More and more ADAS (Advanced Driver Assistance Systems) are added to cars, trucks and even trains. For those vehicles to function optimally, there has to be a lot of development in the field of Artificial Intelligence (AI), sensor processing and image recognition. These applications require stupendous amount of floating-point calculations. The resource-efficient 16-bit floating-point format is to come to widespread use this year.
The pinnacle of ADAS is the autonomous car. Different companies are already testing them, like the self-driving cars tested by companies such as Tesla and Waymo.
A lot of fundamental technology is used to make these vehicles move on their own. They have to have top of the range sensor equipment for spatial awareness, and AI and image processing to map the surroundings of the vehicle to anticipate movements and dangerous situations. One company working on this technology is Ambarella. Ambarella has created an image processing chip that combines the best of RISC and neural networks.
16-bit Floating-Point (FP)
Billions of floating-point calculations form the backbone of image processing and feature extraction. Shaving of any overhead makes a huge difference in speed, cost and power. For many of these calculations, the precision of 32-bit FP is not necessary. 16-bit FP will do, and that is twice as efficient to store and requires a quarter of the power to compute with.
16-bit FP is not so widespread yet, and is mostly limited to dedicated graphics processors. But that is changing and more and more companies are adding 16-bit FP to the CPU core and vector units.
That is why Solid Sands has added 16-bit FP testing to its portfolio. With this development, we are ahead of the curve as the C and C++ languages do not have a primitive type to accommodate this new format yet and tests are not readily available.
ADAS technology takes big leaps every year. By adopting the 16-bit FP format, the cost of those developments and those vehicles decreases, making them more accessible for everybody. But safety has to come first. That is why testing is so important.
Ambarella uses SuperTest to verify the compilers (yes, there is more than one) that are needed to program their image processing chip. At Solid Sands we continue to develop SuperTest to stay up to date with all the new developments.
Want to know more about all those new developments in both technology and testing? Then do not hesitate to contact us or download our Ambarella customer story.
Click here to contact us