Texture Cache Approximation on GPUs
Dept. of Electrical and Computer Engineering, University of Toronto
Workshop on Approximate Computing Across the Stack (WAX), 2015
We present texture cache approximation as a method for using existing hardware on GPUs to eliminate costly global memory accesses. We develop a technique for using a GPU’s texture fetch units to generate approximate values, and argue that this technique is applicable to a wide variety of GPU kernels. Applying texture cache approximation to an image blur kernel on an NVIDIA 780GTX, we obtain a 12% reduction in kernel execution time while only introducing 0.4% output error in the final image.
July 3, 2015 by hgpu
