Intel PXA270, Руководство по оптимизации

Страница: 51 / 144

Intel® PXA27x Processor Family Optimization Guide 4-3
Intel XScale® Microarchitecture & Intel® Wireless MMX™ Technology Optimization
if (a > 10)
return 0;
else
return 1;
}
The code generated for the if-else portion of this code segment using branches is:
cmp r0, #10
ble L1
mov r0, #0
b L2
L1:
mov r0, #1
L2:
This code takes three cycles to execute the else statement and four cycles for the if statement
assuming best case conditions and no branch misprediction penalties. In the case of the Intel
XScale® Microarchitecture, a branch misprediction incurs a penalty of four cycles. If the branch is
mispredicted 50% of the time and if both the if statement and the else statement are equally likely
to be taken, on an average the code above takes 5.5 cycles to execute.
.
Using the Intel XScale® Microarchitecture to execute instructions conditionally, the code
generated for the preceding if - else statement is:
cmp r0, #10
movgt r0, #0
movle r0, #1
The preceding code segment would not incur any branch misprediction penalties and would take
three cycles to execute assuming best case conditions. Using conditional instructions speeds up
execution significantly. However, the use of conditional instructions should be considered carefully
to ensure it improve performance. To decide when to use conditional instructions over branches,
consider this hypothetical code segment:
if (cond)
if_stmt
else
else_stmt
Using the following data:
N1
B
Number of cycles to execute the if_stmt assuming the use of branch instructions
N2
B
Number of cycles to execute the else_stmt assuming the use of branch instructions
50
100
--------- 4
3 4 +
2
------------ +
×
 
 
5.5 = cycles