freepascal.compiler/rtl/mips/mips.inc

{
    This file is part of the Free Pascal run time library.
    Copyright (c) 2006-2007 by David Zhang

    Processor dependent implementation for the system unit for MIPS

    See the file COPYING.FPC, included in this distribution,
    for details about the copyright.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

 **********************************************************************}

{****************************************************************************
                           MIPS specific stuff
****************************************************************************}
function get_fsr : dword;assembler;nostackframe;[public, alias: 'FPC_GETFSR'];
  asm
    cfc1 $2,$31
  end;


procedure set_fsr(fsr : dword);assembler;nostackframe;[public, alias: 'FPC_SETFSR'];
  asm
    ctc1 $4,$31
  end;


function get_got_z : pointer;assembler;nostackframe;[public, alias: 'FPC_GETGOT_Z'];
  asm
    move $2,$28
  end;

const
  { FPU enable exception bits for FCSR register }
  fpu_enable_inexact   =  $80;
  fpu_enable_underflow = $100;
  fpu_enable_overflow  = $200;
  fpu_enable_div_zero  = $400;
  fpu_enable_invalid   = $800;
  fpu_enable_mask      = $F80;
  default_fpu_enable = fpu_enable_div_zero or fpu_enable_invalid;

  fpu_flags_mask = $7C;
  fpu_cause_mask = $3F000;

  { FPU rounding mask and values }
  fpu_rounding_mask    = $3;
  fpu_rounding_nearest = 0;
  fpu_rounding_towards_zero = 1;
  fpu_rounding_plus_inf = 2;
  fpu_rounding_minus_inf = 3;


procedure fpc_cpuinit;
var
  tmp32: longint;
  begin
    { don't let libraries influence the FPU cw set by the host program }
    if not IsLibrary then
      begin
        tmp32 := get_fsr();
        { enable div by 0 and invalid operation fpu exceptions,
          disable the other exceptions }
        tmp32 := (tmp32 and not fpu_enable_mask) or default_fpu_enable;
        { Reset flags and cause }
        tmp32 := tmp32 and not (fpu_flags_mask or fpu_cause_mask);

        { round towards nearest; ieee compliant arithmetics }
        tmp32 := (tmp32 and not fpu_rounding_mask) or fpu_rounding_nearest;

        set_fsr(tmp32);
      end;
  end;


{$define FPC_SYSTEM_HAS_GET_FRAME}
function get_frame:pointer;assembler;nostackframe;
  asm
    { we need to use the information of the .pdr section to do this properly:
       0 proc. start adress
       4 regmask
       8 reg. offset
      12 fmask
      16 foffset
      20 frame size
      24 stack reg
      28 link reg

      Further, we need to know the pc
    }
    // lw $2,0($sp)
    move $2,$30
  end;


{ Try to find previous $fp,$ra register pair
  reset both to nil if failure }
{$define FPC_SYSTEM_HAS_GET_CALLER_STACKINFO}
procedure get_caller_stackinfo(var framebp,addr : pointer);
const
  instr_size = 4;
  MAX_INSTRUCTIONS = 64000;
type
  instr_p = pdword;
  reg_p = ppointer;
var
  instr,stackpos : dword;
  i,LocalSize : longint;
  ra_offset, s8_offset : longint;
  current_ra : pointer;
begin
  { Here we need to use GDB approach,
    starting at addr
    go back to lower $ra values until we find a
    position with ADDIU $sp,$sp,-LocalSize
  }
  if addr=nil then
    begin
	  framebp:=nil;
	  exit;
	end;
  Try
    current_ra:=addr;
    ra_offset:=-1;
    s8_offset:=-1;
    i:=0;
    LocalSize:=0;
    repeat
      inc(i);
      dec(current_ra,4);
      instr:=instr_p(current_ra)^;
      if (instr shr 16 = $27bd) then
        begin
          { we found the instruction,
            local size is the lo part }
          LocalSize:=smallint(instr and $ffff);
          break;
        end;
    until i> MAX_INSTRUCTIONS;
    if LocalSize <> 0 then
      begin
        repeat
          inc(current_ra,4);
          instr:=instr_p(current_ra)^;
          if (instr shr 16 = $afbf) then
            ra_offset:=smallint(instr and $ffff)
          else if (instr shr 16 = $afbe) then
            s8_offset:=smallint(instr and $ffff);
        until (current_ra >= addr)
          or ((ra_offset<>-1) and (s8_offset<>-1));
        if ra_offset<>-1 then
          begin
            stackpos:=dword(framebp+LocalSize+ra_offset);
            addr:=reg_p(stackpos)^;
          end
        else
          addr:=nil;
        if s8_offset<>-1 then
          begin
            stackpos:=dword(framebp+LocalSize+s8_offset);
            framebp:=reg_p(stackpos)^;
          end
        else
          framebp:=nil;
      end;
  Except
    framebp:=nil;
    addr:=nil;
  end;
end;

{$define FPC_SYSTEM_HAS_GET_PC_ADDR}
function get_pc_addr : pointer;assembler;nostackframe;
asm
  move $2,$31
end;

{$define FPC_SYSTEM_HAS_GET_CALLER_ADDR}
function get_caller_addr(framebp:pointer;addr:pointer=nil):pointer;
begin
  get_caller_stackinfo(framebp,addr);
  get_caller_addr:=addr;
end;

{$define FPC_SYSTEM_HAS_GET_CALLER_FRAME}
function get_caller_frame(framebp:pointer;addr:pointer=nil):pointer;
begin
  get_caller_stackinfo(framebp,addr);
  get_caller_frame:=framebp;
end;

{$define FPC_SYSTEM_HAS_SPTR}
function Sptr:Pointer;assembler;nostackframe;
  asm
    move $2,$sp
  end;

{$ifdef USE_MIPS_STK2_ASM}
{$ifndef FPC_SYSTEM_HAS_MOVE}
(* Disabled for now
{$define FPC_SYSTEM_HAS_MOVE}
procedure Move(const source;var dest;count:longint);[public, alias: 'FPC_MOVE'];assembler;
  asm
{
  Registers:
    $7 temp. to do copying
    $8 inc/decrement
    $9/l0/l1/l2 qword move
}

    sw  $4,0($23)
    sw  $5,-4($23)
    sw  $6,-8($23)
    sw  $7,-12($23)
    sw  $8,-16($23)
    sw  $9,-20($23)
    sw  $10,-24($23)
    sw  $11,-28($23)
    sw  $12,-32($23)
    sw  $13,-36($23)
    sw  $14,-40($23)
    addiu  $23,$23,-44


    // count <= 0 ?
    ble $6,$0,.Lmoveexit
    nop

    // source = dest ?
    beq $4,$5,.Lmoveexit
    nop

    // possible overlap?
    bgt $4,$5,.Lnopossibleoverlap
    nop
    // source < dest ....
    addu $7,$6,$4
    // overlap?
    // source+count < dest ?
    blt $7,$5,.Lnopossibleoverlap
    nop

  .Lcopybackward:
    // check alignment of source and dest
    or $2,$4,$5

    // move src and dest to the end of the blocks
    // assuming 16 byte block size
    addiu $3,$6,-1
    addu $4,$4,$3
    addu $5,$5,$3

    b .Lmovebytewise
    li $3,-1

.Lnopossibleoverlap:

    // check alignment of source and dest
    or $2,$4,$5

    // everything 16 byte aligned ?
    andi $13,$2,15

    beq $13,$0,.Lmovetwordwise
    // load direction in delay slot
    li $3,16


    andi $13,$2,7
    beq $13,$0,.Lmoveqwordwise
    li $3,8

    andi $13,$2,3
    beq $13,$0,.Lmovedwordwise
    li $3,4

    andi $13,$2,1
    beq $13,$0,.Lmovewordwise
    li $3,2
    b .Lmovebytewise
    li $3,1

.Lmovetwordwise:
    srl $13,$6,4
    sll $14,$13,4
    beq $14,$0,.Lmoveqwordwise_shift
    nop

.Lmovetwordwise_loop:
    lw $9,0($4)
    lw $10,4($4)
    addiu $13,$13,-1
    lw $11,8($4)
    lw $12,12($4)
    addu $4,$4,$3
    sw $9,0($5)
    sw $10,4($5)
    sw $11,8($5)
    sw $12,12($5)
    addu $5,$5,$3
    bne $13,$0,.Lmovetwordwise_loop
    nop
    subu $6,$6,$14
    beq $6,$0,.Lmoveexit
    nop

.Lmoveqwordwise_shift:
    sra $3,$3,1

.Lmoveqwordwise:
    srl $13,$6,3
    sll $14,$13,3
    beq $14,$0,.Lmovedwordwise_shift
    nop

  .Lmoveqwordwise_loop:
    lw $9,0($4)
    lw $10,4($4)
    addiu $13,$13,-1
    addu $4,$3,$4
    sw $9,0($5)
    sw $10,4($5)
    addu $5,$3,$5
    bne $13,$0,.Lmoveqwordwise_loop
    nop

    subu $6,$6,$14
    beq $6,$0,.Lmoveexit
    nop

  .Lmovedwordwise_shift:
    sra $3,$3,1

  .Lmovedwordwise:
    srl $13,$6,2
    sll $14,$13,2
    beq $14,$0,.Lmovewordwise_shift
    nop

  .Lmovedwordwise_loop:
    lw $9,0($4)
    addiu $13,$13,-1
    addu $4,$4,$3
    sw $9,0($5)
    addu $5,$5,$3
    bne $13,$0,.Lmovedwordwise_loop
    nop

    subu $6,$6,$14
    beq $6,$0,.Lmoveexit
    nop

  .Lmovewordwise_shift:
    sra $3,$3,1
  .Lmovewordwise:
    srl $13,$6,1
    sll $14,$13,1
    beq $14,$0, .Lmovebytewise_shift
    nop

  .Lmovewordwise_loop:
    lhu $9,0($4)
    addiu $13,$13,-1
    addu $4,$4,$3
    sh $9,0($5)
    addu $5,$5,$3
    bne $13,$0,.Lmovewordwise_loop
    nop

    subu $6,$6,$14
    beq $6,$0, .Lmoveexit
    nop

  .Lmovebytewise_shift:
    sra $3,$3,1
  .Lmovebytewise:
    beq $6,$0, .Lmoveexit
    nop

    lbu $9,0($4)
    addiu $6,$6,-1
    addu $4,$4,$3
    sb $9,0($5)
    addu $5,$5,$3
    bne $6,$0,.Lmovebytewise
    nop
  .Lmoveexit:

  addiu  $23,$23,44
  lw  $4,0($23)
  lw  $5,-4($23)
  lw  $6,-8($23)
  lw  $7,-12($23)
  lw  $8,-16($23)
  lw  $9,-20($23)
  lw  $10,-24($23)
  lw  $11,-28($23)
  lw  $12,-32($23)
  lw  $13,-36($23)
  lw  $14,-40($23)

  end;
*)
{$endif FPC_SYSTEM_HAS_MOVE}
{****************************************************************************
                               Integer math
****************************************************************************}

{$define FPC_SYSTEM_HAS_ABS_LONGINT}
function abs(l:longint):longint; assembler;{$ifdef SYSTEMINLINE}inline;{$endif}nostackframe;
asm
    sra $1,$4,31 // $at,$4,31
    xor $2,$4,$1 // $2,$4,$at
    sub $2,$2,$1 // $2,$2,$at
end;

var
  fpc_system_lock : longint; export name 'fpc_system_lock';


{$define FPC_SYSTEM_HAS_DECLOCKED_LONGINT}
function declocked(var l : longint) : boolean;assembler;nostackframe;
{ input:  address of l in $4                                      }
{ output: boolean indicating whether l is zero after decrementing }
asm
  sw  $4,0($23)
  sw  $5,-4($23)
  sw  $6,-8($23)
  sw  $7,-12($23)
  sw  $8,-16($23)
  sw  $9,-20($23)
  sw  $10,-24($23)
  sw  $11,-28($23)
  sw  $12,-32($23)
  sw  $13,-36($23)
  sw  $14,-40($23)
  addiu  $23,$23,-44



.Ldeclocked1:
  lui $5,%hi(fpc_system_lock)
  addiu $5,$5,%lo(fpc_system_lock)
  ll $6,0($5)
  ori $7,$6,1
  beq $7,$6,.Ldeclocked1
  nop
  sc  $7,0($5)

  beq $7,$0,.Ldeclocked1
  nop

  lw $5,0($4)
  addiu $5,$5,-1
  sw $5,0($4)
  seq $2,$5,$0


  { unlock }
  lui $5,%hi(fpc_system_lock)
  addiu $5,$5,%lo(fpc_system_lock)
  sw $0,0($5)

  addiu  $23,$23,44
  lw  $4,0($23)
  lw  $5,-4($23)
  lw  $6,-8($23)
  lw  $7,-12($23)
  lw  $8,-16($23)
  lw  $9,-20($23)
  lw  $10,-24($23)
  lw  $11,-28($23)
  lw  $12,-32($23)
  lw  $13,-36($23)
  lw  $14,-40($23)
end;


{$define FPC_SYSTEM_HAS_INCLOCKED_LONGINT}
procedure inclocked(var l : longint);assembler;nostackframe;
asm
  { usually, we shouldn't lock here so saving the stack frame for these extra intructions is
    worse the effort, especially while waiting :)
  }

  { unlock }

  sw  $4,0($23)
  sw  $5,-4($23)
  sw  $6,-8($23)
  sw  $7,-12($23)
  sw  $8,-16($23)
  sw  $9,-20($23)
  sw  $10,-24($23)
  sw  $11,-28($23)
  sw  $12,-32($23)
  sw  $13,-36($23)
  sw  $14,-40($23)
  addiu  $23,$23,-44


.Ldeclocked1:
  lui $5,%hi(fpc_system_lock)
  addiu $5,$5,%lo(fpc_system_lock)
  ll $6,0($5)
  ori $7,$6,1
  beq $7,$6,.Ldeclocked1
  nop
  sc  $7,0($5)

  beq $7,$0,.Ldeclocked1
  nop

  lw $5,0($4)
  addiu $5,$5,1
  sw $5,0($4)


  { unlock }
  lui $5,%hi(fpc_system_lock)
  addiu $5,$5,%lo(fpc_system_lock)
  sw $0,0($5)

  addiu  $23,$23,44
  lw  $4,0($23)
  lw  $5,-4($23)
  lw  $6,-8($23)
  lw  $7,-12($23)
  lw  $8,-16($23)
  lw  $9,-20($23)
  lw  $10,-24($23)
  lw  $11,-28($23)
  lw  $12,-32($23)
  lw  $13,-36($23)
  lw  $14,-40($23)

end;
{$endif def USE_MIPS_STK2_ASM}

function InterLockedDecrement (var Target: longint) : longint; assembler; nostackframe;
asm
  {$warning FIXME: This implementation of InterLockedDecrement in not yet ThreadSafe }
  // must return value after decrement
  lw  $v0,($a0)
  addi $v0,$v0,-1
  sw  $v0,($a0)
end;


function InterLockedIncrement (var Target: longint) : longint; assembler; nostackframe;
asm
  {$warning FIXME: This implementation of InterLockedIncrement in not yet ThreadSafe }
  // must return value after increment
  lw  $v0,($a0)
  addi $v0,$v0,1
  sw  $v0,($a0)
end;


function InterLockedExchange (var Target: longint;Source : longint) : longint; assembler; nostackframe;
asm
  {$warning FIXME: This implementation of InterLockedExchange in not yet ThreadSafe }
  lw  $v0,($a0)
  sw  $a1,($a0)
end;

function InterLockedExchangeAdd (var Target: longint;Source : longint) : longint; assembler; nostackframe;
asm
  {$warning FIXME: This implementation of InterLockedExchangeAdd in not yet ThreadSafe }
  lw  $v0,($a0)
  add $a1,$v0,$a1
  sw  $a1,($a0)
end;


function InterlockedCompareExchange(var Target: longint; NewValue: longint; Comperand: longint): longint; assembler; nostackframe;
asm
  {$warning FIXME: This implementation of InterLockedCompareAdd in not yet ThreadSafe }
  { put old value of Target into $v0, result register }
  lw  $v0,($a0)
  { copy to t0 register }
  move  $t0,$v0
  move $v1,$a2
  xor  $t0,$t0,$v1
  beq  $t0,$zero,.L1
  b    .L2
.L1:
  {store NewValue (in $a1) into Target in ($(a0)) }
  sw    $a1,($a0)
.L2:
end;