Title :
A 40 MHz 64-bit floating-point co-processor
Author :
Molnar, K. ; Ho, C.-Y. ; Staver, D. ; Davis, B. ; Jerdonek, R.
Author_Institution :
Gen. Electr. Corp. Res. & Dev. Center, Schenectady, NY, USA
Abstract :
An arithmetic coprocessor capable of executing 64-bit double-precision floating-point, 32-bit single-precision floating-point, and 32-bit integer instructions has been integrated onto a 1.0-cm*1.1-cm chip in a 1.2- mu m, single-poly, double-metal bulk CMOS process. The chip contains 17000 transistors and includes a register file, two accumulators, and separate interface, multiplication, and addition subprocessors. The coprocessor which is the arithmetic unit for a multichip microprocessor system, is packaged in a 132-pin leadless ceramic chip carrier. The coprocessor can be issued a new instruction each 25-ns clock cycle, and 64-bit double-precision arithmetic with full IEEE rounding is executed at a peak rate of 26.7 MFLOPs (million floating-point operations per second). The waveforms of a store instruction operating at 40 MHz are shown.<>
Keywords :
CMOS integrated circuits; digital arithmetic; microprocessor chips; 1.2 micron; 26.7 MFLOPS; 32-bit integer instructions; 40 MHz; 64 bit; accumulators; addition subprocessors; arithmetic coprocessor; double-metal bulk CMOS process; double-precision floating-point; floating point coprocessor; interface subprocessor; leadless ceramic chip carrier; multichip microprocessor system; multiplication subprocessor; peak rate; register file; single-precision floating-point; store instruction; waveforms; Adders; Ceramics; Circuits; Clocks; Coprocessors; Floating-point arithmetic; Microprocessors; Packaging; Registers; Research and development;
Conference_Titel :
Solid-State Circuits Conference, 1989. Digest of Technical Papers. 36th ISSCC., 1989 IEEE International
Conference_Location :
New York, NY, USA
DOI :
10.1109/ISSCC.1989.48228