1-2hit |
Kazuhito ITO Kesami HAGIWARA Takashi SHIMIZU Hiroaki KUNIEDA
A further study on a VLSI system compiler, named VEGA (VLSI Embodiment for General Algorithms), is presented. It maps a general digital signal processing algorithm onto a neo-systolic array, which is a VLSI oriented multiprocessor array. Highly complicated mapping problem is divided into subproblems such as modularization, operation grouping, processor placement, scheduling, control logic synthesis, and mask pattern generation. In this paper, the modularization technique is proposed which homogenizes all the operations of the processing algorithm to multiply-add operations. The processor placement algorithm to map processing algorithm onto a neo-systolic array so as to minimize data transfer time is also proposed.
Yasuo SUGURE Seiji TAKEUCHI Yuichi ABE Hiromichi YAMADA Kazuya HIRAYANAGI Akihiko TOMITA Kesami HAGIWARA Takeshi KATAOKA Takanori SHIMURA
A 32-bit embedded RISC microcontroller core targeted for automotive, industrial, and PC-peripheral applications has been developed to offer the smaller code size, lower-latency instruction and interrupt processing needed for next-generation microcontrollers. The 360 MIPS/400MFLOPS/200 MHz core--based on the Harvard bus architecture--uses 0.13/0.15-µm CMOS technology and consists of a CPU, FPU, and register banks. To reduce the size of the control programs, new instructions have been added to the instruction set. These new instructions, as well as an enhanced C compiler, produce object files about 25% smaller than those for a previous designed core. A dual-issue superscalar structure consisting of three- or five-stage pipelines provides instruction processing with low latency. The cycle performance is thus an average of 1.8 times faster than the previous designed core. The superscalar structure is used to save 19 CPU registers in parallel when executing interrupt processing. That is, it saves the 19 CPU registers to the resister bank by accessing four registers at a time. This structure significantly improves interrupt response time from 37 cycles to 6 cycles.