Assume a polygon mesh contains 250000 vertices if a single m
Assume a polygon mesh contains 250,000 vertices, if a single matrix multiplication requires 28 floating point operations, how fast a GPU is needed (floating point operations per second) to produce real-time graphics at the rate of 30 frames per second?
Solution
 A polygon mesh approximates the surface shape
 of an object by specifying a set of points in
 space these points representing vertices of
 various polygonal faces.Matrix multiplication is a fundamental
 building block for scientific computing. Moreover, the algorithmic
 patterns of matrix multiplication are representative.
Many other algorithms share similar optimization techniques as matrix multiplication.
 Therefore, matrix multiplication is one of the most important examples
 in learning parallel programming.
 The ever-increasing power,
 programmability, and precision of
 GPUs has motivated a great deal of
 research on general-purpose computation
 on graphics hardware.Researchers and
 developers use the GPU as a computational
 coprocessor rather than as an
 image-synthesis device.
The GPU’s specialized architecture
 isn’t well suited to every algorithm.
 Many applications are inherently serial
 and are characterized by incoherent
 and unpredictable memory access.
 Many important problems
 require significant computational.
 resources, mapping well to the GPU’s
 many-core arithmetic intensity, or
 they require streaming through large
 quantities of data, mapping well to the
 GPU’s streaming memory subsystem.
 Porting a judiciously chosen algorithm
 to the GPU often produces
 speedups of five to 20 times over
 mature, optimized CPU codes running
 on state-of-the-art CPUs, and speedups
 of more than 100 times have been
 reported for some algorithms that
 map especially well.

