RaySollium99/picodrive
1
0
Fork 0
mirror of https://github.com/RaySollium99/picodrive.git synced 2025-10-27 00:29:39 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions
picodrive/pico/32x/draw_arm.S

515 lines
14 KiB
ArmAsm
Raw Blame History

@*
@* PicoDrive
@* (C) notaz, 2010
@* (C) kub, 2019
@*
@* This work is licensed under the terms of MAME license.
@* See COPYING file in the top-level directory.
@*
#include <pico/arm_features.h>
#include <pico/pico_int_offs.h>
.extern Pico32x
.extern Pico
.equiv P32XV_PRI, (1<< 7)
.text
.align 2
PIC_LDR_INIT()
.macro call_scan_prep cond pico @ &Pico
.if \cond
PIC_LDR(r4, r6, PicoScan32xBegin)
PIC_LDR(r5, r6, PicoScan32xEnd)
add r6, \pico, #OFS_Pico_est
ldr r4, [r4]
ldr r5, [r5]
stmfd sp!, {r4,r5,r6}
.endif
.endm
.macro call_scan_fin_ge cond
.if \cond
addge sp, sp, #4*3
.endif
.endm
.macro call_scan_begin cond
.if \cond
stmfd sp!, {r1-r3}
and r0, r2, #0xff
add r0, r0, r4
mov lr, pc
ldr pc, [sp, #(3+0)*4]
ldr r0, [sp, #(3+2)*4] @ &Pico.est
ldmfd sp!, {r1-r3}
ldr r0, [r0, #OFS_EST_DrawLineDest]
.endif
.endm
.macro call_scan_end cond
.if \cond
stmfd sp!, {r0-r3}
and r0, r2, #0xff
add r0, r0, r4
mov lr, pc
ldr pc, [sp, #(4+1)*4]
ldr r1, [sp, #(4+2)*4] @ &Pico.est
ldr r0, [r1, #OFS_EST_DrawLineDest]
ldr r2, [r1, #OFS_EST_DrawLineDestIncr]
add r0, r0, r2
str r0, [r1, #OFS_EST_DrawLineDest]
ldmfd sp!, {r0-r3}
.endif
.endm
@ direct color
@ unsigned short *dst, unsigned short *dram, int lines_sft_offs, int mdbg
.macro make_do_loop_dc name call_scan do_md
.global \name
\name:
stmfd sp!, {r4-r11,lr}
PIC_LDR(lr, r9, Pico)
PIC_LDR(r10,r9, Pico32x)
ldr r11, [lr, #OFS_Pico_est+OFS_EST_Draw2FB]
ldrh r10,[r10, #0x40] @ Pico32x.vdp_regs[0]
add r9, lr, #OFS_Pico_est+OFS_EST_HighPal @ palmd
and r4, r2, #0xff
mov r5, #328
mov r3, r3, lsl #26 @ mdbg << 26
mla r11,r4,r5,r11 @ r11 = pmd = PicoDraw2FB + offs*328: md data
tst r10,#P32XV_PRI
movne r10,#0
moveq r10,#0x8000 @ r10 = !inv_bit
call_scan_prep \call_scan lr
mov r4, #0 @ line
b 1f @ loop_outer_entry
0: @ loop_outer:
call_scan_end \call_scan
ldr r12, [r9, #OFS_EST_DrawLineDestIncr-OFS_EST_HighPal]
sub r0, r0, #320*2
add r0, r0, r12
add r4, r4, #1
cmp r4, r2, lsr #16
call_scan_fin_ge \call_scan
ldmgefd sp!, {r4-r11,pc}
1: @ loop_outer_entry:
call_scan_begin \call_scan
mov r12,r4, lsl #1
ldrh r12,[r1, r12]
add r11,r11,#8
mov r6, #320/2
add r5, r1, r12, lsl #1 @ p32x = dram + dram[l]
2: @ loop_inner:
@ r4,r6 - counters; r5 - 32x data; r9 - md pal; r10 - inv_prio; r11 - md data
@ r7,r8,r12,lr - temp
ldrh r7, [r5], #2
ldrh r8, [r5], #2
subs r6, r6, #1
blt 0b @ loop_outer
cmp r7, r8
beq 5f @ check_fill
3: @ no_fill:
ldrb r12,[r11], #1 @ MD pixel 0
eor r7, r7, r10
and lr, r7, #0x03e0 @ convert BGR555 -> RGB565
mov r7, r7, ror #5
orr r7, r7, r7, ror #10+11
orr r7, r7, lr, lsl #1+16
eor r8, r8, r10
and lr, r8, #0x03e0
mov r8, r8, ror #5
orr r8, r8, r8, ror #10+11
orr r8, r8, lr, lsl #1+16
ldrb lr, [r11], #1 @ MD pixel 1
.if \do_md
cmp r3, r12, lsl #26
tstne r7, #0x20<<16
movne r12,r12, lsl #1 @ load MD color if no 32X prio and not bg
ldrneh r12,[r9, r12]
moveq r12,r7, lsr #16 @ else replace with 32X color
cmp r3, lr, lsl #26
tstne r8, #0x20<<16
movne lr, lr, lsl #1 @ load MD color if no 32X prio and not bg
ldrneh lr, [r9, lr]
moveq lr, r8, lsr #16 @ else replace with 32X color
orr r12,r12, lr, lsl #16 @ combine 2 pixels to optimize memory bandwidth
str r12,[r0], #4 @ (no write combining on ARM9)
.else
cmp r3, r12, lsl #26 @ replace MD bg info into prio bit
tstne r7, #0x20<<16
moveq r7, r7, lsr #16
streqh r7, [r0, #0]
cmp r3, lr, lsl #26
tstne r8, #0x20<<16
moveq r8, r8, lsr #16
streqh r8, [r0, #2]
add r0, r0, #4 @ store 32x pixels if 32X prio or MD bg
.endif
b 2b @ loop_inner
5: @ check_fill:
@ count pixels, align if needed
ldrh r12,[r5, #0] @ only do this for at least 4 pixels
ldrh lr ,[r5, #2]
cmp r12,r7
cmpeq lr ,r7
bne 3b @ no_fill
add r5, r5, #4 @ adjust for the check above
sub lr, r5, #4+4 @ starting r5 (32x render data start)
add r6, r6, #1 @ restore from dec
6: @ count_loop:
sub r12,r5, lr @ loop checks 2 pixels
ldrh r8, [r5], #2
cmp r12,r6, lsl #2
ldrh r12,[r5], #2
bge 7f @ count_done
cmp r8, r7
cmpeq r12,r7
beq 6b
7: @ count_done:
sub r5, r5, #4 @ undo readahead
sub r8, r5, lr @ pixel count
mov r8, r8, lsr #1
cmp r8, r6, lsl #1 @ limit count to line length
movgt r8, r6, lsl #1
sub r6, r6, r8, lsr #1 @ consume pixels
eor r7, r7, r10
and r12,r7, #0x03e0 @ convert BGR555 -> RGB565
mov r7, r7, ror #5
orr r7, r7, r7, ror #10+11
orr r7, r7, r12,lsl #1+16
lsr r7, #16
tst r7, #0x20 @ check for prio transfer
bne 9f @ bg_loop
add r11,r11,r8 @ consume md pixels (not used)
orr r12,r7, r7, lsl #16
mov r7 ,r12
8: @ 32x_loop:
subs r8, r8, #4 @ store 4 pixels
stmgeia r0!, {r7, r12}
bgt 8b @ 32x_loop
beq 2b @ loop_inner
adds r8, r8, #2
strge r7, [r0], #4 @ store 2 leftover pixels
b 2b @ loop_inner
9: @ bg_loop:
ldrb r12,[r11],#1 @ MD pixel 0,1
ldrb lr, [r11],#1
.if \do_md
cmp r3, r12,lsl #26 @ MD pixel 0 has bg?
mov r12,r12,lsl #1
ldrneh r12,[r9, r12] @ t = palmd[*pmd]
moveq r12,r7
cmp r3, lr, lsl #26 @ MD pixel 1 has bg?
mov lr, lr, lsl #1
ldrneh lr, [r9, lr]
moveq lr, r7
orr r12,r12,lr, lsl #16 @ combine 2 pixels to optimize memory bandwidth
str r12,[r0], #4 @ (no write combining on ARM9)
.else
add r0, r0, #4
cmp r3, r12,lsl #26 @ MD pixel 0 has bg?
streqh r7, [r0, #-4]
cmp r3, lr, lsl #26 @ MD pixel 1 has bg?
streqh r7, [r0, #-2]
.endif
subs r8, r8, #2
bgt 9b @ bg_loop
b 2b @ loop_inner
.endm
@ packed pixel
@ note: this may read a few bytes over the end of PicoDraw2FB and dram,
@ so those should have a bit more alloc'ed than really needed.
@ unsigned short *dst, unsigned short *dram, int lines_sft_offs, int mdbg
.macro make_do_loop_pp name call_scan do_md
.global \name
\name:
stmfd sp!, {r4-r11,lr}
PIC_LDR(lr, r9, Pico)
PIC_LDR(r10,r9, Pico32xMem)
ldr r9, =OFS_PMEM32x_pal_native
ldr r10, [r10]
ldr r11, [lr, #OFS_Pico_est+OFS_EST_Draw2FB]
add r10,r10,r9
add r9, lr, #OFS_Pico_est+OFS_EST_HighPal @ palmd
and r4, r2, #0xff
mov r5, #328
mov r3, r3, lsl #26 @ mdbg << 26
mla r11,r4,r5,r11 @ r11 = pmd = PicoDraw2FB + offs*328: md data
call_scan_prep \call_scan lr
mov r4, #0 @ line
b 1f @ loop_outer_entry
0: @ loop_outer:
call_scan_end \call_scan
ldr r12, [r9, #OFS_EST_DrawLineDestIncr-OFS_EST_HighPal]
sub r0, r0, #320*2
add r0, r0, r12
add r4, r4, #1
cmp r4, r2, lsr #16
call_scan_fin_ge \call_scan
ldmgefd sp!, {r4-r11,pc}
1: @ loop_outer_entry:
call_scan_begin \call_scan
mov r12,r4, lsl #1
ldrh r12,[r1, r12]
add r11,r11,#8
mov r6, #320/2
add r5, r1, r12, lsl #1 @ p32x = dram + dram[l]
and r12,r2, #0x100 @ shift
add r5, r5, r12,lsr #8
2: @ loop_inner:
@ r4,r6 - counters; r5 - 32x data; r9,r10 - md,32x pal; r11 - md data
@ r7,r8,r12,lr - temp
tst r5, #1
ldreqb r8, [r5], #2
ldrb r7, [r5, #-1]
ldrneb r8, [r5, #2]! @ r7,r8 - 32X pixel 0,1
subs r6, r6, #1
blt 0b @ loop_outer
cmp r7, r8
beq 5f @ check_fill
3: @ no_fill:
ldrb r12,[r11], #1 @ MD pixel 0
ldrb lr, [r11], #1 @ MD pixel 1
mov r7, r7, lsl #1
mov r8, r8, lsl #1
ldrh r7, [r10,r7] @ 32X color 0
ldrh r8, [r10,r8] @ 32X color 1
.if \do_md
cmp r3, r12, lsl #26
movne r12,r12, lsl #1 @ load MD color if not bg
ldrneh r12,[r9, r12]
orreq r7, r7, #0x20 @ accumulate MD bg info into prio bit
cmp r3, lr, lsl #26
movne lr, lr, lsl #1
ldrneh lr, [r9, lr]
orreq r8, r8, #0x20
tst r7, #0x20 @ replace 32X with MD color if no prio and not bg
moveq r7, r12
tst r8, #0x20
moveq r8, lr
orr r7, r7, r8, lsl #16 @ combine 2 pixels to optimize memory bandwidth
str r7, [r0], #4 @ (no write combining on ARM9)
.else
cmp r3, r12, lsl #26 @ replace MD bg info into prio bit
orreq r7, r7, #0x20
cmp r3, lr, lsl #26
orreq r8, r8, #0x20
add r0, r0, #4 @ store 32x pixels if 32X prio or MD bg
tst r7, #0x20
strneh r7, [r0, #-4]
tst r8, #0x20
strneh r8, [r0, #-2]
.endif
b 2b @ loop_inner
5: @ check_fill:
@ count pixels, align if needed
bic r12,r5, #1
ldrh r12,[r12, #0] @ only do this for at least 4 pixels
orr lr, r7, r7, lsl #8
cmp r12,lr
bne 3b @ no_fill
add r5, r5, #2 @ adjust for the check above
sub lr, r5, #4 @ starting r5 (32x render data start)
bic r5, r5, #1
add r6, r6, #1 @ restore from dec
orr r7, r7, r7, lsl #8
6: @ count_loop:
sub r12,r5, lr @ loop checks 4 pixels
ldrh r8, [r5], #2
cmp r12,r6, lsl #1
ldrh r12,[r5], #2
bge 7f @ count_done
cmp r8, r7
cmpeq r12,r7
beq 6b
cmp r8, r7
addeq r5, r5, #2 @ adjust if 2 pixels where ok
7: @ count_done:
sub r5, r5, #4 @ undo readahead
tst lr, #1 @ fix alignment and calculate count
subne r5, r5, #1
sub r8, r5, lr
and r7, r7, #0xff @ 32x pixel color
mov r7, r7, lsl #1
ldrh r7, [r10,r7]
cmp r8, r6, lsl #1 @ limit count to line length
movgt r8, r6, lsl #1
sub r6, r6, r8, lsr #1 @ consume pixels
tst r7, #0x20 @ check for prio transfer
beq 9f @ bg_loop
add r11,r11,r8 @ consume md pixels (not used)
orr r12,r7, r7, lsl #16
mov r7 ,r12
8: @ 32x_loop:
subs r8, r8, #4 @ store 4 pixels
stmgeia r0!, {r7, r12}
bgt 8b @ 32x_loop
beq 2b @ loop_inner
adds r8, r8, #2
strge r7, [r0], #4 @ store 2 leftover pixels
b 2b @ loop_inner
9: @ bg_loop:
ldrb r12,[r11],#1 @ MD pixel 0,1
ldrb lr, [r11],#1
.if \do_md
cmp r3, r12,lsl #26 @ MD pixel 0 has bg?
mov r12,r12,lsl #1
ldrneh r12,[r9, r12] @ t = palmd[*pmd]
moveq r12,r7
cmp r3, lr, lsl #26 @ MD pixel 1 has bg?
mov lr, lr, lsl #1
ldrneh lr, [r9, lr]
moveq lr, r7
orr r12,r12,lr, lsl #16 @ combine 2 pixels to optimize memory bandwidth
str r12,[r0], #4 @ (no write combining on ARM9)
.else
add r0, r0, #4
cmp r3, r12,lsl #26 @ MD pixel 0 has bg?
streqh r7, [r0, #-4]
cmp r3, lr, lsl #26 @ MD pixel 1 has bg?
streqh r7, [r0, #-2]
.endif
subs r8, r8, #2
bgt 9b @ bg_loop
b 2b @ loop_inner
.endm
@ run length
@ unsigned short *dst, unsigned short *dram, int lines_sft_offs, int mdbg
.macro make_do_loop_rl name call_scan do_md
.global \name
\name:
stmfd sp!, {r4-r11,lr}
PIC_LDR(lr, r9, Pico)
PIC_LDR(r10,r9, Pico32xMem)
ldr r9, =OFS_PMEM32x_pal_native
ldr r10, [r10]
ldr r11, [lr, #OFS_Pico_est+OFS_EST_Draw2FB]
add r10,r10,r9
add r9, lr, #OFS_Pico_est+OFS_EST_HighPal @ palmd
and r4, r2, #0xff
mov r5, #328
mov r3, r3, lsl #26 @ mdbg << 26
mla r11,r4,r5,r11 @ r11 = pmd = PicoDraw2FB + offs*328: md data
call_scan_prep \call_scan lr
mov r4, #0 @ line
b 1f @ loop_outer_entry
0: @ loop_outer:
call_scan_end \call_scan
ldr r12, [r9, #OFS_EST_DrawLineDestIncr-OFS_EST_HighPal]
sub r0, r0, #320*2
add r0, r0, r12
add r4, r4, #1
cmp r4, r2, lsr #16
call_scan_fin_ge \call_scan
ldmgefd sp!, {r4-r11,pc}
1: @ loop_outer_entry:
call_scan_begin \call_scan
mov r12,r4, lsl #1
ldrh r12,[r1, r12]
add r11,r11,#8
mov r6, #320
add r5, r1, r12, lsl #1 @ p32x = dram + dram[l]
2: @ loop_inner:
ldrh r8, [r5], #2 @ control word
and r12,r8, #0xff
mov r12,r12,lsl #1
ldrh lr, [r10,r12] @ t = 32x pixel
eor lr, lr, #0x20
3: @ loop_innermost:
subs r6, r6, #1
ldrgeb r7, [r11], #1 @ MD pixel
blt 0b @ loop_outer
tst lr, #0x20
cmpne r3, r7, lsl #26 @ MD has bg pixel?
.if \do_md
mov r7, r7, lsl #1
ldrneh r12,[r9, r7] @ t = palmd[*pmd]
streqh lr, [r0], #2
strneh r12,[r0], #2 @ *dst++ = t
.else
streqh lr, [r0]
add r0, r0, #2
.endif
subs r8, r8, #0x100
bge 3b @ loop_innermost
b 2b @ loop_inner
.endm
make_do_loop_dc do_loop_dc, 0, 0
make_do_loop_dc do_loop_dc_md, 0, 1
make_do_loop_dc do_loop_dc_scan, 1, 0
make_do_loop_dc do_loop_dc_scan_md, 1, 1
.pool
make_do_loop_pp do_loop_pp, 0, 0
make_do_loop_pp do_loop_pp_md, 0, 1
make_do_loop_pp do_loop_pp_scan, 1, 0
make_do_loop_pp do_loop_pp_scan_md, 1, 1
.pool
make_do_loop_rl do_loop_rl, 0, 0
make_do_loop_rl do_loop_rl_md, 0, 1
make_do_loop_rl do_loop_rl_scan, 1, 0
make_do_loop_rl do_loop_rl_scan_md, 1, 1
.pool
@ vim:filetype=armasm
Reference in a new issue View git blame Copy permalink