Texas Instruments TMS3320C5515 - Manual (Page 38)

Power Management www.ti.com
Table 1-23. CPU Clock Domain Idle Requirements (continued)
To Idle the Following Module/Port Requirements Before Going to Idle
XPORT
CPU CPUI must also be set.
DPORT
1.5.3.1.3 Clock Configuration Process
The clock configuration indicates which portions of the CPU clock domain will be idle, and which will be
active. The basic steps to the clock configuration process are:
1. To idle MPORT, DMA controller, LCD DMA, and USB CDMA must not be accessing SARMA or
DARAM. If any DMA is in active, wait for completion of the DMA transfer.
2. Write the desired configuration to the idle configuration register (ICR). Make sure that you use a valid
idle configuration (see Section 1.5.3.1.2 ).
3. Apply the new idle configuration by executing the IDLE instruction. The content of ICR is copied to the
idle status register (ISTR). The bits of ISTR are then propagated through the CPU domain system to
enable or disable the specified clocks. If the CPU domain was idled, then program execution will stop
immediately after the idle instruction. If the CPU domain was not idled, then program execution will
continue past the idle instruction but the appropriate domains will be idle.
The IDLE instruction cannot be executed in parallel with another instruction.
The CPU, DPORT, XPORT, and IPORT domains are enabled automatically by any unmasked interrupts.
There is a logic in the DSP core that enables CPU, DPORT, XPORT, and IPORT (clears the bits 0, 5, 6,
and 8 of the ISTR register) asynchronously upon detecting an interrupt signal. Therefore, when an
unmasked interrupt signal reaches the DSP core, these domains are un-idled automatically. Once the
CPU is enabled, it takes 3 CPU cycles to detect the interrupt in the IFR. Note that HWA and MPORT have
to be manually enabled after being disabled.
1.5.3.2 Peripheral Domain Clock Gating
The peripheral clock gating allows software to disable clocks to the DSP peripherals, in order to reduce
the peripheral's active power consumption to zero. Aside from the analog logic, the DSP is designed in
static CMOS; thus, when a peripheral clock stops, the peripheral's state is preserved, and no active
current is consumed. When the clock is restarted the peripheral resumes operating from the stopping
point.
NOTE: Stopping clocks to a peripheral only affects active power consumption; it does not affect
leakage power consumption.
If a peripheral's clock is stopped while being accessed, the access may not occur completely, and could
potentially lock-up the device. To avoid this issue, some peripherals have a clock stop request and
acknowledge protocol that allows software to ask the peripheral when it is safe to stop the clocks. This is
described further in
Section 1.5.3.2.2 . For the peripherals that do not have the request/acknowledge
protocol, the user must ensure that all of the transactions to the peripheral are finished prior to stopping
the clocks.
The procedure to turn peripheral clocks on/off is described in Section 1.5.3.2.3 .
Some peripherals provide additional power saving features by clock gating components within its
peripheral boundary. See the peripheral-specific user's guide for more details on these additional power
saving features.
38 System Control SPRUFX5A – October 2010 – Revised November 2010
Submit Documentation Feedback
Copyright © 2010, Texas Instruments Incorporated