source: rtems/cpukit/score/cpu/epiphany/rtems/score/cpu.h @ 01b32d4

/**
 * @file rtems/score/cpu.h
 */

/*
 *
 * Copyright (c) 2015 University of York.
 * Hesham ALMatary <hmka501@york.ac.uk>
 *
 * COPYRIGHT (c) 1989-1999.
 * On-Line Applications Research Corporation (OAR).
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

#ifndef _EPIPHANY_CPU_H
#define _EPIPHANY_CPU_H

#ifdef __cplusplus
extern "C" {
#endif

#include <rtems/score/epiphany.h> /* pick up machine definitions */
#include <rtems/score/types.h>
#ifndef ASM
#include <rtems/bspIo.h>
#include <stdint.h>
#include <stdio.h> /* for printk */
#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.]
 *
 */

#define CPU_INLINE_ENABLE_DISPATCH       FALSE

/*
 *  Does RTEMS manage a dedicated interrupt stack in software?
 *
 *  If TRUE, then a stack is allocated in _ISR_Handler_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.
 *
 *  Currently, for epiphany port, _ISR_Handler is responsible for switching to
 *  RTEMS dedicated interrupt task.
 *
 */

#define CPU_HAS_SOFTWARE_INTERRUPT_STACK TRUE

/*
 *  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 FALSE

/*
 *  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
 *  or CPU_INSTALL_HARDWARE_INTERRUPT_STACK is TRUE.
 *
 */

#define CPU_ALLOCATE_INTERRUPT_STACK TRUE

/*
 *  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 1

/*
 *  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 "epiphany_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.
 *
 *  The CPU_SOFTWARE_FP is used to indicate whether or not there
 *  is software implemented floating point that must be context
 *  switched.  The determination of whether or not this applies
 *  is very tool specific and the state saved/restored is also
 *  compiler specific.
 *
 *  epiphany Specific Information:
 *
 *  At this time there are no implementations of Epiphany that are
 *  expected to implement floating point.
 */

#define CPU_HARDWARE_FP     FALSE
#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    TRUE

/*
 *  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 (64)))

/*
 *  Define what is required to specify how the network to host conversion
 *  routines are handled.
 *
 *  epiphany Specific Information:
 *
 *  This version of RTEMS is designed specifically to run with
 *  big endian architectures. If you want little endian, you'll
 *  have to make the appropriate adjustments here and write
 *  efficient routines for byte swapping. The epiphany architecture
 *  doesn't do this very well.
 */

#define CPU_HAS_OWN_HOST_TO_NETWORK_ROUTINES     FALSE
#define CPU_BIG_ENDIAN                           FALSE
#define CPU_LITTLE_ENDIAN                        TRUE

/*
 *  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   0x00000001

/*
 *  Processor defined structures required for cpukit/score.
 */

/*
 * 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.
 *
 *
 */
#ifndef ASM

typedef struct {
  uint32_t  r[64];

  uint32_t status;
  uint32_t config;
  uint32_t iret;

#ifdef RTEMS_SMP
    /**
     * @brief On SMP configurations the thread context must contain a boolean
     * indicator to signal if this context is executing on a processor.
     *
     * This field must be updated during a context switch.  The context switch
     * to the heir must wait until the heir context indicates that it is no
     * longer executing on a processor.  The context switch must also check if
     * a thread dispatch is necessary to honor updates of the heir thread for
     * this processor.  This indicator must be updated using an atomic test and
     * set operation to ensure that at most one processor uses the heir
     * context at the same time.
     *
     * @code
     * void _CPU_Context_switch(
     *   Context_Control *executing,
     *   Context_Control *heir
     * )
     * {
     *   save( executing );
     *
     *   executing->is_executing = false;
     *   memory_barrier();
     *
     *   if ( test_and_set( &heir->is_executing ) ) {
     *     do {
     *       Per_CPU_Control *cpu_self = _Per_CPU_Get_snapshot();
     *
     *       if ( cpu_self->dispatch_necessary ) {
     *         heir = _Thread_Get_heir_and_make_it_executing( cpu_self );
     *       }
     *     } while ( test_and_set( &heir->is_executing ) );
     *   }
     *
     *   restore( heir );
     * }
     * @endcode
     */
    volatile bool is_executing;
#endif
} Context_Control;

#define _CPU_Context_Get_SP( _context ) \
  (_context)->r[13]

typedef struct {
  /** FPU registers are listed here */
  double  some_float_register;
} Context_Control_fp;

typedef Context_Control CPU_Interrupt_frame;

/*
 *  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.
 *
 *  epiphany Specific Information:
 *
 */

#define CPU_CONTEXT_FP_SIZE  0
SCORE_EXTERN Context_Control_fp  _CPU_Null_fp_context;

/*
 *  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

/*
 *  Should be large enough to run all RTEMS tests.  This insures
 *  that a "reasonable" small application should not have any problems.
 *
 */

#define CPU_STACK_MINIMUM_SIZE  4096

/*
 *  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 8

/*
 *  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

/*
 *  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 although it should be
 *         a multiple of 2 greater than or equal to 2.  The requirement
 *         to be a multiple of 2 is because the heap uses the least
 *         significant field of the front and back flags to indicate
 *         that a block is in use or free.  So you do not want any odd
 *         length blocks really putting length data in that bit.
 *
 *         On byte oriented architectures, CPU_HEAP_ALIGNMENT normally will
 *         have to be greater or equal to than CPU_ALIGNMENT to ensure that
 *         elements allocated from the heap meet all restrictions.
 *
 */

#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        8

/* ISR handler macros */

/*
 *  Support routine to initialize the RTEMS vector table after it is allocated.
 *
 *  NO_CPU Specific Information:
 *
 *  XXX document implementation including references if appropriate
 */

#define _CPU_Initialize_vectors()

/*
 *  Disable all interrupts for an RTEMS critical section.  The previous
 *  level is returned in _level.
 *
 */

static inline uint32_t epiphany_interrupt_disable( void )
{
  uint32_t sr;
  __asm__ __volatile__ ("movfs %[sr], status \n" : [sr] "=r" (sr):);
  __asm__ __volatile__("gid \n");
  return sr;
}

static inline void epiphany_interrupt_enable(uint32_t level)
{
  __asm__ __volatile__("gie \n");
  __asm__ __volatile__ ("movts status, %[level] \n" :: [level] "r" (level):);
}

#define _CPU_ISR_Disable( _level ) \
    _level = epiphany_interrupt_disable()

/*
 *  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 )  \
  epiphany_interrupt_enable( _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 ) \
  do{ \
      if ( (_level & 0x2) != 0 ) \
        _CPU_ISR_Enable( _level ); \
      epiphany_interrupt_disable(); \
    } while(0)

/*
 *  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.
 *
 */

void _CPU_ISR_Set_level( uint32_t level );

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.
 *
 */

/**
 * @brief Account for GCC red-zone
 *
 * The following macro is used when initializing task's stack
 * to account for GCC red-zone.
 */

#define EPIPHANY_GCC_RED_ZONE_SIZE 128

/**
 * @brief Initializes the CPU context.
 *
 * The following steps are performed:
 *  - setting a starting address
 *  - preparing the stack
 *  - preparing the stack and frame pointers
 *  - setting the proper interrupt level in the context
 *
 * @param[in] context points to the context area
 * @param[in] stack_area_begin is the low address of the allocated stack area
 * @param[in] stack_area_size is the size of the stack area in bytes
 * @param[in] new_level is the interrupt level for the task
 * @param[in] entry_point is the task's entry point
 * @param[in] is_fp is set to @c true if the task is a floating point task
 * @param[in] tls_area is the thread-local storage (TLS) area
 */
void _CPU_Context_Initialize(
  Context_Control *context,
  void *stack_area_begin,
  size_t stack_area_size,
  uint32_t new_level,
  void (*entry_point)( void ),
  bool is_fp,
  void *tls_area
);

/*
 *  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(_source, _error ) \
          printk("Fatal Error %d.%d Halted\n",_source, _error); \
          asm("trap 3" :: "r" (_error)); \
          for(;;)

/* end of Fatal Error manager macros */

/* Bitfield handler macros */

/*
 *  This routine sets _output to the bit number of the first bit
 *  set in _value.  _value is of CPU dependent type Priority_Bit_map_control.
 *  This type may be either 16 or 32 bits wide although only the 16
 *  least significant bits will be used.
 *
 *  There are a number of variables in using a "find first bit" type
 *  instruction.
 *
 *    (1) What happens when run on a value of zero?
 *    (2) Bits may be numbered from MSB to LSB or vice-versa.
 *    (3) The numbering may be zero or one based.
 *    (4) The "find first bit" instruction may search from MSB or LSB.
 *
 *  RTEMS guarantees that (1) will never happen so it is not a concern.
 *  (2),(3), (4) are handled by the macros _CPU_Priority_mask() and
 *  _CPU_Priority_bits_index().  These three form a set of routines
 *  which must logically operate together.  Bits in the _value are
 *  set and cleared based on masks built by _CPU_Priority_mask().
 *  The basic major and minor values calculated by _Priority_Major()
 *  and _Priority_Minor() are "massaged" by _CPU_Priority_bits_index()
 *  to properly range between the values returned by the "find first bit"
 *  instruction.  This makes it possible for _Priority_Get_highest() to
 *  calculate the major and directly index into the minor table.
 *  This mapping is necessary to ensure that 0 (a high priority major/minor)
 *  is the first bit found.
 *
 *  This entire "find first bit" and mapping process depends heavily
 *  on the manner in which a priority is broken into a major and minor
 *  components with the major being the 4 MSB of a priority and minor
 *  the 4 LSB.  Thus (0 << 4) + 0 corresponds to priority 0 -- the highest
 *  priority.  And (15 << 4) + 14 corresponds to priority 254 -- the next
 *  to the lowest priority.
 *
 *  If your CPU does not have a "find first bit" instruction, then
 *  there are ways to make do without it.  Here are a handful of ways
 *  to implement this in software:
 *
 *    - a series of 16 bit test instructions
 *    - a "binary search using if's"
 *    - _number = 0
 *      if _value > 0x00ff
 *        _value >>=8
 *        _number = 8;
 *
 *      if _value > 0x0000f
 *        _value >=8
 *        _number += 4
 *
 *      _number += bit_set_table[ _value ]
 *
 *    where bit_set_table[ 16 ] has values which indicate the first
 *      bit set
 *
 */

  /* #define CPU_USE_GENERIC_BITFIELD_CODE FALSE */
#define CPU_USE_GENERIC_BITFIELD_CODE TRUE
#define CPU_USE_GENERIC_BITFIELD_DATA TRUE

#if (CPU_USE_GENERIC_BITFIELD_CODE == FALSE)

#define _CPU_Bitfield_Find_first_bit( _value, _output ) \
  { \
    (_output) = 0;   /* do something to prevent warnings */ \
  }
#endif

/* end of Bitfield handler macros */

/*
 *  This routine builds the mask which corresponds to the bit fields
 *  as searched by _CPU_Bitfield_Find_first_bit().  See the discussion
 *  for that routine.
 *
 */

#if (CPU_USE_GENERIC_BITFIELD_CODE == FALSE)

#define _CPU_Priority_Mask( _bit_number ) \
    (1 << _bit_number)

#endif

/*
 *  This routine translates the bit numbers returned by
 *  _CPU_Bitfield_Find_first_bit() into something suitable for use as
 *  a major or minor component of a priority.  See the discussion
 *  for that routine.
 *
 */

#if (CPU_USE_GENERIC_BITFIELD_CODE == FALSE)

#define _CPU_Priority_bits_index( _priority ) \
  (_priority)

#endif

typedef struct {
/* There is no CPU specific per-CPU state */
} CPU_Per_CPU_control;
#endif /* ASM */

/**
 * Size of a pointer.
 *
 * This must be an integer literal that can be used by the assembler.  This
 * value will be used to calculate offsets of structure members.  These
 * offsets will be used in assembler code.
 */
#define CPU_SIZEOF_POINTER 4
#define CPU_EXCEPTION_FRAME_SIZE 260
#define CPU_PER_CPU_CONTROL_SIZE 0

#ifndef ASM
typedef uint16_t Priority_bit_map_Word;

typedef struct {
  uint32_t r[62];
  uint32_t status;
  uint32_t config;
  uint32_t iret;
} CPU_Exception_frame;

/**
 * @brief Prints the exception frame via printk().
 *
 * @see rtems_fatal() and RTEMS_FATAL_SOURCE_EXCEPTION.
 */
void _CPU_Exception_frame_print( const CPU_Exception_frame *frame );


/* end of Priority handler macros */

/* functions */

/*
 *  _CPU_Initialize
 *
 *  This routine performs CPU dependent initialization.
 *
 */

void _CPU_Initialize(
  void
);

/*
 *  _CPU_ISR_install_raw_handler
 *
 *  This routine installs a "raw" interrupt handler directly into the
 *  processor's vector table.
 *
 */

void _CPU_ISR_install_raw_handler(
  uint32_t    vector,
  proc_ptr    new_handler,
  proc_ptr   *old_handler
);

/*
 *  _CPU_ISR_install_vector
 *
 *  This routine installs an interrupt vector.
 *
 *  NO_CPU Specific Information:
 *
 *  XXX document implementation including references if appropriate
 */

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_Thread_Idle_body
 *
 *  This routine is the CPU dependent IDLE thread body.
 *
 *  NOTE:  It need only be provided if CPU_PROVIDES_IDLE_THREAD_BODY
 *         is TRUE.
 *
 */

void *_CPU_Thread_Idle_body( uintptr_t ignored );

/*
 *  _CPU_Context_switch
 *
 *  This routine switches from the run context to the heir context.
 *
 *  epiphany Specific Information:
 *
 *  Please see the comments in the .c file for a description of how
 *  this function works. There are several things to be aware of.
 */

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
) RTEMS_NO_RETURN;

/*
 *  _CPU_Context_save_fp
 *
 *  This routine saves the floating point context passed to it.
 *
 */

void _CPU_Context_save_fp(
  void **fp_context_ptr
);

/*
 *  _CPU_Context_restore_fp
 *
 *  This routine restores the floating point context passed to it.
 *
 */

void _CPU_Context_restore_fp(
  void **fp_context_ptr
);

/*  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 insure 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 unsigned int CPU_swap_u32(
  unsigned int value
)
{
  uint32_t   byte1, byte2, byte3, byte4, swapped;

  byte4 = (value >> 24) & 0xff;
  byte3 = (value >> 16) & 0xff;
  byte2 = (value >> 8)  & 0xff;
  byte1 =  value        & 0xff;

  swapped = (byte1 << 24) | (byte2 << 16) | (byte3 << 8) | byte4;
  return( swapped );
}

#define CPU_swap_u16( value ) \
  (((value&0xff) << 8) | ((value >> 8)&0xff))

static inline void _CPU_Context_volatile_clobber( uintptr_t pattern )
{
  /* TODO */
}

static inline void _CPU_Context_validate( uintptr_t pattern )
{
  while (1) {
    /* TODO */
  }
}

typedef uint32_t CPU_Counter_ticks;

CPU_Counter_ticks _CPU_Counter_read( void );

static inline CPU_Counter_ticks _CPU_Counter_difference(
  CPU_Counter_ticks second,
  CPU_Counter_ticks first
)
{
  return second - first;
}

#ifdef RTEMS_SMP
  /**
   * @brief Performs CPU specific SMP initialization in the context of the boot
   * processor.
   *
   * This function is invoked on the boot processor during system
   * initialization.  All interrupt stacks are allocated at this point in case
   * the CPU port allocates the interrupt stacks.  This function is called
   * before _CPU_SMP_Start_processor() or _CPU_SMP_Finalize_initialization() is
   * used.
   *
   * @return The count of physically or virtually available processors.
   * Depending on the configuration the application may use not all processors.
   */
  uint32_t _CPU_SMP_Initialize( void );

  /**
   * @brief Starts a processor specified by its index.
   *
   * This function is invoked on the boot processor during system
   * initialization.
   *
   * This function will be called after _CPU_SMP_Initialize().
   *
   * @param[in] cpu_index The processor index.
   *
   * @retval true Successful operation.
   * @retval false Unable to start this processor.
   */
  bool _CPU_SMP_Start_processor( uint32_t cpu_index );

  /**
   * @brief Performs final steps of CPU specific SMP initialization in the
   * context of the boot processor.
   *
   * This function is invoked on the boot processor during system
   * initialization.
   *
   * This function will be called after all processors requested by the
   * application have been started.
   *
   * @param[in] cpu_count The minimum value of the count of processors
   * requested by the application configuration and the count of physically or
   * virtually available processors.
   */
  void _CPU_SMP_Finalize_initialization( uint32_t cpu_count );

  /**
   * @brief Returns the index of the current processor.
   *
   * An architecture specific method must be used to obtain the index of the
   * current processor in the system.  The set of processor indices is the
   * range of integers starting with zero up to the processor count minus one.
   */
   uint32_t _CPU_SMP_Get_current_processor( void );

  /**
   * @brief Sends an inter-processor interrupt to the specified target
   * processor.
   *
   * This operation is undefined for target processor indices out of range.
   *
   * @param[in] target_processor_index The target processor index.
   */
  void _CPU_SMP_Send_interrupt( uint32_t target_processor_index );

  /**
   * @brief Broadcasts a processor event.
   *
   * Some architectures provide a low-level synchronization primitive for
   * processors in a multi-processor environment.  Processors waiting for this
   * event may go into a low-power state and stop generating system bus
   * transactions.  This function must ensure that preceding store operations
   * can be observed by other processors.
   *
   * @see _CPU_SMP_Processor_event_receive().
   */
  void _CPU_SMP_Processor_event_broadcast( void );

  /**
   * @brief Receives a processor event.
   *
   * This function will wait for the processor event and may wait forever if no
   * such event arrives.
   *
   * @see _CPU_SMP_Processor_event_broadcast().
   */
  static inline void _CPU_SMP_Processor_event_receive( void )
  {
    __asm__ volatile ( "" : : : "memory" );
  }

  /**
   * @brief Gets the is executing indicator of the thread context.
   *
   * @param[in] context The context.
   */
  static inline bool _CPU_Context_Get_is_executing(
    const Context_Control *context
  )
  {
    return context->is_executing;
  }

  /**
   * @brief Sets the is executing indicator of the thread context.
   *
   * @param[in] context The context.
   * @param[in] is_executing The new value for the is executing indicator.
   */
  static inline void _CPU_Context_Set_is_executing(
    Context_Control *context,
    bool is_executing
  )
  {
    context->is_executing = is_executing;
  }
#endif /* RTEMS_SMP */

#endif /* ASM */

#ifdef __cplusplus
}
#endif

#endif