Intel PXA270, Руководство по оптимизации

Страница: 91 / 144

Intel® PXA27x Processor Family Optimization Guide 4-43
Intel XScale® Microarchitecture & Intel® Wireless MMX™ Technology Optimization
4.8.11 Thumb* Instructions
In general, the timing of THUMB* instructions is the same as their equivalent ARM* instructions,
except for these cases:
•
If the equivalent ARM* instruction maps to an entry in Table 4-3 , the “Minimum Issue
Latency with branch misprediction” goes from 5 to 6 cycles. This is due to the branch latency
penalty.
•
If the equivalent ARM* instruction maps to one in Table 4-4 , the “Minimum Issue Latency
when the Branch is Taken” increases by one cycle. This is due to the branch latency penalty.
•
The timings of a THUMB* BL instruction and an ARM* data processing instruction when
H=0 are the same.
The mapping of THUMB* instructions to ARM* instructions can be found in the
ARM*Architecture Reference Manual
4.9 Instruction Latencies for Intel® Wireless MMX™
Technology
The issue cycle and result latency of all the PXA27x processor instructions is shown in Table 4-18 .
In this table, the issue cycle is the number of cycles that an instruction takes to leave the register
file. The result latency is the number of cycles required to calculate the result and make it
available to the bypassing logic. A result latency of 1 indicates that the value is available
immediately to the following instruction. Table 4-18 shows the best case result latency that can be
degraded by data or resource hazards.
Table 4-18. Issue Cycle and Result Latency of the PXA27x processor Instructions (Sheet 1 of
2)
Instructions Issue Cycle Result Latency
WADD 1 1
WSUB 1 1
WCMPEQ 1 2
WCMPGT 1 2
WAND 1 1
WANDN 1 1
WOR 1 1
WXOR 1 1
WAVG2 1 1
WMAX 1 2
WMIN 1 2
WSAD 1 1
WACC 1 1
WMUL 1 1
WMADD 1 1