}{dup 255 eq{pop 127}{dup 2 idiv S 128 and or}ifelse}ifelse put 1 adv}B /clr{rw cp 2 index string putinterval adv}B /set{rw cp fillstr 0 4 index getinterval putinterval adv}B /fillstr 18 string 0 1 17{2 copy 255 put pop}for N /pl[{adv 1 chg} {adv 1 chg nd}{1 add chg}{1 add chg nd}{adv lsh}{adv lsh nd}{adv rsh}{ adv rsh nd}{1 add adv}{/rc X nd}{1 add set}{1 add clr}{adv 2 chg}{adv 2 chg nd}{pop nd}]dup{bind pop}forall N /D{/cc X dup type /stringtype ne{] }if nn /base get cc ctr put nn /BitMaps get S ctr S sf 1 ne{dup dup length 1 sub dup 2 index S get sf div put}if put /ctr ctr 1 add N}B /I{ cc 1 add D}B /bop{userdict /bop-hook known{bop-hook}if /SI save N @rigin 0 0 moveto /V matrix currentmatrix dup 1 get dup mul exch 0 get dup mul add .99 lt{/QV}{/RV}ifelse load def pop pop}N /eop{SI restore userdict /eop-hook known{eop-hook}if showpage}N /@start{userdict /start-hook known{start-hook}if pop /VResolution X /Resolution X 1000 div /DVImag X /IE 256 array N 0 1 255{IE S 1 string dup 0 3 index put cvn put}for 65781.76 div /vsize X 65781.76 div /hsize X}N /p{show}N /RMat[1 0 0 -1 0 0]N /BDot 260 string N /rulex 0 N /ruley 0 N /v{/ruley X /rulex X V}B /V {}B /RV statusdict begin /product where{pop product dup length 7 ge{0 7 getinterval dup(Display)eq exch 0 4 getinterval(NeXT)eq or}{pop false} ifelse}{false}ifelse end{{gsave TR -.1 .1 TR 1 1 scale rulex ruley false RMat{BDot}imagemask grestore}}{{gsave TR -.1 .1 TR rulex ruley scale 1 1 false RMat{BDot}imagemask grestore}}ifelse B /QV{gsave newpath transform round exch round exch itransform moveto rulex 0 rlineto 0 ruley neg rlineto rulex neg 0 rlineto fill grestore}B /a{moveto}B /delta 0 N /tail {dup /delta X 0 rmoveto}B /M{S p delta add tail}B /b{S p tail}B /c{-4 M} B /d{-3 M}B /e{-2 M}B /f{-1 M}B /g{0 M}B /h{1 M}B /i{2 M}B /j{3 M}B /k{ 4 M}B /w{0 rmoveto}B /l{p -4 w}B /m{p -3 w}B /n{p -2 w}B /o{p -1 w}B /q{ p 1 w}B /r{p 2 w}B /s{p 3 w}B /t{p 4 w}B /x{0 S rmoveto}B /y{3 2 roll p a}B /bos{/SS save N}B /eos{SS restore}B end %%EndProcSet %%BeginProcSet: special.pro TeXDict begin /SDict 200 dict N SDict begin /@SpecialDefaults{/hs 612 N /vs 792 N /ho 0 N /vo 0 N /hsc 1 N /vsc 1 N /ang 0 N /CLIP 0 N /rwiSeen false N /rhiSeen false N /letter{}N /note{}N /a4{}N /legal{}N}B /@scaleunit 100 N /@hscale{@scaleunit div /hsc X}B /@vscale{@scaleunit div /vsc X}B /@hsize{/hs X /CLIP 1 N}B /@vsize{/vs X /CLIP 1 N}B /@clip{ /CLIP 2 N}B /@hoffset{/ho X}B /@voffset{/vo X}B /@angle{/ang X}B /@rwi{ 10 div /rwi X /rwiSeen true N}B /@rhi{10 div /rhi X /rhiSeen true N}B /@llx{/llx X}B /@lly{/lly X}B /@urx{/urx X}B /@ury{/ury X}B /magscale true def end /@MacSetUp{userdict /md known{userdict /md get type /dicttype eq{userdict begin md length 10 add md maxlength ge{/md md dup length 20 add dict copy def}if end md begin /letter{}N /note{}N /legal{} N /od{txpose 1 0 mtx defaultmatrix dtransform S atan/pa X newpath clippath mark{transform{itransform moveto}}{transform{itransform lineto} }{6 -2 roll transform 6 -2 roll transform 6 -2 roll transform{ itransform 6 2 roll itransform 6 2 roll itransform 6 2 roll curveto}}{{ closepath}}pathforall newpath counttomark array astore /gc xdf pop ct 39 0 put 10 fz 0 fs 2 F/|______Courier fnt invertflag{PaintBlack}if}N /txpose{pxs pys scale ppr aload pop por{noflips{pop S neg S TR pop 1 -1 scale}if xflip yflip and{pop S neg S TR 180 rotate 1 -1 scale ppr 3 get ppr 1 get neg sub neg ppr 2 get ppr 0 get neg sub neg TR}if xflip yflip not and{pop S neg S TR pop 180 rotate ppr 3 get ppr 1 get neg sub neg 0 TR}if yflip xflip not and{ppr 1 get neg ppr 0 get neg TR}if}{noflips{TR pop pop 270 rotate 1 -1 scale}if xflip yflip and{TR pop pop 90 rotate 1 -1 scale ppr 3 get ppr 1 get neg sub neg ppr 2 get ppr 0 get neg sub neg TR}if xflip yflip not and{TR pop pop 90 rotate ppr 3 get ppr 1 get neg sub neg 0 TR}if yflip xflip not and{TR pop pop 270 rotate ppr 2 get ppr 0 get neg sub neg 0 S TR}if}ifelse scaleby96{ppr aload pop 4 -1 roll add 2 div 3 1 roll add 2 div 2 copy TR .96 dup scale neg S neg S TR}if}N /cp {pop pop showpage pm restore}N end}if}if}N /normalscale{Resolution 72 div VResolution 72 div neg scale magscale{DVImag dup scale}if 0 setgray} N /psfts{S 65781.76 div N}N /startTexFig{/psf$SavedState save N userdict maxlength dict begin /magscale true def normalscale currentpoint TR /psf$ury psfts /psf$urx psfts /psf$lly psfts /psf$llx psfts /psf$y psfts /psf$x psfts currentpoint /psf$cy X /psf$cx X /psf$sx psf$x psf$urx psf$llx sub div N /psf$sy psf$y psf$ury psf$lly sub div N psf$sx psf$sy scale psf$cx psf$sx div psf$llx sub psf$cy psf$sy div psf$ury sub TR /showpage{}N /erasepage{}N /copypage{}N /p 3 def @MacSetUp}N /doclip{ psf$llx psf$lly psf$urx psf$ury currentpoint 6 2 roll newpath 4 copy 4 2 roll moveto 6 -1 roll S lineto S lineto S lineto closepath clip newpath moveto}N /endTexFig{end psf$SavedState restore}N /@beginspecial{SDict begin /SpecialSave save N gsave normalscale currentpoint TR @SpecialDefaults count /ocount X /dcount countdictstack N}N /@setspecial {CLIP 1 eq{newpath 0 0 moveto hs 0 rlineto 0 vs rlineto hs neg 0 rlineto closepath clip}if ho vo TR hsc vsc scale ang rotate rwiSeen{rwi urx llx sub div rhiSeen{rhi ury lly sub div}{dup}ifelse scale llx neg lly neg TR }{rhiSeen{rhi ury lly sub div dup scale llx neg lly neg TR}if}ifelse CLIP 2 eq{newpath llx lly moveto urx lly lineto urx ury lineto llx ury lineto closepath clip}if /showpage{}N /erasepage{}N /copypage{}N newpath }N /@endspecial{count ocount sub{pop}repeat countdictstack dcount sub{ end}repeat grestore SpecialSave restore end}N /@defspecial{SDict begin} N /@fedspecial{end}B /li{lineto}B /rl{rlineto}B /rc{rcurveto}B /np{ /SaveX currentpoint /SaveY X N 1 setlinecap newpath}N /st{stroke SaveX SaveY moveto}N /fil{fill SaveX SaveY moveto}N /ellipse{/endangle X /startangle X /yrad X /xrad X /savematrix matrix currentmatrix N TR xrad yrad scale 0 0 1 startangle endangle arc savematrix setmatrix}N end %%EndProcSet TeXDict begin 40258431 52099146 1000 600 600 (temp.dvi) @start /Fa 43 118 df<121C127FEAFF80A213C0A3127F121C1200A412011380A21203 13005A1206120E5A5A5A12600A1979B917>39 D<121C127FEAFF80A213C0A3127F121C12 00A412011380A2120313005A1206120E5A5A5A12600A19798817>44 DI<150C151E153EA2153C157CA2157815F8A215F01401A215E0 1403A215C01407A21580140FA215005CA2141E143EA2143C147CA2147814F8A25C1301A2 5C1303A2495AA25C130FA291C7FC5BA2131E133EA2133C137CA2137813F8A25B1201A25B 1203A25B1207A25B120FA290C8FC5AA2121E123EA2123C127CA2127812F8A25A12601F53 7BBD2A>47 DIIII<15 38A2157815F8A2140114031407A2140F141F141B14331473146314C313011483EB030313 071306130C131C131813301370136013C01201EA038013005A120E120C5A123812305A12 E0B712F8A3C73803F800AB4A7E0103B512F8A325397EB82A>I<0006140CD80780133C90 38F003F890B5FC5D5D158092C7FC14FC38067FE090C9FCABEB07F8EB3FFE9038780F8039 07E007E090388003F0496C7E12066E7EC87EA28181A21680A4123E127F487EA490C71300 485C12E000605C12700030495A00385C6C1303001E495A6C6C485A3907E03F800001B5C7 FC38007FFCEB1FE0213A7CB72A>II56 D<1538A3157CA315FEA34A7EA34A6C7EA202077FEC063FA2020E7FEC0C1FA2 021C7FEC180FA202387FEC3007A202707FEC6003A202C07F1501A2D901807F81A249C77F 167FA20106810107B6FCA24981010CC7121FA2496E7EA3496E7EA3496E7EA213E0707E12 01486C81D80FFC02071380B56C90B512FEA3373C7DBB3E>65 DI<91 3A01FF800180020FEBE003027F13F8903A01FF807E07903A03FC000F0FD90FF0EB039F49 48EB01DFD93F80EB00FF49C8127F01FE153F12014848151F4848150FA248481507A2485A 1703123F5B007F1601A35B00FF93C7FCAD127F6DED0180A3123F7F001F160318006C7E5F 6C7E17066C6C150E6C6C5D00001618017F15386D6C5CD91FE05C6D6CEB03C0D903FCEB0F 80902701FF803FC7FC9039007FFFFC020F13F002011380313D7BBA3C>IIIII II75 DIIIII82 D I<003FB812E0A3D9C003EB001F273E0001FE130348EE01F00078160000701770A3006017 30A400E01738481718A4C71600B3B0913807FF80011FB612E0A335397DB83C>II<007FB590383FFFFCA3C601F801071380D97F E0D903FCC7FC013FEC01F06D6C5C5F6D6C5C6D6C13034CC8FC6D6C1306160E6D6C5B6DEB 8018163891387FC0306E6C5A16E06E6C5A91380FF18015FB6EB4C9FC5D14036E7EA26E7F 6F7EA24B7E15DF9138019FF09138038FF8150F91380607FC91380E03FE140C4A6C7EEC38 000230804A6D7E14E04A6D7E49486D7E130391C76C7E01066E7E130E010C6E7E011C1401 013C8101FE822607FF80010713E0B500E0013FEBFF80A339397EB83E>88 DI91 D93 D97 D101 D<147E903803FF8090380FC1E0EB1F8790383F0FF013 7EA213FCA23901F803C091C7FCADB512FCA3D801F8C7FCB3AB487E387FFFF8A31C3B7FBA 19>I108 D<3807E01F00FFEB7FC09038E1E3E09038E387F0380FE707EA03E613EE9038EC03E09038 FC0080491300A45BB3A2487EB512F0A31C257EA421>114 DI<1318A51338A31378A313F812011203 1207001FB5FCB6FCA2D801F8C7FCB215C0A93800FC011580EB7C03017E13006D5AEB0FFE EB01F81A347FB220>II E /Fb 2 117 df104 D<1360A413E0A312011203A21207121FB512F0A23803E000AF1418A714383801F0301470 3800F860EB3FE0EB0F80152A7FA81B>116 D E /Fc 9 117 df<387FFFFEA3B5FCA21705 799521>45 D<147E49B47E903907C1C38090391F80EFC090383F00FF017E137F49148048 48133F485AA248481400120F5B001F5C157E485AA215FE007F5C90C7FCA21401485C5AA2 1403EDF0385AA21407EDE078020F1370127C021F13F0007E013F13E0003E137FECF3E126 1F01E313C03A0F8781E3803A03FF00FF00D800FC133E252977A72E>97 D101 D103 D<1478EB01FCA21303A314F8EB00E01400AD137C48B4FC38038F80EA0707000E13C0121E 121CEA3C0F1238A2EA781F00701380A2EAF03F140012005B137E13FE5BA212015BA21203 5B1438120713E0000F1378EBC070A214F0EB80E0A2EB81C01383148038078700EA03FEEA 00F8163E79BC1C>105 D108 D110 D 114 D116 D E /Fd 52 122 df<121C127FEAFF80B3EA7F00B2123EC7FCA8121C 127FA2EAFF80A3EA7F00A2121C09396DB830>33 D<1438147C14FCA4EB03FF011F13E090 B512FC4880000780481580261FFEFD13C09039F0FC3FE0D83FC0131FD87F80EB0FF00100 1307007E15F800FE14035A1507A36CEC03F0A2007F91C7FC138013C0EA3FF0EA1FFE13FF 6C13FF6C14E0000114F86C6C7F011F7F01037F0100148002FD13C09138FC7FE0151FED0F F015070018EC03F8127E1501B4FCA35AA26CEC03F07E01801307ED0FE0D83FC0131F01F0 EB7FC0D81FFEB512806CB612006C5C6C5CC614F0013F13C0D907FEC7FCEB00FCA5147C14 3825477BBE30>36 D39 D<141E147F14FF5BEB03FEEB07FCEB0FF0EB1FE0EB3FC0EB7F80EBFF00485A5B12035B48 5A120F5BA2485AA2123F5BA2127F90C7FCA412FEAD127FA47F123FA27F121FA26C7EA27F 12076C7E7F12017F6C7EEB7F80EB3FC0EB1FE0EB0FF0EB07FCEB03FEEB01FF7F147F141E 184771BE30>I<127812FE7E7F6C7E6C7EEA0FF06C7E6C7E6C7E6C7EEB7F80133F14C013 1FEB0FE014F01307A2EB03F8A214FC1301A214FE1300A4147FAD14FEA4130114FCA21303 14F8A2EB07F0A2130F14E0EB1FC0133F1480137FEBFF00485A485A485A485AEA3FE0485A 485A90C7FC5A1278184778BE30>I<14E0497E497EA60038EC0380007EEC0FC0D8FF83EB 3FE001C3137F9038F3F9FF267FFBFB13C06CB61280000FECFE00000314F86C5C6C6C13C0 011F90C7FC017F13C048B512F04880000F14FE003FECFF80267FFBFB13C026FFF3F913E0 9038C3F87F0183133FD87E03EB0FC00038EC0380000091C7FCA66D5A6D5A23277AAE30> I<143EA2147FAF007FB7FCA2B81280A36C1600A2C76CC8FCAF143EA229297DAF30>II<007FB612F0A2B712F8A36C15 F0A225077B9E30>I<120FEA3FC0EA7FE0A2EAFFF0A4EA7FE0A2EA3FC0EA0F000C0C6E8B 30>I<14FE903807FFC0497F013F13F8497F90B57E48EB83FF4848C6138049137F4848EB 3FC04848EB1FE049130F001F15F0491307A24848EB03F8A290C712014815FCA400FEEC00 FEAD6C14016C15FCA36D1303003F15F8A26D1307001F15F0A26D130F6C6CEB1FE0A26C6C EB3FC06C6CEB7F806D13FF2601FF8313006CEBFFFE6D5B6D5B010F13E06D5BD900FEC7FC 273A7CB830>48 DII II<000FB612804815 C05AA316800180C8FCAEEB83FF019F13C090B512F015FC8181D9FE0313809039F0007FC0 49133F0180EB1FE06CC7120F000E15F0C81207A216F81503A31218127EA2B4FC150716F0 48140F6C15E06C141F6DEB3FC06D137F3A3FE001FF80261FFC0F13006CB55A6C5C6C5C6C 14E06C6C1380D90FFCC7FC25397BB730>II<127CB712FC16FEA416FC48C7EA0FF8 16F0ED1FE0007CEC3FC0C8EA7F80EDFF00A24A5A4A5A5D14075D140F5D4A5AA24A5AA24A C7FCA25C5C13015CA213035CA213075CA4495AA6131F5CA96D5A6DC8FC273A7CB830>I< 49B4FC010F13E0013F13F890B57E4880488048010113803A0FFC007FC0D81FF0EB3FE048 48131F49EB0FF048481307A290C7EA03F85A4815FC1501A416FEA37E7E6D130315076C7E 6C6C130F6D133FD80FFC13FF6CB6FC7E6C14FE6C14F9013FEBE1FC010F13819038006001 1400ED03F8A2150716F0150F000F15E0486C131F486CEB3FC0157FEDFF804A1300EC07FE 391FF01FFC90B55A6C5C6C5C6C1480C649C7FCEB3FF0273A7CB830>57 D<120FEA3FC0EA7FE0A2EAFFF0A4EA7FE0A2EA3FC0EA0F00C7FCAF120FEA3FC0EA7FE0A2 EAFFF0A4EA7FE0A2EA3FC0EA0F000C276EA630>I<007FB7FCA2B81280A36C16006C5DCB FCA7003FB612FE4881B81280A36C1600A229157DA530>61 D<1278127EB4FC13C07FEA7F F813FEEA1FFF6C13C000037F6C13F86C6C7EEB1FFF6D7F010313E06D7F9038007FFC6E7E 91380FFF806E13C0020113F080ED3FF8151F153FEDFFF05C020713C04A138091383FFE00 4A5A903801FFF0495B010F13804990C7FCEB7FFC48485A4813E0000F5B4890C8FCEA7FFE 13F8EAFFE05B90C9FC127E1278252F7BB230>I<147F4A7EA2497FA4497F14F7A401077F 14E3A3010F7FA314C1A2011F7FA490383F80FEA590387F007FA4498049133F90B6FCA348 81A39038FC001F00038149130FA4000781491307A2D87FFFEB7FFFB56CB51280A46C496C 130029397DB830>65 D<007FB512F0B612FE6F7E82826C813A03F8001FF815076F7E1501 A26F7EA615015EA24B5A1507ED1FF0ED7FE090B65A5E4BC7FC6F7E16E0829039F8000FF8 ED03FC6F7E1500167FA3EE3F80A6167F1700A25E4B5A1503ED1FFC007FB6FCB75A5E16C0 5E6C02FCC7FC29387EB730>I<91387F803C903903FFF03E49EBFC7E011F13FE49EBFFFE 5B9038FFE07F48EB801F3903FE000F484813075B48481303A2484813015B123F491300A2 127F90C8FC167C16005A5AAC7E7EA2167C6D14FE123FA27F121F6D13016C6C14FCA26C6C EB03F86D13076C6CEB0FF03901FF801F6C9038E07FE06DB512C06D14806D1400010713FC 6D13F09038007FC0273A7CB830>I<003FB512E04814FCB67E6F7E6C816C813A03F8007F F0ED1FF8150F6F7E6F7E15016F7EA2EE7F80A2163F17C0161FA4EE0FE0AC161F17C0A316 3F1780A2167F17005E4B5A15034B5A150F4B5AED7FF0003FB65A485DB75A93C7FC6C14FC 6C14E02B387FB730>I<007FB7FCB81280A47ED803F8C7123FA8EE1F0093C7FCA4157C15 FEA490B5FCA6EBF800A4157C92C8FCA5EE07C0EE0FE0A9007FB7FCB8FCA46C16C02B387E B730>I<003FB712804816C0B8FCA27E7ED801FCC7121FA8EE0F8093C7FCA5153E157FA4 90B6FCA69038FC007FA4153E92C8FCAE383FFFF8487FB5FCA27E6C5B2A387EB730>I<02 FF13F00103EBC0F8010F13F1013F13FD4913FF90B6FC4813C1EC007F4848133F4848131F 49130F485A491307121F5B123F491303A2127F90C7FC6F5A92C8FC5A5AA892B5FC4A1480 5CA26C7F6C6D1400ED03F8A27F003F1407A27F121F6D130F120F7F6C6C131FA2D803FE13 3F6C6C137FECC1FF6C90B5FC7F6D13FB010F13F30103EBC1F0010090C8FC293A7DB830> I<3B3FFF800FFFE0486D4813F0B56C4813F8A26C496C13F06C496C13E0D803F8C7EAFE00 B290B6FCA601F8C7FCB3A23B3FFF800FFFE0486D4813F0B56C4813F8A26C496C13F06C49 6C13E02D387FB730>I<007FB6FCB71280A46C1500260007F0C7FCB3B3A8007FB6FCB712 80A46C1500213879B730>I<49B512F04914F85BA27F6D14F090C7EAFE00B3B3123C127E B4FCA24A5A1403EB8007397FF01FF86CB55A5D6C5C00075C000149C7FC38003FF025397A B730>II<383FFFF8487FB57EA26C5B6C5BD801FCC9FCB3B0EE0F80EE1FC0A900 3FB7FC5AB8FCA27E6C16802A387EB730>III<90383FFFE048B512FC000714FF481580 4815C04815E0EBF80001E0133FD87F80EB0FF0A290C71207A44815F8481403B3A96C1407 A26C15F0A36D130FA26D131F6C6CEB3FE001F813FF90B6FC6C15C06C15806C1500000114 FCD8003F13E0253A7BB830>I<007FB512F0B612FE6F7E16E0826C813903F8003FED0FFC ED03FE15016F7EA2821780163FA6167F17005EA24B5A1503ED0FFCED3FF890B6FC5E5E16 804BC7FC15F001F8C9FCB0387FFFC0B57EA46C5B29387EB730>I<003FB57E4814F0B612 FC15FF6C816C812603F8017F9138003FF0151F6F7E15071503821501A515035E1507150F 4B5A153F4AB45A90B65A5E93C7FC5D8182D9F8007FED3FE0151F150F821507A817F8EEF1 FCA53A3FFF8003FB4801C0EBFFF8B56C7E17F06C496C13E06C49EB7FC0C9EA1F002E397F B730>82 D<90390FF803C0D97FFF13E048B512C74814F74814FF5A381FF80F383FE00149 7E4848137F90C7123F5A48141FA2150FA37EED07C06C91C7FC7F7FEA3FF0EA1FFEEBFFF0 6C13FF6C14E0000114F86C80011F13FF01031480D9003F13C014019138007FE0151FED0F F0A2ED07F8A2007C140312FEA56C140716F07F6DEB0FE06D131F01F8EB3FC001FF13FF91 B51280160000FD5CD8FC7F13F8D8F81F5BD878011380253A7BB830>I<003FB712C04816 E0B8FCA43AFE003F800FA8007CED07C0C791C7FCB3B1011FB5FC4980A46D91C7FC2B387E B730>I<3B7FFFC007FFFCB56C4813FEA46C496C13FCD803F8C7EA3F80B3B16D147F0001 1600A36C6C14FE6D13016D5CEC800390393FE00FF890391FF83FF06DB55A6D5C6D5C6D91 C7FC9038007FFCEC1FF02F3980B730>I87 D<3A3FFF01FFF84801837F02C77FA202835B6C01015B3A01FC00 7F806D91C7FC00005C6D5BEB7F01EC81FCEB3F8314C3011F5B14E7010F5B14FF6D5BA26D 5BA26D5BA26D90C8FCA4497FA2497FA2815B81EB0FE781EB1FC381EB3F8181EB7F008149 7F49800001143F49800003141F49800007140FD87FFEEB7FFFB590B5128080A25C6C486D 130029387DB730>II<001FB612FC4815FE5AA490C7EA03FCED07F816F0150FED1F E016C0153FED7F80003E1500C85A4A5A5D14034A5A5D140F4A5A5D143F4A5A92C7FC5C49 5A5C1303495A5C130F495A5C133F495A91C8FC5B4848147C4914FE1203485A5B120F485A 5B123F485A90B6FCB7FCA46C15FC27387CB730>I<007FB612F0A2B712F8A36C15F0A225 077B7D30>95 D97 D<02FC137E3B7FC3FF01FF80D8FFEF0187 7F90B500CF7F15DF92B57E6C010F13872607FE07EB03F801FC13FE9039F803FC01A201F0 13F8A301E013F0B3A23C7FFE0FFF07FF80B548018F13C0A46C486C01071380322881A730 >109 D112 D114 D<3B3FFFC07FFF80486DB512C0B515E0A26C16C06C496C13803B01FC 0003F000A2000014076D5C137E150F017F5C7F151FD91F805BA214C0010F49C7FCA214E0 0107137EA2EB03F0157C15FCEB01F85DA2EB00F9ECFDF0147D147FA26E5AA36E5AA35DA2 143F92C8FCA25C147EA2000F13FE486C5AEA3FC1EBC3F81387EB8FF0EBFFE06C5B5C6C90 C9FC6C5AEA01F02B3C7EA630>121 D E /Fe 31 122 df44 D46 D<90381FFF8090B512F000 0314FC000F14FF261FF8071380263FC00113C0018014E0486C7E486C14F07FA56C5A4A13 E0EA3FC0260F000313C0C714804A13004A5A4A5A15F04A5A4A5A5D92C7FC14FE5C13015C A25CAA90C9FCA8EB03F0497E497E497E497EA66D5A6D5A6D5A6D5A24407ABF31>63 D<16FCA24B7EA24B7EA34B7FA24B7FA34B7FA24B7FA34B7F157C03FC7FEDF87FA2020180 EDF03F0203804B7E02078115C082020F814B7E021F811500824A81023E7F027E81027C7F A202FC814A147F49B77EA34982A2D907E0C7001F7F4A80010F835C83011F8391C87E4983 133E83017E83017C81B500FC91B612FCA5463F7CBE4F>65 D<922607FFC0130E92B500FC 131E020702FF133E023FEDC07E91B7EAE1FE01039138803FFB499039F80003FF4901C013 00013F90C8127F4948151FD9FFF8150F48491507485B4A1503481701485B18004890CAFC 197E5A5B193E127FA349170012FFAC127F7F193EA2123FA27F6C187E197C6C7F19FC6C6D 16F86C6D150119F06C6D15036C6DED07E0D97FFEED0FC06D6CED3F80010F01C0ECFF006D 01F8EB03FE6D9039FF801FFC010091B55A023F15E002071580020002FCC7FC030713C03F 407ABE4C>67 DII79 DI82 D<903A03FFC001C0011FEBF803017FEBFE 0748B6128F4815DF48010013FFD80FF8130F48481303497F4848EB007F127F49143F161F 12FF160FA27F1607A27F7F01FC91C7FCEBFF806C13F8ECFFC06C14FCEDFF806C15E016F8 6C816C816C816C16806C6C15C07F010715E0EB007F020714F0EC003F1503030013F8167F 163F127800F8151FA2160FA27EA217F07E161F6C16E06D143F01E015C001F8EC7F8001FE EB01FF9026FFE00713004890B55A486C14F8D8F81F5CD8F00314C027E0003FFEC7FC2D40 7ABE3A>I<003FB912FCA5903BFE003FFE003FD87FF0EE0FFE01C0160349160190C71500 197E127EA2007C183EA400FC183F48181FA5C81600B3AF010FB712F8A5403D7CBC49>I< 903807FFC0013F13F848B6FC48812607FE037F260FF8007F6DEB3FF0486C806F7EA36F7E A26C5A6C5AEA01E0C8FC153F91B5FC130F137F3901FFFE0F4813E0000F1380381FFE0048 5A5B485A12FF5BA4151F7F007F143F6D90387BFF806C6C01FB13FE391FFF07F36CEBFFE1 00031480C6EC003FD91FF890C7FC2F2B7DA933>97 D99 DII< 903A03FF8007F0013F9038F83FF8499038FCFFFC48B712FE48018313F93A07FC007FC348 48EB3FE1001FEDF1FC4990381FF0F81700003F81A7001F5DA26D133F000F5D6C6C495A3A 03FF83FF8091B5C7FC4814FC01BF5BD80F03138090CAFCA2487EA27F13F06CB6FC16F016 FC6C15FF17806C16C06C16E01207001F16F0393FE000034848EB003F49EC1FF800FF150F 90C81207A56C6CEC0FF06D141F003F16E001F0147FD81FFC903801FFC02707FF800F1300 6C90B55AC615F8013F14E0010101FCC7FC2F3D7DA834>103 D<13FFB5FCA512077EAFED 1FF8EDFFFE02036D7E4A80DA0FE07F91381F007F023C805C4A6D7E5CA25CA35CB3A4B5D8 FE0FB512E0A5333F7CBE3A>II<13FFB5FCA512077EB3B3AFB512 FCA5163F7CBE1D>108 D<01FFD91FF8ECFFC0B590B5010713F80203DAC01F13FE4A6E48 7FDA0FE09026F07F077F91261F003FEBF8010007013EDAF9F0806C0178ECFBC04A6DB448 6C7FA24A92C7FC4A5CA34A5CB3A4B5D8FE07B5D8F03FEBFF80A551297CA858>I<01FFEB 1FF8B5EBFFFE02036D7E4A80DA0FE07F91381F007F0007013C806C5B4A6D7E5CA25CA35C B3A4B5D8FE0FB512E0A533297CA83A>II<01FFEBFFE0B5000713FC021FEBFF80027F80DAFF8113F09139FC007FF8000301F0 6D7E4A6D7E4A130F4A6D7E1880A27013C0A38218E0AA4C13C0A318805E18005E6E5C6E49 5A6E495A02FCEBFFF0DAFF035B92B55A029F91C7FC028713FC028113C00280C9FCACB512 FEA5333B7DA83A>I<3901FE01FE00FF903807FF804A13E04A13F0EC3F1F91387C3FF800 0713F8000313F0EBFFE0A29138C01FF0ED0FE091388007C092C7FCA391C8FCB3A2B6FCA5 25297DA82B>114 D<90383FFC1E48B512BE000714FE5A381FF00F383F800148C7FC007E 147EA200FE143EA27E7F6D90C7FC13F8EBFFE06C13FF15C06C14F06C806C806C806C80C6 1580131F1300020713C014000078147F00F8143F151F7EA27E16806C143F6D140001E013 FF9038F803FE90B55A15F0D8F87F13C026E00FFEC7FC222B7DA929>IIII120 DI E /Ff 48 122 df<143F147E14FCEB01F81303EB07F0EB0FE014C0131FEB 3F80A2EB7F00A213FE12015B12035BA212075B120FA25B121FA25BA2123FA35BA2127FA6 90C7FC5AB3A27E7FA6123FA27FA3121FA27FA2120F7FA212077F1203A27F12017F120013 7FA2EB3F80A2EB1FC0130F14E0EB07F0EB03F81301EB00FC147E143F186478CA26>40 D<12FC127E7E6C7E7F6C7E6C7E12037F6C7EA26C7EA2137F1480133F14C0131FA214E013 0F14F0A2130714F8A21303A214FCA31301A214FEA6130014FFB3A214FE1301A614FCA213 03A314F8A21307A214F0130FA214E0131F14C0A2133F1480137F140013FEA2485AA2485A 5B1207485A485A5B48C7FC127E5A18647BCA26>I44 D<12FFA8080877871B>46 D<14FF010713E0011F13F8497F90B6FC48158014813A03FE007FC04848EB3FE049131F48 48EB0FF0491307001F15F8491303003F15FCA2491301A2007F15FEA390C8FCA34815FFB3 A36C15FE6D1301A5003F15FC6D1303A36C6CEB07F8A26C6CEB0FF0A26C6CEB1FE06D133F 6C6CEB7FC03A01FF81FF8091B5FC6C1500013F13FC6D5B010713E0010090C7FC28457CC2 31>48 D<1418147814F81303130F137FB5FCA413F71387EA0007B3B3AD007FB61280A621 4378C231>II<49B4FC010F13 E0013F13F890B512FE48800007158048010113C03A1FF8007FE0D83FE0133F49EB1FF048 C7120F123E16F8001C140712181208C8FC150FA216F0A2151F16E0153F16C0157FEDFF80 02031300EC1FFE90380FFFFC5D15E08115FC15FFD9000113809138007FC0ED1FE0ED0FF0 16F8ED07FCA2ED03FEA3ED01FFA80040EC03FE1260A200F0EC07FC7E00FEEC0FF8127FD8 3FC0EB1FF0D81FF0EB7FE0390FFE01FF6CB612806C15006C5C6C6C13F8011F13E0010190 C7FC28457CC231>II<000FB612F8A601 F0C8FCB1EC3FC09038F1FFF001F313FC90B57EEDFF8016C002E013E0EC003F49EB1FF049 EB0FF84913074914FC15034914FEC8FCA2150116FFAAED03FE12101230003815FC007C14 07007E15F800FE140FD87F80EB1FF06DEB3FE0D83FF0137F3A1FFE03FFC06CB612800003 15006C14FC6C6C5B011F13C0D903FEC7FC28437DC031>II< EDFF804A7FA34A7FA3913807EFF0A215E7020F7F15C7A2021F7F15C391383F83FEA21581 027F7F1501814A805CA201016E7E5CA201036E7E5CA201076E7E5CA2010F6E7E5C011F81 16075C013F8116035C017F818291B7FC90B87EA34883A201FCC8127F0003707EA2484882 171F5B000F83170F5B001F8317075B003F8317035B007F838390C9FC48701380A239457D C440>65 DIIIIIIII75 D IIIII82 DIII< D87FE0ED03FE6C6C15076D4B5A6C6C5E000F161F6C6C4B5A6D5E6C167F6C6D4A5A6C6D5D 4C90C7FC6D6C5B6D6C5C011F4A5A6E130F6D6C5C6D6C495A163F6D6C5C6D6D485A6D14FF 03C15BDA7FE190C8FC91383FE3FEEC1FF7EDFFFC6E5B805E6E5B806E5B5E824A7F5C824A 7F5CEDF7FC91381FF3FE91383FE1FF15C1DA7FC07F4A486C7E8349EB003F49486D7E0107 814A130F49486D7E011F814A6D7E49487F017F824A6D7F49486E7E5A91C86C7E48486F7E 1207496F7E48486F7E001F83484881496F1380007F7013C0484817E049167F3B457EC440 >88 DI<003FB81280A518 00C9EA07FEA24C5A5F161F4C5AA24C5A4C5AA24B5B4B90C7FCA24B5A4B5AA24B5AA24B5A 4B5AA24B5A4A5BA24A90C8FC4A5AA24A5A5D141F4A5AA24A5A4A5AA2495B4990C9FCA249 5A495AA2495AA2495A495AA2495A485BA24890CAFC485AA2485A5B121F485AA248B81280 B9FCA531457BC43C>I97 D99 DII103 D105 D110 DII<903901FE01FE90380FFF81013F13E14913F990B512FD4814FF48EBE0 7F48EB001F4848130F01F81307485A1503485A491301127F5BA348C7FCAA7F127FA27F12 3F7F15036C7E6D13076C6C130F6C6C133F9038FFC0FF6C90B5FC6C14FD6C14F1013F13E1 6D1381903803FC0190C7FCB227407DAC32>I<141F00FE13FF13035B131F5B5BEBFFF014 8038FFFE005B5B5B5B5BA25BA390C7FCB3A8182D79AC21>II117 D119 D121 D E /Fg 19 90 df<121EEA3F80EA7FC012FFA41380EA7F00123C0A0A788919>46 D48 D<157015F014011407143F903803FFE0137FEBFFCFEBF80F1300141F15C0A5 143F1580A5147F1500A55C5CA513015CA513035CA513075CA5130F5CA3131F497EB612F8 A31D3D78BC2D>I<17E016011603831607A2160FA2161F83163FA2167F167716F7EEE7FC ED01E316C3150316831507EE03FEED0F01150E151E151C153C03387FED7800157015F05D 4A4880177F4A5AA24AC7FCA2020E81173F5C021FB6FC5CA20270C7EA3FE0171F5CA2495A A2494881170F49C8FCA2130EA24982013C1507A2137CD801FE4B7E2607FF80EC3FFEB500 F00107B512FC19F85E3E417DC044>65 D67 D<013FB7FC18E018F8903B007FF0000FFE6E48EB01FF9438007FC04B6E7E180F85 727E727E147F4B6E7EA2727EA302FF178092C9FCA54918C05CA41A8013034A5DA41A0013 074A5DA261A24E5A130F4A5E180F61181F61011F4C5A5C4E5A4EC7FC4D5A4D5A013F4B5A 4A4A5AEF3FE0EF7F80017F4A48C8FC01FFEC1FFCB812F0178004FCC9FC423E7DBD45>I< 013FB812F8A39026007FF0C7127F6E48140F18034B14011800A31978147F4B1570A502FF 143892C7FCA3190017784915704A14F016011603160F91B6FC495DA29138FC001F160716 03160101075D5CA2197019F019E0010F4A5A4A90C7120119C0A218031980011F16075CF0 0F00A260181E013F163E4A157E4D5A1703017F150F01FFEDFFF8B9FCA2603D3E7DBD3E> I<4BB46C1370031F01F013F0037F9038FC01E0913A03FF807E03913A0FF8000F83DA1FE0 EB07C7DA7F80EB01EF4AC812FFD903FE16C04948157F4948153F495A4948151F495A4948 168091C9120F5A485AA2485A000F18004982121FA248485EA295C7FC485AA412FF5BA604 3FB512E05BA29339001FFC00715AA2607F127FA2171F123F6D5EA2121F7F000F163F6C7E 6C6C4B5A7F6C6C15FF6C6DEB01EFD93FC0EB07C7D91FF0EB1F87D907FE9038FE03800101 B5EAF8016D6C01E0C8FCDA07FEC9FC3C4276BF47>71 D<011FB512FC5BA29039003FF800 6E5AA25DA5143F5DA5147F5DA514FF92C7FCA55B5CA513035CA513075CA5130F5CA5131F 5CA3133F497E007FB512F0A2B6FC263E7EBD21>73 D<90263FFFF0933807FFFE5013FC62 9026007FF8EFFC00023F4D5AA2023BEF77F0A2DA39FC16E7A2F101CF0279EE038FDA70FE 5FF1070FA2190E1A1FDAF07F151C02E060193819706F7EF1E03F130102C0DB01C05BA26F 6CEB0380A2953807007F0103160E4A6C6C93C7FC60A2606201076D6C5B02005F60A26F6C 485A94380380015B010EDB07005BA2923801FC0EA24D1303131E011C6D6C485C5FA25F19 07013CEC7FC0013860013C5D137C01FE6EC7120F2607FF80013E4A7EB500FC031FB512F8 043C5E4A131C573E7DBD53>77 D<90263FFFE0023FB5FC6F16FEA29026003FF8020313C0 021F030013004A6C157C023B163C6F15381439810238167802787FDA707F157082153F82 031F15F002F07FDAE00F5D8215078203031401010180DAC0015D82811780047F13030103 15C04A013F5C17E0161F17F0040F1307010715F891C7000791C7FC17FC160317FE04015B 4915FF010E6E130E188E177F18CEEF3FDE011E16FE011C6F5AA2170FA21707133C01386F 5A133C017C150113FE2607FF801400B512FC18705C483E7DBD44>I<923803FF80031F13 F09238FE01FE913903F0003FDA0FC0EB1FC0DA3F80EB07E0027EC76C7E49486E7E494881 49486E7E4948157F495A013F17804948ED3FC049C9FCA24848EE1FE012035B000718F05B 120FA2485A19F8123F5BA2127FA219F04848163FA5F07FE0A35BF0FFC0A219805F19007F 4D5A127F4D5A60003F160F6D5E001F4C5A4D5A6C6C4B5A95C7FC6C6C15FE00034B5A6C6C 4A5A6C6C4A5A017FEC1FC06D6C495AD90FE001FEC8FC903903F807F80100B512C0DA0FFC C9FC3D4276BF47>I<013FB612FEEFFFE018F8903B007FF0000FFC6E48EB01FF7113804B EC7FC0183F19E0F01FF0A2147F5D19F8A402FFED3FF092C8FCA219E0A2F07FC05B4AEDFF 8019004D5A4D5AEF0FF80103ED3FE04A903801FF8091B648C7FC17F002FCCAFCA213075C A5130F5CA5131F5CA5133F5CA3137F497EB612E0A25D3D3E7DBD3E>I<923803FF80031F 13F09238FE01FE913903F8003FDA0FE0EB1FC0DA3F806D7E4AC7EA03F0D901FC8149486E 7E49486E7E010F82494881494816804948ED3FC013FF91C9FC484817E00003171F5B0007 18F0A2485AA2485A19F8123FA25B127FA219F04848163FA519E0187F5BA219C018FF1980 A24D1300A24D5A6C7E4D5A60003F160F037C5C6C6C48B4495A913A0783803FC0000F9026 0E01C05B9026F00C0049C7FC0007011CEBE0FE2603F818EB61FCD801FCEC73F8D800FEEC 77F0017FEC7FC0D93F985CD90FFC01FEC8FC902703FE07F813030100B5FC91260FFC3C5B 91C7FC180E163E181E043F137CEF81FC17FF60A260A2705B60705B7048C7FCEE01F83D52 76BF47>I<013FB612F017FF18E0903B007FF0003FF86E48EB07FCEF01FE4B6D7EF07F80 19C0183F19E0147F4B15F0A502FFED7FE092C8FCA219C0F0FF80A2494B13004A5D4D5AEF 0FF04D5AEF7F800103DA07FEC7FC91B612F017809139FC0007E0EE03F8EE00FC0107814A 147F717EA284A2130F5CA484011F157F5CA41902013F17075CA2F0F00F017F170E496C14 3FB600E0011F131C94380FF83C4B01071378CA3801FFE09438003F8040407DBD43>I<92 39FF8003800207EBF007021F9038FC0F0091387F00FE02FCEB1F1FD903F0EB07BF49486D B4FC49487F4A6D5A49C8FC49157E133E137E173E49153CA57F1738A26D92C7FC808080EB 7FFEECFFE06D13FEEDFFC06D14F06D14FC010380010080143F020380DA003F7F15031500 707E163F161FA2160F121CA31607160F003C5EA35F003E151F94C7FC007E5D007F153E6D 5C16FC01E0495AD87DF0495AD8FCFCEB0FC03AF87F803F8027F01FFFFEC8FCD8E00713F8 39C0007FC031427BBF33>I<0007B912F0A33C0FFE000FF8003F01F0160F01C04A130348 48160190C7FC121EF000E048141F5E1238A212781270153F5E5AA3C81600157F5EA515FF 93C9FCA55C5DA514035DA514075DA5140F5DA3141FEC7FFC0003B7FCA33C3D76BC42>I< B600E090B512FC4B15F8A2000101C0C7000F13006C49EC03FCEF01F091C9FC60A317015A 495EA417031203495EA4170712074993C7FCA45F120F49150EA4171E121F49151CA4173C 123F491538A31778177017F05F001F15015F16036D4A5A000F93C8FC5E6C6C141E6C6C5C 000115F86C6C495A017FEB07C090393FC03F8090260FFFFEC9FC010313F89038007FC03E 4073BD44>I89 D E /Fh 10 58 df<13FF000313C0380781E0380F00F0001E137848 133CA248131EA400F8131FAD0078131EA2007C133E003C133CA26C13786C13F0380781E0 3803FFC0C6130018227DA01E>48 D<13E01201120712FF12F91201B3A7487EB512C0A212 217AA01E>II<13FF000313C0380F03E0381C 00F014F8003E13FC147CA2001E13FC120CC712F8A2EB01F0EB03E0EB0FC03801FF00A238 0003E0EB00F01478147C143E143F1230127812FCA2143E48137E0060137C003813F8381E 03F0380FFFC00001130018227DA01E>I<14E01301A213031307A2130D131D1339133113 6113E113C1EA01811203EA07011206120C121C12181230127012E0B6FCA2380001E0A6EB 03F0EB3FFFA218227DA11E>I<00101330381E01F0381FFFE014C01480EBFE00EA1BF000 18C7FCA513FE381BFF80381F03C0381C01E0381800F014F8C71278A2147CA21230127812 F8A214784813F8006013F0387001E01238381E07803807FF00EA01F816227CA01E>II<1230123C003FB5FCA248 13FE14FC3860001C143814704813E014C0EA0001EB0380EB07001306130E5BA25BA21378 A35BA41201A76C5A18237CA11E>I<137F3803FFC0380781E0380E00704813380018131C 1238A3123C003F1338381FC078EBE0F0380FF9E03807FF80120114C0000713F0380F0FF8 381C03FC383801FE3870007E141F48130F1407A314060070130E0078130C6C1338001F13 F03807FFC0C6130018227DA01E>I<13FE3803FFC0380781E0380E0070481378003C1338 48133CA200F8131EA3141FA40078133FA26C137F121C380F01DF3807FF9F3803FE1EC7FC A2143E143C001C1338003E13781470003C13E0381801C0381C0780380FFE00EA03F81822 7DA01E>I E /Fi 2 110 df<121EEA7F80A2EAFFC0A4EA7F80A2EA1E000A0A798919>58 D109 D E /Fj 6 117 df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k 1 14 df13 D E /Fl 71 123 df<4AB4EB0FE0021F9038E03FFC913A7F00F8FC1ED901FC90383FF03FD907F090397FE0 7F80494801FF13FF4948485BD93F805C137F0200ED7F00EF003E01FE6D91C7FC82ADB97E A3C648C76CC8FCB3AE486C4A7E007FD9FC3FEBFF80A339407FBF35>11 D<4AB4FC021F13C091387F01F0903901FC0078D907F0131C4948133E494813FF49485A13 7F1400A213FE6F5A163893C7FCAA167FB8FCA33900FE00018182B3AC486CECFF80007FD9 FC3F13FEA32F407FBF33>I<001E130F397F803FC000FF137F01C013E0A201E013F0A300 7F133F391E600F3000001300A401E01370491360A3000114E04913C00003130101001380 481303000EEB070048130E0018130C0038131C003013181C1C7DBE2D>34 D<121EEA7F8012FF13C0A213E0A3127FEA1E601200A413E013C0A312011380120313005A 120E5A1218123812300B1C79BE19>39 D<1430147014E0EB01C0EB03801307EB0F00131E 133E133C5B13F85B12015B1203A2485AA2120F5BA2121F90C7FCA25AA3123E127EA6127C 12FCB2127C127EA6123E123FA37EA27F120FA27F1207A26C7EA212017F12007F13787F13 3E131E7FEB07801303EB01C0EB00E014701430145A77C323>I<12C07E12707E7E121E7E 6C7E7F12036C7E7F12007F1378137CA27FA2133F7FA21480130FA214C0A3130714E0A613 0314F0B214E01307A614C0130FA31480A2131F1400A25B133EA25BA2137813F85B12015B 485A12075B48C7FC121E121C5A5A5A5A145A7BC323>I<121EEA7F8012FF13C0A213E0A3 127FEA1E601200A413E013C0A312011380120313005A120E5A1218123812300B1C798919 >44 DI<121EEA7F80A2EAFFC0A4EA7F80A2EA1E000A0A798919> IIIIII<150E151E153EA2157EA215FE1401A21403EC077E14 06140E141CA214381470A214E0EB01C0A2EB0380EB0700A2130E5BA25B5BA25B5B120148 5A90C7FC5A120E120C121C5AA25A5AB8FCA3C8EAFE00AC4A7E49B6FCA3283E7EBD2D>I< 00061403D80780131F01F813FE90B5FC5D5D5D15C092C7FC14FCEB3FE090C9FCACEB01FE 90380FFF8090383E03E090387001F8496C7E49137E497F90C713800006141FC813C0A216 E0150FA316F0A3120C127F7F12FFA416E090C7121F12FC007015C012780038EC3F80123C 6CEC7F00001F14FE6C6C485A6C6C485A3903F80FE0C6B55A013F90C7FCEB07F8243F7CBC 2D>II<1238123C123F90B612FCA3 16F85A16F016E00078C712010070EC03C0ED078016005D48141E151C153C5DC8127015F0 4A5A5D14034A5A92C7FC5C141EA25CA2147C147814F8A213015C1303A31307A3130F5CA2 131FA6133FAA6D5A0107C8FC26407BBD2D>III<121EEA7F 80A2EAFFC0A4EA7F80A2EA1E00C7FCB3121EEA7F80A2EAFFC0A4EA7F80A2EA1E000A2779 A619>I<121EEA7F80A2EAFFC0A4EA7F80A2EA1E00C7FCB3121E127FEAFF80A213C0A412 7F121E1200A412011380A3120313005A1206120E120C121C5A1230A20A3979A619>I<15 074B7EA34B7EA34B7EA34B7EA34B7E15E7A2913801C7FC15C3A291380381FEA34AC67EA3 020E6D7EA34A6D7EA34A6D7EA34A6D7EA34A6D7EA349486D7E91B6FCA249819138800001 A249C87EA24982010E157FA2011E82011C153FA2013C820138151FA2017882170F13FC00 034C7ED80FFF4B7EB500F0010FB512F8A33D417DC044>65 DI< DB3FF01306912603FFFE130E020F9038FF801E913A3FF007E03E9139FF8000F8D903FEC7 EA7C7ED907F8EC1EFE4948140FD93FE0140749481403495A91C812014848150012034848 167E5B000F173EA24848161EA2123F5B180E127FA349160012FFAC127F7F180EA2123FA2 7F001F171E181C6C7EA20007173C6D16386C6C1678000117706C6C16F06EEC01E06D6C15 C06D6C1403D90FF0EC07806D6CEC1F00D903FE143E902600FF8013F891393FF007F0020F B512C0020391C7FC9138003FF037427BBF42>IIIIIII<011FB512FCA3D9000713006E5A1401B3B3A6123FEA7F80EAFFC0A44A5A1380D87F 005B007C130700385C003C495A6C495A6C495A2603E07EC7FC3800FFF8EB3FC026407CBD 2F>III< B500C093383FFFF0A300016D93387FF800D8007F18E0D977F016EFA3D973F8ED01CFA2D9 71FCED038FA3D970FEED070FA26E150E80A26E6C141CA36E6C1438A26E6C1470A36E6C14 E0A26E6CEB01C0A36E6CEB0380A36E6CEB0700A2037F130EA36F6C5AA26F6C5AA36F6C5A A25FED07F0A2923803F9C0A36FB45AA26F90C7FCA213F8486C147ED807FFEF3FF8B500F8 013C011FB512F0A34C3E7DBD53>IIII82 DI<003FB91280A3903AF0007FE001018090393FC0003F48C7ED1FC0007E1707 127C00781703A300701701A548EF00E0A5C81600B3B14B7E4B7E0107B612FEA33B3D7DBC 42>II<007FB5D8C0 03B512E0A3C649C7EBFC00D93FF8EC3FE06D48EC1F806D6C92C7FC171E6D6C141C6D6C14 3C5F6D6C14706D6D13F04C5ADA7FC05B023F13036F485ADA1FF090C8FC020F5BEDF81E91 3807FC1C163C6E6C5A913801FF7016F06E5B6F5AA26F7E6F7EA28282153FED3BFEED71FF 15F103E07F913801C07F0203804B6C7EEC07004A6D7E020E6D7E5C023C6D7E02386D7E14 784A6D7E4A6D7F130149486E7E4A6E7E130749C86C7E496F7E497ED9FFC04A7E00076DEC 7FFFB500FC0103B512FEA33F3E7EBD44>88 DI<003FB712F8A391C7EA1FF013F801E0EC3FE00180EC7FC090C8FC003EEDFF80 A2003C4A1300007C4A5A12784B5A4B5AA200704A5AA24B5A4B5AA2C8485A4A90C7FCA24A 5A4A5AA24A5AA24A5A4A5AA24A5A4A5AA24990C8FCA2495A4948141CA2495A495AA2495A 495A173C495AA24890C8FC485A1778485A484815F8A24848140116034848140F4848143F ED01FFB8FCA32E3E7BBD38>I<486C13C00003130101001380481303000EEB070048130E 0018130C0038131C003013180070133800601330A300E01370481360A400CFEB678039FF C07FE001E013F0A3007F133FA2003F131F01C013E0390F0007801C1C73BE2D>92 D97 D I<49B4FC010F13E090383F00F8017C131E4848131F4848137F0007ECFF80485A5B121FA2 4848EB7F00151C007F91C7FCA290C9FC5AAB6C7EA3003FEC01C07F001F140316806C6C13 076C6C14000003140E6C6C131E6C6C137890383F01F090380FFFC0D901FEC7FC222A7DA8 28>IIII<167C903903F801FF903A1FFF078F8090397E0FDE1F90 38F803F83803F001A23B07E000FC0600000F6EC7FC49137E001F147FA8000F147E6D13FE 00075C6C6C485AA23901F803E03903FE0FC026071FFFC8FCEB03F80006CAFC120EA3120F A27F7F6CB512E015FE6C6E7E6C15E06C810003813A0FC0001FFC48C7EA01FE003E140048 157E825A82A46C5D007C153E007E157E6C5D6C6C495A6C6C495AD803F0EB0FC0D800FE01 7FC7FC90383FFFFC010313C0293D7EA82D>III107 DI<2701F801FE14FF 00FF902707FFC00313E0913B1E07E00F03F0913B7803F03C01F80007903BE001F87000FC 2603F9C06D487F000101805C01FBD900FF147F91C75B13FF4992C7FCA2495CB3A6486C49 6CECFF80B5D8F87FD9FC3F13FEA347287DA74C>I<3901F801FE00FF903807FFC091381E 07E091387803F000079038E001F82603F9C07F0001138001FB6D7E91C7FC13FF5BA25BB3 A6486C497EB5D8F87F13FCA32E287DA733>I<14FF010713E090381F81F890387E007E01 F8131F4848EB0F804848EB07C04848EB03E0000F15F04848EB01F8A2003F15FCA248C812 FEA44815FFA96C15FEA36C6CEB01FCA3001F15F86C6CEB03F0A26C6CEB07E06C6CEB0FC0 6C6CEB1F80D8007EEB7E0090383F81FC90380FFFF0010090C7FC282A7EA82D>I<3901FC 03FC00FF90381FFF8091387C0FE09039FDE003F03A03FFC001FC6C496C7E91C7127F49EC 3F805BEE1FC017E0A2EE0FF0A3EE07F8AAEE0FF0A4EE1FE0A2EE3FC06D1580EE7F007F6E 13FE9138C001F89039FDE007F09039FC780FC0DA3FFFC7FCEC07F891C9FCAD487EB512F8 A32D3A7EA733>I<02FF131C0107EBC03C90381F80F090397F00387C01FC131CD803F813 0E4848EB0FFC150748481303121F485A1501485AA448C7FCAA6C7EA36C7EA2001F14036C 7E15076C6C130F6C7E6C6C133DD8007E137990383F81F190380FFFC1903801FE0190C7FC AD4B7E92B512F8A32D3A7DA730>I<3901F807E000FFEB1FF8EC787CECE1FE3807F9C100 031381EA01FB1401EC00FC01FF1330491300A35BB3A5487EB512FEA31F287EA724>I<90 383FC0603901FFF8E03807C03F381F000F003E1307003C1303127C0078130112F81400A2 7E7E7E6D1300EA7FF8EBFFC06C13F86C13FE6C7F6C1480000114C0D8003F13E0010313F0 EB001FEC0FF800E01303A214017E1400A27E15F07E14016C14E06CEB03C0903880078039 F3E01F0038E0FFFC38C01FE01D2A7DA824>I<131CA6133CA4137CA213FCA21201120312 07001FB512C0B6FCA2D801FCC7FCB3A215E0A912009038FE01C0A2EB7F03013F13809038 1F8700EB07FEEB01F81B397EB723>IIIII I<001FB61280A2EBE0000180140049485A001E495A121C4A5A003C495A141F00385C4A5A 147F5D4AC7FCC6485AA2495A495A130F5C495A90393FC00380A2EB7F80EBFF005A5B4848 13071207491400485A48485BA248485B4848137F00FF495A90B6FCA221277EA628>I E /Fm 23 122 df<121EEA7F8012FF13C0A213E0A3127FEA1E601200A413E013C0A31201 1380120313005A1206120E5A5A5A12600B1D78891B>44 D<143014F013011303131F13FF B5FC13E713071200B3B3B0497E497E007FB6FCA3204278C131>49 DI 54 D<14FF010713E0011F13F890387F80FC9038FC007E48487F4848EB1F804848EB0FC0 000FEC07E0485AED03F0485A16F8007F140190C713FCA25AA216FE1500A516FFA46C5CA3 6C7E5D121F7F000F5C6C6C1306150E6C6C5B6C6C5BD8007C5B90383F01E090390FFF80FE 903801FE0090C8FC150116FCA4ED03F8A216F0D80F801307486C14E0486C130F16C0ED1F 80A249EB3F0049137E001EC75A001C495A000F495A3907E01FE06CB51280C649C7FCEB1F F028447CC131>57 D70 D72 D<010FB512FEA3D90003 13806E130080B3B3AB123F487E487EA44A5A13801300006C495A00705C6C13076C5C6C49 5A6CEB1F802603E07FC7FC3800FFFCEB1FE027467BC332>74 D80 D97 D99 D101 DIIII109 D<3901FC01FE00FF903807FFC091381E07F091383801F8000701707F0003EBE0002601FD C07F5C01FF147F91C7FCA25BA35BB3A8486CECFF80B5D8F83F13FEA32F2C7DAB36>II<3903F803F000FFEB1FFCEC3C3EEC707F0007 EBE0FF3803F9C000015B13FBEC007E153C01FF13005BA45BB3A748B4FCB512FEA3202C7D AB26>114 D<1306A5130EA4131EA3133E137EA213FE12011207001FB512F0B6FCA2C648 C7FCB3A4150CAA017E131C017F1318A26D133890381F8030ECC070903807E0E0903801FF C09038007F001E3E7EBC26>116 DI121 D E /Fn 20 118 df12 D70 D72 D76 D80 D83 D97 D<4AB47E020F13F8023F13FE9139FF007F 80D903FCEB07E0D907F0EB01F0D91FE0EB007849488049488049C87E48485D4915FF0003 4B138048485CA2485AA2485AA2003F6F130049EC007C94C7FC127FA35B12FFAD127F7FA4 123F7FA2001FEE01C07F000F16036D168012076C6C15076D160000015E6C6C151E6D6C5C 6D6C5C6D6C5CD90FF8495AD903FCEB07C0903A00FF803F8091263FFFFEC7FC020F13F802 01138032417CBF3A>99 D101 DIII<133C13FF487F487FA66C5B6C90C7FC133C90C8FCB3A2EB03C0EA07FF127FA41201 EA007FA2133FB3B3AC497E497EB612E0A41B5F7DDE23>I 109 DIII<9039078003F8D807FFEB0FFFB5013F13 C092387C0FE0913881F01F9238E03FF00001EB838039007F8700148FEB3F8E029CEB1FE0 EE0FC00298EB030002B890C7FCA214B014F0A25CA55CB3B0497EEBFFF8B612FCA42C3F7C BE33>114 D<1438A71478A414F8A31301A31303A21307130F131FA2137F13FF1203000F 90B6FCB8FCA3260007F8C8FCB3AE17E0AE6D6CEB01C0A316036D6C148016076D6C14006E 6C5A91383FC01E91381FF07C6EB45A020313E09138007F802B597FD733>116 DI E /Fo 5 85 df<1A1E1A1F6262A262A261A26161A2618761A26161197B19FB19F3F001E3 A2F003C318071983180F1903061E80A2183C85187818F818F0EF01E0A2EF03C0A2EF0780 170F1800171E875FA24D7F17F85F4C5AA24C5AA24C5A160F94C8FC5E161E4C82A25E04F8 157F5E15015E4BB8FCA25DA24BC9127F5D151E4B83A25D1A3F5D14015D4A5AA24A5AA24A CAFC5C141E5C875C14F81A1F495A13031307010F183F131FD97FF84D7E2603FFFE0403B5 7E007FD9FFE092B7FC4E1680B61A006C5C596678E568>65 D<92B912F04A18FEF3FFC06E 19F0DB007F90C700017F70489138003FFE7048ED0FFF043F04037F757F4D6F7F767E047F 173F767E4D707EA204FF717EA24D707EA24B1A80885FA24B1AC0A25F885D1EE094CBFCA2 5D645EA2151FA25EA2153F645E1EC0157FA24C5FA215FF1E804C5FA25C1E004C5FA24A62 A24C173F655C525A93CBFCA24A4F5AA24B4D5BA2021F62515B5D5190C7FC023F611B0F4B 4D5A64027F183F515A4B60515A02FF4D5B5090C8FC4B4C5A1A0F49F01FF8505A4B4C5A50 5A494D5B070790C9FC4BED1FFC4F5A49933801FFE04904071380013FDC7FFECAFC007FB9 12F8BA12E096CBFC18F0636276E16A>68 D<92BB12FC5CA280DB007F90C8120F70481500 7048EE3FF8043F170F1C074D1603A2167F1C014D17F0A216FFA25FA25D1DE05FA25DA25F 1C034B19C0A294CAFC1AE04B4B6C14805090C7FC5EA2031F1503625E1907153F624C140F 191F157F077FC9FC4C5C180F92B8FC61A34A9038E0000FF001FC4C1300A25C19785E19F8 5C6193C8FCA24A1501615DA2021F5E95CBFC5DA2143FA25DA2147FA25DA214FFA25DA25B A25DA25BA25DA25B5B013F13F0007FB612FEB7FCA35E6276E15C>70 D<92B812FCF2FFC04A18F86E18FEDB007F90C7383FFF80704802037F7048020013F0043F EE3FF8757E4D6F7E1B07047F837513805F1DC016FFA25F1DE05DA25FA25D5113C05FA25D 51138094C9FC1D004B5F644C4C5AA2031F4D5A644C4C5A505B033F4C5B5048C7FC4C4B5A F23FF0037F4C5A963801FF804CDA0FFEC8FCF1FFF892B812C04FC9FC619339E00003FF4A 9238007FC0737E4C6E7E737E4A1607864C8119034A83A293C8FCA24A84A25DA2021F5EA2 4B94C8FCA2023F5EA25DA2027F5EA25DA202FF5E625DA2491BE01C015DA2491A031DC04B 18071D80491A0F49051F1500013F01E060007FB600F0020F143EB76F6C5B735C4C6EEB83 F0CD6CB45A081F1380E003FEC7FC5B6576E166>82 D<017FBB12FC1CFE90BCFCA292C727 7FFC000113FC4801F06E48EB001F0280027F150791C81603D803FC4C14015B4903FF16F8 49190000075F5B5E48481AF06090C8FC484B1601A2001E5F003E1BE0003C5DA2007C94C8 FC00781A034C17C012F8485EA2161F481B80C94992C7FCA2163FA25FA2167FA25FA216FF A25FA25DA25FA25DA25FA25DA294CBFCA25DA25EA2151FA25EA2153FA25EA2157FA25EA2 15FFA25EA25CA25EA25CA25EA25CA293CCFC5CA25C49B512E0001FB8FC5AA25F576166E0 64>84 D E end %%EndProlog %%BeginSetup %%Feature: *Resolution 600dpi TeXDict begin %%EndSetup %%Page: 0 1 0 0 bop 1597 1915 a Fo(D)50 b(R)h(A)f(F)g(T)1194 2098 y Fn(High)45 b(P)l(erformance)e(F)-11 b(ortran)1291 2281 y(Language)43 b(Sp)t(eci\014cation)1234 2670 y Fm(High)32 b(P)m(erformance)h(F)-8 b(ortran)31 b(F)-8 b(orum)1611 2922 y(Jan)m(uary)33 b(2,)f(1996)p eop %%Page: 1 2 1 1 bop 3725 -200 a Fl(i)332 2911 y(This)37 b(is)h(the)h(result)f(of)i (a)f(LaT)-8 b(eX)40 b(run)e(of)h(a)g(draft)g(of)g(a)h(single)d(c)m (hapter)j(of)f(the)g(HPFF)h(Final)150 3024 y(Rep)s(ort)30 b(do)s(cumen)m(t.)357 4934 y(c)332 4937 y Fk(\015)p Fl(1992)k(Rice)f (Univ)m(ersit)m(y)-8 b(,)33 b(Houston)f(T)-8 b(exas.)48 b(P)m(ermission)31 b(to)i(cop)m(y)h(without)d(fee)i(all)f(or)g(part)h (of)150 5050 y(this)c(material)h(is)f(gran)m(ted,)j(pro)m(vided)d(the)h (Rice)g(Univ)m(ersit)m(y)g(cop)m(yrigh)m(t)h(notice)f(and)g(the)h (title)f(of)g(this)150 5163 y(do)s(cumen)m(t)g(app)s(ear,)g(and)g (notice)h(is)e(giv)m(en)h(that)h(cop)m(ying)g(is)e(b)m(y)h(p)s (ermission)e(of)i(Rice)h(Univ)m(ersit)m(y)-8 b(.)p eop %%Page: 2 3 2 2 bop 150 763 a Fj(Con)-6 b(ten)g(ts)286 1208 y Fl(0.1)94 b(Mapping)30 b(Inquiry)e(Pro)s(cedures)90 b Fi(:)46 b(:)g(:)f(:)h(:)g (:)f(:)h(:)g(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:) g(:)f(:)h(:)184 b Fl(1)495 1321 y(0.1.1)106 b(New)31 b(in)e(HPF)i(2.0)84 b Fi(:)46 b(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)f(:)h(:)g (:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:) 184 b Fl(1)495 1433 y(0.1.2)106 b(HPF)p 980 1433 28 4 v 33 w(ALIGNMENT\(ALIGNEE,)22 b(LB,)e(UB,)h(STRIDE,)f(AXIS)p 3049 1433 V 33 w(MAP)-8 b(,)21 b(IDEN-)786 1546 y(TITY)p 1024 1546 V 32 w(MAP)-8 b(,)32 b(D)m(YNAMIC,)g(NCOPIES\))24 b Fi(:)46 b(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f (:)h(:)184 b Fl(1)495 1659 y(0.1.3)106 b(HPF)p 980 1659 V 33 w(TEMPLA)-8 b(TE\(ALIGNEE,)21 b(TEMPLA)-8 b(TE)p 2551 1659 V 32 w(RANK,)21 b(LB,)g(UB,)g(AXIS)p 3468 1659 V 32 w(TYPE,)786 1772 y(AXIS)p 1011 1772 V 63 w(INF)m(O,)31 b(NUMBER)p 1770 1772 V 34 w(ALIGNED,)g(D)m(YNAMIC\))90 b Fi(:)45 b(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)184 b Fl(3)495 1885 y(0.1.4)106 b(HPF)p 980 1885 V 33 w(DISTRIBUTION\(DISTRIBUTEE,)39 b(AXIS)p 2662 1885 V 32 w(TYPE,)h(AXIS)p 3237 1885 V 32 w(INF)m(O,)786 1998 y(PR)m(OCESSORS)p 1407 1998 V 51 w(RANK,)20 b(PR)m(OCESSORS)p 2393 1998 V 31 w(SHAPE,)g(PLB,)g(PUB,)h (PSTRIDE\))95 b(5)495 2111 y(0.1.5)106 b(MAP)p 1004 2111 V 33 w(ARRA)-8 b(Y\(DISTRIBUTEE,)31 b(DIM\))51 b Fi(:)45 b(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)184 b Fl(8)495 2224 y(0.1.6)106 b(BLOCK)p 1120 2224 V 32 w(SIZES\(DISTRIBUTEE,)29 b(DIM\))85 b Fi(:)46 b(:)g(:)f(:)h(:)g(:)g(:)f (:)h(:)g(:)f(:)h(:)g(:)g(:)f(:)h(:)184 b Fl(8)84 5438 y @beginspecial @setspecial -3 0 moveto 3 0 lineto 3 -720 lineto -3 -720 lineto closepath 1 setgray fill @endspecial 3934 92 a Fh(1)3934 205 y(2)3934 318 y(3)3934 431 y(4)3934 544 y(5)3934 657 y(6)3934 770 y(7)3934 882 y(8)3934 995 y(9)3919 1108 y(10)3919 1221 y(11)3919 1334 y(12)3919 1447 y(13)3919 1560 y(14)3919 1673 y(15)3919 1786 y(16)3919 1899 y(17)3919 2012 y(18)3919 2124 y(19)3919 2237 y(20)3919 2350 y(21)3919 2463 y(22)3919 2576 y(23)3919 2689 y(24)3919 2802 y(25)3919 2915 y(26)3919 3028 y(27)3919 3141 y(28)3919 3254 y(29)3919 3366 y(30)3919 3479 y(31)3919 3592 y(32)3919 3705 y(33)3919 3818 y(34)3919 3931 y(35)3919 4044 y(36)3919 4157 y(37)3919 4270 y(38)3919 4383 y(39)3919 4496 y(40)3919 4609 y(41)3919 4721 y(42)3919 4834 y(43)3919 4947 y(44)3919 5060 y(45)3919 5173 y(46)3919 5286 y(47)3919 5399 y(48)p eop %%Page: 1 4 1 3 bop 150 -200 a Fg(0.1.)72 b(MAPPING)31 b(INQUIR)-8 b(Y)31 b(PR)m(OCEDURES)1792 b Fl(1)150 99 y Ff(0.1)97 b(Mapping)32 b(Inquiry)g(Pro)s(cedures)150 302 y Fl(Revised)e(Jan)g(2,) h(1996,)h(RSS.)150 547 y Ff(0.1.1)97 b(New)34 b(in)e(HPF)h(2.0)150 719 y Fl(Changes:)261 908 y(1.)46 b(Remo)m(v)m(e)32 b(the)f(stricture)f (against)g(structure)g(comp)s(onen)m(ts.)261 1097 y(2.)46 b(Add)30 b(new)g(DISTRIBUTION)f(t)m(yp)s(e:)41 b(INDIRECT.)261 1286 y(3.)46 b(Add)30 b(in)m(uiry)e(functions)h(MAP)p 1450 1286 28 4 v 33 w(ARRA)-8 b(Y)31 b(and)f(BLOCK)p 2350 1286 V 32 w(SIZES)261 1475 y(4.)46 b(Add)22 b(LB,)h(UB,)h(and)e (STRIDE)f(optional)h(argumen)m(ts)h(to)h(DISTRIBUTION)d(to)j(handle)d (sections)377 1588 y(of)31 b(pro)s(cessors)f(arrangemen)m(ts)h(in)e (ONTO)g(clause.)150 1833 y Ff(0.1.2)97 b(HPF)p 639 1833 30 4 v 36 w(ALIGNMENT\(ALIGNEE,)43 b(LB,)f(UB,)g(STRIDE,)g(AXIS)p 2735 1833 V 36 w(MAP)-8 b(,)43 b(IDENTITY)p 3492 1833 V 34 w(MAP)-8 b(,)448 1949 y(D)m(YNAMIC,)34 b(NCOPIES\))423 2121 y Fe(Optional)22 b(Argumen)m(ts.)37 b Fd(LB)p Fl(,)20 b Fd(UB)p Fl(,)g Fd(STRIDE)p Fl(,)f Fd(AXIS)p 2225 2121 29 4 v 33 w(MAP)p Fl(,)h Fd(IDENTITY)p 2831 2121 V 32 w(MAP)p Fl(,)f Fd(DYNAMIC)p Fl(,)g Fd(NCOPIES)423 2335 y Fe(Description.)50 b Fl(Returns)32 b(information)f(regarding)i(the)g (corresp)s(ondence)g(of)g(a)h(v)-5 b(ariable)32 b(and)423 2448 y(the)e Fc(align-tar)-5 b(get)32 b Fl(\(arra)m(y)f(or)f (template\))i(to)f(whic)m(h)e(it)h(is)f(ultimately)g(aligned.)423 2662 y Fe(Class.)40 b Fl(Mapping)29 b(inquiry)f(subroutine.)423 2876 y Fe(Argumen)m(ts.)423 3064 y Fd(ALIGNEE)744 b Fl(ma)m(y)30 b(b)s(e)g(of)g(an)m(y)g(t)m(yp)s(e.)41 b(It)30 b(ma)m(y)h(b)s(e)e (scalar)h(or)g(arra)m(y)g(v)-5 b(alued.)40 b(It)1503 3177 y(m)m(ust)22 b(not)h(b)s(e)f(an)g(assumed-size)g(arra)m(y)-8 b(.)39 b(If)22 b(it)g(is)g(a)h(mem)m(b)s(er)e(of)i(an)1503 3290 y(aggregate)41 b(v)-5 b(ariable)36 b(group,)k(then)d(it)g(m)m(ust) h(b)s(e)f(an)h(aggregate)1503 3403 y(co)m(v)m(er)k(of)e(the)h(group.)70 b(\(See)41 b(Section)f Fe(??)h Fl(for)f(the)g(de\014nitions)1503 3516 y(of)29 b(\\aggregate)k(v)-5 b(ariable)28 b(group")h(and)g (\\aggregate)k(co)m(v)m(er."\))43 b(It)1503 3629 y(m)m(ust)28 b(not)h(b)s(e)f(a)h(p)s(oin)m(ter)e(that)i(is)f(disasso)s(ciated)g(or)g (an)h(allo)s(cat-)1503 3742 y(able)39 b(arra)m(y)i(that)f(is)f(not)h (allo)s(cated.)69 b(It)40 b(is)f(an)h Fd(INTENT)46 b(\(IN\))1503 3855 y Fl(argumen)m(t.)1503 4002 y(If)28 b Fd(ALIGNEE)e Fl(is)i(a)h(p)s(oin)m(ter,)f(information)f(ab)s(out)h(the)h(alignmen)m (t)1503 4114 y(of)46 b(its)g(target)i(is)d(returned.)88 b(The)46 b(target)i(m)m(ust)e(not)g(b)s(e)g(an)1503 4227 y(assumed-size)19 b(dumm)m(y)h(argumen)m(t)g(or)h(a)f(section)h(of)f (an)g(assumed-)1503 4340 y(size)38 b(dumm)m(y)f(argumen)m(t.)64 b(If)37 b(the)h(target)i(is)d(\(a)i(section)f(of)7 b(\))39 b(a)1503 4453 y(mem)m(b)s(er)c(of)i(an)f(aggregate)k(v)-5 b(ariable)35 b(group,)i(then)f(the)h(mem-)1503 4566 y(b)s(er)29 b(m)m(ust)h(b)s(e)g(an)g(aggregate)k(co)m(v)m(er)e(of)e(the)h(group.) 423 4730 y Fd(LB)e Fl(\(optional\))561 b(m)m(ust)31 b(b)s(e)g(of)h(t)m (yp)s(e)g(default)e(in)m(teger)i(and)f(of)h(rank)f(one.)45 b(Its)31 b(size)1503 4843 y(m)m(ust)36 b(b)s(e)h(at)g(least)g(equal)g (to)h(the)f(rank)f(of)h Fd(ALIGNEE)p Fl(.)e(It)i(is)f(an)1503 4956 y Fd(INTENT)46 b(\(OUT\))21 b Fl(argumen)m(t.)39 b(The)23 b(\014rst)f(elemen)m(t)h(of)h(the)f(i)3498 4923 y Fb(th)3595 4956 y Fl(axis)1503 5068 y(of)f Fd(ALIGNEE)d Fl(is)i(ultimately)f(aligned)h(to)h(the)g Fd(LB)p Fl(\(i\))3211 5035 y Fb(th)3303 5068 y Fc(align-tar)-5 b(get)1503 5181 y Fl(elemen)m(t)27 b(along)g(the)g(axis)f(of)h(the)g Fc(align-tar)-5 b(get)37 b Fl(asso)s(ciated)27 b(with)1503 5294 y(the)43 b(i)1698 5261 y Fb(th)1796 5294 y Fl(axis)f(of)i Fd(ALIGNEE)p Fl(.)d(If)i(the)g(i)2811 5261 y Fb(th)2909 5294 y Fl(axis)g(of)g Fd(ALIGNEE)e Fl(is)i(a)1503 5407 y(collapsed)29 b(axis,)h Fd(LB)p Fl(\(i\))g(is)f(implemen)m(tation)g (dep)s(enden)m(t.)p eop %%Page: 2 5 2 4 bop 150 -200 a Fl(2)3039 b Fg(CONTENTS)423 99 y Fd(UB)29 b Fl(\(optional\))561 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s(e)g(default) e(in)m(teger)i(and)f(of)h(rank)f(one.)45 b(Its)31 b(size)1503 211 y(m)m(ust)36 b(b)s(e)h(at)g(least)g(equal)g(to)h(the)f(rank)f(of)h Fd(ALIGNEE)p Fl(.)e(It)i(is)f(an)1503 324 y Fd(INTENT)46 b(\(OUT\))24 b Fl(argumen)m(t.)39 b(The)25 b(last)h(elemen)m(t)f(of)h (the)g(i)3498 291 y Fb(th)3595 324 y Fl(axis)1503 437 y(of)c Fd(ALIGNEE)d Fl(is)i(ultimately)f(aligned)h(to)h(the)g Fd(UB)p Fl(\(i\))3211 404 y Fb(th)3303 437 y Fc(align-tar)-5 b(get)1503 550 y Fl(elemen)m(t)27 b(along)g(the)g(axis)f(of)h(the)g Fc(align-tar)-5 b(get)37 b Fl(asso)s(ciated)27 b(with)1503 663 y(the)43 b(i)1698 630 y Fb(th)1796 663 y Fl(axis)f(of)i Fd(ALIGNEE)p Fl(.)d(If)i(the)g(i)2811 630 y Fb(th)2909 663 y Fl(axis)g(of)g Fd(ALIGNEE)e Fl(is)i(a)1503 776 y(collapsed)29 b(axis,)h Fd(UB)p Fl(\(i\))g(is)f(implemen)m(tation)g (dep)s(enden)m(t.)423 936 y Fd(STRIDE)f Fl(\(optional\))370 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s(e)g(default)e(in)m(teger)i(and)f (of)h(rank)f(one.)45 b(Its)31 b(size)1503 1049 y(m)m(ust)36 b(b)s(e)h(at)g(least)g(equal)g(to)h(the)f(rank)f(of)h Fd(ALIGNEE)p Fl(.)e(It)i(is)f(an)1503 1162 y Fd(INTENT)46 b(\(OUT\))38 b Fl(argumen)m(t.)69 b(The)39 b(i)2807 1129 y Fb(th)2905 1162 y Fl(elemen)m(t)h(of)g Fd(STRIDE)e Fl(is)1503 1275 y(set)27 b(to)g(the)g(stride)e(used)h(in)f(aligning)g (the)h(elemen)m(ts)h(of)g Fd(ALIGNEE)1503 1388 y Fl(along)h(its)g(i) 1894 1355 y Fb(th)1992 1388 y Fl(axis.)40 b(If)28 b(the)h(i)2482 1355 y Fb(th)2580 1388 y Fl(axis)f(of)h Fd(ALIGNEE)d Fl(is)i(a)h(collapsed)1503 1500 y(axis,)h Fd(STRIDE)p Fl(\(i\))e(is)i(zero.)423 1660 y Fd(AXIS)p 621 1660 29 4 v 33 w(MAP)f Fl(\(optional\))288 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s (e)g(default)e(in)m(teger)i(and)f(of)h(rank)f(one.)45 b(Its)31 b(size)1503 1773 y(m)m(ust)36 b(b)s(e)h(at)g(least)g(equal)g (to)h(the)f(rank)f(of)h Fd(ALIGNEE)p Fl(.)e(It)i(is)f(an)1503 1886 y Fd(INTENT)46 b(\(OUT\))27 b Fl(argumen)m(t.)41 b(The)28 b(i)2757 1853 y Fb(th)2855 1886 y Fl(elemen)m(t)i(of)f Fd(AXIS)p 3490 1886 V 33 w(MAP)f Fl(is)1503 1999 y(set)h(to)h(the)f Fc(align-tar)-5 b(get)39 b Fl(axis)29 b(asso)s(ciated)g(with)f(the)h(i) 3395 1966 y Fb(th)3493 1999 y Fl(axis)g(of)1503 2112 y Fd(ALIGNEE)p Fl(.)39 b(If)i(the)g(i)2198 2079 y Fb(th)2296 2112 y Fl(axis)g(of)g Fd(ALIGNEE)e Fl(is)h(a)i(collapsed)e(axis,)1503 2225 y Fd(AXIS)p 1701 2225 V 33 w(MAP\()p Fl(i)p Fd(\))28 b Fl(is)i(0.)423 2385 y Fd(IDENTITY)p 813 2385 V 32 w(MAP)f Fl(\(optional\))97 b(m)m(ust)23 b(b)s(e)g(scalar)g(and)g(of)h(t)m(yp)s (e)g(default)e(logical.)38 b(It)24 b(is)e(an)i Fd(INTENT)1503 2498 y(\(OUT\))35 b Fl(argumen)m(t.)60 b(It)36 b(is)g(set)h(to)g(true)g (if)e(the)i(ultimate)f Fc(align-)1503 2611 y(tar)-5 b(get)45 b Fl(asso)s(ciated)36 b(with)e Fd(ALIGNEE)f Fl(has)i(a)h(shap)s(e)e (iden)m(tical)g(to)1503 2724 y Fd(ALIGNEE)p Fl(,)43 b(the)h(axes)i(are) f(mapp)s(ed)e(using)g(the)i(iden)m(tit)m(y)f(p)s(er-)1503 2836 y(m)m(utation,)35 b(and)e(the)i(strides)d(are)j(all)e(p)s(ositiv)m (e)g(\(and)g(therefore)1503 2949 y(equal)28 b(to)i(1,)g(b)s(ecause)f (of)h(the)f(shap)s(e)f(constrain)m(t\);)j(otherwise)d(it)1503 3062 y(is)23 b(set)i(to)g(false.)39 b(If)23 b(a)i(v)-5 b(ariable)23 b(has)h(not)h(app)s(eared)f(as)g(an)h Fc(aligne)-5 b(e)1503 3175 y Fl(in)28 b(an)i Fd(ALIGN)f Fl(or)h Fd(REALIGN)e Fl(directiv)m(e,)i(and)f(do)s(es)h(not)g(ha)m(v)m(e)i(the)1503 3288 y Fd(INHERIT)j Fl(attribute,)j(then)e Fd(IDENTITY)p 2894 3288 V 33 w(MAP)f Fl(m)m(ust)i(b)s(e)f(true;)k(it)1503 3401 y(can)30 b(b)s(e)g(true)g(in)f(other)i(circumstances)f(as)h(w)m (ell.)423 3561 y Fd(DYNAMIC)d Fl(\(optional\))322 b(m)m(ust)23 b(b)s(e)g(scalar)g(and)g(of)h(t)m(yp)s(e)g(default)e(logical.)38 b(It)24 b(is)e(an)i Fd(INTENT)1503 3674 y(\(OUT\))42 b Fl(argumen)m(t.)83 b(It)44 b(is)f(set)i(to)g(true)f(if)f Fd(ALIGNEE)f Fl(has)i(the)1503 3787 y Fd(DYNAMIC)24 b Fl(attribute;)k(otherwise)e(it)f(is)h(set)h(to)g(false.)39 b(If)26 b Fd(ALIGNEE)1503 3900 y Fl(has)20 b(the)g(p)s(oin)m(ter)f (attribute,)k(then)d(the)g(result)f(applies)g(to)i Fd(ALIGN)p Fl(-)1503 4013 y Fd(EE)29 b Fl(itself)h(rather)g(than)g(its)g(target.) 423 4173 y Fd(NCOPIES)e Fl(\(optional\))322 b(m)m(ust)21 b(b)s(e)g(scalar)h(and)f(of)h(t)m(yp)s(e)g(default)f(in)m(teger.)38 b(It)22 b(is)f(an)h Fd(INTENT)1503 4285 y(\(OUT\))42 b Fl(argumen)m(t.)81 b(It)44 b(is)f(set)h(to)h(the)f(n)m(um)m(b)s(er)e (of)i(copies)g(of)1503 4398 y Fd(ALIGNEE)30 b Fl(that)j(are)g (ultimately)e(aligned)g(to)i Fc(align-tar)-5 b(get)9 b Fl(.)48 b(F)-8 b(or)1503 4511 y(a)30 b(non-replicated)g(v)-5 b(ariable,)29 b(it)h(is)f(set)i(to)g(one.)423 4715 y Fe(Examples.)50 b Fl(If)34 b Fd(ALIGNEE)e Fl(is)h(scalar,)j(then)e(no)g (elemen)m(ts)g(of)g Fd(LB)p Fl(,)g Fd(UB)p Fl(,)g Fd(STRIDE)p Fl(,)e(or)j Fd(AXIS)p 3580 4715 V 33 w(MAP)423 4828 y Fl(are)30 b(set.)423 4977 y(Giv)m(en)g(the)h(declarations)527 5181 y Fd(REAL)47 b(PI)g(=)h(3.1415927)527 5294 y(POINTER)e (P_TO_A\(:\))527 5407 y(DIMENSION)g(A\(10,10\),B\(20,30\),C\(20,)o(40,) o(10\),)o(D\(40)o(\))p eop %%Page: 3 6 3 5 bop 150 -200 a Fg(0.1.)72 b(MAPPING)31 b(INQUIR)-8 b(Y)31 b(PR)m(OCEDURES)1792 b Fl(3)241 99 y Fd(!HPF$)46 b(TEMPLATE)g(T\(40,20\))241 211 y(!HPF$)g(DYNAMIC)g(A)241 324 y(!HPF$)g(ALIGN)h(A\(I,:\))f(WITH)g(T\(1+3*I,2:20:2\))241 437 y(!HPF$)g(ALIGN)h(C\(I,*,J\))e(WITH)i(T\(J,21-I\))241 550 y(!HPF$)f(ALIGN)h(D\(I\))f(WITH)h(T\(I,4\))241 663 y(!HPF$)f(PROCESSORS)f(PROCS\(4,2\),)g(SCALARPROC)241 776 y(!HPF$)h(DISTRIBUTE)f(T\(BLOCK,BLOCK\))f(ONTO)j(PROCS)241 889 y(!HPF$)f(DISTRIBUTE)f(B\(CYCLIC,BLOCK\))f(ONTO)j(PROCS)241 1002 y(!HPF$)f(DISTRIBUTE)f(ONTO)i(SCALARPROC)e(::)i(PI)527 1115 y(P_TO_A)f(=>)i(A\(3:9:2,)d(6\))150 1330 y Fl(assuming)19 b(that)i(the)f(actual)h(mappings)d(are)j(as)f(the)h(directiv)m(es)f(sp) s(ecify)-8 b(,)21 b(the)g(results)e(of)h Fd(HPF)p 3294 1330 29 4 v 34 w(ALIGNMENT)150 1443 y Fl(are:)p 599 1532 2702 4 v 597 1631 4 100 v 1346 1631 V 1498 1601 a Fa(A)p 1709 1631 V 303 w(B)p 2072 1631 V 404 w(C)p 2637 1631 V 351 w(D)p 2896 1631 V 148 w(P)p 3010 1601 25 4 v 29 w(TO)p 3164 1601 V 30 w(A)p 3299 1631 4 100 v 599 1635 2702 4 v 599 1651 V 597 1751 4 100 v 649 1721 a(LB)p 1346 1751 V 679 w([4,)28 b(2])p 1709 1751 V 182 w([1,)f(1])p 2072 1751 V 155 w([20,)g(N/A,)g(1])p 2637 1751 V 149 w([1])p 2896 1751 V 223 w([10])p 3299 1751 V 599 1754 2702 4 v 597 1854 4 100 v 649 1824 a(UB)p 1346 1854 V 628 w([31,)g(20])p 1709 1854 V 98 w([20,)g(30])p 2072 1854 V 99 w([)h(1,)f(N/A,)h(10])p 2637 1854 V 114 w([40])p 2896 1854 V 201 w([28])p 3299 1854 V 599 1857 2702 4 v 597 1957 4 100 v 649 1927 a(STRIDE)p 1346 1957 V 473 w([3,)g(2])p 1709 1957 V 182 w([1,)f(1])p 2072 1957 V 224 w([-1,)g(0,)g(1])p 2637 1957 V 218 w([1])p 2896 1957 V 230 w([)g(6])p 3299 1957 V 599 1960 2702 4 v 597 2060 4 100 v 649 2030 a(AXIS)p 854 2030 25 4 v 31 w(MAP)p 1346 2060 4 100 v 364 w([1,)h(2])p 1709 2060 V 182 w([1,)f(2])p 2072 2060 V 238 w([2,)g(0,)g(1])p 2637 2060 V 232 w([1])p 2896 2060 V 230 w([)g(1])p 3299 2060 V 599 2063 2702 4 v 597 2163 4 100 v 649 2133 a(IDENTITY)p 1078 2133 25 4 v 30 w(MAP)p 1346 2163 4 100 v 152 w(false)p 1709 2163 V 208 w(true)p 2072 2163 V 310 w(false)p 2637 2163 V 252 w(false)p 2896 2163 V 171 w(false)p 3299 2163 V 599 2166 2702 4 v 597 2266 4 100 v 649 2236 a(D)n(YNAMIC)p 1346 2266 V 393 w(true)p 1709 2266 V 209 w(false)p 2072 2266 V 304 w(false)p 2637 2266 V 252 w(false)p 2896 2266 V 171 w(false)p 3299 2266 V 599 2269 2702 4 v 597 2368 4 100 v 649 2339 a(NCOPIES)p 1346 2368 V 483 w(1)p 1709 2368 V 320 w(1)p 2072 2368 V 422 w(1)p 2637 2368 V 371 w(1)p 2896 2368 V 288 w(1)p 3299 2368 V 599 2372 2702 4 v 150 2525 a Fl(where)f(\\N/A")i(denotes)e(a)h(implemen)m(tation-dep) s(enden)m(t)d(result.)39 b(T)-8 b(o)26 b(illustrate)f(the)h(use)g(of)h Fd(NCOPIES)p Fl(,)150 2638 y(consider:)527 2828 y Fd(LOGICAL)46 b(BOZO\(20,20\),RONALD_MCDONA)o(LD\()o(20\))241 2941 y(!HPF$)g(TEMPLATE)g(EMMETT_KELLY\(100,100\))241 3054 y(!HPF$)g(ALIGN)h(RONALD_MCDONALD\(I\))42 b(WITH)47 b(BOZO\(I,*\))241 3167 y(!HPF$)f(ALIGN)h(BOZO\(J,K\))e(WITH)i(EMMETT_KELLY\(J,5*K\))150 3357 y(CALL)g(HPF)p 539 3357 29 4 v 33 w(ALIGNMENT\(RONALD)p 1340 3357 V 30 w(MCDONALD,)e(NCOPIES)h(=)i(NC\))29 b Fl(sets)i Fd(NC)f Fl(to)h(20.)41 b(No)m(w)31 b(consider:)527 3547 y Fd(LOGICAL)46 b(BOZO\(20,20\),RONALD_MCDONA)o(LD\()o(20\))241 3660 y(!HPF$)g(TEMPLATE)g(WILLIE_WHISTLE\(100\))241 3773 y(!HPF$)g(ALIGN)h(RONALD_MCDONALD\(I\))42 b(WITH)47 b(BOZO\(I,*\))241 3885 y(!HPF$)f(ALIGN)h(BOZO\(J,*\))e(WITH)i(WILLIE_WHISTLE\(5*J\))150 4075 y(CALL)g(HPF)p 539 4075 V 33 w(ALIGNMENT\(RONALD)p 1340 4075 V 30 w(MCDONALD,)e(NCOPIES)h(=)i(NC\))29 b Fl(sets)i Fd(NC)f Fl(to)h(one.)150 4322 y Ff(0.1.3)97 b(HPF)p 639 4322 30 4 v 36 w(TEMPLA)-8 b(TE\(ALIGNEE,)43 b(TEMPLA)-8 b(TE)p 2188 4322 V 34 w(RANK,)44 b(LB,)f(UB,)h(AXIS)p 3154 4322 V 36 w(TYPE,)f(AXIS)p 3720 4322 V 448 4438 a(INF)m(O,)33 b(NUMBER)p 1138 4438 V 35 w(ALIGNED,)g(D)m(YNAMIC\))423 4610 y Fe(Optional)h(Argumen)m(ts.)40 b Fd(LB)p Fl(,)30 b Fd(UB)p Fl(,)g Fd(AXIS)p 1928 4610 29 4 v 33 w(TYPE)p Fl(,)g Fd(AXIS)p 2400 4610 V 33 w(INFO)p Fl(,)g Fd(NUMBER)p 2968 4610 V 32 w(ALIGNED)p Fl(,)423 4723 y Fd(TEMPLATE)p 813 4723 V 32 w(RANK)p Fl(,)f Fd(DYNAMIC)423 4938 y Fe(Description.)51 b Fl(The)32 b Fd(HPF)p 1367 4938 V 34 w(TEMPLATE)f Fl(subroutine)h (returns)g(information)g(regarding)g(the)i(ul-)423 5051 y(timate)43 b Fc(align-tar)-5 b(get)52 b Fl(asso)s(ciated)44 b(with)d(a)i(v)-5 b(ariable;)48 b Fd(HPF)p 2509 5051 V 34 w(TEMPLATE)40 b Fl(returns)i(information)423 5164 y(concerning)33 b(the)g(v)-5 b(ariable)33 b(from)g(the)g(template's)h (p)s(oin)m(t)f(of)h(view)e(\(assuming)h(the)h(alignmen)m(t)423 5277 y(is)29 b(to)i(a)f(template)h(rather)f(than)g(to)h(an)f(arra)m (y\),)h(while)d Fd(HPF)p 2486 5277 V 34 w(ALIGNMENT)g Fl(returns)h(information)423 5390 y(from)h(the)g(v)-5 b(ariable's)29 b(p)s(oin)m(t)h(of)g(view.)p eop %%Page: 4 7 4 6 bop 150 -200 a Fl(4)3039 b Fg(CONTENTS)423 99 y Fe(Class.)40 b Fl(Mapping)29 b(inquiry)f(subroutine.)423 340 y Fe(Argumen)m(ts.)423 557 y Fd(ALIGNEE)744 b Fl(ma)m(y)30 b(b)s(e)g(of)g(an)m(y)g(t)m(yp)s (e.)41 b(It)30 b(ma)m(y)h(b)s(e)e(scalar)h(or)g(arra)m(y)g(v)-5 b(alued.)40 b(It)1503 670 y(m)m(ust)22 b(not)h(b)s(e)f(an)g (assumed-size)g(arra)m(y)-8 b(.)39 b(If)22 b(it)g(is)g(a)h(mem)m(b)s (er)e(of)i(an)1503 783 y(aggregate)41 b(v)-5 b(ariable)36 b(group,)k(then)d(it)g(m)m(ust)h(b)s(e)f(an)h(aggregate)1503 895 y(co)m(v)m(er)k(of)e(the)h(group.)70 b(\(See)41 b(Section)f Fe(??)h Fl(for)f(the)g(de\014nitions)1503 1008 y(of)29 b(\\aggregate)k(v)-5 b(ariable)28 b(group")h(and)g(\\aggregate)k(co)m (v)m(er."\))43 b(It)1503 1121 y(m)m(ust)28 b(not)h(b)s(e)f(a)h(p)s(oin) m(ter)e(that)i(is)f(disasso)s(ciated)g(or)g(an)h(allo)s(cat-)1503 1234 y(able)39 b(arra)m(y)i(that)f(is)f(not)h(allo)s(cated.)69 b(It)40 b(is)f(an)h Fd(INTENT)46 b(\(IN\))1503 1347 y Fl(argumen)m(t.)1503 1508 y(If)28 b Fd(ALIGNEE)e Fl(is)i(a)h(p)s(oin)m (ter,)f(information)f(ab)s(out)h(the)h(alignmen)m(t)1503 1621 y(of)46 b(its)g(target)i(is)d(returned.)88 b(The)46 b(target)i(m)m(ust)e(not)g(b)s(e)g(an)1503 1734 y(assumed-size)19 b(dumm)m(y)h(argumen)m(t)g(or)h(a)f(section)h(of)f(an)g(assumed-)1503 1846 y(size)38 b(dumm)m(y)f(argumen)m(t.)64 b(If)37 b(the)h(target)i (is)d(\(a)i(section)f(of)7 b(\))39 b(a)1503 1959 y(mem)m(b)s(er)c(of)i (an)f(aggregate)k(v)-5 b(ariable)35 b(group,)i(then)f(the)h(mem-)1503 2072 y(b)s(er)29 b(m)m(ust)h(b)s(e)g(an)g(aggregate)k(co)m(v)m(er)e(of) e(the)h(group.)423 2257 y Fd(TEMPLATE)p 813 2257 29 4 v 32 w(RANK)e Fl(\(optional\))49 b(m)m(ust)21 b(b)s(e)g(scalar)h(and)f (of)h(t)m(yp)s(e)g(default)f(in)m(teger.)38 b(It)22 b(is)f(an)h Fd(INTENT)1503 2370 y(\(OUT\))40 b Fl(argumen)m(t.)74 b(It)42 b(is)e(set)i(to)h(the)e(rank)g(of)h(the)g(ultimate)1503 2483 y Fc(align-tar)-5 b(get)9 b Fl(.)57 b(This)34 b(can)i(b)s(e)f (di\013eren)m(t)g(from)g(the)h(rank)f(of)h(the)1503 2596 y Fd(ALIGNEE)p Fl(,)28 b(due)i(to)h(collapsing)e(and)h(replicating.)423 2780 y Fd(LB)f Fl(\(optional\))561 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s (e)g(default)e(in)m(teger)i(and)f(of)h(rank)f(one.)45 b(Its)31 b(size)1503 2893 y(m)m(ust)i(b)s(e)f(at)i(least)f(equal)g(to)h (the)f(rank)g(of)g(the)g Fc(align-tar)-5 b(get)43 b Fl(to)1503 3006 y(whic)m(h)h Fd(ALIGNEE)f Fl(is)h(ultimately)g(aligned;)52 b(this)44 b(is)g(the)h(v)-5 b(alue)1503 3119 y(returned)26 b(in)h Fd(TEMPLATE)p 2363 3119 V 32 w(RANK)p Fl(.)g(It)h(is)f(an)h Fd(INTENT)46 b(\(OUT\))27 b Fl(argu-)1503 3232 y(men)m(t.)44 b(The)31 b(i)1985 3199 y Fb(th)2083 3232 y Fl(elemen)m(t)h(of)g Fd(LB)f Fl(con)m(tains)g(the)h(declared)f Fc(align-)1503 3345 y(tar)-5 b(get)40 b Fl(lo)m(w)m(er)30 b(b)s(ound)f(for)h(the)g(i) 2605 3312 y Fb(th)2703 3345 y Fl(template)h(axis.)423 3529 y Fd(UB)e Fl(\(optional\))561 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s (e)g(default)e(in)m(teger)i(and)f(of)h(rank)f(one.)45 b(Its)31 b(size)1503 3642 y(m)m(ust)i(b)s(e)f(at)i(least)f(equal)g(to)h (the)f(rank)g(of)g(the)g Fc(align-tar)-5 b(get)43 b Fl(to)1503 3755 y(whic)m(h)h Fd(ALIGNEE)f Fl(is)h(ultimately)g(aligned;)52 b(this)44 b(is)g(the)h(v)-5 b(alue)1503 3868 y(returned)26 b(in)h Fd(TEMPLATE)p 2363 3868 V 32 w(RANK)p Fl(.)g(It)h(is)f(an)h Fd(INTENT)46 b(\(OUT\))27 b Fl(argu-)1503 3981 y(men)m(t.)44 b(The)31 b(i)1985 3948 y Fb(th)2083 3981 y Fl(elemen)m(t)h(of)g Fd(UB)f Fl(con)m(tains)g(the)h(declared)f Fc(align-)1503 4094 y(tar)-5 b(get)40 b Fl(upp)s(er)28 b(b)s(ound)h(for)h(the)g(i)2628 4061 y Fb(th)2726 4094 y Fl(template)h(axis.)423 4278 y Fd(AXIS)p 621 4278 V 33 w(TYPE)e Fl(\(optional\))240 b(m)m(ust)38 b(b)s(e)f(a)i(rank)f(one)g(arra)m(y)h(of)f(t)m(yp)s(e)h (default)e(c)m(haracter.)66 b(It)1503 4391 y(ma)m(y)48 b(b)s(e)f(of)h(an)m(y)g(length,)k(although)47 b(it)h(m)m(ust)f(b)s(e)h (of)g(length)1503 4504 y(at)40 b(least)g(10)h(in)d(order)h(to)i(con)m (tain)f(the)g(complete)g(v)-5 b(alue.)68 b(Its)1503 4617 y(elemen)m(ts)48 b(are)g(set)g(to)g(the)g(v)-5 b(alues)47 b(b)s(elo)m(w)f(as)i(if)f(b)m(y)g(a)h(c)m(har-)1503 4730 y(acter)c(in)m(trinsic)d(assignmen)m(t)i(statemen)m(t.)82 b(Its)43 b(size)g(m)m(ust)h(b)s(e)1503 4843 y(at)f(least)f(equal)g(to)h (the)g(rank)e(of)i(the)f Fc(align-tar)-5 b(get)52 b Fl(to)43 b(whic)m(h)1503 4956 y Fd(ALIGNEE)30 b Fl(is)i(ultimately)e(aligned;)j (this)e(is)h(the)g(v)-5 b(alue)32 b(returned)1503 5068 y(in)23 b Fd(TEMPLATE)p 1993 5068 V 32 w(RANK)p Fl(.)g(It)h(is)f(an)i Fd(INTENT)46 b(\(OUT\))22 b Fl(argumen)m(t.)39 b(The)1503 5181 y(i)1529 5148 y Fb(th)1626 5181 y Fl(elemen)m(t)32 b(of)f Fd(AXIS)p 2265 5181 V 33 w(TYPE)e Fl(con)m(tains)i(information)e (ab)s(out)i(the)1503 5294 y(i)1529 5261 y Fb(th)1626 5294 y Fl(axis)k(of)f(the)h Fc(align-tar)-5 b(get)9 b Fl(.)55 b(The)34 b(follo)m(wing)f(v)-5 b(alues)34 b(are)h(de-)1503 5407 y(\014ned)23 b(b)m(y)i(HPF)g(\(implemen)m(tations)f(ma)m(y)h (de\014ne)f(other)h(v)-5 b(alues\):)p eop %%Page: 5 8 5 7 bop 150 -200 a Fg(0.1.)72 b(MAPPING)31 b(INQUIR)-8 b(Y)31 b(PR)m(OCEDURES)1792 b Fl(5)1503 99 y Fd('NORMAL')43 b Fl(The)37 b Fc(align-tar)-5 b(get)48 b Fl(axis)37 b(has)g(an)h(axis)f (of)h Fd(ALIGNEE)1673 211 y Fl(aligned)j(to)j(it.)78 b(F)-8 b(or)43 b(elemen)m(ts)h(of)f Fd(AXIS)p 3160 211 29 4 v 33 w(TYPE)f Fl(assigned)1673 324 y(this)36 b(v)-5 b(alue,)39 b(the)e(corresp)s(onding)f(elemen)m(t)h(of)h Fd(AXIS)p 3532 324 V 33 w(INFO)1673 437 y Fl(is)c(set)i(to)g(the)g(n)m (um)m(b)s(er)e(of)h(the)h(axis)e(of)i Fd(ALIGNEE)d Fl(aligned)1673 550 y(to)e(this)e Fc(align-tar)-5 b(get)40 b Fl(axis.)1503 678 y Fd('REPLICATED')i(ALIGNEE)21 b Fl(is)h(replicated)g(along)h(this) f Fc(align-tar-)1673 791 y(get)51 b Fl(axis.)78 b(F)-8 b(or)44 b(elemen)m(ts)f(of)g Fd(AXIS)p 2970 791 V 33 w(TYPE)f Fl(assigned)g(this)1673 904 y(v)-5 b(alue,)28 b(the)h(corresp)s(onding)e(elemen)m(t)i(of)f Fd(AXIS)p 3302 904 V 33 w(INFO)g Fl(is)f(set)1673 1017 y(to)j(the)h(n)m(um)m(b)s (er)d(of)i(copies)g(of)g Fd(ALIGNEE)e Fl(along)i(this)f Fc(align-)1673 1130 y(tar)-5 b(get)40 b Fl(axis.)1503 1258 y Fd('SINGLE')j(ALIGNEE)28 b Fl(is)i(aligned)f(with)g(one)i(co)s (ordinate)f(of)h(the)1673 1371 y Fc(align-tar)-5 b(get)32 b Fl(axis.)37 b(F)-8 b(or)23 b(elemen)m(ts)g(of)f Fd(AXIS)p 3180 1371 V 34 w(TYPE)f Fl(assigned)1673 1484 y(this)h(v)-5 b(alue,)24 b(the)g(corresp)s(onding)d(elemen)m(t)j(of)f Fd(AXIS)p 3447 1484 V 34 w(INFO)f Fl(is)1673 1597 y(set)35 b(to)g(the)g Fc(align-tar)-5 b(get)44 b Fl(co)s(ordinate)34 b(to)i(whic)m(h)d Fd(ALIGNEE)1673 1710 y Fl(is)c(aligned.)423 1876 y Fd(AXIS)p 621 1876 V 33 w(INFO)g Fl(\(optional\))240 b(m)m(ust)31 b(b)s(e)g(of)h(t)m(yp)s(e)g(default)e(in)m(teger)i(and)f (of)h(rank)f(one.)45 b(Its)31 b(size)1503 1989 y(m)m(ust)i(b)s(e)f(at)i (least)f(equal)g(to)h(the)f(rank)g(of)g(the)g Fc(align-tar)-5 b(get)43 b Fl(to)1503 2102 y(whic)m(h)h Fd(ALIGNEE)f Fl(is)h(ultimately)g(aligned;)52 b(this)44 b(is)g(the)h(v)-5 b(alue)1503 2215 y(returned)26 b(in)h Fd(TEMPLATE)p 2363 2215 V 32 w(RANK)p Fl(.)g(It)h(is)f(an)h Fd(INTENT)46 b(\(OUT\))27 b Fl(argu-)1503 2328 y(men)m(t.)41 b(See)30 b(the)h(description)d(of)j Fd(AXIS)p 2853 2328 V 33 w(TYPE)e Fl(ab)s(o)m(v)m(e.)423 2488 y Fd(NUMBER)p 717 2488 V 32 w(ALIGNED)g Fl(\(optional\))45 b(m)m(ust)h(b)s(e)f(scalar)g(and)g (of)h(t)m(yp)s(e)g(default)f(in)m(teger.)87 b(It)45 b(is)g(an)1503 2601 y Fd(INTENT)h(\(OUT\))37 b Fl(argumen)m(t.)67 b(It)39 b(is)f(set)h(to)h(the)f(total)g(n)m(um)m(b)s(er)1503 2714 y(of)33 b(v)-5 b(ariables)33 b(aligned)f(to)i(the)g(ultimate)f Fc(align-tar)-5 b(get)9 b Fl(.)51 b(This)32 b(is)1503 2827 y(the)26 b(n)m(um)m(b)s(er)e(of)i(v)-5 b(ariables)24 b(that)i(are)h(mo)m(v)m(ed)f(if)f(the)h Fc(align-tar)-5 b(get)1503 2940 y Fl(is)29 b(redistributed.)423 3099 y Fd(DYNAMIC)f Fl(\(optional\))322 b(m)m(ust)23 b(b)s(e)g(scalar)g(and) g(of)h(t)m(yp)s(e)g(default)e(logical.)38 b(It)24 b(is)e(an)i Fd(INTENT)1503 3212 y(\(OUT\))32 b Fl(argumen)m(t.)52 b(It)34 b(is)f(set)i(to)g(true)e(if)g(the)h Fc(align-tar)-5 b(get)44 b Fl(has)1503 3325 y(the)30 b Fd(DYNAMIC)f Fl(attribute,)h (and)g(to)h(false)f(otherwise.)423 3529 y Fe(Example.)64 b Fl(Giv)m(en)39 b(the)g(declarations)f(in)f(the)i(example)g(of)g (Section)f(0.1.2,)43 b(and)38 b(assuming)423 3642 y(that)c(the)g (actual)g(mappings)d(are)j(as)g(the)g(directiv)m(es)f(sp)s(ecify)-8 b(,)34 b(the)g(results)e(of)i Fd(HPF)p 3341 3642 V 33 w(TEMPLATE)423 3755 y Fl(are:)p 646 3858 2609 4 v 644 3957 4 100 v 1542 3957 V 1798 3927 a Fa(A)p 2112 3957 V 509 w(C)p 2682 3957 V 508 w(D)p 3252 3957 V 646 3961 2609 4 v 646 3977 V 644 4077 4 100 v 696 4047 a(LB)p 1542 4077 V 932 w([1,)27 b(1])p 2112 4077 V 390 w([1,)g(1])p 2682 4077 V 390 w([1,)g(1])p 3252 4077 V 646 4080 2609 4 v 644 4180 4 100 v 696 4150 a(UB)p 1542 4180 V 880 w([40,)g(20])p 2112 4180 V 307 w([40,)f(20])p 2682 4180 V 307 w([40,)g(20])p 3252 4180 V 646 4183 2609 4 v 644 4283 4 100 v 696 4253 a(AXIS)p 901 4253 25 4 v 30 w(TYPE)p 1542 4283 4 100 v 432 w(['NORMAL',)p 2112 4283 V 100 w(['NORMAL',)p 2682 4283 V 100 w(['NORMAL',)p 3252 4283 V 644 4382 V 1542 4382 V 1605 4353 a('NORMAL'])p 2112 4382 V 123 w('NORMAL'])p 2682 4382 V 157 w('SINGLE'])p 3252 4382 V 646 4386 2609 4 v 644 4485 4 100 v 696 4455 a(AXIS)p 901 4455 25 4 v 30 w(INF)n(O)p 1542 4485 4 100 v 604 w([1,)h(2])p 2112 4485 V 390 w([3,)g(1])p 2682 4485 V 390 w([1,)g(4])p 3252 4485 V 646 4489 2609 4 v 644 4588 4 100 v 696 4558 a(NUMBER)p 1078 4558 25 4 v 30 w(ALIGNED)p 1542 4588 4 100 v 314 w(3)p 2112 4588 V 528 w(3)p 2682 4588 V 528 w(3)p 3252 4588 V 646 4592 2609 4 v 644 4691 4 100 v 696 4661 a(TEMPLA)-7 b(TE)p 1175 4661 25 4 v 28 w(RANK)p 1542 4691 4 100 v 360 w(2)p 2112 4691 V 528 w(2)p 2682 4691 V 528 w(2)p 3252 4691 V 646 4695 2609 4 v 644 4794 4 100 v 696 4764 a(D)n(YNAMIC)p 1542 4794 V 640 w(false)p 2112 4794 V 410 w(false)p 2682 4794 V 410 w(false)p 3252 4794 V 646 4798 2609 4 v 150 5006 a Ff(0.1.4)97 b(HPF)p 639 5006 30 4 v 36 w (DISTRIBUTION\(DISTRIBUTEE,)19 b(AXIS)p 2263 5006 V 35 w(TYPE,)j(AXIS)p 2807 5006 V 36 w(INF)m(O,)g(PROCESSORS)p 3725 5006 V 448 5123 a(RANK,)33 b(PROCESSORS)p 1390 5123 V 36 w(SHAPE,)h(PLB,)e(PUB,)h(PSTRIDE\))423 5294 y Fe(Optional)h (Argumen)m(ts.)40 b Fd(AXIS)p 1626 5294 29 4 v 33 w(TYPE)p Fl(,)30 b Fd(AXIS)p 2098 5294 V 33 w(INFO)p Fl(,)g Fd(PROCESSORS)p 2858 5294 V 31 w(RANK)p Fl(,)423 5407 y Fd(PROCESSORS)p 909 5407 V 31 w(SHAPE)p Fl(,)g Fd(PLB)p Fl(,)f Fd(PUB)p Fl(,)h Fd(PSTRIDE)p Fl(,)p eop %%Page: 6 9 6 8 bop 150 -200 a Fl(6)3039 b Fg(CONTENTS)423 99 y Fe(Description.)40 b Fl(The)26 b Fd(HPF)p 1350 99 29 4 v 34 w(DISTRIBUTION)d Fl(subroutine)i(returns)g(information)g(regarding)h(the)423 211 y(distribution)g(of)31 b(the)g(ultimate)e Fc(align-tar)-5 b(get)40 b Fl(asso)s(ciated)31 b(with)e(a)i(v)-5 b(ariable.)423 435 y Fe(Class.)40 b Fl(Mapping)29 b(inquiry)f(subroutine.)423 659 y Fe(Argumen)m(ts.)423 858 y Fd(DISTRIBUTEE)552 b Fl(ma)m(y)30 b(b)s(e)g(of)g(an)m(y)g(t)m(yp)s(e.)41 b(It)30 b(ma)m(y)h(b)s(e)e(scalar)h(or)g(arra)m(y)g(v)-5 b(alued.)40 b(It)1503 971 y(m)m(ust)22 b(not)h(b)s(e)f(an)g(assumed-size)g(arra)m (y)-8 b(.)39 b(If)22 b(it)g(is)g(a)h(mem)m(b)s(er)e(of)i(an)1503 1084 y(aggregate)41 b(v)-5 b(ariable)36 b(group,)k(then)d(it)g(m)m(ust) h(b)s(e)f(an)h(aggregate)1503 1196 y(co)m(v)m(er)k(of)e(the)h(group.)70 b(\(See)41 b(Section)f Fe(??)h Fl(for)f(the)g(de\014nitions)1503 1309 y(of)29 b(\\aggregate)k(v)-5 b(ariable)28 b(group")h(and)g (\\aggregate)k(co)m(v)m(er."\))43 b(It)1503 1422 y(m)m(ust)28 b(not)h(b)s(e)f(a)h(p)s(oin)m(ter)e(that)i(is)f(disasso)s(ciated)g(or)g (an)h(allo)s(cat-)1503 1535 y(able)39 b(arra)m(y)i(that)f(is)f(not)h (allo)s(cated.)69 b(It)40 b(is)f(an)h Fd(INTENT)46 b(\(IN\))1503 1648 y Fl(argumen)m(t.)1503 1800 y(If)35 b Fd(DISTRIBUTEE)e Fl(is)h(a)i(p)s(oin)m(ter,)h(information)d(ab)s(out)h(the)h(dis-)1503 1913 y(tribution)d(of)j(its)f(target)i(is)e(returned.)55 b(The)35 b(target)i(m)m(ust)f(not)1503 2026 y(b)s(e)d(an)i (assumed-size)e(dumm)m(y)h(argumen)m(t)g(or)g(a)h(section)g(of)f(an) 1503 2139 y(assumed-size)j(dumm)m(y)g(argumen)m(t.)64 b(If)37 b(the)h(target)h(is)e(\(a)i(sec-)1503 2251 y(tion)c(of)7 b(\))36 b(a)g(mem)m(b)s(er)g(of)g(an)f(aggregate)k(v)-5 b(ariable)34 b(group,)j(then)1503 2364 y(the)30 b(mem)m(b)s(er)g(m)m (ust)g(b)s(e)g(an)g(aggregate)k(co)m(v)m(er)e(of)e(the)h(group.)423 2536 y Fd(AXIS)p 621 2536 V 33 w(TYPE)e Fl(\(optional\))240 b(m)m(ust)38 b(b)s(e)f(a)i(rank)f(one)g(arra)m(y)h(of)f(t)m(yp)s(e)h (default)e(c)m(haracter.)66 b(It)1503 2648 y(ma)m(y)48 b(b)s(e)f(of)h(an)m(y)g(length,)k(although)47 b(it)h(m)m(ust)f(b)s(e)h (of)g(length)1503 2761 y(at)c(least)g(9)g(in)e(order)h(to)h(con)m(tain) g(the)f(complete)h(v)-5 b(alue.)80 b(Its)1503 2874 y(elemen)m(ts)48 b(are)g(set)g(to)g(the)g(v)-5 b(alues)47 b(b)s(elo)m(w)f(as)i(if)f(b)m (y)g(a)h(c)m(har-)1503 2987 y(acter)c(in)m(trinsic)d(assignmen)m(t)i (statemen)m(t.)82 b(Its)43 b(size)g(m)m(ust)h(b)s(e)1503 3100 y(at)f(least)f(equal)g(to)h(the)g(rank)e(of)i(the)f Fc(align-tar)-5 b(get)52 b Fl(to)43 b(whic)m(h)1503 3213 y Fd(DISTRIBUTEE)34 b Fl(is)i(ultimately)g(aligned;)k(this)c(is)g(the)i (v)-5 b(alue)37 b(re-)1503 3326 y(turned)29 b(b)m(y)h Fd(HPF)p 2072 3326 V 34 w(TEMPLATE)e Fl(in)h Fd(TEMPLATE)p 3008 3326 V 32 w(RANK)p Fl(\).)h(It)g(is)g(an)1503 3439 y Fd(INTENT)46 b(\(OUT\))34 b Fl(argumen)m(t.)57 b(Its)35 b(i)2734 3406 y Fb(th)2832 3439 y Fl(elemen)m(t)h(con)m(tains)g(infor-) 1503 3552 y(mation)h(on)h(the)g(distribution)d(of)j(the)g(i)2921 3519 y Fb(th)3019 3552 y Fl(axis)f(of)h(that)h Fc(align-)1503 3665 y(tar)-5 b(get)9 b Fl(.)45 b(The)32 b(follo)m(wing)e(v)-5 b(alues)31 b(are)h(de\014ned)e(b)m(y)i(HPF)g(\(imple-)1503 3778 y(men)m(tations)e(ma)m(y)h(de\014ne)f(other)h(v)-5 b(alues\):)1503 3957 y Fd('BLOCK')43 b Fl(The)29 b(axis)g(is)g (distributed)d Fd(BLOCK)p Fl(.)j(The)g(corresp)s(ond-)1673 4070 y(ing)g(elemen)m(t)i(of)g Fd(AXIS)p 2462 4070 V 33 w(INFO)e Fl(con)m(tains)i(the)f(blo)s(c)m(k)g(size.)1503 4202 y Fd('GEN)p 1701 4202 V 33 w(BLOCK')44 b Fl(The)22 b(axis)g(is)g(distributed)e Fd(BLOCK\(array\))p Fl(.)34 b(The)1673 4315 y(v)-5 b(alue)37 b(of)h(the)g(corresp)s(onding)d (elemen)m(t)j(of)g Fd(AXIS)p 3433 4315 V 33 w(INFO)f Fl(is)1673 4428 y(implemen)m(tation)29 b(dep)s(enden)m(t.)1503 4560 y Fd('COLLAPSED')42 b Fl(The)28 b(axis)h(is)e(collapsed)h (\(distributed)e(with)i(the)1673 4673 y(\\)p Fd(*)p Fl(")41 b(sp)s(eci\014cation\).)71 b(The)41 b(v)-5 b(alue)40 b(of)h(the)g(corresp)s(onding)1673 4786 y(elemen)m(t)31 b(of)f Fd(AXIS)p 2310 4786 V 33 w(INFO)g Fl(is)f(implemen)m(tation)g (dep)s(enden)m(t.)1503 4918 y Fd('CYCLIC')43 b Fl(The)20 b(axis)g(is)f(distributed)e Fd(CYCLIC)p Fl(.)i(The)h(corresp)s(ond-) 1673 5031 y(ing)29 b(elemen)m(t)i(of)g Fd(AXIS)p 2462 5031 V 33 w(INFO)e Fl(con)m(tains)i(the)f(blo)s(c)m(k)g(size.)1503 5164 y Fd('INDIRECT')42 b Fl(The)20 b(axis)g(is)f(distributed)f Fd(INDIRECT\(map-array\))p Fl(.)1673 5277 y(The)23 b(v)-5 b(alue)24 b(of)g(the)g(corresp)s(onding)e(elemen)m(t)j(of)f Fd(AXIS)p 3532 5277 V 33 w(INFO)1673 5390 y Fl(is)29 b(implemen)m(tation)g(dep)s(enden)m(t.)p eop %%Page: 7 10 7 9 bop 150 -200 a Fg(0.1.)72 b(MAPPING)31 b(INQUIR)-8 b(Y)31 b(PR)m(OCEDURES)1792 b Fl(7)423 99 y Fd(AXIS)p 621 99 29 4 v 33 w(INFO)29 b Fl(\(optional\))240 b(m)m(ust)25 b(b)s(e)g(a)i(rank)e(one)h(arra)m(y)g(of)g(t)m(yp)s(e)g(default)f(in)m (teger,)i(and)e(size)1503 211 y(at)43 b(least)f(equal)g(to)h(the)g (rank)e(of)i(the)f Fc(align-tar)-5 b(get)52 b Fl(to)43 b(whic)m(h)1503 324 y Fd(DISTRIBUTEE)27 b Fl(is)i(ultimately)g(aligned) g(\(whic)m(h)g(is)g(returned)g(b)m(y)1503 437 y Fd(HPF)p 1653 437 V 33 w(TEMPLATE)22 b Fl(in)i Fd(TEMPLATE)p 2577 437 V 32 w(RANK)p Fl(\).)g(It)g(is)g(an)g Fd(INTENT)46 b(\(OUT\))1503 550 y Fl(argumen)m(t.)88 b(The)46 b(i)2221 517 y Fb(th)2318 550 y Fl(elemen)m(t)h(of)f Fd(AXIS)p 2987 550 V 34 w(INFO)e Fl(con)m(tains)j(the)1503 663 y(blo)s(c)m(k)23 b(size)h(in)f(the)h(blo)s(c)m(k)g(or)g(cyclic)g (distribution)c(of)k(the)h(i)3498 630 y Fb(th)3595 663 y Fl(axis)1503 776 y(of)33 b(the)f(ultimate)g Fc(align-tar)-5 b(get)43 b Fl(of)33 b Fd(DISTRIBUTEE)p Fl(;)d(if)h(that)i(axis)1503 889 y(is)e(a)i(collapsed)f(axis,)h(then)f(the)h(v)-5 b(alue)32 b(is)g(implemen)m(tation)f(de-)1503 1002 y(p)s(enden)m(t.)423 1164 y Fd(PROCESSORS)p 909 1164 V 31 w(RANK)f Fl(\(optional\))45 b(m)m(ust)40 b(b)s(e)g(scalar)g(and)g(of)g(t)m(yp)s(e)h(default)f(in)m (teger.)71 b(It)40 b(is)f(set)1503 1276 y(to)49 b(the)h(rank)e(of)h (the)h(pro)s(cessor)e(arrangemen)m(t)i(on)m(to)g(whic)m(h)1503 1389 y Fd(DISTRIBUTEE)35 b Fl(is)i(distributed.)61 b(It)38 b(is)f(an)h Fd(INTENT)46 b(\(OUT\))37 b Fl(ar-)1503 1502 y(gumen)m(t.)423 1664 y Fd(PROCESSORS)p 909 1664 V 31 w(SHAPE)29 b Fl(\(optional\))46 b(m)m(ust)31 b(b)s(e)f(a)i(rank)e(one)i (arra)m(y)f(of)g(t)m(yp)s(e)h(default)e(in)m(teger)i(and)1503 1777 y(of)26 b(size)g(at)g(least)g(equal)g(to)h(the)f(v)-5 b(alue,)26 b Fi(m)p Fl(,)h(returned)e(in)f Fd(PROCES)p Fl(-)1503 1890 y Fd(SORS)p 1701 1890 V 33 w(RANK)p Fl(.)31 b(It)g(is)g(an)h Fd(INTENT)46 b(\(OUT\))30 b Fl(argumen)m(t.)45 b(Its)31 b(\014rst)g Fi(m)1503 2003 y Fl(elemen)m(ts)38 b(are)h(set)g(to)g(the)g(shap)s(e)e(of)i(the)f(pro)s(cessor)g(arrange-) 1503 2116 y(men)m(t)j(on)m(to)g(whic)m(h)e Fd(DISTRIBUTEE)f Fl(is)h(mapp)s(ed.)70 b(\(It)41 b(ma)m(y)g(b)s(e)1503 2228 y(necessary)31 b(to)h(call)e Fd(HPF)p 2335 2228 V 34 w(DISTRIBUTION)e Fl(t)m(wice,)k(the)f(\014rst)f(time)1503 2341 y(to)h(obtain)e(the)i(v)-5 b(alue)30 b(of)g Fd(PROCESSORS)p 2877 2341 V 32 w(RANK)f Fl(in)g(order)g(to)i(allo-)1503 2454 y(cate)h Fd(PROCESSORS)p 2181 2454 V 31 w(SHAPE)p Fl(.\))423 2616 y Fd(PLB)d Fl(\(optional\))513 b(m)m(ust)34 b(b)s(e)g(a)g(rank)g(one)h(arra)m(y)g(of)f(t)m(yp)s(e)h(default)e(in)m (teger)i(and)f(of)1503 2729 y(size)j(at)i(least)f(equal)g(to)g(the)g (rank)g(of)g(DISTRIBUTEE.)f(It)h(is)1503 2842 y(an)43 b Fd(INTENT)j(\(OUT\))c Fl(argumen)m(t.)79 b(The)43 b(i)2964 2809 y Fb(th)3062 2842 y Fl(elemen)m(t)g(is)f(set)i(to)1503 2955 y(the)g(smallest)g(pro)s(cessor)g(index)f(ONTO)g(whic)m(h)g(the)i (i)3498 2922 y Fb(th)3595 2955 y Fl(axis)1503 3068 y(of)38 b(DISTRIBUTEE)g(is)g(mapp)s(ed;)j(if)d(that)h(axis)f(is)g(collapsed,) 1503 3181 y(then)21 b(the)g(corresp)s(onding)e(elemen)m(t)j(of)g(PLB)f (is)f(implemen)m(tation)1503 3293 y(dep)s(enden)m(t.)423 3455 y Fd(PUB)29 b Fl(\(optional\))513 b(m)m(ust)34 b(b)s(e)g(a)g(rank) g(one)h(arra)m(y)g(of)f(t)m(yp)s(e)h(default)e(in)m(teger)i(and)f(of) 1503 3568 y(size)j(at)i(least)f(equal)g(to)g(the)g(rank)g(of)g (DISTRIBUTEE.)f(It)h(is)1503 3681 y(an)24 b Fd(INTENT)46 b(\(OUT\))24 b Fl(argumen)m(t.)39 b(The)24 b(i)2868 3648 y Fb(th)2966 3681 y Fl(elemen)m(t)h(is)f(set)h(to)h(the)1503 3794 y(largest)k(pro)s(cessor)f(index)f(ONTO)h(whic)m(h)f(the)h(i)3182 3761 y Fb(th)3280 3794 y Fl(axis)g(of)h(DIS-)1503 3907 y(TRIBUTEE)25 b(is)g(mapp)s(ed;)h(if)f(that)i(axis)e(is)g(collapsed,)h (then)g(the)1503 4020 y(corresp)s(onding)h(elemen)m(t)j(of)f(PUB)h(is)e (implemen)m(tation)g(dep)s(en-)1503 4133 y(den)m(t.)423 4294 y Fd(PSTRIDE)g Fl(\(optional\))322 b(m)m(ust)34 b(b)s(e)g(a)g(rank)g(one)h(arra)m(y)g(of)f(t)m(yp)s(e)h(default)e(in)m (teger)i(and)f(of)1503 4407 y(size)j(at)i(least)f(equal)g(to)g(the)g (rank)g(of)g(DISTRIBUTEE.)f(It)h(is)1503 4520 y(an)43 b Fd(INTENT)j(\(OUT\))c Fl(argumen)m(t.)79 b(The)43 b(i)2964 4487 y Fb(th)3062 4520 y Fl(elemen)m(t)g(is)f(set)i(to)1503 4633 y(the)28 b(in)m(terpro)s(cessor)f(stride)g(in)g(the)h(ONTO)g (clause)f(with)g(whic)m(h)1503 4746 y(the)32 b(i)1687 4713 y Fb(th)1785 4746 y Fl(axis)g(of)g(DISTRIBUTEE)g(is)f(mapp)s(ed;)i (if)e(that)i(axis)f(is)1503 4859 y(collapsed,)37 b(then)e(the)i (corresp)s(onding)d(elemen)m(t)j(of)f(PSTRIDE)1503 4972 y(is)29 b(set)i(to)g(zero.)423 5181 y Fe(Example.)36 b Fl(Giv)m(en)23 b(the)g(declarations)f(in)f(the)i(example)f(of)h (Section)g(0.1.2,)j(and)c(assuming)f(that)423 5294 y(the)34 b(actual)h(mappings)e(are)i(as)f(the)h(directiv)m(es)f(sp)s(ecify)-8 b(,)34 b(the)h(results)e(of)i Fd(HPF)p 3150 5294 V 33 w(DISTRIBUTION)423 5407 y Fl(are:)p eop %%Page: 8 11 8 10 bop 150 -200 a Fl(8)3039 b Fg(CONTENTS)p 462 11 2976 4 v 460 110 4 100 v 1434 110 V 1849 80 a Fa(A)p 2322 110 V 840 w(B)p 3235 110 V 484 w(PI)p 3436 110 V 462 113 2976 4 v 462 130 V 460 230 4 100 v 512 200 a(AXIS)p 717 200 25 4 v 31 w(TYPE)p 1434 230 4 100 v 507 w(['BLOCK',)26 b('BLOCK'])p 2322 230 V 99 w(['CYCLIC',)i('BLOCK'])p 3235 230 V 98 w([)56 b(])p 3436 230 V 462 233 2976 4 v 460 333 4 100 v 512 303 a(AXIS)p 717 303 25 4 v 31 w(INF)n(O)p 1434 333 4 100 v 797 w([10,)26 b(10])p 2322 333 V 658 w([1,)h(15])p 3235 333 V 394 w([)56 b(])p 3436 333 V 462 336 2976 4 v 460 436 4 100 v 512 406 a(PR)n(OCESSORS)p 1079 406 25 4 v 28 w(SHAPE)p 1434 436 4 100 v 404 w([4,)27 b(2])p 2322 436 V 720 w([4,)h(2])p 3235 436 V 415 w([)56 b(])p 3436 436 V 462 439 2976 4 v 460 539 4 100 v 512 509 a(PR)n(OCESSORS)p 1079 509 25 4 v 28 w(RANK)p 1434 539 4 100 v 507 w(2)p 2322 539 V 859 w(2)p 3235 539 V 514 w(0)p 3436 539 V 462 542 2976 4 v 150 754 a Ff(0.1.5)97 b(MAP)p 665 754 30 4 v 35 w(ARRA)-8 b(Y\(DISTRIBUTEE,)32 b(DIM\))423 927 y Fe(Optional)i(Argumen)m(ts.)40 b Fl(DIM.)423 1141 y Fe(Description.)h Fl(Returns)30 b(the)g(map)g(arra)m(y)h(for)f (a)h(sp)s(eci\014ed)e(axis)h(of)g(an)g(arra)m(y)-8 b(.)423 1355 y Fe(Class.)40 b Fl(Mapping)29 b(inquiry)f(function.)423 1570 y Fe(Argumen)m(ts.)423 1759 y Fd(DISTRIBUTEE)552 b Fl(ma)m(y)30 b(b)s(e)g(of)g(an)m(y)g(t)m(yp)s(e.)41 b(It)30 b(ma)m(y)h(b)s(e)e(scalar)h(or)g(arra)m(y)g(v)-5 b(alued.)40 b(It)1503 1872 y(m)m(ust)22 b(not)h(b)s(e)f(an)g (assumed-size)g(arra)m(y)-8 b(.)39 b(If)22 b(it)g(is)g(a)h(mem)m(b)s (er)e(of)i(an)1503 1985 y(aggregate)41 b(v)-5 b(ariable)36 b(group,)k(then)d(it)g(m)m(ust)h(b)s(e)f(an)h(aggregate)1503 2098 y(co)m(v)m(er)k(of)e(the)h(group.)70 b(\(See)41 b(Section)f Fe(??)h Fl(for)f(the)g(de\014nitions)1503 2211 y(of)29 b(\\aggregate)k(v)-5 b(ariable)28 b(group")h(and)g (\\aggregate)k(co)m(v)m(er."\))43 b(It)1503 2323 y(m)m(ust)28 b(not)h(b)s(e)f(a)h(p)s(oin)m(ter)e(that)i(is)f(disasso)s(ciated)g(or)g (an)h(allo)s(cat-)1503 2436 y(able)39 b(arra)m(y)i(that)f(is)f(not)h (allo)s(cated.)69 b(It)40 b(is)f(an)h Fd(INTENT)46 b(\(IN\))1503 2549 y Fl(argumen)m(t.)1503 2696 y(If)35 b Fd(DISTRIBUTEE)e Fl(is)h(a)i(p)s(oin)m(ter,)h(information)d(ab)s(out)h(the)h(dis-)1503 2809 y(tribution)d(of)j(its)f(target)i(is)e(returned.)55 b(The)35 b(target)i(m)m(ust)f(not)1503 2922 y(b)s(e)d(an)i (assumed-size)e(dumm)m(y)h(argumen)m(t)g(or)g(a)h(section)g(of)f(an) 1503 3035 y(assumed-size)j(dumm)m(y)g(argumen)m(t.)64 b(If)37 b(the)h(target)h(is)e(\(a)i(sec-)1503 3148 y(tion)c(of)7 b(\))36 b(a)g(mem)m(b)s(er)g(of)g(an)f(aggregate)k(v)-5 b(ariable)34 b(group,)j(then)1503 3261 y(the)30 b(mem)m(b)s(er)g(m)m (ust)g(b)s(e)g(an)g(aggregate)k(co)m(v)m(er)e(of)e(the)h(group.)423 3425 y Fd(DIM)936 b Fl(m)m(ust)25 b(b)s(e)f(scalar)h(and)f(of)h(t)m(yp) s(e)g(default)f(in)m(teger.)40 b(Its)25 b(v)-5 b(alue)24 b(m)m(ust)1503 3538 y(b)s(e)33 b(b)s(et)m(w)m(een)h(one)g(and)e(the)i (rank)f(of)h(DISTRIBUTEE,)f(unless)1503 3651 y(DISTRIBUTEE)k(is)g (scalar,)j(in)d(whic)m(h)g(case)i(it)f(m)m(ust)g(not)h(b)s(e)1503 3764 y(presen)m(t.)423 3978 y Fe(Result)f(T)m(yp)s(e,)h(T)m(yp)s(e)f(P) m(arameter,)f(and)h(Shap)s(e.)49 b Fl(The)33 b(result)f(is)g(a)h(rank)g (one)g(arra)m(y)h(of)423 4091 y(t)m(yp)s(e)e(default)f(in)m(teger)i (and)f(size)g(equal)f(to)i(the)g(exten)m(t)g(of)g(axis)e(DIM)i(of)f (DISTRIBUTEE;)g(if)423 4204 y(DISTRIBUTEE)d(is)h(scalar,)g(then)g(the)h (result)e(is)h(of)g(size)g(one.)423 4418 y Fe(Result)e(v)-6 b(alue.)39 b Fl(Elemen)m(t)24 b(i)f(of)i(the)f(result)g(is)f(the)h(pro) s(cessor)g(index)f(to)i(whic)m(h)e(the)i(i)3347 4385 y Fb(th)3444 4418 y Fl(elemen)m(t)423 4531 y(of)38 b(DISTRIBUTEE)g (along)g(axis)g(DIM)g(is)g(mapp)s(ed.)63 b(If)38 b(axis)f(DIM)i(of)g (DISTRIBUTEE)e(is)423 4644 y(collapsed,)29 b(then)h(all)g(elemen)m(ts)g (of)h(the)g(result)e(ha)m(v)m(e)i(the)g(v)-5 b(alue)30 b(one.)150 4890 y Ff(0.1.6)97 b(BLOCK)p 773 4890 V 35 w(SIZES\(DISTRIBUTEE,)31 b(DIM\))423 5062 y Fe(Optional)j(Argumen)m (ts.)40 b Fl(DIM.)423 5277 y Fe(Description.)h Fl(Returns)30 b(the)g(arra)m(y)h(of)f(blo)s(c)m(k)g(sizes)g(for)g(a)h(sp)s(eci\014ed) d(blo)s(c)m(k-distribtued)f(axis)423 5390 y(of)j(an)h(arra)m(y)-8 b(.)p eop %%Page: 9 12 9 11 bop 150 -200 a Fg(0.1.)72 b(MAPPING)31 b(INQUIR)-8 b(Y)31 b(PR)m(OCEDURES)1792 b Fl(9)423 99 y Fe(Class.)40 b Fl(Mapping)29 b(inquiry)f(function.)423 311 y Fe(Argumen)m(ts.)423 499 y Fd(DISTRIBUTEE)552 b Fl(ma)m(y)26 b(b)s(e)f(of)h(an)m(y)g(t)m(yp) s(e.)39 b(It)26 b(m)m(ust)g(b)s(e)f(arra)m(y)h(v)-5 b(alued.)38 b(It)26 b(m)m(ust)g(not)1503 612 y(b)s(e)31 b(an)i(assumed-size)e(arra) m(y)-8 b(.)48 b(If)31 b(it)h(is)f(a)i(mem)m(b)s(er)f(of)g(an)g(aggre-) 1503 724 y(gate)25 b(v)-5 b(ariable)23 b(group,)h(then)g(it)f(m)m(ust)h (b)s(e)f(an)h(aggregate)i(co)m(v)m(er)g(of)1503 837 y(the)33 b(group.)47 b(\(See)33 b(Section)g Fe(??)g Fl(for)g(the)g (de\014nitions)d(of)j(\\aggre-)1503 950 y(gate)27 b(v)-5 b(ariable)25 b(group")h(and)f(\\aggregate)30 b(co)m(v)m(er."\))42 b(It)26 b(m)m(ust)g(not)1503 1063 y(b)s(e)j(a)i(p)s(oin)m(ter)e(that)i (is)f(disasso)s(ciated)f(or)h(an)g(allo)s(catable)g(arra)m(y)1503 1176 y(that)h(is)e(not)i(allo)s(cated.)40 b(It)31 b(is)e(an)i Fd(INTENT)46 b(\(IN\))29 b Fl(argumen)m(t.)1503 1322 y(If)35 b Fd(DISTRIBUTEE)e Fl(is)h(a)i(p)s(oin)m(ter,)h(information)d (ab)s(out)h(the)h(dis-)1503 1435 y(tribution)d(of)j(its)f(target)i(is)e (returned.)55 b(The)35 b(target)i(m)m(ust)f(not)1503 1548 y(b)s(e)d(an)i(assumed-size)e(dumm)m(y)h(argumen)m(t)g(or)g(a)h (section)g(of)f(an)1503 1661 y(assumed-size)j(dumm)m(y)g(argumen)m(t.) 64 b(If)37 b(the)h(target)h(is)e(\(a)i(sec-)1503 1774 y(tion)c(of)7 b(\))36 b(a)g(mem)m(b)s(er)g(of)g(an)f(aggregate)k(v)-5 b(ariable)34 b(group,)j(then)1503 1887 y(the)30 b(mem)m(b)s(er)g(m)m (ust)g(b)s(e)g(an)g(aggregate)k(co)m(v)m(er)e(of)e(the)h(group.)423 2050 y Fd(DIM)936 b Fl(m)m(ust)25 b(b)s(e)f(scalar)h(and)f(of)h(t)m(yp) s(e)g(default)f(in)m(teger.)40 b(Its)25 b(v)-5 b(alue)24 b(m)m(ust)1503 2162 y(b)s(e)d(b)s(et)m(w)m(een)i(one)f(and)f(the)i (rank)e(of)h(DISTRIBUTEE.)g(The)f(cor-)1503 2275 y(resp)s(onding)26 b(axis)i(of)g(DISTRIBUTEE)g(m)m(ust)g(ha)m(v)m(e)i(a)f(BLOCK)1503 2388 y(distribution.)423 2601 y Fe(Result)46 b(T)m(yp)s(e,)j(T)m(yp)s (e)d(P)m(arameter,)h(and)f(Shap)s(e.)70 b Fl(The)39 b(result)g(is)g(a)i (rank)e(one)h(arra)m(y)423 2714 y(of)e(t)m(yp)s(e)g(default)e(in)m (teger)j(and)e(size)g(equal)h(to)g(the)g(exten)m(t)h(of)f(axis)f(DIM)i (of)f(the)f(pro)s(cessors)423 2827 y(arrangemen)m(t)31 b(on)m(to)h(whic)m(h)d(DISTRIBUTEE)g(is)h(distributed.)423 3039 y Fe(Result)j(v)-6 b(alue.)41 b Fl(The)28 b(result)g(is)g(the)i(v) m(ector)g(of)g(blo)s(c)m(k)e(sizes)h(used)f(to)i(map)f(the)g(sp)s (eci\014ed)e(axis)423 3152 y(of)36 b(DISTRIBUTEE.)f(If)g(the)i (indicated)d(axis)h(of)h(DISTRIBUTEE)f(do)s(es)h(not)g(ha)m(v)m(e)h(a)f (blo)s(c)m(k)423 3265 y(mapping,)29 b(the)h(result)g(is)f(implemen)m (tation)g(dep)s(enden)m(t.)p eop %%Trailer end userdict /end-hook known{end-hook}if %%EOF --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Fri Jan 5 10:47:20 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id KAA26621 for hpff-distribute-out; Fri, 5 Jan 1996 10:47:20 -0600 (CST) Received: from beech.soton.ac.uk (beech.soton.ac.uk [152.78.128.78]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id KAA26601 for ; Fri, 5 Jan 1996 10:47:07 -0600 (CST) Received: from bright.ecs.soton.ac.uk (bright.ecs.soton.ac.uk [152.78.64.201]) by beech.soton.ac.uk (8.6.12/hub-8.5a) with SMTP id QAA00124 for ; Fri, 5 Jan 1996 16:47:04 GMT Received: from diana.ecs.soton.ac.uk by bright.ecs.soton.ac.uk; Fri, 5 Jan 96 16:48:18 GMT From: John Merlin Message-Id: <3678.199601051647@diana.ecs.soton.ac.uk> Subject: hpff-distribute: Thoughts on HPF2 and 2000 To: hpff-distribute@cs.rice.edu Date: Fri, 5 Jan 1996 16:47:34 +0000 (GMT) Cc: jhm@ecs.soton.ac.uk (John Merlin) X-Mailer: ELM [version 2.4 PL24] Content-Type: text Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Hi, I recently learnt of the HPF2.0 and HPF2000 proposals, and thought I'd present some thoughts on them as input for the impending discussions. I'll start with the higher denomination standard first. HPF2000 ------- My first impression was -- Wow! To define HPF2000 in 1996, four years early, is certainly a bold and far-sighted move! I can see the advantage. It's nice to get things done ahead of schedule (at least I imagine so -- I don't speak from personal experience you understand). Despite this, I'd like to suggest an alternative idea. (Sorry to be a spoilsport.) That is, don't present this language as a standard at all, but instead put everything that isn't in HPF2 in one or more official HPF technical reports as proposals for the standard to follow HPF2. This would be in line with the procedure that I believe has been adopted for Fortran evolution by the ANSI Fortran committee. Roughly speaking (Jerry will have the full picture) the idea is to use official TRs for features that if put into F96 would unduly delay its publication and implementation, but which are felt to be too important to forget altogether until F2000. This way vendors have guidelines on useful extensions to F96, and they can be officially standardised in F2000 with the benefit of experience. Similarly for HPF. Putting the new features in TRs means that users and the marketplace can test their usefulness, efficiency, implementability, technical correctness, etc. The HPFF need be less concerned about getting every detail perfect, because implementation and use would iron out any rough edges. Then, come 2000, they can be incorporated in an official HPF2000 standard with the benefit of experience, just as Fortran and C generally standardise existing practice rather than defining totally new things. If any of the proposed features aren't implemented by 2000, then clearly they aren't needed or are technically flawed, so they shouldn't go in the standard. To paraphrase a paraphrase of a well-known saying: "Premature optimisaton is the root of all evil". As for the name of this language, I suggest something like Extended or Extra HPF, or HPF+, etc -- don't put a date on it. The argument against this approach is that it's nice to put everything in a standard so that users can rely on them being universally available. However, the name HPF2000 is an open admission that this won't be the case for another four years. Given that, the TR approach seems sensible. HPF2.0 ------ I can understand the desire to make the next full HPF standard some sort of subset of HPF1.1, so that users have something they can rely on being implemented and universally available. However, I was alarmed by the suggestion that the offical 'Subset HPF' would somehow cease to exist. I don't know how you can undefine a published standard, even an unofficial one, and it's unfair on users and implementors alike, and damages credibility, to shift the goalposts like this. My chagrin was heightened by the fact that we've recently announced a public domain implementation of Subset HPF (see the advert at the end of this message!), and were proud of the fact that we've implemented the data mapping features in virtually their full generality. I didn't expect all the interesting bits to be eliminated from the next HPF standard, given that this is the core and most important part of HPF. Given that HPF2 will effectively be the real standard for some years to come, so realistically users will have to restrict themselves to what's in it, I believe it shouldn't be too spartan and should include at least all of the alignment and distribution features in Subset HPF. This also reasonable given that Subset HPF is what most users and implementors have been targetting so far. I'll now try to give an even more cogent argument for this. Firstly, I don't think the proposed criterion for including features in HPF2 (that they "give high performance on all platforms") is a good one. It seems like a cop out by vendors. A better criterion is to include the minimal set of features necessary to allow programs to execute with the highest (reasonably achievable) performance. E.g., although CYCLIC (N) distributions may be slower than BLOCK in a direct comparison, there will be applications for which the use of CYCLIC (N) will give better performance than being forced to use a simple BLOCK distribution. The HPF philosophy is that the user has to experiment to get optimum efficiency; I doubt if the user will use relatively inefficient features like strided alignment unecessarily, so I don't see why we should protect him/her from it. On the basis of this argument, and the fact that these features have already been widely implemented so obviously aren't too hard to provide, I'd like to argue for the re-introduction of general alignments and distributions. Some justifications: -- Alignment offsets and strides are needed to describe the distribution of an array section when it is passed as an actual argument into a procedure. In this way no data movement is required, and the mapping is still explicitly specified inside the procedure. Without this facility the argument would either have to be remapped on entry and exit, so that it has a simple mapping that can be described without alignment, or the mapping would have to be inherited (which is probably also disallowed, and is certainly less efficient than explicitly specifying the mapping). I gave an example of this for Gauss elimination in a Sci. Prog. paper about HPF a while back. -- Alignment axis permutation is needed to efficiently implement references like: FORALL (i=...,j=...) a (i,j) = b (j,i) Anyway, I don't understand what complication or inefficieny this facility could introduce. Can someone explain? -- Embedded alignment, e.g.: ALIGN s WITH a(m,n) is needed to allow fine-scale control of where data is stored and thus where computations are performed, and can avoid unecessary communications. For example, in Gaussian elimination with pivotting, intermediate values of the pivot element and value can be stored and computed just on the processor owning the column in question, and the final values can then be broadcast by assigning them to globally replicated scalars. If one could not embed the variables holdinh the intermediate values on a particular processor, they would be globally broadcast at each step of the pivot search, with massive overhead. Using an 'ON HOME' directive doesn't solve this problem; though the computation can be localised, the intermediate results are still broadcast if assigned to a globally replicated variable, as they must be to preserve HPF semantics. (I got this example from a paper I reviewed, which found that the embedding solution gave immensely better performance.) -- '*'s in the alignment are needed to specify partial replication, e.g. replicating a vector over dimension 1 of a 2d processor array and distributing it over dim 2. Surely there is no debate over the usefulness of this. Thus, I argue for keeping Subset HPF's general alignment and distribution features in HPF2. I have just two other comments on features for inclusion in HPF2: -- Why is it necessary for HPF2's base language to be full F90 (or F96), rather than the F90 subset selected for Subset HPF? True, I expect most large vendors will wish to provide full F96 along with HPF, but they could do this anyway even if HPF2 only required Subset F90. (Perhaps they could distinguish that their HPF implementation supports full F96 by calling it HPF96 rather than just HPF2, for instance). The problem I percieve with requiring full F96 as a basis is that it tends to exclude small-scale research and academic efforts, which may well provide public domain implementations (e.g. shpf and ADAPTOR), from being legitamately calling themselves HPF. I think public domain implementations can greatly help to prime the market for HPF, by providing free development and testing systems to get people interested, and for education purposes, etc. Thus I think the HPFF should support and facilitate such efforts rather than putting big hurdles in the way. Also, some vendors may wish to expend more effort on the HPF features important for performance (like general data mapping support!) rather than nice but relatively unimportant F90 features. Finally, F96 does introduce some extra complications for HPF, like pointers, arrays-of-arrays, etc. I'm not convinced these have been well sorted-out, nor do I think they are trivial. -- I agree with the removal of sequence and storage association, and the SEQUENCE directive. This is one area where Subset HPF provided features that were guaranteed to give low performance, and were extremely hard to implement in full generality. Also, of course, I support the introduction of PURE procedures to HPF2! -- Finally, since we seem to be in the business of tweaking and correcting the last HPF standard, I'd like to plead again for common blocks to be disallowed inside pure procedures. (Please, don't sigh so loud Chuck! :-)). In fact they only cause problems if they contain distributed data (much like sequence and storage association), so could be allowed otherwise, but for regularity it's probably prefereble to disallow them altogether, in HPF2 at least. Without common blocks, pure procedures are trivial to implement, and very useful for getting high efficiency in many circumstances as they permit functiional parallelism in FORALLs (or indeed in array syntax, using F96 ELEMENTAL procedures). However, by allowing them to access external data that may be stored on a different processor to that executing the pure procedure, they violate one of the rules of Kernel HPF, namely, that features should "allow high performance across all platforms", because they become extremely hard (or perhaps virtually impossible) to implement on distributed memory architectures, which is still an important architecural class for HPF (e.g. workstation networks). Has anybody actually implemented pure procedures in their full generality, with global data accesses allowed, on any DM platforms yet? Furthermore, I notice that the desire for 'purer than pure' procedures has been voiced in several contexts at recent HPF meetings. Perhaps this is the time to bite the bullet and add this constraint. That's all for now, you'll be pleased to know. Wishing you all a Happy and High Performance New Year, John. P.S. And now for that advert I promised. 'shpf', a Subset High Performance Fortran compilation system, is now freely available by anonymous ftp from: ftp.hpcc.soton.ac.uk in directory /pub/packages/shpf. It supports arbitrary alignments and distributions, in any number of dimensions, in virtually any context (for what it's worth!). Full information is (or shortly will be) available on the WWW at URL: http://www.hpcc.ecs.soton.ac.uk/shpf/shpf.html. Please spread the good news to your friends and colleagues, and enjoy it! ----------------------------------------------------------- John Merlin, jhm@ecs.soton.ac.uk Dept of Electronics and Computer Science, tel: +44 1703 593943 University of Southampton, fax: +44 1703 593045 Southampton, SO17 1BJ, U.K. --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Mon Jan 8 06:49:29 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id GAA18284 for hpff-distribute-out; Mon, 8 Jan 1996 06:49:29 -0600 (CST) Received: from beech.soton.ac.uk (beech.soton.ac.uk [152.78.128.78]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id GAA18158 for ; Mon, 8 Jan 1996 06:46:28 -0600 (CST) Received: from bright.ecs.soton.ac.uk (bright.ecs.soton.ac.uk [152.78.64.201]) by beech.soton.ac.uk (8.6.12/hub-8.5a) with SMTP id MAA00814 for ; Mon, 8 Jan 1996 12:46:11 GMT Received: from landlord.ecs.soton.ac.uk by bright.ecs.soton.ac.uk; Mon, 8 Jan 96 12:41:47 GMT From: John Merlin Received: from bacchus.ecs.soton.ac.uk by landlord.ecs.soton.ac.uk; Mon, 8 Jan 96 12:41:21 GMT Message-Id: <707.9601081241@bacchus.ecs.soton.ac.uk> Subject: Re: hpff-distribute: Thoughts on HPF2 and 2000 To: hpff-distribute@cs.rice.edu Date: Mon, 8 Jan 1996 12:41:20 +0000 (GMT) In-Reply-To: <3678.199601051647@diana.ecs.soton.ac.uk> from "John Merlin" at Jan 5, 96 04:47:34 pm X-Mailer: ELM [version 2.4 PL0] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- I just noticed a terrible mistake in my my last message... > To paraphrase a paraphrase of a well-known saying: > "Premature optimisaton is the root of all evil". As for the ^^^^^^^^^^^ I meant 'standardisation' of course. Sorry for any confusion. John. --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Mon Jan 15 18:49:41 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id SAA18634 for hpff-distribute-out; Mon, 15 Jan 1996 18:49:41 -0600 (CST) Received: from hplms26.hpl.hp.com (hplms26.hpl.hp.com [15.255.168.31]) by cs.rice.edu (8.7.1/8.7.1) with ESMTP id SAA18625; Mon, 15 Jan 1996 18:49:38 -0600 (CST) Received: from hplpp3.hpl.hp.com by hplms26.hpl.hp.com with ESMTP ($Revision: 1.36.108.11 $/15.5+ECS 3.3+HPL1.1S) id AA190723376; Mon, 15 Jan 1996 16:49:37 -0800 Received: by hplpp3.hpl.hp.com (1.37.109.14/15.5+ECS 3.3+HPL1.1) id AA095263381; Mon, 15 Jan 1996 16:49:41 -0800 Date: Mon, 15 Jan 1996 16:49:41 -0800 From: Rob Schreiber Message-Id: <199601160049.AA095263381@hplpp3.hpl.hp.com> To: hpff-interpret@cs.rice.edu Subject: hpff-distribute: pointers, mapping etc Cc: hpff-distribute@cs.rice.edu Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- There seems to be a lot of confusion about what mappings pointers can have. I hope the distribution group will tackle this and produce something we all understand and IBM will implement. The issues are these: 1. There seems to be no appetite anymore for any ambiguity as to mapping. That is why realign and redistribute and maybe inherit are on there way out. Thus, pointers that assume the mapping of their targets are viewed as bad. Since pointers must agree with their targets as to type and rank, the obvious solution is to give pointers mappings statically (at the declaration) and to require that the target have an identical mapping. This would apply to the allocated object if the pointer is used in an allocate. Why is this tricky? 1. What if the pointer and target have imprecise mappings, with something left to the compiler: real, pointer :: p(:,:) real x(100,100) !hpf$ distribute (block, block) :: x, p The compiler might choose different processors arrangements. 2. How do I generate and map linked structures one cell at a time? A reasonable HPF answer might be "Use some other programming language." 3. I dont see distribution ranges as the right answer here. In fact, with these rules and without dynamic mapping (and inherit?) there is no need for them. --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Mon Jan 22 15:16:10 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id PAA06169 for hpff-distribute-out; Mon, 22 Jan 1996 15:16:10 -0600 (CST) Received: from fontainebleau.ensmp.fr (root@fontainebleau.ensmp.fr [192.54.148.100]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id PAA06152 for ; Mon, 22 Jan 1996 15:16:02 -0600 (CST) Received: from cri.ensmp.fr (chailly.ensmp.fr) by fontainebleau.ensmp.fr with SMTP id AA25199 (5.67a8/IDA-1.5 for ); Mon, 22 Jan 1996 22:15:36 +0100 Received: from provins.caii by cri.ensmp.fr (4.1/SMI-4.0) id AA07995; Mon, 22 Jan 96 22:15:35 +0100 Date: Mon, 22 Jan 96 22:15:35 +0100 Message-Id: <9601222115.AA07995@cri.ensmp.fr> Received: by provins.caii (4.1/SMI-4.1) id AA00727; Mon, 22 Jan 96 22:15:29 +0100 From: Fabien COELHO To: schreiber@hpl.hp.com Subject: hpff-distribute: ASSUME directive proposal Cc: hpff-distribute@cs.rice.edu Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Dear Rob, > There seems to be no appetite anymore for any ambiguity as to mapping. This is just the point I want to address in this proposal! great! > That is why realign and redistribute and maybe inherit are on there way > out. I do think that realign and redistribute can be kept and be useful by disallowing mapping ambiguity. But this is another story. I would like the following proposal to be discussed by the Forum for possible addition to HPF-later and/or the kernel. Any comment from the list is welcome, for sure. The idea is to allow the user to specify several descriptive mappings for a given argument to a subroutine. I already sent the suggestion once. This one version is more formal. Design and implementation issues are discussed more precisely. It is a possible alternative to some use of the inherit directive. This proposal may replace the inherit directive in the kernel... I talked with Thomas Brandes (Adaptor) last friday about implementation issues of such a feature, and he agreed that it is not big a problem. I guess all implementors would agree on that point. Thus the question is more about the usefulness of the feature. Fabien. -- Fabien COELHO __ http://www.cri.ensmp.fr/~coelho __ coelho@cri.ensmp.fr CRI, ENSMP, 35, rue Saint Honoré, 77305 Fontainebleau Cedex, France phone: 33 1 64 69 {voice: 48 52, fax: 47 09, standard: 47 08} ________ All opinions expressed here are mine ________ %%%%%%%%%% PROPOSAL: ASSUME (or whatever name you might prefer) directive -------------------------------------------------------------- The aim of this proposal is NOT to add a new feature or directive to HPF, but rather to promote good software engineering in HPF and to provide more knowledge to the compiler about the mapping of data, in order to improve the performances. This is in the spirit of HPF. The proposal does not add any new "semantics" to the language. It is just designed to help the compiler simply and efficiently. Outline: -------- The idea is to allow to specify several descriptive mappings for arguments at subroutine interfaces. It is an alternative to the inherit directive as a mean to switch beetween several subroutines, depending on the mapping of their arguments, to do the same job. It does not prevent separate compilation. The compiler can take advantage of maximum static knowledge when compiling the subroutines. The user is not required to clone its own procedure by hand to specify several mappings for a subroutine, and or to use the inherit stuff. Syntax: ------- assume-directive is ASSUME list-of-or-separated-static-mapping-directives END ASSUME list-of-or-separated-static-mapping-directives is list-of-static-mapping-directives [ OR list-of-or-separated-static-mapping-directives ] Constraints: ------------ C1: declaration: The assume-directive may appear within the declarations of the subroutine. If so, it MUST also appear in the subroutine interface. Notes: ------ N1: the list of mapping directives might be empty, meaning that non distributed arrays may be used as arguments. N2: there may be other mapping directives outside of the ASSUME scope. Discussion: ----------- Here are discussed some issues that could be switched to constraints... D1: same order... ? If C1, it must be the SAME order in the subroutine interface and declarations. Rationale: this allow a simple naming convention for direct implementation of the feature to be managed at compile/link-time in a cloning-based approach. If the order is not a constraint, then a compile/link-time management should be based on the mapping directives by adding some computed suffix to the name of the subroutines compiled with the different assumptions, and all names must be different... Thus requiring D4(a). This issue is extensivelly discussed hereafter. D2: several/nested ASSUME directives? Might be allowed. It would then specify the cross product of mappings as possible assumptions to be managed by the compiler. On the one hand, this could shorten the mapping description for some subroutines. On the other hand, it would allow the user to specify many cases to the compiler without making him/her feel the cost of these many different assumptions. Should the user be trusted? I guess so:-) D3: mapping of local variables? The mapping of local variables may be linked to the mapping of the arguments, thus local variables may be specified a mapping within the ASSUME directive. An align outside the assume scope might not be necessarily sufficient. There is no significant impact on the clone-based implementation. If local variables are allowed within the assume scope in a subroutine, they must not appear in the subroutine interface... D4: distinct descriptive mappings, or not? (a) The different lists of mapping directives should contain distinct descriptive mappings about the arguments of the subroutine, so that there can be no ambiguity about the right "version" of the subroutine to call. (b) Or no such constraints, and the first convenient version is chosen? This has some impact on the possible implementations: It would be desirable to reject the next example because only the first set of directives would be used (would be the rationale for (a))" subroutine sub(A) real B ! B is local to sub chpf$ assume chpf$ distribute A(block) chpf$ distribute B(block) chpf$ or chpf$ distribute A(block) chpf$ distribute B(cyclic) ! inconsistent with the previous one... chpf$ end assume However (a) would prevent the example in D5 and a default prescriptive case when non distributed data are also to be considered (see D6). Thus I would prefer (b), but it requires D1 as a constraint, for simple naming convention management and compile/link time handling. D5: prescriptive mappings? prescriptive mappings for some arguments of the subroutine may be allowed: For instance: subroutine sub(A(n),B(n)) chpf$ assume chpf$ distribute A(block) chpf$ distribute * B(cyclic) ! requires a remapping chpf$ or chpf$ distribute * A(cyclic) ! idem chpf$ distribute B(block) chpf$ end assume However you would not like this case to be rejected because the compiler cannot decide the right version to call if both A and B are block distributed. The "take the first one" option would make sense here. The nature of the ambiguity here is quite similar to the one in D4, but here it is not as useless... D6: default? The last option in the assume, under D4(b), can be used as a default case with prescriptive arguments: chpf$ assume chpf$ or chpf$ distribute A(cyclic) chpf$ or chpf$ distribute * A(block) chpf$ end assume Ths would mean that A may be local or cyclic distributed, otherwise please block-distribute it for the routine. Note that maybe it would not make much sense to add a descriptive assumption about A as block-distributed, because the default case should perform equally well (the remapping should not be performed at runtime... maybe a simple copy... and the body of the subroutine would be compiled with equal assumptions? I guess it depends on the way prescriptive mappings are handled by the compiler). Advices to users: ----------------- The directive may be expensive at compile time because it may require to compile several times the routine with the different assumptions. Advices to implementers: ------------------------ To summarize the main issue in the discussion above: (1) the declaration order (in the assume or-list) is important: => very simple naming convention for compile/link-time handling, based on the occurence number. (2) the declaration is not important AND distinct descriptive mappings => possible naming convention for compile/link-time handling a` la C++, but disallow interesting use of the feature (D5/D6)... (3) the declaration order is not important => no simple compile/link-time handling. dynamic switching required:-( My opinion: best is (1) [D1/D5/D6 and D4(b)]. It does not prevent some stupid set of directives to be supplied by the user. Well, this is the problem of the user, it is not the main issue of a language to prevent stupid use of features. It allows a direct and simple compile/link-time implementation, what is highly desirable. Thus the directive *can* (this is not mandatory) be implemented as follow: (1) Rewrite subroutine sub(...) chpf$ assume CASE 1 chpf$ or CASE 2 chpf$ end assume ... end sub As (this is a cloning) subroutine sub_1(...) CASE 1 ... subroutine sub_2(...) CASE 2 ... Compile the subroutines. When such a subroutine is called, insert the call to the right version (which is simply derived under constraint D1: it is sub_"minimum occurrence number compatible with the arguments"). (2) Other possible implementation, without constraint D1, but requiring constraint D4(a): the routine is rewritten as subroutine sub_"name_from_case(CASE X)"(...) CASE X... name_from_case should depend only on the descriptive mapping of arguments. This is how (C++) templates can be compiled in C++, as far as I remmeber. (3) Other possible implementation, without D1 and D4(a), the switching must be performed at runtime, between the cloned versions, thus: subroutine sub if CASE X call sub_x ... else if CASE Y call sub_y ... Example: -------- The computation and distributions in this example do not make much sense. Note the align directive outside the assume/end assume. sub callee(a, c) parameter(n=100) real a(n,n), c ! arguments real b(n,n) ! local array chpf$ align b with a chpf$ assume chpf$ distribute a(block,*) chpf$ or chpf$ distribute a(block,block) chpf$ end assume chpf$ independent(i,j) forall(i=1:n,j=1:n) b(i,j) = c*sqrt(i*i+j*j) chpf$ independent(i,j) forall(i=1:n,j=1:n) a(i,j)=a(i,j)/b(i,j) end callee sub caller() parameter(n=100) real x(n,n),y(n,n) chpf$ distribute x(block,*) chpf$ distribute y(block,block) ... call callee(x, 0.1) call callee(y, 0.25) ... end caller --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-distribute Tue Jan 23 04:23:10 1996 Received: (from daemon@localhost) by cs.rice.edu (8.7.1/8.7.1) id EAA20871 for hpff-distribute-out; Tue, 23 Jan 1996 04:23:10 -0600 (CST) Received: from fontainebleau.ensmp.fr (root@fontainebleau.ensmp.fr [192.54.148.100]) by cs.rice.edu (8.7.1/8.7.1) with SMTP id EAA20866 for ; Tue, 23 Jan 1996 04:22:51 -0600 (CST) Received: from cri.ensmp.fr (chailly.ensmp.fr) by fontainebleau.ensmp.fr with SMTP id AA29682 (5.67a8/IDA-1.5 for ); Tue, 23 Jan 1996 11:22:22 +0100 Received: from episy.caii by cri.ensmp.fr (4.1/SMI-4.0) id AA18337; Tue, 23 Jan 96 11:22:22 +0100 Date: Tue, 23 Jan 96 11:22:22 +0100 Message-Id: <9601231022.AA18337@cri.ensmp.fr> Received: by episy.caii (4.1/SMI-4.1) id AA17955; Tue, 23 Jan 96 11:22:05 +0100 From: Fabien COELHO To: hpff-distribute@cs.rice.edu Subject: hpff-distribute: ASSUME directive proposal - erratum Cc: coelho@chailly.ensmp.fr Sender: owner-hpff-distribute Precedence: bulk --------------------------------------------------------------------------- hpff-distribute@cs.rice.edu is a mailing list for discussion of data distribution in High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- > (2) the declaration is not important AND distinct descriptive mappings > => possible naming convention for compile/link-time handling > a` la C++, but disallow interesting use of the feature (D5/D6)... > > (2) Other possible implementation, without constraint D1, but requiring > constraint D4(a): the routine is rewritten as > > subroutine sub_"name_from_case(CASE X)"(...) > CASE X... > > name_from_case should depend only on the descriptive mapping of > arguments. This is how (C++) templates can be compiled in C++, as far as I > remmeber. Well, I'm wrong. The analogy is with the management of C++ polymorphism by simple linkers. The problem with C++ templates is to compile only once the routines, and some compilers keep a database of already compiled template instances to avoid several compilations. The polymorphism problem is to differentiate int my_function(int) and int my_function(float), if the linker does only want to know about the function name and not its type. The suggested solution is to add some suffix which depends on the function type, generating my_function_$#%$@ and my_function_43%$21 so that the linker will be happy with different function names. However I think that such techniques should not be necessary for an HPF assume-like directive if care is taken in the definition of the directive. Fabien. -- Fabien COELHO __ http://www.cri.ensmp.fr/~coelho __ coelho@cri.ensmp.fr CRI, ENSMP, 35, rue Saint Honoré, 77305 Fontainebleau Cedex, France phone: 33 1 64 69 {voice: 48 52, fax: 47 09, standard: 47 08} ________ All opinions expressed here are mine ________ --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-distribute-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe ---------------------------------------------------------------------------