Skip to main content

Table 1 OpenCL kernel performance for RS with different lengths

From: Coupling SIMD and SIMT architectures to boost performance of a phylogeny-aware alignment kernel

  Execution times GCUPS Speedup GPU vs
Reference length Seq SSE(4) GPU Seq SSE(4) GPU Seq SSE(4)
500 40.79 1.11 1.89 0.49 18.01 8.4 21.58 0.59
1000 90.45 2.58 2.5 0.44 15.52 14.4 36.18 1.03
5000 494.65 14.30 9.23 0.40 14.30 21.2 53.59 1.55
10000 1006.1 29.27 18.31 0.40 13.66 21.6 54.95 1.60
50000 5103.4 319.92 90.95 0.39 6.25 21.9 56.11 3.52
100000 10369 1785.8 181.31 0.38 2.24 22.0 57.19 9.85
500000 51448 9005.3 906.21 0.39 2.22 22.1 56.77 9.94
  1. Execution times (in seconds) to align 1250 QS to 320 RS.