Excerpt from those years when we won FPGA

The essence that spld, CPLD, and FPGA can implement any logic is that any logic can be expressed (or approached) by polynomials ). Polynomials are nothing more than multiplication and addition operations. And exactly, match the door, or match the door plus.

FPGA extended architecture: the on-chip programmable system. There are two types: one is the pure FPGA architecture, and the other is the system. Generally, the built-in multiplier, multiplier, and guitar bit transmission module can greatly optimize the algorithm that uses logic to achieve high resource consumption and slow speed. Another is the FPGA-based multi-core architecture. Including FPGA + processor (CPU, arm ......), FPGA + DSP, fpna, etc.

Generally, DRAM is used in the memory. The difference between DRAM and SRAM is that DRAM needs to be refreshed once for a period of time, and an additional timed refresh circuit is required because the hardware structure is capacitor charging and discharging. In terms of small-scale dram, the consumption of refresh circuit is not worth the candle, which is too complicated for FPGA. Therefore, FPGA uses SRAM and does not need to be dynamically refreshed. Once the data is loaded, it will maintain continuous power.

The FPGA Implementation Machine Based on LUT (query table) technology converts all possible inputs through the synthesizer to obtain all possible results (truth table) in advance ), the calculation process of the truth table is actually a comprehensive process in the development process. Then load the results into the LUT (SRAM) storage unit and index the results with different inputs. It should be noted that the LUT hardware itself has a two-choice, multi-path device structure. Try to understand it.

For a query table with N input, it can implement any N input combination function. At present, most of them are based on 4-input LUT. To make full use of LUT, if LUT is 4 input, the logical signal width should be a multiple of 4. For example, if it is 23 bits, 24 bits are used. One is to avoid wasting resources, and the other is to prevent counting overflow.

The concept of the fuse and the anti-fuse are compared.

MUX is a two-way multi-path selector. MUX-based FPGA is actually based on anti-fuse technology.

FPGA applications: image processing, communication, digital signal processing, CPU, and so on. General software provides related IP cores (look back ).

Embedded generally refers to an embedded processor. So I understand that single-chip microcomputer and FPGA are not very strict embedded systems.

The hard core is to directly put the microprocessor into the FPGA chip. The soft core is to configure a group of logical resource blocks into a microprocessor, which is a little slower but simple. You can configure multiple soft Cores when resources are allowed.

Top-down refers to dividing the system into several subsystems according to certain principles based on the functions of the system, and then dividing the system into several functional modules until it is divided into many basic modules. Modular modularization serves as the connection between layers.

Excerpt from those years when we won FPGA