ADLINK CM3-GF - Manual (Page 30)

TME-104P-CMx-GF-1V7 Rev 1.7 22 (81)
 Shader instruction store, using an advance caching system
 Advanced shader design, with ultra-threading sequencer for high-efficiency
operations
 Advanced, high-performance branching support, including static and dynamic
branching
 16-bit and 32-bit floating point components for high dynamic range computations
 Full anti-aliasing on render surfaces up to and including 128-bit floating point
formats
 Support for OpenCL™ 1.1
 Support for OpenGL 3.2 and 2.1
 Anti-Aliasing Filtering:*
o
* Support for anti-aliasing filtering is dependent on application
o 2x/4x/8x modes
o Sparse multi-sample algorithm with gamma correction, programmable sample patterns, and
centroid sampling
o Temporal anti-aliasing
o Custom filter anti-aliasing with up to 12 samples per pixel
o Adaptive anti-aliasing mode
o Lossless color compression (up to 8:1) at all resolutions, up to and including widescreen
HDTV
 Anisotropic Filtering:*
o * Support for anisotropic filtering is dependent on application
o 2x/4x/8x/16x modes
o Up to 128-tap texture filtering
o Anisotropic biasing to allow trading quality for performance
o Improved quality mode due to improved sub-pixel precision, and higher precision LOD
computations
o Advanced texture compression (3Dc+)
o High quality 4:1 compression for normal maps and luminance maps
o Angle-invariant algorithm for improved quality
o Works with single-channel or two-channel data format
 HW support to overcome "Small batch" issues in CPU limited applications
 3D resources virtualized to a 32-bit addressing space, for support of large numbers of render targets
and textures
 Support for up to 16k x 16k textures, including 128-bit/pixel textures
 Software-upgradeable, programmable arbitration logic maximizing memory efficiency
 Fully associative texture, color, and Z cache design
 Hierarchical Z and stencil buffers with early Z Test
 Lossless Z-buffer compression for both Z and stencil
 Fast Z-buffer clear
 Fast color-buffer clear
 Z cache optimized for real-time shadow rendering
 Z and color compression resources virtualized to a 32-bit addressing space, for simultaneous support
of multiple render targets and textures