X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=Pico%2FDraw.s;h=e31ba0baa6911ad14782bf7823d97d4b6051c6af;hb=32826a1a22fd3e6203310bba855d8c2b6f3c403a;hp=749aac538517a27893105412c77fcccdca561dab;hpb=cc68a136aa179a5f32fe40208371eb9c2b0aadae;p=picodrive.git

diff --git a/Pico/Draw.s b/Pico/Draw.s
index 749aac5..e31ba0b 100644
--- a/Pico/Draw.s
+++ b/Pico/Draw.s
@@ -1,7 +1,9 @@
+@ vim:filetype=armasm
+
 @ assembly "optimized" version of some funtions from draw.c
 @ this is highly specialized, be careful if changing related C code!
 
-@ (c) Copyright 2006, notaz
+@ (c) Copyright 2007, Grazvydas "notaz" Ignotas
 @ All Rights Reserved
 
 
@@ -12,7 +14,6 @@
 .extern HighSprZ
 .extern rendstatus
 .extern DrawLineDest
-.extern DrawStripVSRam
 .extern DrawStripInterlace
 
 
@@ -68,7 +69,7 @@
     streqb  r4, [r1,#\offs]
 .endm
 
-@ TileNorm (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: register with helper pattern 0xf, touches r3 high bits
+@ TileNormShHP (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: register with helper pattern 0xf, touches r3 high bits
 .macro TileNormShHP
     TilePixelShHP 12, 0         @ #0x0000f000
     TilePixelShHP  8, 1         @ #0x00000f00
@@ -80,7 +81,7 @@
     TilePixelShHP 16, 7         @ #0x000f0000
 .endm
 
-@ TileFlip (r1=pdest, r2=pixels8, r3=pal) r4: scratch, pat: register with helper pattern 0xf
+@ TileFlipShHP (r1=pdest, r2=pixels8, r3=pal) r4: scratch, pat: register with helper pattern 0xf
 .macro TileFlipShHP
     TilePixelShHP 16, 0         @ #0x000f0000
     TilePixelShHP 20, 1         @ #0x00f00000
@@ -210,9 +211,9 @@
 @   int cells;   // 0x14
 @ };
 
-@ int DrawLayer(int plane, int *hcache, int maxcells, int sh)
+@ void DrawLayer(int plane_sh, int *hcache, int cellskip, int maxcells);
 
-.global DrawLayer @ int plane, int *hcache, int maxcells, int sh
+.global DrawLayer
 
 DrawLayer:
     stmfd   sp!, {r4-r11,lr}
@@ -220,10 +221,11 @@ DrawLayer:
     ldr     r11, =(Pico+0x22228)  @ Pico.video
     mov     r8, #1
 
-    ldrb    r7, [r11, #16]        @ ??hh??ww
+    ldrb    r7, [r11, #16]        @ ??vv??hh
 
     mov     r6, r1                @ hcache
-    orr     r9, r2, r3, lsl #31   @ r9=maxcells|(sh<<31)
+    orr     r9, r3, r0, lsl #30
+    orr     r9, r9, r2, lsl #8    @ r9=sh[31]|cellskip[15:8]|maxcells[7:0]  (tmp)
 
     mov     r1, r7, lsl #4
     orr     r1, r1, #0x00ff
@@ -243,7 +245,7 @@ DrawLayer:
     sub     r5, r5, #1            @ r5=xmask
 
     @ Find name table:
-    tst     r0,  r0
+    ands    r0,  r0, #1
     ldreqb  r12, [r11, #2]
     ldrneb  r12, [r11, #4]
 
@@ -291,7 +293,6 @@ DrawLayer:
     add     r12, r12, r4, lsl r10  @ nametab+=(ts.line>>3)<<shift[width];
 
     @ ldmia   r0, {r1,r2,r3,r5,r6,r9} @ r2=line, r3=ts->hscroll, r5=ts->xmask, r6=ts->hc, r9=ts->cells
-@    mov     r12,r1,  lsl #1 @ r12=(ts->nametab<<1) (halfword compliant)
 
     and     r10,r2,  #7
     mov     r10,r10, lsl #1 @ r10=ty=(ts->line&7)<<1;
@@ -306,14 +307,18 @@ DrawLayer:
 
     tst     r9, #1<<31
     mov     r3, #0
-    orrne   r10,r10, #1<<23 @ r10=(cells<<24|sh<<23|hi_not_empty<<22|ty)
+    orrne   r10,r10, #1<<23 @ r10=(cells<<24|sh<<23|hi_not_empty<<22|had_output<<21|ty)
     movne   r3, #0x40       @ default to shadowed pal on sh mode
 
-    mvn     r9, #0          @ r9=prevcode=-1
-
     cmp     r7, #8
     addne   r10,r10, #0x01000000 @ we will loop cells+1 times if there is scroll
 
+    and     r9, r9, #0xff00
+    add     r8, r8, r9, lsr #8   @ tilex+=cellskip
+    add     r7, r7, r9, lsr #5   @ dx+=cellskip<<3;
+    sub     r10,r10,r9, lsl #16  @ cells-=cellskip
+    mvn     r9, #0               @ r9=prevcode=-1
+
     @ cache some stuff to avoid mem access
     ldr     r11,=HighCol
     mov     r0, #0xf
@@ -342,6 +347,7 @@ DrawLayer:
     beq     .DrawStrip_samecode @ we know stuff about this tile already
 
     mov     r9, r7          @ remember code
+    orr     r10, r10, #1<<21 @ seen non hi-prio tile
 
     movs    r2, r9, lsl #20 @ if (code&0x1000)
     mov     r2, r2, lsl #1
@@ -363,16 +369,17 @@ DrawLayer:
     beq     .DrawStrip_SingleColor @ tileline singlecolor 
 
     tst     r9, #0x0800
-    beq     .DrawStrip_TileNorm
+    bne     .DrawStrip_TileFlip
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r0: helper pattern
-    TileFlip r0
-    b       .dsloop
-
 .DrawStrip_TileNorm:
     TileNorm r0
     b       .dsloop
 
+.DrawStrip_TileFlip:
+    TileFlip r0
+    b       .dsloop
+
 .DrawStrip_SingleColor:
     and     r4, r2, #0xf
     orr     r4, r3, r4
@@ -386,6 +393,20 @@ DrawLayer:
     strneb  r4, [r1], #1       @ have a remaining unaligned pixel?
     b       .dsloop_subr1
 
+.DrawStrip_hiprio_maybempt:
+    cmp     r7, r9
+    beq     .dsloop         @ must've been empty, otherwise we wouldn't get here
+    movs    r2, r7, lsl #20 @ if (code&0x1000)
+    mov     r2, r2, lsl #1
+    add     r2, r2, r10, lsl #17
+    mov     r2, r2, lsr #17
+    eorcs   r2, r2, #0x0e   @ if (code&0x1000) addr^=0xe;
+    ldr     r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
+    mov     r9, r7          @ remember code
+    tst     r2, r2
+    beq     .dsloop
+    orr     r10, r10, #1<<22
+
 .DrawStrip_hiprio:
     tst     r10, #0x00c00000
     beq     .DrawStrip_hiprio_maybempt
@@ -398,9 +419,163 @@ DrawLayer:
     mov     r0, #0xf
     b       .dsloop
 
-.DrawStrip_hiprio_maybempt:
+.dsloop_exit:
+    tst     r10, #1<<21 @ seen non hi-prio tile
+    ldreq   r1, =rendstatus
+    mov     r0, #0
+    ldreq   r2, [r1]
+    str     r0, [r6]    @ terminate the cache list
+    orreq   r2, r2, #0x40 @ had a layer with all hi-prio tiles
+    streq   r2, [r1]
+
+    ldmfd   sp!, {r4-r11,lr}
+    bx      lr
+
+@ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
+
+.DrawStrip_vsscroll:
+    rsb     r8, r3, #0
+    mov     r8, r8, lsr #3        @ r8=tilex=(-ts->hscroll)>>3
+    bic     r8, r8, #0xff000000
+    orr     r8, r8, r5, lsl #25   @ r8=(xmask[31:25]|had_output[24]|tilex[23:0])
+
+    ldr     r4, =Scanline
+    orr     r5, r1, r10, lsl #24
+    ldr     r4, [r4]
+    sub     r1, r3, #1
+    orr     r5, r5, r4, lsl #16   @ r5=(shift_width[31:24]|scanline[23:16]|ymask[15:0])
+    and     r1, r1, #7
+    add     r7, r1, #1            @ r7=dx=((ts->hscroll-1)&7)+1
+
+    mov     r10,r9, lsl #16
+    tst     r0, #1
+    orrne   r10,r10, #0x8000
+    tst     r9, #1<<31
+    mov     r3, #0
+    orr     r10,r10, #0xff000000 @ will be adjusted on entering loop
+    orrne   r10,r10, #1<<23 @ r10=(cell[31:24]|sh[23]|hi_not_empty[22]|cells_max[21:16]|plane[15]|ty[14:0])
+    movne   r3, #0x40       @ default to shadowed pal on sh mode
+
+    cmp     r7, #8
+    subne   r10,r10, #0x01000000 @ have hscroll, start with negative cell
+
+    and     r9, r9, #0xff00
+    add     r8, r8, r9, lsr #8   @ tilex+=cellskip
+    add     r7, r7, r9, lsr #5   @ dx+=cellskip<<3;
+    add     r10,r10,r9, lsl #16  @ cell+=cellskip
+    mvn     r9, #0               @ r9=prevcode=-1
+
+    @ cache some stuff to avoid mem access
+    ldr     r11,=HighCol
+    mov     r0, #0xf
+    add     r1, r11, r7         @ r1=pdest
+
+
+    @ r4 & r7 are scratch in this loop
+.dsloop_vs_subr1:
+    sub     r1, r1, #8
+.dsloop_vs: @ 40-41 times
+    add     r10,r10, #0x01000000
+    and     r4, r10, #0x003f0000
+    cmp     r4, r10, asr #8
+    ble     .dsloop_vs_exit
+
+    @ calc offset and read tileline code to r7, also calc ty
+    add     r7, lr, #0x012000
+    add     r7, r7, #0x000180     @ r7=Pico.vsram (Pico+0x22180)
+    add     r7, r7, r10,asr #23   @ vsram + ((cell&~1)<<1)
+    bic     r7, r7, #3
+    tst     r10,#0x8000           @ plane1?
+    addne   r7, r7, #2
+    ldrh    r7, [r7]              @ r7=vscroll
+
+    bic     r10,r10,#0xff         @ clear old ty
+    and     r4, r5, #0xff0000
+    add     r4, r4, r7, lsl #16
+    and     r4, r4, r5, lsl #16   @ r4=line<<16
+    and     r7, r4, #0x70000
+    orr     r10,r10,r7, lsr #15   @ new ty
+
+    mov     r4, r4, lsr #19
+    mov     r7, r5, lsr #24
+    mov     r4, r4, lsl r7        @ nametabadd
+
+    and     r7, r8, r8, lsr #25
+    add     r7, lr, r7, lsl #1    @ Pico.vram+((tilex&ts->xmask) as halfwords)
+    add     r7, r7, r4, lsl #1
+    ldrh    r7, [r7, r12]         @ r7=code (int, but from unsigned, no sign extend)
+
+    add     r1, r1, #8
+    add     r8, r8, #1
+
+    tst     r7, #0x8000
+    bne     .DrawStrip_vs_hiprio
+
     cmp     r7, r9
-    beq     .dsloop         @ must've been empty, otherwise we wouldn't get here
+    beq     .DrawStrip_vs_samecode @ we know stuff about this tile already
+
+    mov     r9, r7          @ remember code
+    orr     r8, r8, #1<<24  @ seen non hi-prio tile
+
+    movs    r2, r9, lsl #20 @ if (code&0x1000)
+    mov     r2, r2, lsl #1
+    add     r2, r2, r10, lsl #17
+    mov     r2, r2, lsr #17
+    eorcs   r2, r2, #0x0e   @ if (code&0x1000) addr^=0xe;
+
+    ldr     r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
+
+    bic     r7, r3, #0x3f
+    and     r3, r9, #0x6000
+    add     r3, r7, r3, lsr #9 @ r3=pal=((code&0x6000)>>9);
+
+.DrawStrip_vs_samecode:
+    tst     r2, r2
+    beq     .dsloop_vs              @ tileline blank
+
+    cmp     r2, r2, ror #4
+    beq     .DrawStrip_vs_SingleColor @ tileline singlecolor 
+
+    tst     r9, #0x0800
+    bne     .DrawStrip_vs_TileFlip
+
+    @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r0: helper pattern
+.DrawStrip_vs_TileNorm:
+    TileNorm r0
+    b       .dsloop_vs
+
+.DrawStrip_vs_TileFlip:
+    TileFlip r0
+    b       .dsloop_vs
+
+.DrawStrip_vs_SingleColor:
+    and     r4, r2, #0xf
+    orr     r4, r3, r4
+    orr     r4, r4, r4, lsl #8
+    tst     r1, #1             @ not aligned?
+    strneb  r4, [r1], #1
+    streqh  r4, [r1], #2
+    strh    r4, [r1], #2
+    strh    r4, [r1], #2
+    strh    r4, [r1], #2
+    strneb  r4, [r1], #1       @ have a remaining unaligned pixel?
+    b       .dsloop_vs_subr1
+
+.DrawStrip_vs_hiprio:
+    tst     r10, #0x00c00000
+    beq     .DrawStrip_vs_hiprio_maybempt
+    sub     r0, r1, r11
+    orr     r7, r7, r0,  lsl #16
+    orr     r7, r7, r10, lsl #25 @ (ty<<25)
+    tst     r7, #0x1000
+    eorne   r7, r7, #7<<26  @ if(code&0x1000) cval^=7<<26;
+    str     r7, [r6], #4    @ cache hi priority tile
+    mov     r0, #0xf
+    b       .dsloop_vs
+
+.DrawStrip_vs_hiprio_maybempt:
+    cmp     r7, r9
+    beq     .dsloop_vs         @ must've been empty, otherwise we wouldn't get here
     movs    r2, r7, lsl #20 @ if (code&0x1000)
     mov     r2, r2, lsl #1
     add     r2, r2, r10, lsl #17
@@ -410,41 +585,23 @@ DrawLayer:
     mov     r9, r7          @ remember code
     tst     r2, r2
     orrne   r10, r10, #1<<22
-    bne     .DrawStrip_hiprio
-    b       .dsloop
+    bne     .DrawStrip_vs_hiprio
+    b       .dsloop_vs
 
-.dsloop_exit:
+.dsloop_vs_exit:
+    tst     r8, #1<<24 @ seen non hi-prio tile
+    ldreq   r1, =rendstatus
     mov     r0, #0
+    ldreq   r2, [r1]
     str     r0, [r6]    @ terminate the cache list
+    orreq   r2, r2, #0x40 @ had a layer with all hi-prio tiles
+    streq   r2, [r1]
 
     ldmfd   sp!, {r4-r11,lr}
     bx      lr
 
 
-.DrawStrip_vsscroll:
-    @ shit, we have 2-cell column based vscroll
-    @ let the c code handle this (for now)
-
-    @   int nametab; // 0x00
-    @   int line;    // 0x04
-    @   int hscroll; // 0x08
-    @   int xmask;   // 0x0C
-    @   int *hc;     // 0x10 (pointer to cache buffer)
-    @   int cells;   // 0x14
-
-    sub     sp, sp, #6*4
-    orr     r2, r1, r10, lsl #24 @ ts.line=ymask|(shift[width]<<24); // save some stuff instead of line
-    mov     r1, r0               @ plane
-    mov     r0, r12, lsr #1      @ halfwords
-    and     r9, r9, #0xff
-    stmia   sp, {r0,r2,r3,r5,r6,r9}
-
-    mov     r0, sp
-    bl      DrawStripVSRam @ struct TileStrip *ts, int plane
-
-    add     sp, sp, #6*4
-    ldmfd   sp!, {r4-r11,lr}
-    bx      lr
+@ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
 
 @ interlace mode 2? Sonic 2?
 .DrawStrip_interlace:
@@ -517,26 +674,28 @@ BackFill:
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
 
 
-.global DrawTilesFromCache @ int *hc, int sh
+.global DrawTilesFromCache @ int *hc, int sh, int rlim
 
 DrawTilesFromCache:
     stmfd   sp!, {r4-r8,r11,lr}
 
-    mvn     r5, #0         @ r5=prevcode=-1
-    mov     r8, r1
-
     @ cache some stuff to avoid mem access
     ldr     r11,=HighCol
     ldr     lr, =(Pico+0x10000) @ lr=Pico.vram
     mov     r12,#0xf
 
+    mvn     r5, #0         @ r5=prevcode=-1
+    ands    r8, r1, #1
+    orr     r8, r8, r2, lsl #1
+    bne     .dtfc_check_rendflags
+
     @ scratch: r4, r7
 .dtfc_loop:
     ldr     r6, [r0], #4    @ read code
     movs    r1, r6, lsr #16 @ r1=dx;
     ldmeqfd sp!, {r4-r8,r11,pc} @ dx is never zero, this must be a terminator, return
-    bic     r1, r1, #0xfe00
-    add     r1, r11, r1     @ r1=pdest
+    bic     r4, r1, #0xfe00
+    add     r1, r11, r4     @ r1=pdest
 
     mov     r7, r6, lsl #16
     cmp     r5, r7, lsr #16
@@ -554,7 +713,10 @@ DrawTilesFromCache:
     ldr     r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
 
 .dtfc_samecode:
-    tst     r8, r8
+    rsbs    r4, r4, r8, lsr #1
+    bmi     .dtfc_cut_tile
+
+    tst     r8, #1
     bne     .dtfc_shadow
 
     tst     r2, r2
@@ -564,16 +726,17 @@ DrawTilesFromCache:
     beq     .dtfc_SingleColor @ tileline singlecolor 
 
     tst     r5, #0x0800
-    beq     .dtfc_TileNorm
+    bne     .dtfc_TileFlip
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlip r12
-    b       .dtfc_loop
-
 .dtfc_TileNorm:
     TileNorm r12
     b       .dtfc_loop
 
+.dtfc_TileFlip:
+    TileFlip r12
+    b       .dtfc_loop
+
 .dtfc_SingleColor:
     and     r4, r2, #0xf
     orr     r4, r3, r4
@@ -595,16 +758,17 @@ DrawTilesFromCache:
     beq     .dtfc_SingleColor @ tileline singlecolor 
 
     tst     r5, #0x0800
-    beq     .dtfc_TileNormShHP
+    bne     .dtfc_TileFlipShHP
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlipShHP
-    b       .dtfc_loop
-
 .dtfc_TileNormShHP:
     TileNormShHP
     b       .dtfc_loop
 
+.dtfc_TileFlipShHP:
+    TileFlipShHP
+    b       .dtfc_loop
+
 .dtfc_shadow_blank:
     ldrb    r4, [r1]        @ 1ci
     ldrb    r12,[r1,#1]
@@ -641,6 +805,64 @@ DrawTilesFromCache:
     mov     r12, #0xf
     b       .dtfc_loop
 
+.dtfc_cut_tile:
+    add     r4, r4, #7      @ 0-6
+    mov     r4, r4, lsl #2
+    mov     r12,#0xf<<28
+    mov     r12,r12,asr r4
+    mov     r2, r2, ror #16
+    tst     r5, #0x0800     @ flipped?
+    mvnne   r12,r12
+    and     r2, r2, r12
+    mov     r2, r2, ror #16
+    mov     r12,#0xf
+    tst     r8, #1
+    bne     .dtfc_shadow
+    tst     r2, r2
+    beq     .dtfc_loop
+    tst     r5, #0x0800
+    beq     .dtfc_TileNorm
+    b       .dtfc_TileFlip
+
+@ check if we have detected layer covered with hi-prio tiles:
+.dtfc_check_rendflags:
+    ldr     r1, =rendstatus
+    ldr     r2, [r1]
+    tst     r2, #0xc0
+    beq     .dtfc_loop
+    bic     r8, r8, #1      @ sh/hi mode off
+    tst     r2, #0x80
+    bne     .dtfc_loop      @ already processed
+    orr     r2, r2, #0x80
+    str     r2, [r1]
+
+    add     r1, r11,#8
+    mov     r3, #320/4
+    mov     r7, #0x80
+    orr     r7, r7, r7, lsl #8
+    orr     r7, r7, r7, lsl #16
+    mov     r6, #0x3f
+    orr     r6, r6, r6, lsl #8
+    orr     r6, r6, r6, lsl #16
+.dtfc_loop_shprep:
+    subs    r3, r3, #1
+    bmi     .dtfc_loop      @ done
+    ldr     r2, [r1]
+    tst     r2, r7
+    andeq   r2, r2, r6
+    streq   r2, [r1], #4
+    beq     .dtfc_loop_shprep
+    tst     r2,     #0x80000000
+    biceq   r2, r2, #0xc0000000
+    tst     r2,     #0x00800000
+    biceq   r2, r2, #0x00c00000
+    tst     r2,     #0x00008000
+    biceq   r2, r2, #0x0000c000
+    tst     r2,     #0x00000080
+    biceq   r2, r2, #0x000000c0
+    str     r2, [r1], #4
+    b       .dtfc_loop_shprep
+
 .pool
 
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@@ -717,16 +939,17 @@ DrawSpritesFromCache:
     beq     .dsfc_SingleColor @ tileline singlecolor 
 
     tst     r9, #0x10000
-    beq     .dsfc_TileNorm
+    bne     .dsfc_TileFlip
 
     @ TileFlip (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlip r12
-    b       .dsfc_inloop
-
 .dsfc_TileNorm:
     TileNorm r12
     b       .dsfc_inloop
 
+.dsfc_TileFlip:
+    TileFlip r12
+    b       .dsfc_inloop
+
 .dsfc_SingleColor:
     tst     r0, #1              @ not aligned?
     and     r4, r2, #0xf
@@ -745,16 +968,17 @@ DrawSpritesFromCache:
     beq     .dsfc_singlec_sh
 
     tst     r9, #0x10000
-    beq     .dsfc_TileNorm_sh
+    bne     .dsfc_TileFlip_sh
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlipSh
-    b       .dsfc_inloop
-
 .dsfc_TileNorm_sh:
     TileNormSh
     b       .dsfc_inloop
 
+.dsfc_TileFlip_sh:
+    TileFlipSh
+    b       .dsfc_inloop
+
 .dsfc_singlec_sh:
     cmp     r2, #0xe0000000
     bcc     .dsfc_SingleColor   @ normal singlecolor tileline (carry inverted in ARM)
@@ -861,17 +1085,18 @@ DrawSprite:
     beq     .dspr_SingleColor @ tileline singlecolor 
 
     tst     r9, #0x0800
-    beq     .dspr_TileNorm
+    bne     .dspr_TileFlip
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlip r12
-    b       .dspr_loop
-
 @ scratch: r4, r7
 .dspr_TileNorm:
     TileNorm r12
     b       .dspr_loop
 
+.dspr_TileFlip:
+    TileFlip r12
+    b       .dspr_loop
+
 .dspr_SingleColor:
     and     r4, r2, #0xf
     orr     r4, r3, r4
@@ -890,16 +1115,17 @@ DrawSprite:
     beq     .dspr_singlec_sh
 
     tst     r9, #0x0800
-    beq     .dspr_TileNorm_sh
+    bne     .dspr_TileFlip_sh
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r12: helper pattern
-    TileFlipSh
-    b       .dspr_loop
-
 .dspr_TileNorm_sh:
     TileNormSh
     b       .dspr_loop
 
+.dspr_TileFlip_sh:
+    TileFlipSh
+    b       .dspr_loop
+
 .dspr_singlec_sh:
     cmp     r2, #0xe0000000
     bcc     .dspr_SingleColor   @ normal tileline
@@ -1027,16 +1253,17 @@ DrawWindow:
     beq     .dw_SingleColor @ tileline singlecolor 
 
     tst     r9, #0x0800
-    beq     .dw_TileNorm
+    bne     .dw_TileFlip
 
     @ (r1=pdest, r2=pixels8, r3=pal) r4: scratch, r0: helper pattern
-    TileFlip r0
-    b       .dwloop
-
 .dw_TileNorm:
     TileNorm r0
     b       .dwloop
 
+.dw_TileFlip:
+    TileFlip r0
+    b       .dwloop
+
 .dw_SingleColor:
     and     r4, r0, r2         @ #0x0000000f
     orr     r4, r3, r4
@@ -1285,35 +1512,21 @@ vidConvCpyRGB565: @ void *to, void *from, int pixels
 
 FinalizeLineRGB555:
     stmfd   sp!, {r4-r8,lr}
-    ldr     r5, =(Pico+0x22228)  @ Pico.video
+    ldr     r8, =(Pico+0x22228)  @ Pico.video
     ldr     r4, =HighPal
 
-    ldrb    r7, [r5, #-0x1a]     @ 0x2220e ~ dirtyPal
+    ldrb    r7, [r8, #-0x1a]     @ 0x2220e ~ dirtyPal
     mov     r6, r0
     mov     r1, #0
     tst     r7, r7
     beq     .fl_noconvRGB555
-    strb    r1, [r5, #-0x1a]
-    sub     r1, r5, #0x128       @ r1=Pico.cram
+    strb    r1, [r8, #-0x1a]
+    sub     r1, r8, #0x128       @ r1=Pico.cram
     mov     r0, r4
     mov     r2, #0x40
     bl      vidConvCpyRGB565
 
 .fl_noconvRGB555:
-    ldrb    r12, [r5, #12]
-    ldr     r0, =DrawLineDest
-    ldr     r0, [r0]
-
-    tst     r12, #1
-    movne   r2, #320/8           @ len
-    bne     .fl_no32colRGB555
-    ldr     r3, =PicoOpt
-    mov     r2, #256/8
-    ldr     r3, [r3]
-    tst     r3, #0x100
-    addeq   r0, r0, #32*2
-
-.fl_no32colRGB555:
     mov     r3, r4
     tst     r6, r6
     beq     .fl_noshRGB555
@@ -1348,12 +1561,27 @@ FinalizeLineRGB555:
 
     sub     r3, r3, #0x40*2
 
-
 .fl_noshRGB555:
+    ldr     r0, =DrawLineDest
     ldr     r1, =(HighCol+8)
+    ldr     r0, [r0]
+
+    ldrb    r12, [r8, #12]
     mov     lr, #0xff
     mov     lr, lr, lsl #1
 
+    tst     r12, #1
+    movne   r2, #320/8           @ len
+    bne     .fl_no32colRGB555
+    ldr     r4, =PicoOpt
+    mov     r2, #256/8
+    ldr     r4, [r4]
+    tst     r4, #0x4000
+    bne     .fl_32scale_RGB555
+    tst     r4, #0x0100
+    addeq   r0, r0, #32*2
+
+.fl_no32colRGB555:
 .fl_loopRGB555:
 
     ldr     r12, [r1], #4
@@ -1387,9 +1615,71 @@ FinalizeLineRGB555:
     stmia   r0!, {r4,r5,r8,r12}
     bne     .fl_loopRGB555
 
+    ldmfd   sp!, {r4-r8,lr}
+    bx      lr
+
+
+.fl_32scale_RGB555:
+    stmfd   sp!, {r9,r10}
+    mov     r9, #0x3900 @ f800 07e0 001f | e000 0780 001c | 3800 01e0 0007
+    orr     r9, r9, #0x00e7
+
+.fl_loop32scale_RGB555:
+    ldr     r12, [r1], #4
+    ldr     r7,  [r1], #4
+
+    and     r4, lr, r12,lsl #1
+    ldrh    r4, [r3, r4]
+    and     r5, lr, r12,lsr #7
+    ldrh    r5, [r3, r5]
+    and     r4, r4, r9, lsl #2
+    orr     r4, r4, r4, lsl #14       @ r4[31:16] = 1/4 pix_s 0
+    and     r5, r5, r9, lsl #2
+    sub     r6, r5, r5, lsr #2        @ r6 = 3/4 pix_s 1
+    add     r4, r4, r6, lsl #16       @ pix_d 0, 1
+    and     r6, lr, r12,lsr #15
+    ldrh    r6, [r3, r6]
+    and     r12,lr, r12,lsr #23
+    ldrh    r12,[r3, r12]
+    and     r6, r6, r9, lsl #2
+    add     r5, r5, r6
+    mov     r5, r5, lsr #1
+    sub     r6, r6, r6, lsr #2        @ r6 = 3/4 pix_s 2
+    orr     r5, r5, r6, lsl #16
+
+    and     r6, lr, r7, lsl #1
+    ldrh    r6, [r3, r6]
+    and     r12,r12,r9, lsl #2
+    add     r5, r5, r12,lsl #14       @ pix_d 2, 3
+    and     r6, r6, r9, lsl #2
+    orr     r6, r12,r6, lsl #16       @ pix_d 4, 5
+
+    and     r12,lr, r7, lsr #7
+    ldrh    r12,[r3, r12]
+    and     r10,lr, r7, lsr #15
+    ldrh    r10,[r3, r10]
+    and     r12,r12,r9, lsl #2
+    sub     r8, r12,r12,lsr #2        @ r8 = 3/4 pix_s 1
+    add     r8, r8, r6, lsr #18
+    and     r7, lr, r7, lsr #23
+    ldrh    r7, [r3, r7]
+    and     r10,r10,r9, lsl #2
+    orr     r8, r8, r10,lsl #15
+    add     r8, r8, r12,lsl #15       @ pix_d 6, 7
+    sub     r10,r10,r10,lsr #2        @ r10= 3/4 pix_s 2
+    and     r7, r7, r9, lsl #2
+    add     r10,r10,r7, lsr #2        @ += 1/4 pix_s 3
+    orr     r10,r10,r7, lsl #16       @ pix_d 8, 9
+
+    subs    r2, r2, #1
+
+    stmia   r0!, {r4,r5,r6,r8,r10}
+    bne     .fl_loop32scale_RGB555
 
+    ldmfd   sp!, {r9,r10}
     ldmfd   sp!, {r4-r8,lr}
-    bx lr
+    bx      lr
+
 
 @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@