Loop nesting: why can large loops inside and small loops outside improve the running efficiency of programs

A lot of people think , Loop nesting can be nested arbitrarily , Regardless of the number of cycles ;

But that's not the case ;

Programs are for computers , Eventually, it compiles and generates instructions that the computer can recognize ;

Today's computers in order to improve efficiency , The instruction execution mechanism will have branches of preprocessing instructions ;

In the case of loops, there are branch instructions , For example, a 100 The cycle of time , Except for the first and last prediction instruction errors , Other predicted instructions can predict the instructions in the loop ; It avoids the waste of time caused by fetching new instructions again when executing branch instructions .

For the two A and B For two cycles , When A Far greater than B When , Loops can be nested in two ways

Case one ：A Outside ,B , , Every time you cycle ,B There will be two errors , Co cycle A Time ;A The loop itself also has two mispredictions , So the final instruction prediction error is A*2 +2 ;

The second case ：B Outside ,A , , Each cycle will occur 2 Time prediction error , Co cycle B Time ;B There are also two errors in the loop itself , therefore , The resulting instruction prediction error is B*2 + 2;

because A Far greater than B, So finally, the first method predicts the number of errors A*2 +2＞B*2 + 2; In the first case, fetching new instructions causes more time waste ;

therefore , When dealing with loop nesting ,“ The big cycle is inside , The small loop is out ” It's a good choice ;