Abstract
Efficient generation of communication sets and local index sets is important for evaluation of array expressions in scientific languages such as Fortran-90 and High Performance Fortran implemented on distributed-memory machines. We show that for arrays affinely aligned with templates that are distributed on multiple processors with a block-cyclic distribution, the local memory access sequence and communication sets can be efficiently enumerated using closed forms. First, closed form solutions are presented for arrays that are aligned with identity templates that are distributed using block or cyclic distributions. These closed forms are then used with a uirtual processor approach to give an efficient solution for arrays with block-cyclic distributions. These results are extended to arrays affinely aligned to arbitrary templates that have regular distributions. We present performance results on an iPSC/860 processor, that demonstrate the low runtime overhead of this scheme.
| Original language | English (US) |
|---|---|
| Article number | 4134228 |
| Pages (from-to) | 301-305 |
| Number of pages | 5 |
| Journal | Proceedings of the International Conference on Parallel Processing |
| Volume | 2 |
| DOIs | |
| State | Published - Jan 1 1993 |
| Externally published | Yes |
| Event | 1993 International Conference on Parallel Processing, ICPP 1993 - Syracuse, United States Duration: Aug 16 1993 → Aug 20 1993 |
Fingerprint
ASJC Scopus subject areas
- Software
- Mathematics(all)
- Hardware and Architecture
Cite this
On compiling array expressions for Efficient execution on distributed-memory machines. / Gupta, Sandeep; Kaushik, S. D.; Mufti, S.; Sharma, S.; Huang, C. H.; Sadayappan, P.
In: Proceedings of the International Conference on Parallel Processing, Vol. 2, 4134228, 01.01.1993, p. 301-305.Research output: Contribution to journal › Conference article
}
TY - JOUR
T1 - On compiling array expressions for Efficient execution on distributed-memory machines
AU - Gupta, Sandeep
AU - Kaushik, S. D.
AU - Mufti, S.
AU - Sharma, S.
AU - Huang, C. H.
AU - Sadayappan, P.
PY - 1993/1/1
Y1 - 1993/1/1
N2 - Efficient generation of communication sets and local index sets is important for evaluation of array expressions in scientific languages such as Fortran-90 and High Performance Fortran implemented on distributed-memory machines. We show that for arrays affinely aligned with templates that are distributed on multiple processors with a block-cyclic distribution, the local memory access sequence and communication sets can be efficiently enumerated using closed forms. First, closed form solutions are presented for arrays that are aligned with identity templates that are distributed using block or cyclic distributions. These closed forms are then used with a uirtual processor approach to give an efficient solution for arrays with block-cyclic distributions. These results are extended to arrays affinely aligned to arbitrary templates that have regular distributions. We present performance results on an iPSC/860 processor, that demonstrate the low runtime overhead of this scheme.
AB - Efficient generation of communication sets and local index sets is important for evaluation of array expressions in scientific languages such as Fortran-90 and High Performance Fortran implemented on distributed-memory machines. We show that for arrays affinely aligned with templates that are distributed on multiple processors with a block-cyclic distribution, the local memory access sequence and communication sets can be efficiently enumerated using closed forms. First, closed form solutions are presented for arrays that are aligned with identity templates that are distributed using block or cyclic distributions. These closed forms are then used with a uirtual processor approach to give an efficient solution for arrays with block-cyclic distributions. These results are extended to arrays affinely aligned to arbitrary templates that have regular distributions. We present performance results on an iPSC/860 processor, that demonstrate the low runtime overhead of this scheme.
UR - http://www.scopus.com/inward/record.url?scp=84956696333&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84956696333&partnerID=8YFLogxK
U2 - 10.1109/ICPP.1993.171
DO - 10.1109/ICPP.1993.171
M3 - Conference article
AN - SCOPUS:84956696333
VL - 2
SP - 301
EP - 305
JO - Proceedings of the International Conference on Parallel Processing
JF - Proceedings of the International Conference on Parallel Processing
SN - 0190-3918
M1 - 4134228
ER -