Automatic vectorization

Automatic vectorization, in <a href="/facts/Parallel_computing/nQzcpzQt">parallel computing</a>, is a special case of <a href="/facts/Automatic_parallelization/t40BDoWG">automatic parallelization</a>, where a <a href="/facts/Computer_program/e38jH9N9">computer program</a> is converted from a <a href="/facts/Scalar_(computing)/VoCKtwml">scalar</a> implementation, which processes a single pair of <a href="/facts/Operand/HSuUzrKz">operands</a> at a time, to a <a href="/facts/Vector_(data_structure)/BKGvF8AM">vector</a> implementation, which processes one operation on multiple pairs of operands at once. For example, modern conventional computers, including specialized <a href="/facts/Supercomputers/HT9NwGAN">supercomputers</a>, typically have <a href="/facts/Vector_processing/nYAsIKT0">vector operations</a> that simultaneously perform operations such as the following four additions (via <a href="/facts/SIMD/gQHWQSpo">SIMD</a> or <a href="/facts/SPMD/SEDIrB0l">SPMD</a> hardware):

c
                  
                    1
                  
                
              
              
                
                =
                
                  a
                  
                    1
                  
                
                +
                
                  b
                  
                    1
                  
                
              
            
            
              
                
                  c
                  
                    2
                  
                
              
              
                
                =
                
                  a
                  
                    2
                  
                
                +
                
                  b
                  
                    2
                  
                
              
            
            
              
                
                  c
                  
                    3
                  
                
              
              
                
                =
                
                  a
                  
                    3
                  
                
                +
                
                  b
                  
                    3
                  
                
              
            
            
              
                
                  c
                  
                    4
                  
                
              
              
                
                =
                
                  a
                  
                    4
                  
                
                +
                
                  b
                  
                    4
                  
                
              
            
          
        
      
    
    {\displaystyle {\begin{aligned}c_{1}&=a_{1}+b_{1}\\c_{2}&=a_{2}+b_{2}\\c_{3}&=a_{3}+b_{3}\\c_{4}&=a_{4}+b_{4}\end{aligned}}}

However, in most <a href="/facts/Programming_language/cWTYbgWa">programming languages</a> one typically writes loops that sequentially perform additions of many numbers. Here is an example of such a loop, written in <a href="/facts/C_(programming_language)/Ky2No763">C</a>:

for (i = 0; i < n; i++)
 c[i] = a[i] + b[i];

A vectorizing <a href="/facts/Compiler/WNfCDFJe">compiler</a> transforms such loops into sequences of vector operations. These vector operations perform additions on blocks of elements from the arrays a, b and c. Automatic vectorization is a major research topic in computer science.

Automatic vectorization open-in-new

Automatic vectorization