An Improved Distance Matrix Computation Algorithm for Multicore Clusters

Algorithm 1

DistVect1: Hybrid DistVect Algorithm.

Input: Number of sequences N, average sequences’ length L, sequences’ dataset

𝑆={𝑆₁,𝑆₂,…,𝑆_𝑁}, substitution matrix sbt, gab cost g, Number of Nodes M,

Number of cores P

Output: Distance Vector DV

(1) Let ℎ=0

(2) Let 𝑛_{𝑆𝑝𝑙𝑖𝑡}←[𝑁(𝑁−1)/2𝑀]

(3)                ▹ Distributing splits over M nodes

(4) for 𝑁𝑜𝑑𝑒𝑠←1,𝑀 do in parallel

(5)  Identify Split

(6)  for 𝑃𝑎𝑖𝑟𝑠←1,𝑛_{𝑆𝑝𝑙𝑖𝑡} do

(7)   Get 𝑆_𝑖 and 𝑆_𝑗 from Split

(8)   Initialize 𝑉¹₁,𝑉²₁,𝑁¹₁ and 𝑁²₁

(9)   for 𝑥←1,𝐿+𝑃 do

(10)          ▹ Distributing distance computation over P cores

(11)    for 𝑘←1,𝑃 do in parallel

(12)     Compute 𝑉^𝑥(𝑘) using (3)

(13)     if 𝑆_𝑖(𝑘)=𝑆_𝑗(𝑥−𝑘+1) then Let 𝑚(𝑘)=1

(14)     else Let 𝑚(𝑘)=0

(15)     end if

(16)     Compute 𝑁^𝑥_𝑣(𝑘) using (4)

(17)    end for 𝑘

(18)    Let 𝑉^𝑥₁=𝑉^𝑥₂ and 𝑉^𝑥₂=𝑉^𝑥

(19)    Let 𝑁^𝑥₁=𝑁^𝑥₂ and 𝑁^𝑥₂=𝑁^𝑥

(20)   end for 𝑥

(21)   Identify 𝑘_max

(22)   Let DV(ℎ)=1−𝑘_max/𝐿

(23)   Increment ℎ

(24) end for Pairs

(25) end for Nodes

(26) return DV