Efficient CSR-Based Sparse Matrix-Vector Multiplication on GPU

Algorithm 2

Kernel II in PCSR.

Input: , ;

Output: res;

()    ;

()    ;

()    ;

()    ;

()    ;

()    if()

       ;

()    ();

()    ; ; ;

()    for to with   +=   do

()    ;

()    ();

       //Load temp into the shared memory temp_

()    for to do

()      if then

()       ;

()         +=  ;

()      end

()    done

()    ();

       //Perform a scalar-style reduction

()    if ) is false then

()      ;

()      ;

()      for to do

()         +=  ;

()      done

()    end

()    ;

() done