• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.29A no.9
• /
• pp.76-84
• /
• 1992

The DSP silicon compiler consists of language compiler, module generator, placement tool, router, layout generation tools, and simulator. In this paper, The language compiler, the module generator, placement tool, and simulator were developed and provided for the system designer. The language compiler translates the designer's system description language into the intermediate form file. The intermediate form file expresses the interconnections and specifications of the cells in the cell library. The simulator was developed and provided for the behavioral verification of the DSP system. For its implementation, the event-driven technique and the C$^{++}$ task library was used. The module generator was developed for the layout of the verified DSP system, and generates the functional block to be used in the DSP chip. And then the placement tool determines the appropriate positions of the cells in the DSP chip. In this paper, the placement tool was implemented by Min-Cut and Simulated Annealing algorithm. The placement process can be controlled by the several conditions input by the system designer.

• Proceedings of the KIEE Conference
• /
• 1988.07a
• /
• pp.521-524
• /
• 1988

This paper describes BSDL(Behavioral/Structural Description Language), CDTF(Control Data Text File) and state synthesizer built for use in ICSC(InCheon Silicon Compiler). BSDL describes structral and behaviral specifications of an ASIC(Application Specific IC) for digital system design. ICSC's paser generates CDTF consists of if-then-else, arithmetic and data transfer statement according to each BSDL statement. State synthesizer generates CCG(Control Constraint Graph) in consideration of execution of statement and generates VCG (Variable Constraint Graph) in consideration use of variable generation and use of variable. Also, it involves allocating algorithm operation nodes in the data path and the control path to machine states with minimum state number and as small area as possible.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.2
• /
• pp.228-234
• /
• 2006

This paper describes a systematic process which can generate a complete circuit specification efficiently for a given recursive DSP algorithm based on an optimal multiprocessor scheduler. The process is composed of two states: scheduling and circuit synthesis. The scheduling part accepts a fully specified flow graph(FSFG) as an input, and generates an optimal synchronous multiprocessor schedule. Then the circuit synthesis part translates the modified schedule into a complete circuit diagram including a control specification. The circuit diagram can be applied to a silicon compiler for VLSI layout generation. This paper illustrates the whole process with an example of a second order Gray-Market lattice filter.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.29A no.4
• /
• pp.64-73
• /
• 1992

This paper presents the design of multi-level logic optimization algorithm and the development of the SMILE system based on the algorithm. Considering the fanin constraints in algorithmic level, SMILE performs global and local optimization in a predefined sequence using heuristic information. Designed under the Sogang Silicon Compiler design environment, SMILE takes the SLIF netlist or Berkeley equation formats obtained from high-level synthesis process, and generates the optimized circuits in the same format. Experimental results show that SMILE produces the promising results for some circuits from MCNC benchmarks, comparable to the popularly used multi-level logic optimization system, MIS.

• Proceedings of the KIEE Conference
• /
• 1987.07b
• /
• pp.850-853
• /
• 1987

A VLSI digital filter design using a semi-custom method is described. The digital filters composed of TDM/FDM Transmultiplexer are designed. Using the polyphase network approach a filter bank composed of only all-pass digital filter sections was designed. The use of all-pass filters as basic building blocks is shown to provide a transmultiplexer structure that has low computational requirements, low quantization noise, and high modularity. The silicon compiler systems is used to reduce the design time and increase the credibility of designed filters. A design of 1st order and 2nd order all pass filters is done using CMOS 2um N-well double metal cell.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.33A no.11
• /
• pp.152-161
• /
• 1996

This paper descrisbes the implementation of a 20D DCT that can be used for video conference, JPEG, and MPEG-related applications. The implemented DCT consists of two 1-D DCTs and a transposed memory between them, and uses ROM-based multipliers instead of conventional ones. As the system bit length, the minimum bit length that satisfies the accuracy specified by the ITU standard H.261 was chosen through the simulations using the C language. The proposed design uses a dual port RAM for the transposed memory, and processes two bits of input-pixel data simultaneously t ospeed up addition process using two sets of ROMs. The basic system architecture was designed using th Synopsys schematic editor, and internal modules were described in VHDL and synthesized to logic level after simulation. Then, the compass silicon compiler was used to create the final lyout with 0.8um CMOS libraries, using the standard cell approach. The final layout contains about 110, 000 transistors and has a die area of 4.68mm * 4.96mm, and the system has the processing speed of about 50M pixels/sec.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.2
• /
• pp.227-232
• /
• 1988

A semicustom VLSI design fo digital filters used in TDM/FDM transmultiplexer is described. A filter bank composed of only all-pass digital filter sections are implemented with the polyphase network. The use of all-pass filters as basic building blocks is shown to provide a trans-multiplexer structure that has low computational requirements, low quantization noise, and high modularity. The silicon compiler system is used to reduce the design time and to increase the credibility of designed filters. As a prototype, 1st and 2nd order all pass filter are designed, using CMOS N-well double metal technology. The chip sizes of first order filter and the second order filter are 2652 x 533\ulcorner\ulcorner 5334x4300\ulcorner\ulcorner respectively.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.10
• /
• pp.1252-1262
• /
• 1988

This paper describes a functional-level design and simulation of Graphics Processor(GP) which can be used in various graphics systems. GP is divided into two parts: One is CPU, and the other is the interface to I/O peripherals. In order to achieve fast execution of graphics instructions, the CPU has special ALU, barrel shifter and window comparator and a FIFO for instruction prefetch. I/O part controls the DRAM and VRAM which constitute the GP's local memory, generates the signals to drive monitor, and communicates with the host processor. The functional simulation of CPU was done on Daisy workstation while the I/O part was designed using GENESIL, a silicon compiler.

• Journal of the Institute of Convergence Signal Processing
• /
• v.8 no.1
• /
• pp.73-77
• /
• 2007

In this paper, we propose a VLSI architecture of an LMS filter based on a Cyclo-static scheduler for fast computation of LMS filteing algorithm which is widely used in adptive filtering area. This process is composed of two steps: scheduling and circuit synthesis. The scheduling step accepts a fully specified flow graph(FSFG) as an input, and generates an optimal Cyclo-static schedule in the sense of the sampling rate, the number of processors, and the input-output delay. Then the generated schedule is transformed so that the number of communication edges between the processors. The circuit synthesis part translates the modified schedule into a complete circuit diagram by performing resource allocations. The VLSI layout generation can be performed easily by an existing silicon compiler.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.3
• /
• pp.355-366
• /
• 2022

This paper proposes a microcode-based neural network accelerator controller for artificial intelligence accelerators that can be reconstructed using a programmable architecture and provide the advantages of low-power and ultra-small chip size. In order for the target accelerator to support various neural network models, the neural network model can be converted into microcode through microcode compiler and mounted on accelerator to control the operators of the accelerator such as datapath and memory access. While the proposed controller and accelerator can run various CNN models, in this paper, we tested them using the YOLOv2-Tiny CNN model. Using a system clock of 200 MHz, the Controller and accelerator achieved an inference time of 137.9 ms/image for VOC 2012 dataset to detect object, 99.5ms/image for mask detection dataset to detect wearing mask. When implementing an accelerator equipped with the proposed controller as a silicon chip, the gate count is 618,388, which corresponds to 65.5% reduction in chip area compared with an accelerator employing a CPU-based controller (RISC-V).