Squeeze: Efficient compact fractals for tensor core GPUs

This work presents Squeeze, an efficient compact fractal processing scheme for tensor core GPUs. By combining discrete-space transformations between and expanded forms, one can do data-parallel computation on a with neighborhood access without needing to expand the in memory. The space are formulated as two GPU tensor-core accelerated thread maps, ?(?) ?(?), which act compact-to-expanded expanded-to-compact functions, respectively. cost of maps is O(log2logs(n)) time, n being side n×n embedding its form, s linear scaling factor. proposed approach works any that belongs Non-overlapping-Bounding-Boxes (NBB) class discrete fractals, be extended three dimensions well. Experimental results using Sierpinski Triangle case study shows up ?12× speedup memory reduction factor ?315× respect GPU-based expanded-space bounding box approach. These show will allow scientific community efficiently tackle problems now could not fit into