MATLAB COMPILER RELEASE NOTES Guide de l'utilisateur télécharger pdf (Page 175)

Optimizing Loops

6-7

Optimizing Loops

Simple Indexing

(array_indexing) T his optim ization improves the performance of simple one-

and two-dimensional array index expressions. Without this optimization, all

array indexing uses the fully general array indexing function, which is not

optimized for one- and two-dimensional indexi ng. With this optimization

enabled, indexing uses faster routines that are optimized for simple indexing.

For example,

function y = test(x,i1,i2);

y = x(i1,i2);

If you compile this with the -O none option, you get

...

mlfAssign(

&y,

mlfIndexRef(mclVsa(x, "x"), "(?,?)", mclVsa(i1, "i1"),

mclVsa(i2, "i2")));

...

Compiling wit h -O none -O array_indexing:on gives

...

mlfAssign(

&y, mclArrayRef2(mclVsa(x, "x"), mclVsa(i1, "i1"),

mclVsa(i2,"i2")));

...

The mclArrayRef2 function is optimized for two-dimensional indexing.

mclArrayRef1 is used for one-dimensional i ndexing.

Loop Simplification

(optimize_integer_for_loops) This optimization detects when a loop starts

and increments with int egers and replaces the loop with a much simpler loop

that uses C integer variables instead of array valued variables. The

performance improvements with this optimization can be dramatic.

1 2 ... 170 171 172 173 174 175 176 177 178 179 180 ... 263 264

Pas de commentaire

MATLAB COMPILER RELEASE NOTES Guide de l'utilisateur Page 175