Table 1.
eQTL scan runtimes for R/qtl and LiteQTL
| Dataset | Precision | LiteQTL |
R/qtl | LiteQTL speedup vs R/qtl |
|||
|---|---|---|---|---|---|---|---|
| CPU only | CPU and GPU | GPU speedup | CPU only | CPU and GPU | |||
| Spleen (35K traits) | Single | 0.56 s | 0.04 s | 14.0x | — | — | — |
| Double | 0.83 s | 0.06 s | 13.8x | 40.02 s | 48.2x | 667.0x | |
| Hippocampus (1.2M traits) | Single | 12.36 s | 1.66 s | 7.4x | — | — | — |
| Double | 16.57 s | 2.82 s | 5.4x | 2070.71 s | 124.9x | 734.2x | |
We show the time taken in seconds to perform eQTL genome scans with LiteQTL and R/qtl. LiteQTL times are shown by precision (single vs double), and whether the GPU was used or not. The GPU speedup column computes the speedup for LiteQTL using the CPU and GPU vs using the CPU only. LiteQTL was used with the maximum LOD output option to reduce data transfer time.