新思科技:AI对EDA工具产生了怎样的影响?

时间:2018-11-09 17:19:50 Sassine Ghazi 新思科技 芯片设计事业部联席总经理

新思科技:AI对EDA工具产生了怎样的影响?

受访者:Sassine Ghazi, 新思科技芯片设计事业部联席总经理

采访者:Echo Zhao, Aspencore中国区主分析师


在2018全球CEO峰会期间,记者有幸采访到了新思科技芯片设计事业部联席总经理Sassine Ghazi先生,以下是访谈实录。

Q: I noticed that you have very rich experience in both engineering and marketing, could you share some opinions in your angle about the requirements from this generation of IC designers, and the response of EDA companies?

问: 我注意到您在工程和市场营销领域都有着非常丰富的经验,您能否就这一代IC设计师的需求以及EDA公司的反应发表一些看法?

A: Actually if you look at the last 3 decades, the major disruptors have been computes, internet, and mobility. And if you look ahead for the next 3 decades, I strongly believe that digitalization is gona be the next disruptor. Which is driven by the factor that everything is gona be intelligent, connected, and sensory. If you think about the requirements to achieve that, it will need to have a significant processing innovation, and connectivity. Therefore, if I look ahead for the next generation IC designers, beside the fundamental basics, which is hard basics not simple, of achieving best power, performance, area, schedule, it’s time for them to think about the next architecture, the AI architecture, to enable faster computes, and connectivity which most likely to be on the cloud.

答:事实上,如果你回顾过去30年,产业的主要颠覆因素是计算机、互联网和移动性。如果你展望未来30年,我坚信数字化将成为下一个颠覆者。这个驱动力来自:所有的东西都是智能的,联网的,传感的。如果您考虑实现这一目标的需求,那么它将需要有明显的处理能力创新、以及连接性。因此,对于下一代IC设计师来说,除了复杂的基础设计,实现最佳功率、性能、面积、设计时间,这些都是时候让他们思考未来架构,即AI架构,来获得更快的计算、和连接性(最有可能在云上)。

We are putting significant investment on innovation around AI, inside our product and AI in order to support our customer developing AI chips. 我们在EDA产品中加入AI,从而支持客户开发AI芯片,在AI创新方面投入了大量资金。

Q: AI EDA tools to design AI product.

A: exactly. Both in our EDA tools and develop AI products.

问:以AI的EDA工具来设计AI产品。

答:没错。将AI放入EDA工具并以此开发AI产品。

Q: Can AI technology be as disruptive to chip design as it has been in other applications, like image processing or speech recognition? If so, could you give us some examples?

问: 人工智能技术对芯片设计的颠覆性,会不会像在图像处理或语音识别等其他应用中一样?如果有,您能给我们举几个例子吗?

A: Yes, actually it’s a huge opportunity for EDA. Because the value of EDA is automation as you know. Recently, one of our costumers represent our user group, that customer come from the No.1 company in the AI industry today in the world. They spend about 70% of the human effort to develop their SoC, that not unique to them. That actually very common. That’s a huge opportunity for EDA to reduce the human effort involved in developing SoC. So we have 2 significant investment, one we call it machine learning inside the tools, where we are changing our algorithms to take advantage of the machine learning, that allow the tool to provide our customer better design flow and methodology using AI.
是的,事实上这对EDA来说是一个巨大的机会。因为EDA的价值在于自动化。最近,我们的一个客户代表我们的用户群,(这个客户来自当今世界人工智能行业排名第一的公司),他们花了大约70%的人力来开发他们的SoC——而且他们不是个例,这在行业内其实很普遍。所以这对EDA来说是一个巨大的机会,因为可以减少开发SoC所涉及的人力资源。因此,我们有两项重大投资,一项是工具内部的机器学习,我们正在改变我们的算法来利用机器学习,使用AI让工具提供给客户更好的设计流程和方法。

Two quick example, one example is we recently introduce a significant (power ECU?) through machine learning. We provide our customer 5x productivity improvement, from weeks to hours. The other one is in the verification domain. We have 4 verification improvement through machine learning technics, gain 10x productivity improvement. So again, there are a lot of opportunities to improve, both inside the products and alone the product.
两个简单的例子,一个是我们最近通过机器学习推出一个重要的(power ECU?),把客户的生产力提升了5倍,设计周期从几星期缩短到几小时。另一个案例是在验证域。我们通过机器学习技术提升了验证能力,从而提高了10倍的生产率。因此,我们有很多机会可以改进。

With less human touch, the machine can do it faster, and in many cases, it can optimize better. 用更少的人力,机器可以做得更快,在很多情况下,它还可以优化得更好。

Q: What does “Fusion technology” help SoC designers?

问:“Fusion technology”对SoC设计师有什么帮助?

A: we recently announce Fusion complier, which is our new product. We strongly believe that it transform the way our customer do design today. Our industry is really split today between front end designers and back end designers. And handoff is really painful.

答:我们最近发布了Fusion complier,这是我们的新产品。我们坚信,它改变了我们客户今天的设计方式。现在,我们的行业被分裂成前端设计师和后端设计师。前后端工程师的交接是非常痛苦的。

The reason we do it this way is historical. It’s incredibly difficult to have one solution build on the same infrusture, same data model, etc. so Synopsys make significant investment and effort to blur the lines between front end and back end by sharing algorithms. The highest value to our customer is the ability to achieve best quality results, and most convergent flow. So you can get the result much faster, and in a less interation between the different phases of design. So we incredibly excited about this innovation. We have engaged a number of customer and they have seen significant improvement of the traditional way to do the design by using Fusion technology.
这个现状是有历史遗留因素的。以前,把一个解决方案建立在相同的架构、相同的数据模型上是非常困难的。所以Synopsys进行了大量的投资和努力,通过共享算法来模糊前端和后端之间的界限。对我们的客户来说,最大的价值是能够达到最好的结果,以及最收敛的设计流程。工程师可以更快地得到结果,在不同设计阶段之间的交互也更少。我们对这项创新感到无比兴奋。我们吸引了很多客户,他们看到了Fusion technology技术对传统的设计方式的显著改进。

It will actually make designer spend most of their energy on optimizing what their products require, versus interfacing between different phases of the design. 这会让设计师把大部分精力花在优化产品需求上,而不是在设计的不同阶段之间进行交互。

Thank you for sharing.

谢谢您的分享。

热门视频
相关推荐