Just as an example…when the CPU and GPU exchange information in a generic PC, the CPU inputs information, and the GPU needs to read the information and clear the cache, initially. When returning the results, the GPU needs to clear the cache, then return the result to the CPU. We’ve created a cache bypass. The GPU can return the result using this bypass directly. By using this design, we can send data directly from the main memory to the GPU shader core. Essentially, we can bypass the GPU L1 and L2 cache. Of course, this isn’t just for data read, but also for write. Because of this, we have an extremely high bandwidth of 10GB/sec.
Also, we’ve also added a little tag to the L2 cache. We call this the VOLATILE tag. We are able to control data in the cache based on whether the data is marked with VOLATILE or not. If this tag is used, this data can be written directly to the memory. As a result, the entirety of the cache can be used efficiently for graphics processing.
This function allows for harmonization of graphics processing and computing, and allows for efficient function of both. Essentially “Harmony” in Japanese. We’re trying to replicate the SPU Runtime System (SPURS) of the PS3 by heavily customizing the cache and bus. SPURS is designed to virtualize and independently manage SPU resources. For the PS4 hardware, the GPU can also be used in an analogous manner as x86-64 to use resources at various levels. This idea has 8 pipes and each pipe(?) has 8 computation queues. Each queue can execute things such as physics computation middle ware, and other prioprietarily designed workflows. This, while simultaneously handling graphics processing.
This type of functionality isn’t used widely in the launch titles. However, I expect this to be used widely in many games throughout the life of the console and see this becoming an extremely important feature.
The properties of CPU and GPU are quite difference, so in the current stage, if you were to use an unified architecture such as HSA, it will be difficult to efficiently use the CPU and GPU. However, once the CPU and GPU are able to use the same APIs, development efficiency should increase exponentially. This will be rather huge. Thus, we expect to see this as somewhat of a long-term goal.
:v:
