The OpenGL Performance Characterization (OPC) project group has released its GLperf benchmark to the public and reported a second round of performance results.
GLperf tests graphics performance based on the exact same objects and attributes, which allows comparisons across different OpenGL-based systems. Standardized testing makes GLperf a superior alternative to primitive-level benchmarks such as triangles per second, which are often based on different attributes.
A simple input file format within GLperf allows users to specify the OpenGL primitives they want to time. The benchmark is flexible to allow testing for a variety of objects (line strips, polygons, triangles) with different attributes (line width, number of lights, texture functions, etc.). The ability to specify detailed GLperf testing parameters enables vendors and users to closely inspect performance under different scenarios.
The performance-testing flexibility of GLperf is matched by the reporting and display options available to the user. A user can specify, for example, whether timing is to be reported as objects per second, pixels per second, or microseconds per object. A sophisticated set of Java applets are used in the GPC News Web site to read results from the server’s database and allow tabular and graphical comparisons on the client’s Web browser.
GLperf complements Viewperf, the OPC group’s popular benchmark that measures performance for 3D graphics paths and models used by specific commercial applications.
Questions about GLperf and OPC project group activities can be addressed to the OPC e-mail alias: gpcopc-info@spec.org.
For more detailed information on GLperf, go to the GLperf FAQ document
Download GLperf via anonymous ftp: /dist/gpc/opc/glperf