Table 1.
Similarity matrix construction throughput for CPU (SSE4 population count algorithm with row-major and Morton-order) vs GPU (GeForce GTX 480), 1024-bit fingerprints. Similarity matrix for N molecules had shape N×N.
# molecules | Method | Time (s) | Throughput (Tan/s * 106) |
---|---|---|---|
SSE4 | 6.36 | 174.3 | |
32,768 | SSE4-Morton | 5.2 | 214.8 |
GPU | 1.19 | 1088 | |
SSE4 | 124.1 | 139.4 | |
131,072 | SSE4-Morton | 79.98 | 217.0 |
GPU | 15.6 | 1157 |