From owner-hpff-interpret Fri Jan 10 06:17:15 1997 Received: (from daemon@localhost) by cs.rice.edu (8.8.4/8.7.1) id GAA14793 for hpff-interpret-out; Fri, 10 Jan 1997 06:17:15 -0600 (CST) Received: from TYO9.gate.nec.co.jp (TYO9.gate.nec.co.jp [203.180.98.35]) by cs.rice.edu (8.8.4/8.7.1) with ESMTP id GAA14787 for ; Fri, 10 Jan 1997 06:17:11 -0600 (CST) Received: from mailsv.nec.co.jp ([133.200.254.203]) by TYO9.gate.nec.co.jp (8.8.4+2.7Wbeta2/3.4Wb-NEC-TYO9) with ESMTP id VAA18321 for ; Fri, 10 Jan 1997 21:17:09 +0900 (JST) Received: from mailsv.bs1.fc.nec.co.jp (venus.d2.bs1.fc.nec.co.jp [133.203.2.164]) by mailsv.nec.co.jp (8.8.4+2.7Wbeta4/3.4W-96122618) with ESMTP id VAA16505 for ; Fri, 10 Jan 1997 21:17:08 +0900 (JST) Received: from c1mail.lang1.bs1.fc.nec.co.jp by mailsv.bs1.fc.nec.co.jp (8.7.3+2.6Wbeta5/6.4JAIN) id VAA07133; Fri, 10 Jan 1997 21:16:48 +0900 (JST) Received: from c1mpc.lang1.bs1.fc.nec.co.jp (c1mpc) by c1mail.lang1.bs1.fc.nec.co.jp (5.67+1.6W/6.4J.6) id AA26810; Fri, 10 Jan 97 21:17:47 JST Received: by c1mpc.lang1.bs1.fc.nec.co.jp (5.67+1.6W/6.4J.6) id AA03800; Fri, 10 Jan 97 21:19:00 JST Date: Fri, 10 Jan 97 21:19:00 JST From: hayashi@lang1.bs1.fc.nec.co.jp (Yasuharu Hayashi) Message-Id: <9701101219.AA03800@c1mpc.lang1.bs1.fc.nec.co.jp> To: hpff-interpret@cs.rice.edu Cc: jahpf-tech@csl.cl.nec.co.jp Subject: hpff-interpret: Is the owner of "shadow elements" a real owner? Sender: owner-hpff-interpret Precedence: bulk --------------------------------------------------------------------------- hpff-interpret@cs.rice.edu is a mailing list for corrections, clarifications, and interpretations related to High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- Dear HPFF: I have a question about the interpretation of SHADOW attribute. REAL A(10),B(10) !HPF$ PROCESSORS P(2) !HPF$ DISTRIBUTE (BLOCK) ONTO P :: A,B !HPF$ SHADOW A(1:0) !HPF$ INDEPENDENT DO I=1,10 !HPF$ ON HOME(A(I)) B(I) = A(I) ENDDO In this case, the ON directive indicates both P(1) and P(2) execute the assignment statement B(5) = A(5) in the loop, because P(2) is the owner of the "shadow element" A(5) and of course P(1) is the owner of A(5). Is this interpretation correct? Thanks, Yasuharu Hayashi ------------------------------------------------------------------- # Yasuharu Hayashi # NEC Corporation # hayashi@lang1.bs1.fc.nec.co.jp ------------------------------------------------------------------- --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-interpret-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe --------------------------------------------------------------------------- From owner-hpff-interpret Fri Jan 10 07:19:18 1997 Received: (from daemon@localhost) by cs.rice.edu (8.8.4/8.7.1) id HAA16097 for hpff-interpret-out; Fri, 10 Jan 1997 07:19:18 -0600 (CST) Received: from mail12.digital.com (mail12.digital.com [192.208.46.20]) by cs.rice.edu (8.8.4/8.7.1) with ESMTP id HAA16091 for ; Fri, 10 Jan 1997 07:19:15 -0600 (CST) Received: from mpsg.hpc.pko.dec.com by mail12.digital.com (8.7.5/UNX 1.5/1.0/WV) id IAA25591; Fri, 10 Jan 1997 08:13:43 -0500 (EST) Received: by mpsg.hpc.pko.dec.com; id AA07398; Fri, 10 Jan 1997 08:13:10 -0500 From: offner@hpc.pko.dec.com (Carl Offner) Received: by hardy.hpc.pko.dec.com; (5.65v3.2/1.1.8.2/01Nov94-0839AM) id AA30098; Fri, 10 Jan 1997 08:13:38 -0500 Date: Fri, 10 Jan 1997 08:13:38 -0500 Message-Id: <9701101313.AA30098@hardy.hpc.pko.dec.com> To: hpff-interpret@cs.rice.edu Subject: hpff-interpret: Re: owners of "shadow elements" Cc: offner@hpc.pko.dec.com Sender: owner-hpff-interpret Precedence: bulk --------------------------------------------------------------------------- hpff-interpret@cs.rice.edu is a mailing list for corrections, clarifications, and interpretations related to High Performance Fortran. Instructions for adding or deleting yourself from this list appear at the bottom of this message. --------------------------------------------------------------------------- The purpose of the SHADOW directive is simply to give information to the compiler. It does not change the meaning of which processor owns which element of an array. Shadow elements are not really visible to the programmer, and are managed completely by the compiler. So for instance in the example given: REAL A(10),B(10) !HPF$ PROCESSORS P(2) !HPF$ DISTRIBUTE (BLOCK) ONTO P :: A,B !HPF$ SHADOW A(1:0) !HPF$ INDEPENDENT DO I=1,10 !HPF$ ON HOME(A(I)) B(I) = A(I) ENDDO There is no requirement that the compiler keep the shadow elements up to date. The shadow elements are simply there for the compiler to use if and when a nearest-neighbor computation occurs (and in the example given, there isn't one). When the compiler needs to use shadow elements, of course it is responsible for initializing them correctly. So the short answer is no, the proposed interpretation is not correct, and SHADOW does not affect the meaning of ON, at least in this case. --Carl Offner --------------------------------------------------------------------------- To (un)subscribe to this list, send mail to hpff-interpret-request@cs.rice.edu. Leave the subject line blank, and in the body put the line (un)subscribe ---------------------------------------------------------------------------