Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: CUDAMPF: a multi-tiered parallel framework for accelerating protein sequence search in HMMER on CUDA-enabled GPU

Fig. 4

Comparison of alignment with sequential and striped layout. An example of MSV score alignment with HMM model of L hmm = 382. (a) 32 threads of w a r p1 concurrently calculate 128 8-bit scores at each iteration, and every 4 scores are stored together as one 32-bit datatype. Orange-colored boundaries indicate private regions of different threads, where threads only take charge of cells within their own regions. (b) a warp of threads process model states with striped interval Q=3 that is also the number of iteration required for the current row. Yellow-colored cells represent the result after one 8-bit shifting from q=Q−1=2, which will be used to calculate q=0 for next residue

Back to article page