Table 4 Performance and resource usage comparison for the different OpenCL kernel implementations
From: SWIFOLD: Smith-Waterman implementation on FPGA with OpenCL for long DNA sequences
Kernel | int_bw256 | int_bw512 | int_bw1024 | int_bw1152 | short_bw512 | short_bw1024 | short_bw1536 | char_bw512 | char_bw1024 | char_bw1536 | |
|---|---|---|---|---|---|---|---|---|---|---|---|
Integer type | int (32 bits) | short (16 bits) | char (8 bits) | ||||||||
Maximum value | 2147483647 | 32767 | 127 | ||||||||
BW | 256 | 512 | 1024 | 1152 | 512 | 1024 | 1536 | 512 | 1024 | 1536 | |
Resource | ALMs | 29% | 49% | 87% | 94% | 32% | 52% | 73% | 21% | 31% | 41% |
usage | Regs | 3% | 3% | 4% | 4% | 3% | 4% | 5% | 3% | 4% | 4% |
RAM | 8% | 8% | 20% | 22% | 7% | 18% | 27% | 7% | 18% | 23% | |
DSPs | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | |
Matrix size (cells) | Performance (GCUPS) | ||||||||||
100K | 24.15 | 31.57 | 44.99 | 49.81 | 48.00 | 52.35 | 56.92 | - | - | - | |
3M | 34.94 | 61.59 | 101.89 | 105.14 | 80.71 | 122.72 | 160.44 | 93.03 | 152.75 | 223.1 | |
28M | 36.70 | 68.11 | 119.15 | 122.91 | 85.96 | 146.80 | 186.74 | 102.50 | 173.23 | 255.49 | |
291M | 37.32 | 69.23 | 122.32 | 126.95 | 87.18 | 149.90 | 195.17 | 105.14 | 181.16 | 268.83 | |
1G | 37.42 | 70.13 | 124.93 | 129.44 | - | - | - | - | - | - | |
9G | 37.84 | 70.80 | 126.96 | 131.45 | 88.40 | 155.85 | 202.56 | - | - | - | |
25G | 37.91 | 70.92 | 127.49 | 131.96 | - | - | - | - | - | - | |
35G | 37.93 | 70.94 | 127.47 | 131.98 | 88.71 | 156.43 | 203.51 | - | - | - | |
100G | 37.98 | 70.99 | 127.68 | 132.15 | - | - | - | - | - | - | |
575G | 38.03 | 71.09 | 127.85 | 132.33 | 88.87 | 156.83 | 204.06 | - | - | - | |