/**
* @file rtems/score/cpu.h
*/
/*
* $Id$
*
* This include file contains information pertaining to the ARM
* processor.
*
* Copyright (c) 2007 Ray Xu <Rayx.cn@gmail.com>
*
* Copyright (c) 2006 OAR Corporation
*
* Copyright (c) 2002 Advent Networks, Inc.
* Jay Monkman <jmonkman@adventnetworks.com>
*
* COPYRIGHT (c) 2000 Canon Research Centre France SA.
* Emmanuel Raguet, mailto:raguet@crf.canon.fr
*
* The license and distribution terms for this file may be
* found in the file LICENSE in this distribution or at
* http://www.rtems.com/license/LICENSE.
*
*/
/* FIXME: finish commenting/cleaning up this file */
#ifndef _RTEMS_SCORE_CPU_H
#define _RTEMS_SCORE_CPU_H
#ifdef __cplusplus
extern "C" {
#endif
#include <rtems/score/arm.h> /* pick up machine definitions */
#ifndef ASM
#include <rtems/score/types.h>
#endif
/* conditional compilation parameters */
/*
* Should the calls to _Thread_Enable_dispatch be inlined?
*
* If TRUE, then they are inlined.
* If FALSE, then a subroutine call is made.
*
* Basically this is an example of the classic trade-off of size
* versus speed. Inlining the call (TRUE) typically increases the
* size of RTEMS while speeding up the enabling of dispatching.
* [NOTE: In general, the _Thread_Dispatch_disable_level will
* only be 0 or 1 unless you are in an interrupt handler and that
* interrupt handler invokes the executive.] When not inlined
* something calls _Thread_Enable_dispatch which in turns calls
* _Thread_Dispatch. If the enable dispatch is inlined, then
* one subroutine call is avoided entirely.]
*/
#if defined(__thumb__)
#define CPU_INLINE_ENABLE_DISPATCH FALSE
#else
#define CPU_INLINE_ENABLE_DISPATCH TRUE
#endif
/*
* Should the body of the search loops in _Thread_queue_Enqueue_priority
* be unrolled one time? In unrolled each iteration of the loop examines
* two "nodes" on the chain being searched. Otherwise, only one node
* is examined per iteration.
*
* If TRUE, then the loops are unrolled.
* If FALSE, then the loops are not unrolled.
*
* The primary factor in making this decision is the cost of disabling
* and enabling interrupts (_ISR_Flash) versus the cost of rest of the
* body of the loop. On some CPUs, the flash is more expensive than
* one iteration of the loop body. In this case, it might be desirable
* to unroll the loop. It is important to note that on some CPUs, this
* code is the longest interrupt disable period in RTEMS. So it is
* necessary to strike a balance when setting this parameter.
*/
#define CPU_UNROLL_ENQUEUE_PRIORITY TRUE
/*
* Does RTEMS manage a dedicated interrupt stack in software?
*
* If TRUE, then a stack is allocated in _Interrupt_Manager_initialization.
* If FALSE, nothing is done.
*
* If the CPU supports a dedicated interrupt stack in hardware,
* then it is generally the responsibility of the BSP to allocate it
* and set it up.
*
* If the CPU does not support a dedicated interrupt stack, then
* the porter has two options: (1) execute interrupts on the
* stack of the interrupted task, and (2) have RTEMS manage a dedicated
* interrupt stack.
*
* If this is TRUE, CPU_ALLOCATE_INTERRUPT_STACK should also be TRUE.
*
* Only one of CPU_HAS_SOFTWARE_INTERRUPT_STACK and
* CPU_HAS_HARDWARE_INTERRUPT_STACK should be set to TRUE. It is
* possible that both are FALSE for a particular CPU. Although it
* is unclear what that would imply about the interrupt processing
* procedure on that CPU.
*/
#define CPU_HAS_SOFTWARE_INTERRUPT_STACK FALSE
/*
* Does this CPU have hardware support for a dedicated interrupt stack?
*
* If TRUE, then it must be installed during initialization.
* If FALSE, then no installation is performed.
*
* If this is TRUE, CPU_ALLOCATE_INTERRUPT_STACK should also be TRUE.
*
* Only one of CPU_HAS_SOFTWARE_INTERRUPT_STACK and
* CPU_HAS_HARDWARE_INTERRUPT_STACK should be set to TRUE. It is
* possible that both are FALSE for a particular CPU. Although it
* is unclear what that would imply about the interrupt processing
* procedure on that CPU.
*/
#define CPU_HAS_HARDWARE_INTERRUPT_STACK TRUE
/*
* Does RTEMS allocate a dedicated interrupt stack in the Interrupt Manager?
*
* If TRUE, then the memory is allocated during initialization.
* If FALSE, then the memory is allocated during initialization.
*
* This should be TRUE is CPU_HAS_SOFTWARE_INTERRUPT_STACK is TRUE.
*/
#define CPU_ALLOCATE_INTERRUPT_STACK FALSE
/*
* Does the RTEMS invoke the user's ISR with the vector number and
* a pointer to the saved interrupt frame (1) or just the vector
* number (0)?
*/
#define CPU_ISR_PASSES_FRAME_POINTER 0
/*
* Does the CPU have hardware floating point?
*
* If TRUE, then the RTEMS_FLOATING_POINT task attribute is supported.
* If FALSE, then the RTEMS_FLOATING_POINT task attribute is ignored.
*
* If there is a FP coprocessor such as the i387 or mc68881, then
* the answer is TRUE.
*
* The macro name "ARM_HAS_FPU" should be made CPU specific.
* It indicates whether or not this CPU model has FP support. For
* example, it would be possible to have an i386_nofp CPU model
* which set this to false to indicate that you have an i386 without
* an i387 and wish to leave floating point support out of RTEMS.
*/
#if ( ARM_HAS_FPU == 1 )
#define CPU_HARDWARE_FP TRUE
#else
#define CPU_HARDWARE_FP FALSE
#endif
#define CPU_SOFTWARE_FP FALSE
/*
* Are all tasks RTEMS_FLOATING_POINT tasks implicitly?
*
* If TRUE, then the RTEMS_FLOATING_POINT task attribute is assumed.
* If FALSE, then the RTEMS_FLOATING_POINT task attribute is followed.
*
* If CPU_HARDWARE_FP is FALSE, then this should be FALSE as well.
*/
#define CPU_ALL_TASKS_ARE_FP FALSE
/*
* Should the IDLE task have a floating point context?
*
* If TRUE, then the IDLE task is created as a RTEMS_FLOATING_POINT task
* and it has a floating point context which is switched in and out.
* If FALSE, then the IDLE task does not have a floating point context.
*
* Setting this to TRUE negatively impacts the time required to preempt
* the IDLE task from an interrupt because the floating point context
* must be saved as part of the preemption.
*/
#define CPU_IDLE_TASK_IS_FP FALSE
/*
* Should the saving of the floating point registers be deferred
* until a context switch is made to another different floating point
* task?
*
* If TRUE, then the floating point context will not be stored until
* necessary. It will remain in the floating point registers and not
* disturned until another floating point task is switched to.
*
* If FALSE, then the floating point context is saved when a floating
* point task is switched out and restored when the next floating point
* task is restored. The state of the floating point registers between
* those two operations is not specified.
*
* If the floating point context does NOT have to be saved as part of
* interrupt dispatching, then it should be safe to set this to TRUE.
*
* Setting this flag to TRUE results in using a different algorithm
* for deciding when to save and restore the floating point context.
* The deferred FP switch algorithm minimizes the number of times
* the FP context is saved and restored. The FP context is not saved
* until a context switch is made to another, different FP task.
* Thus in a system with only one FP task, the FP context will never
* be saved or restored.
*/
#define CPU_USE_DEFERRED_FP_SWITCH FALSE
/*
* Does this port provide a CPU dependent IDLE task implementation?
*
* If TRUE, then the routine _CPU_Thread_Idle_body
* must be provided and is the default IDLE thread body instead of
* _CPU_Thread_Idle_body.
*
* If FALSE, then use the generic IDLE thread body if the BSP does
* not provide one.
*
* This is intended to allow for supporting processors which have
* a low power or idle mode. When the IDLE thread is executed, then
* the CPU can be powered down.
*
* The order of precedence for selecting the IDLE thread body is:
*
* 1. BSP provided
* 2. CPU dependent (if provided)
* 3. generic (if no BSP and no CPU dependent)
*/
#define CPU_PROVIDES_IDLE_THREAD_BODY FALSE
/*
* Does the stack grow up (toward higher addresses) or down
* (toward lower addresses)?
*
* If TRUE, then the grows upward.
* If FALSE, then the grows toward smaller addresses.
*/
#define CPU_STACK_GROWS_UP FALSE
/*
* The following is the variable attribute used to force alignment
* of critical RTEMS structures. On some processors it may make
* sense to have these aligned on tighter boundaries than
* the minimum requirements of the compiler in order to have as
* much of the critical data area as possible in a cache line.
*
* The placement of this macro in the declaration of the variables
* is based on the syntactically requirements of the GNU C
* "__attribute__" extension. For example with GNU C, use
* the following to force a structures to a 32 byte boundary.
*
* __attribute__ ((aligned (32)))
*
* NOTE: Currently only the Priority Bit Map table uses this feature.
* To benefit from using this, the data must be heavily
* used so it will stay in the cache and used frequently enough
* in the executive to justify turning this on.
*/
#define CPU_STRUCTURE_ALIGNMENT __attribute__ ((aligned (32)))
/*
* Define what is required to specify how the network to host conversion
* routines are handled.
*/
#if defined(__ARMEL__)
#define CPU_BIG_ENDIAN FALSE
#define CPU_LITTLE_ENDIAN TRUE
#elif defined(__ARMEB__)
#define CPU_BIG_ENDIAN TRUE
#define CPU_LITTLE_ENDIAN FALSE
#else
#error "Unknown endianness"
#endif
/*
* The following defines the number of bits actually used in the
* interrupt field of the task mode. How those bits map to the
* CPU interrupt levels is defined by the routine _CPU_ISR_Set_level().
*/
#define CPU_MODES_INTERRUPT_MASK 0x000000c0
/*
* Processor defined structures required by cpukit/score.
*/
/* may need to put some structures here. */
/*
* Contexts
*
* Generally there are 2 types of context to save.
* 1. Interrupt registers to save
* 2. Task level registers to save
*
* This means we have the following 3 context items:
* 1. task level context stuff:: Context_Control
* 2. floating point task stuff:: Context_Control_fp
* 3. special interrupt level context :: Context_Control_interrupt
*
* On some processors, it is cost-effective to save only the callee
* preserved registers during a task context switch. This means
* that the ISR code needs to save those registers which do not
* persist across function calls. It is not mandatory to make this
* distinctions between the caller/callee saves registers for the
* purpose of minimizing context saved during task switch and on interrupts.
* If the cost of saving extra registers is minimal, simplicity is the
* choice. Save the same context on interrupt entry as for tasks in
* this case.
*
* Additionally, if gdb is to be made aware of RTEMS tasks for this CPU, then
* care should be used in designing the context area.
*
* On some CPUs with hardware floating point support, the Context_Control_fp
* structure will not be used or it simply consist of an array of a
* fixed number of bytes. This is done when the floating point context
* is dumped by a "FP save context" type instruction and the format
* is not really defined by the CPU. In this case, there is no need
* to figure out the exact format -- only the size. Of course, although
* this is enough information for RTEMS, it is probably not enough for
* a debugger such as gdb. But that is another problem.
*/
typedef struct {
uint32_t register_cpsr;
uint32_t register_r4;
uint32_t register_r5;
uint32_t register_r6;
uint32_t register_r7;
uint32_t register_r8;
uint32_t register_r9;
uint32_t register_r10;
uint32_t register_fp;
uint32_t register_sp;
uint32_t register_lr;
uint32_t register_pc;
} Context_Control;
typedef struct {
double some_float_register;
} Context_Control_fp;
typedef struct {
uint32_t register_r0;
uint32_t register_r1;
uint32_t register_r2;
uint32_t register_r3;
uint32_t register_ip;
uint32_t register_lr;
} CPU_Exception_frame;
typedef void (*cpuExcHandlerType) (CPU_Exception_frame*);
extern cpuExcHandlerType _currentExcHandler;
extern void rtems_exception_init_mngt();
/*
* The following structure defines the set of information saved
* on the current stack by RTEMS upon receipt of each interrupt
* that will lead to re-enter the kernel to signal the thread.
*/
typedef CPU_Exception_frame CPU_Interrupt_frame;
/*
* The following table contains the information required to configure
* the XXX processor specific parameters.
*/
typedef struct {
uint32_t interrupt_stack_size;
} rtems_cpu_table;
/*
* This variable is optional. It is used on CPUs on which it is difficult
* to generate an "uninitialized" FP context. It is filled in by
* _CPU_Initialize and copied into the task's FP context area during
* _CPU_Context_Initialize.
*/
SCORE_EXTERN Context_Control_fp _CPU_Null_fp_context;
/*
* The size of the floating point context area. On some CPUs this
* will not be a "sizeof" because the format of the floating point
* area is not defined -- only the size is. This is usually on
* CPUs with a "floating point save context" instruction.
*/
#define CPU_CONTEXT_FP_SIZE sizeof( Context_Control_fp )
/*
* Amount of extra stack (above minimum stack size) required by
* MPCI receive server thread. Remember that in a multiprocessor
* system this thread must exist and be able to process all directives.
*/
#define CPU_MPCI_RECEIVE_SERVER_EXTRA_STACK 0
/*
* This defines the number of entries in the ISR_Vector_table managed
* by RTEMS.
*/
#define CPU_INTERRUPT_NUMBER_OF_VECTORS 8
#define CPU_INTERRUPT_MAXIMUM_VECTOR_NUMBER (CPU_INTERRUPT_NUMBER_OF_VECTORS - 1)
/*
* This is defined if the port has a special way to report the ISR nesting
* level. Most ports maintain the variable _ISR_Nest_level.
*/
#define CPU_PROVIDES_ISR_IS_IN_PROGRESS FALSE
/*
* Should be large enough to run all RTEMS tests. This ensures
* that a "reasonable" small application should not have any problems.
*/
#define CPU_STACK_MINIMUM_SIZE (1024*4)
/*
* CPU's worst alignment requirement for data types on a byte boundary. This
* alignment does not take into account the requirements for the stack.
*/
#define CPU_ALIGNMENT 4
/*
* This number corresponds to the byte alignment requirement for the
* heap handler. This alignment requirement may be stricter than that
* for the data types alignment specified by CPU_ALIGNMENT. It is
* common for the heap to follow the same alignment requirement as
* CPU_ALIGNMENT. If the CPU_ALIGNMENT is strict enough for the heap,
* then this should be set to CPU_ALIGNMENT.
*
* NOTE: This does not have to be a power of 2. It does have to
* be greater or equal to than CPU_ALIGNMENT.
*/
#define CPU_HEAP_ALIGNMENT CPU_ALIGNMENT
/*
* This number corresponds to the byte alignment requirement for memory
* buffers allocated by the partition manager. This alignment requirement
* may be stricter than that for the data types alignment specified by
* CPU_ALIGNMENT. It is common for the partition to follow the same
* alignment requirement as CPU_ALIGNMENT. If the CPU_ALIGNMENT is strict
* enough for the partition, then this should be set to CPU_ALIGNMENT.
*
* NOTE: This does not have to be a power of 2. It does have to
* be greater or equal to than CPU_ALIGNMENT.
*/
#define CPU_PARTITION_ALIGNMENT CPU_ALIGNMENT
/*
* This number corresponds to the byte alignment requirement for the
* stack. This alignment requirement may be stricter than that for the
* data types alignment specified by CPU_ALIGNMENT. If the CPU_ALIGNMENT
* is strict enough for the stack, then this should be set to 0.
*
* NOTE: This must be a power of 2 either 0 or greater than CPU_ALIGNMENT.
*/
#define CPU_STACK_ALIGNMENT 4
/* ISR handler macros */
/*
* Support routine to initialize the RTEMS vector table after it is allocated.
*/
#define _CPU_Initialize_vectors()
/*
* Disable all interrupts for an RTEMS critical section. The previous
* level is returned in _level.
*/
#if (defined(__THUMB_INTERWORK__) || defined(__thumb__))
extern uint32_t _CPU_ISR_Disable_Thumb(void) __attribute__ ((naked));
extern void _CPU_ISR_Enable_Thumb( int ) __attribute__ ((naked));
extern void _CPU_ISR_Flash_Thumb(int) __attribute__ ((naked));
extern void _CPU_ISR_Set_level_Thumb(int ) __attribute__ ((naked));
extern uint32_t _CPU_ISR_Get_level_Thumb(void ) __attribute__ ((naked));
#define _CPU_ISR_Disable(_level) \
(_level) = _CPU_ISR_Disable_Thumb()
#define _CPU_ISR_Enable(a) _CPU_ISR_Enable_Thumb(a)
#define _CPU_ISR_Flash(a) _CPU_ISR_Flash_Thumb(a)
#define _CPU_ISR_Set_level(a) _CPU_ISR_Set_level_Thumb(a)
#define _CPU_ISR_Get_level(a) _CPU_ISR_Get_level_Thumb(a)
#else /*For ARM mode*/
#define _CPU_ISR_Disable( _level ) \
{ \
int reg; \
asm volatile ("MRS %0, cpsr \n" \
"ORR %1, %0, #0xc0 \n" \
"MSR cpsr, %1 \n" \
: "=&r" (_level), "=&r" (reg)); \
}
/*
* Enable interrupts to the previous level (returned by _CPU_ISR_Disable).
* This indicates the end of an RTEMS critical section. The parameter
* _level is not modified.
*/
#define _CPU_ISR_Enable( _level ) \
{ \
asm volatile ("MSR cpsr, %0 \n" \
: : "r" (_level)); \
}
/*
* This temporarily restores the interrupt to _level before immediately
* disabling them again. This is used to divide long RTEMS critical
* sections into two or more parts. The parameter _level is not
* modified.
*/
#define _CPU_ISR_Flash( _level ) \
{ \
int reg; \
asm volatile ("MRS %0, cpsr \n" \
"MSR cpsr, %1 \n" \
"MSR cpsr, %0 \n" \
: "=&r" (reg) \
: "r" (_level)); \
}
/*
* Map interrupt level in task mode onto the hardware that the CPU
* actually provides. Currently, interrupt levels which do not
* map onto the CPU in a generic fashion are undefined. Someday,
* it would be nice if these were "mapped" by the application
* via a callout. For example, m68k has 8 levels 0 - 7, levels
* 8 - 255 would be available for bsp/application specific meaning.
* This could be used to manage a programmable interrupt controller
* via the rtems_task_mode directive.
*
* The get routine usually must be implemented as a subroutine.
*/
#define _CPU_ISR_Set_level( new_level ) \
{ \
int reg = 0; /* to avoid warning */ \
asm volatile ("MRS %0, cpsr \n" \
"BIC %0, %0, #0xc0 \n" \
"ORR %0, %0, %2 \n" \
"MSR cpsr_c, %0 \n" \
: "=r" (reg) \
: "0" (reg), "r" (new_level)); \
}
#endif /*(defined(__THUMB_INTERWORK__) || defined(__thumb__))*/
uint32_t _CPU_ISR_Get_level( void );
/* end of ISR handler macros */
/* Context handler macros */
/*
* Initialize the context to a state suitable for starting a
* task after a context restore operation. Generally, this
* involves:
*
* - setting a starting address
* - preparing the stack
* - preparing the stack and frame pointers
* - setting the proper interrupt level in the context
* - initializing the floating point context
*
* This routine generally does not set any unnecessary register
* in the context. The state of the "general data" registers is
* undefined at task start time.
*
* NOTE: This is_fp parameter is TRUE if the thread is to be a floating
* point thread. This is typically only used on CPUs where the
* FPU may be easily disabled by software such as on the SPARC
* where the PSR contains an enable FPU bit.
*/
void _CPU_Context_Initialize(
Context_Control *the_context,
uint32_t *stack_base,
uint32_t size,
uint32_t new_level,
void *entry_point,
boolean is_fp
);
/*
* This routine is responsible for somehow restarting the currently
* executing task. If you are lucky, then all that is necessary
* is restoring the context. Otherwise, there will need to be
* a special assembly routine which does something special in this
* case. Context_Restore should work most of the time. It will
* not work if restarting self conflicts with the stack frame
* assumptions of restoring a context.
*/
#define _CPU_Context_Restart_self( _the_context ) \
_CPU_Context_restore( (_the_context) );
/*
* The purpose of this macro is to allow the initial pointer into
* a floating point context area (used to save the floating point
* context) to be at an arbitrary place in the floating point
* context area.
*
* This is necessary because some FP units are designed to have
* their context saved as a stack which grows into lower addresses.
* Other FP units can be saved by simply moving registers into offsets
* from the base of the context area. Finally some FP units provide
* a "dump context" instruction which could fill in from high to low
* or low to high based on the whim of the CPU designers.
*/
#define _CPU_Context_Fp_start( _base, _offset ) \
( (void *) _Addresses_Add_offset( (_base), (_offset) ) )
/*
* This routine initializes the FP context area passed to it to.
* There are a few standard ways in which to initialize the
* floating point context. The code included for this macro assumes
* that this is a CPU in which a "initial" FP context was saved into
* _CPU_Null_fp_context and it simply copies it to the destination
* context passed to it.
*
* Other models include (1) not doing anything, and (2) putting
* a "null FP status word" in the correct place in the FP context.
*/
#define _CPU_Context_Initialize_fp( _destination ) \
{ \
*(*(_destination)) = _CPU_Null_fp_context; \
}
/* end of Context handler macros */
/* Fatal Error manager macros */
/*
* This routine copies _error into a known place -- typically a stack
* location or a register, optionally disables interrupts, and
* halts/stops the CPU.
*/
#define _CPU_Fatal_halt( _error ) \
do { \
int _level; \
_CPU_ISR_Disable( _level ); \
asm volatile ("mov r0, %0\n" \
: "=r" (_error) \
: "0" (_error) \
: "r0" ); \
while(1) ; \
} while(0);
/* end of Fatal Error manager macros */
/* Bitfield handler macros */
/*
* If we had a particularly fast function for finding the first
* bit set in a word, it would go here. Since we don't (*), we'll
* just use the universal macros.
*
* (*) On ARM V5 and later, there's a CLZ function which could be
* used to implement much quicker than the default macro.
*/
#define CPU_USE_GENERIC_BITFIELD_CODE TRUE
#define CPU_USE_GENERIC_BITFIELD_DATA TRUE
/* functions */
/*
* _CPU_Initialize
*
* This routine performs CPU dependent initialization.
*/
void _CPU_Initialize(
rtems_cpu_table *cpu_table,
void (*thread_dispatch)
);
typedef enum {
ARM_EXCEPTION_RESET = 0,
ARM_EXCEPTION_UNDEF = 1,
ARM_EXCEPTION_SWI = 2,
ARM_EXCEPTION_PREF_ABORT = 3,
ARM_EXCEPTION_DATA_ABORT = 4,
ARM_EXCEPTION_RESERVED = 5,
ARM_EXCEPTION_IRQ = 6,
ARM_EXCEPTION_FIQ = 7,
MAX_EXCEPTIONS = 8
} Arm_symbolic_exception_name;
/*
* _CPU_ISR_install_vector
*
* This routine installs an interrupt vector.
*/
void _CPU_ISR_install_vector(
uint32_t vector,
proc_ptr new_handler,
proc_ptr *old_handler
);
/*
* _CPU_Install_interrupt_stack
*
* This routine installs the hardware interrupt stack pointer.
*
* NOTE: It need only be provided if CPU_HAS_HARDWARE_INTERRUPT_STACK
* is TRUE.
*/
void _CPU_Install_interrupt_stack( void );
/*
* _CPU_Context_switch
*
* This routine switches from the run context to the heir context.
*/
void _CPU_Context_switch(
Context_Control *run,
Context_Control *heir
);
/*
* _CPU_Context_restore
*
* This routine is generally used only to restart self in an
* efficient manner. It may simply be a label in _CPU_Context_switch.
*
* NOTE: May be unnecessary to reload some registers.
*/
void _CPU_Context_restore(
Context_Control *new_context
);
#if (ARM_HAS_FPU == 1)
/*
* _CPU_Context_save_fp
*
* This routine saves the floating point context passed to it.
*/
void _CPU_Context_save_fp(
Context_Control_fp **fp_context_ptr
);
/*
* _CPU_Context_restore_fp
*
* This routine restores the floating point context passed to it.
*/
void _CPU_Context_restore_fp(
Context_Control_fp **fp_context_ptr
);
#endif /* (ARM_HAS_FPU == 1) */
/* The following routine swaps the endian format of an unsigned int.
* It must be static because it is referenced indirectly.
*
* This version will work on any processor, but if there is a better
* way for your CPU PLEASE use it. The most common way to do this is to:
*
* swap least significant two bytes with 16-bit rotate
* swap upper and lower 16-bits
* swap most significant two bytes with 16-bit rotate
*
* Some CPUs have special instructions which swap a 32-bit quantity in
* a single instruction (e.g. i486). It is probably best to avoid
* an "endian swapping control bit" in the CPU. One good reason is
* that interrupts would probably have to be disabled to ensure that
* an interrupt does not try to access the same "chunk" with the wrong
* endian. Another good reason is that on some CPUs, the endian bit
* endianness for ALL fetches -- both code and data -- so the code
* will be fetched incorrectly.
*/
static inline uint32_t CPU_swap_u32(
uint32_t value
)
{
uint32_t tmp = value; /* make compiler warnings go away */
asm volatile ("EOR %1, %0, %0, ROR #16\n"
"BIC %1, %1, #0xff0000\n"
"MOV %0, %0, ROR #8\n"
"EOR %0, %0, %1, LSR #8\n"
: "=r" (value), "=r" (tmp)
: "0" (value), "1" (tmp));
return value;
}
static inline uint16_t CPU_swap_u16(uint16_t value)
{
uint16_t lower;
uint16_t upper;
value = value & (uint16_t ) 0xffff;
lower = (value >> 8) ;
upper = (value << 8) ;
return (lower | upper);
}
#ifdef __cplusplus
}
#endif
#endif