TY - GEN
T1 - A high performance distributed processing system
AU - Kwong, S.
AU - Chan, Y. K.
PY - 1991
Y1 - 1991
N2 - The authors describe an i860 based prototype add-on card as a compute server in a distributed system. Preliminary scalar oriented performance indicates that the card is about 35% (200%) faster than SUN SPARC station 1 rated at 1.7 MFLOPS (15.8 MIPS) running at 25 MHz. The most notable feature of the system is that it provides a run-time environment which is similar to running the Turbo C program under a DOS session. This feature increases portability and reduces the inertia for application porting. A low-cost high-performance networked server using i860 is being implemented. In this system, only one or two i860 cards are connected to a PC network of about 10 to 20 PC/ATs. The compute code of the Turbo-C program will run on the i860 server via an Ethernet network and the I/O code will be executed on the local PC/AT node. For the example of a 70 × 70 matrix floating point multiplication program, it is found that the system is about 55 times faster than executing the program on the local 12 MHz 286/10 MHz 287 AT node.
AB - The authors describe an i860 based prototype add-on card as a compute server in a distributed system. Preliminary scalar oriented performance indicates that the card is about 35% (200%) faster than SUN SPARC station 1 rated at 1.7 MFLOPS (15.8 MIPS) running at 25 MHz. The most notable feature of the system is that it provides a run-time environment which is similar to running the Turbo C program under a DOS session. This feature increases portability and reduces the inertia for application porting. A low-cost high-performance networked server using i860 is being implemented. In this system, only one or two i860 cards are connected to a PC network of about 10 to 20 PC/ATs. The compute code of the Turbo-C program will run on the i860 server via an Ethernet network and the I/O code will be executed on the local PC/AT node. For the example of a 70 × 70 matrix floating point multiplication program, it is found that the system is about 55 times faster than executing the program on the local 12 MHz 286/10 MHz 287 AT node.
UR - https://www.scopus.com/pages/publications/0026364317
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-0026364317&origin=recordpage
M3 - RGC 32 - Refereed conference paper (with host publication)
VL - 3
SP - 1873
EP - 1877
BT - IECON Proceedings (Industrial Electronics Conference)
PB - IEEE
T2 - Proceedings of the 1991 International Conference on Industrial Electronics, Control and Instrumentation - IECON '91
Y2 - 28 October 1991 through 1 November 1991
ER -