SRB2/src/tmap_asm.s
2021-05-07 17:45:56 +02:00

322 lines
8 KiB
ArmAsm

// SONIC ROBO BLAST 2
//-----------------------------------------------------------------------------
// Copyright (C) 1998-2000 by DooM Legacy Team.
// Copyright (C) 1999-2021 by Sonic Team Junior.
//
// This program is free software distributed under the
// terms of the GNU General Public License, version 2.
// See the 'LICENSE' file for more details.
//-----------------------------------------------------------------------------
/// \file tmap_asm.s
/// \brief ???
//.comm _dc_colormap,4
//.comm _dc_x,4
//.comm _dc_yl,4
//.comm _dc_yh,4
//.comm _dc_iscale,4
//.comm _dc_texturemid,4
//.comm _dc_source,4
//.comm _ylookup,4
//.comm _columnofs,4
//.comm _loopcount,4
//.comm _pixelcount,4
.data
_pixelcount:
.long 0x00000000
_loopcount:
.long 0x00000000
.align 8
_mmxcomm:
.long 0x00000000
.text
.align 4
.globl _R_DrawColumn8_NOMMX
_R_DrawColumn8_NOMMX:
pushl %ebp
pushl %esi
pushl %edi
pushl %ebx
movl _dc_yl,%edx
movl _dc_yh,%eax
subl %edx,%eax
leal 1(%eax),%ebx
testl %ebx,%ebx
jle rdc8ndone
movl _dc_x,%eax
movl _ylookup, %edi
movl (%edi,%edx,4),%esi
movl _columnofs, %edi
addl (%edi,%eax,4),%esi
movl _dc_iscale,%edi
movl %edx,%eax
imull %edi,%eax
movl _dc_texturemid,%ecx
addl %eax,%ecx
movl _dc_source,%ebp
xorl %edx, %edx
subl $0x12345678, %esi
.globl rdc8nwidth1
rdc8nwidth1:
.align 4,0x90
rdc8nloop:
movl %ecx,%eax
shrl $16,%eax
addl %edi,%ecx
andl $127,%eax
addl $0x12345678,%esi
.globl rdc8nwidth2
rdc8nwidth2:
movb (%eax,%ebp),%dl
movl _dc_colormap,%eax
movb (%eax,%edx),%al
movb %al,(%esi)
decl %ebx
jne rdc8nloop
rdc8ndone:
popl %ebx
popl %edi
popl %esi
popl %ebp
ret
//
// Optimised specifically for P54C/P55C (aka Pentium with/without MMX)
// By ES 1998/08/01
//
.globl _R_DrawColumn_8_Pentium
_R_DrawColumn_8_Pentium:
pushl %ebp
pushl %ebx
pushl %esi
pushl %edi
movl _dc_yl,%eax // Top pixel
movl _dc_yh,%ebx // Bottom pixel
movl _ylookup, %edi
movl (%edi,%ebx,4),%ecx
subl %eax,%ebx // ebx=number of pixels-1
jl rdc8pdone // no pixel to draw, done
jnz rdc8pmany
movl _dc_x,%edx // Special case: only one pixel
movl _columnofs, %edi
addl (%edi,%edx,4),%ecx // dest pixel at (%ecx)
movl _dc_iscale,%esi
imull %esi,%eax
movl _dc_texturemid,%edi
addl %eax,%edi // texture index in edi
movl _dc_colormap,%edx
shrl $16, %edi
movl _dc_source,%ebp
andl $127,%edi
movb (%edi,%ebp),%dl // read texture pixel
movb (%edx),%al // lookup for light
movb %al,0(%ecx) // write it
jmp rdc8pdone // done!
.align 4, 0x90
rdc8pmany: // draw >1 pixel
movl _dc_x,%edx
movl _columnofs, %edi
movl (%edi,%edx,4),%edx
leal 0x12345678(%edx, %ecx), %edi // edi = two pixels above bottom
.globl rdc8pwidth5
rdc8pwidth5: // DeadBeef = -2*SCREENWIDTH
movl _dc_iscale,%edx // edx = fracstep
imull %edx,%eax
shll $9, %edx // fixme: Should get 7.25 fix as input
movl _dc_texturemid,%ecx
addl %eax,%ecx // ecx = frac
movl _dc_colormap,%eax // eax = lighting/special effects LUT
shll $9, %ecx
movl _dc_source,%esi // esi = source ptr
imull $0x12345678, %ebx // ebx = negative offset to pixel
.globl rdc8pwidth6
rdc8pwidth6: // DeadBeef = -SCREENWIDTH
// Begin the calculation of the two first pixels
leal (%ecx, %edx), %ebp
shrl $25, %ecx
movb (%esi, %ecx), %al
leal (%edx, %ebp), %ecx
shrl $25, %ebp
movb (%eax), %dl
// The main loop
rdc8ploop:
movb (%esi,%ebp), %al // load 1
leal (%ecx, %edx), %ebp // calc frac 3
shrl $25, %ecx // shift frac 2
movb %dl, 0x12345678(%edi, %ebx)// store 0
.globl rdc8pwidth1
rdc8pwidth1: // DeadBeef = 2*SCREENWIDTH
movb (%eax), %al // lookup 1
movb %al, 0x12345678(%edi, %ebx)// store 1
.globl rdc8pwidth2
rdc8pwidth2: // DeadBeef = 3*SCREENWIDTH
movb (%esi, %ecx), %al // load 2
leal (%ebp, %edx), %ecx // calc frac 4
shrl $25, %ebp // shift frac 3
movb (%eax), %dl // lookup 2
addl $0x12345678, %ebx // counter
.globl rdc8pwidth3
rdc8pwidth3: // DeadBeef = 2*SCREENWIDTH
jl rdc8ploop // loop
// End of loop. Write extra pixel or just exit.
jnz rdc8pdone
movb %dl, 0x12345678(%edi, %ebx)// Write odd pixel
.globl rdc8pwidth4
rdc8pwidth4: // DeadBeef = 2*SCREENWIDTH
rdc8pdone:
popl %edi
popl %esi
popl %ebx
popl %ebp
ret
//
// MMX asm version, optimised for K6
// By ES 1998/07/05
//
.globl _R_DrawColumn_8_K6_MMX
_R_DrawColumn_8_K6_MMX:
pushl %ebp
pushl %ebx
pushl %esi
pushl %edi
movl %esp, %eax // Push 8 or 12, so that (%esp) gets aligned by 8
andl $7,%eax
addl $8,%eax
movl %eax, _mmxcomm // Temp storage in mmxcomm: (%esp) is used instead
subl %eax,%esp
movl _dc_yl,%edx // Top pixel
movl _dc_yh,%ebx // Bottom pixel
movl _ylookup, %edi
movl (%edi,%ebx,4),%ecx
subl %edx,%ebx // ebx=number of pixels-1
jl 0x12345678 // no pixel to draw, done
.globl rdc8moffs1
rdc8moffs1:
jnz rdc8mmany
movl _dc_x,%eax // Special case: only one pixel
movl _columnofs, %edi
addl (%edi,%eax,4),%ecx // dest pixel at (%ecx)
movl _dc_iscale,%esi
imull %esi,%edx
movl _dc_texturemid,%edi
addl %edx,%edi // texture index in edi
movl _dc_colormap,%edx
shrl $16, %edi
movl _dc_source,%ebp
andl $127,%edi
movb (%edi,%ebp),%dl // read texture pixel
movb (%edx),%al // lookup for light
movb %al,0(%ecx) // write it
jmp rdc8mdone // done!
.globl rdc8moffs2
rdc8moffs2:
.align 4, 0x90
rdc8mmany: // draw >1 pixel
movl _dc_x,%eax
movl _columnofs, %edi
movl (%edi,%eax,4),%eax
leal 0x12345678(%eax, %ecx), %esi // esi = two pixels above bottom
.globl rdc8mwidth3
rdc8mwidth3: // DeadBeef = -2*SCREENWIDTH
movl _dc_iscale,%ecx // ecx = fracstep
imull %ecx,%edx
shll $9, %ecx // fixme: Should get 7.25 fix as input
movl _dc_texturemid,%eax
addl %edx,%eax // eax = frac
movl _dc_colormap,%edx // edx = lighting/special effects LUT
shll $9, %eax
leal (%ecx, %ecx), %edi
movl _dc_source,%ebp // ebp = source ptr
movl %edi, 0(%esp) // Start moving frac and fracstep to MMX regs
imull $0x12345678, %ebx // ebx = negative offset to pixel
.globl rdc8mwidth5
rdc8mwidth5: // DeadBeef = -SCREENWIDTH
movl %edi, 4(%esp)
leal (%eax, %ecx), %edi
movq 0(%esp), %mm1 // fracstep:fracstep in mm1
movl %eax, 0(%esp)
shrl $25, %eax
movl %edi, 4(%esp)
movzbl (%ebp, %eax), %eax
movq 0(%esp), %mm0 // frac:frac in mm0
paddd %mm1, %mm0
shrl $25, %edi
movq %mm0, %mm2
psrld $25, %mm2 // texture index in mm2
paddd %mm1, %mm0
movq %mm2, 0(%esp)
.globl rdc8mloop
rdc8mloop: // The main loop
movq %mm0, %mm2 // move 4-5 to temp reg
movzbl (%ebp, %edi), %edi // read 1
psrld $25, %mm2 // shift 4-5
movb (%edx,%eax), %cl // lookup 0
movl 0(%esp), %eax // load 2
addl $0x12345678, %ebx // counter
.globl rdc8mwidth2
rdc8mwidth2: // DeadBeef = 2*SCREENWIDTH
movb %cl, (%esi, %ebx) // write 0
movb (%edx,%edi), %ch // lookup 1
movb %ch, 0x12345678(%esi, %ebx) // write 1
.globl rdc8mwidth1
rdc8mwidth1: // DeadBeef = SCREENWIDTH
movl 4(%esp), %edi // load 3
paddd %mm1, %mm0 // frac 6-7
movzbl (%ebp, %eax), %eax // lookup 2
movq %mm2, 0(%esp) // store texture index 4-5
jl rdc8mloop
jnz rdc8mno_odd
movb (%edx,%eax), %cl // write the last odd pixel
movb %cl, 0x12345678(%esi)
.globl rdc8mwidth4
rdc8mwidth4: // DeadBeef = 2*SCREENWIDTH
rdc8mno_odd:
.globl rdc8mdone
rdc8mdone:
emms
addl _mmxcomm, %esp
popl %edi
popl %esi
popl %ebx
popl %ebp
ret
// Need some extra space to align run-time
.globl R_DrawColumn_8_K6_MMX_end
R_DrawColumn_8_K6_MMX_end:
nop;nop;nop;nop;nop;nop;nop;nop;
nop;nop;nop;nop;nop;nop;nop;nop;
nop;nop;nop;nop;nop;nop;nop;nop;
nop;nop;nop;nop;nop;nop;nop;