Research Article
Transparent Runtime Migration of Loop-Based Traces of Processor Instructions to Reconfigurable Processing Units
Table 4
Execution times for several implementations of the pattern detector for megablocks.
|
# Addrs | Execution times (ms) | | HM@50 MHz | MB@50 MHz | Cortex-A8@1 GHz |
Speedup (HM versus MB/HM versus A8) |
| 12 | 0.0002 | 2.7 | 0.6 | 11,251/2,500 | 24 | 0.0005 | 5.7 | 1.3 | 11,963/2,708 | 48 | 0.0010 | 14.0 | 2.8 | 14,594/2,917 | 96 | 0.0019 | 30.8 | 5.9 | 16,036/3,073 | 192 | 0.0038 | 64.3 | 12.5 | 16,757/3,255 | 384 | 0.0077 | 131.5 | 24.8 | 17,118/3,229 | 768 | 0.0154 | 265.7 | 78.7 | 17,298/5,124 |
|
|