FPGA knowledge Comb (a) a little testimonials to the FPGA industry

Today I want to start the FPGA industry knowledge point to do a big finishing, from personal feelings, to grammar, to the device base, difficult to conquer, to project application. Make a comprehensive review of what you've been exposed to in the past few years and see how you've traveled over the years.

Life is impermanent, a few years of stumbling barely in the FPGA industry gate. has always been to see other people's speech, stroll other people's forum notes. Maybe you should have a breakthrough.

I often compare engineers to peasant uncles, the same place has two points, first they are quiet and honest hard workers. Second, the technology of the world like farmland, you can not opportunistic, there is no shortcut, all the knowledge points need you to ground to open up. You work a little more to repay a little more, know a little more, quick.

Today, I just want to talk about my personal views and the problems that bothered me. Now I think I figured it out and take it out and talk about it.

Question one: Will the FPGA die?

Wandering for some time to learn the FPGA when the time to think about this problem, in case this industry is opencl such software system to replace, then I do not tragedy.

Now I think so, the FPGA industry is divided into two major points of knowledge. 1: Programming language. 2, device. The industry will die, and these two parts are replaced by something else. We'll discuss it separately.

1, programming language

The most common development language of FPGA is HDL, I am, no contact with other, like SystemC and so on. So the HDL language.

In recent years, both Altera and Xilinx have been pushing for HDL development with C, but there has been little progress. There seems to be no disruptive results, so it has been a stage for gimmicks and each other to steal the limelight (now Altera was acquired by Intel). Some people worry (I also thought) HDL will not be replaced by C? As long as there is such an editor, you can. My answer is no. The reason is to start with the programming language feature.

We all know that the programming language development is also very fast, from the original assembly, to C, java,c++, C#,HTML5 ...

Here is a misunderstanding, we think the assembly was killed, in fact, no, now understand the assembly of engineers is a popular. The application scenario has two places, one, small MCU with the emergence of smart home, some product development is the need to assemble. Second, IC design company, like the development of processor MCU, must use the assembly for instruction testing, to achieve full coverage of instructions, this is not the alternative language. A little bit more advanced than the assembly is the C language, and C language has always been classic, language development to the advanced can not completely replace C and assembly, why? Because they are a process-oriented development language, which is to write data processing, and JAVA,C++,C#,HTML5 ... are object-oriented programming languages, and of course their existence has its own characteristics, but the way of thinking is similar.

Say so much, just want to say: HDL is a hardware description language (Hdl:hardware Description Language). C is process-oriented and cannot be fully compatible at all. HDL has always been driven by the idea of a hardware circuit as a design, not a data flow. For example, the C language is to achieve sin. Just write the a=sinb directly. But HDL is going to achieve a sin. That's an IP. And HDL writes a shift operation that is well understood and well realized in the hardware. C will write several lines. The language is not good or bad, they each have characteristics, so want to use C thinking mode to guide the design of hardware circuit, this road seems too long.

Speaking of which we have to mention a language systemverilog. People who know the SV know that the SV will replace the HDL sooner or later, in fact it is not called an alternative, it is an update. Because SV is a combination of VERILOGHDL and C plus some of their own expansion into the formation of the language, which means you learn the Verilog is actually part of the SV. Now the SV is mainly used to do IC verification, forming an independent verification methodology, which is the development of SV characteristics. The combination of C in SV and its own other features are mainly used to write testbench, that is, in the Integrated Code section, or by VERILOGHDL. At present there is no hurry to learn SV, because it is not to do IC verification, using SV to do a project a little kill chicken with the slaughter sledgehammer, of course, more understanding of better.

Therefore, HDL will upgrade to SV, but will not die.

2, device

Devices include device features and development tools (QUARTUSII and Xilinx as a primary).

Device, that's what the process needs to think about, we just have to follow the original line. They will tell us what the characteristics of the device are different from the previous one, and whether the problem is the process or the design factor. As for the tool, an FPGA original want to promote their products must have the relevant technical training, that is, the least technical content but the novice seems to be the most capable of bluffing skills. The recent domestic FPGA momentum is very fierce, no use not to express their views.

These are the same things that the company uses whose products we use, the devices and the tools.

Issue two: Market and technology status of FPGA

Say well is high place is cold-high-end solution, general company No, that stuff expensive AH. The mass production is using ASIC, very few useful FPGA. Threshold relative to the software is a little higher than the threshold of the IC a little lower.

When I was in school, Altera and Xilinx both pushed Sopc to embed the original CPU in the FPGA. Later estimates also know that their CPU performance is too poor, so now all force push soc--since their CPU does not, then use others-embedded in the hard core arm. So the future development is software and hardware this technical barrier breakthrough more and more, learning FPGA also to learn arm, carrot. SV software and hardware language has been fused (although it seems to be some independent feeling at present), the knowledge of course also requires us to learn the Siege Lions.

It's so much more than that. Thanks to the family and friends who have come all the way, but also thank themselves for this strong heart.

FPGA knowledge Comb (a) a little testimonials to the FPGA industry