Below we present videos demonstrating how LegUp can speed up a Mandelbrot fractal
computation. There are three cases as described below.
Please play all three videos at the same time for a visual comparison.
As you will see, the final LegUp optimized version (last video) is significantly faster than the others.
We targeted and ran these designs on the Altera
DE1-SoC development and education board
(with a Cyclone V FPGA and on-chip ARM Cortex-A9 processor).
Baseline software-only implementation running on an Arm-A9 on-chip processor (800MHz).
Hardware-only FPGA implementation synthesized automatically using LegUp.
Hardware-only FPGA implementation synthesized automatically using LegUp.
But with additional optimizations including local memories, loop transformations, and loop pipelining.