Exploring Trade-Offs between Specialized Dataflow Kernels and a Reusable Overlay in a Stereo Matching Case Study

<table class="table-group" id="tab6"><tr><td><table class="table"><tr><td class="thead-hr" colspan="3"><hr/></td></tr><tr class="thead"><td align="left">  Architecture </td><td align="center"> All sizes </td><td align="center"> SXGA-high </td></tr><tr><td class="thead-hr" colspan="3"><hr/></td></tr><tr><td align="left"> MaxKern </td><td align="center"> 6.59 </td><td align="center"> 9.86 </td></tr><tr><td align="left">MaxFused </td><td align="center"> 5.20 </td><td align="center"> 7.76 </td></tr><tr><td align="left">CnyVecTouch </td><td align="center"> 5.80 </td><td align="center"> 8.37 </td></tr><tr class="table-tr"><td colspan="3"><hr class="tbody-hr"/></td></tr></table></td></tr></table>

Kernel-Ratios (geometric mean of all kernel speedups) relative to CPU1 for the SXGA image with high disparity and for the geometric mean over all sizes.

International Journal of Reconfigurable Computing

tab6

Table 6

Table 6: Exploring Trade-Offs between Specialized Dataflow Kernels and a Reusable Overlay in a Stereo Matching Case Study 