Benchmark Design

Mirada contains three application tests and two general performance measurement tests. Complete system performance is dependent on a wide range of elements in the system and depends on such variables as the CPU efficiency and cache, bus speed, memory latency and complete system load among others. The benchmark tests mostly only raw GPU efficiency. As such, it is mainly intended as a tool whose output can be used as an objective measure to communicate the GPU-component performance and as an additional testing tool in product pipeline during development and upgrade/maintenance cycle.

Benchmark uses an in-house modern 3D engine written in ANSI-C for maximum portability and efficiency. The engine is cross-platform compatible and easy to port to new architectures. The benchmark is a standalone application that is portable on any system with an ANSI-C compatible c-compiler and standards compliant OpenGL ES 2.0 interface. Porting consists of implementing a set of system interface functions defined in the product documentation. While running, the program does not register any user input but runs a predetermined test load.

Target for the application tests is 100,000 triangles per frame on average. Test content will utilize multiple render buffers and full screen filtering. Triangle loads will be optimized.

Source Code Approach

With minimal UI, a command line interface or a simple text configuration file, source code is ready to be easily ported and compiled for benchmarking practically any development platform with OpenGL ES 2.0 capabilities. Futuremark engineers can guide licensees’ developers through the implementation.

Application Tests:

• Application Test 1: Dashboard
• Application Test 2: User Interface / Infotainment
• Application test 3: Navigation

Feature Tests

• Polygon Throughput
• Fillrate