core vdp, arm rendering speed optimisation

This commit is contained in:
kub 2024-07-18 21:36:43 +02:00
parent 1fad746a1f
commit 11a1966bf3
2 changed files with 85 additions and 96 deletions

View file

@ -310,7 +310,7 @@ TileFlipMakerAS(TileFlipSH_AS_and, pix_sh_as_and)
\
pack = 0; \
if (code != blank) { \
/* Get tile address/2: */\
/* Get tile address/2: */ \
u32 addr = ((code&0x7ff)<<4) + ty; \
if (code & 0x1000) addr ^= 0xe; /* Y-flip */ \
\

View file

@ -433,7 +433,7 @@ DrawLayer:
movs r3, r9, lsl #1 @ (force[31]|sh[30]) << 1
mov r3, #0
orrmi r10,r10, #1<<23 @ r10=cells[31:24]|sh[23]|hi_not_empty[22]
@ orrcc r10,r10, #1<<20 @ |had_output[21]|!force[20]|hscroll[19:17]|ty[15:0]
@ orrcc r10,r10, #1<<20 @ |had_output[21]|!force[20]|hscroll[18:16]|ty[15:0]
movmi r3, #0x80 @ default to shadowed pal on sh mode
and r4, r7, #7
@ -452,7 +452,7 @@ DrawLayer:
mvn r9, #0 @ r9=prevcode=-1
add r1, r11, r7 @ r1=pdest
@ r10=cells[31:24]|sh[23]|hi_not_empty[22]|had_output[21]|!force[20]|hscroll[19:17]|ty[15:0]
@ r10=cells[31:24]|sh[23]|hi_not_empty[22]|had_output[21]|!force[20]|hscroll[18:16]|ty[15:0]
@ r1=pd+dx r2=pack r3=pal r5=xmask r6=hc r8=tilex r9=prevcode r11=HighCol r12=nametab lr=vram
@ r4 & r7 are scratch in this loop
@ -467,21 +467,19 @@ DrawLayer:
add r8, r8, #1
movs r2, r9, lsl #20 @ if (code&0x1000)
mov r2, r2, lsl #1
tst r9, #0x1000 @ if (code&0x1000)
mov r2, r9, lsl #21
add r2, r2, r10, lsl #17
mov r2, r2, lsr #17
eorcs r2, r2, #0x0e @ if (code&0x1000) addr^=0xe;
eorne r2, r2, #0xe<<17 @ if (code&0x1000) addr^=0xe;
ldr r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
ldr r2, [lr, r2, lsr #16] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
mvn r7, #0
mov r4, r4, lsr #16-2 @ (dx&7)*4
tst r9, #0x0800
moveq r7, r7, lsl r4 @ mask = ~0 [shift] (dx&7)*4
movne r7, r7, lsr r4
mvn r7, r7, ror #16
and r2, r2, r7 @ pack&mask
bic r2, r2, r7, ror #16 @ pack&~mask
orr r9, r9, #0x80000000 @ invalidate oldcode since pack is masked
b .DrawStrip_samecode
@ -504,13 +502,12 @@ DrawLayer:
mov r9, r7 @ remember code
movs r2, r9, lsl #20 @ if (code&0x1000)
mov r2, r2, lsl #1
tst r9, #0x1000 @ if (code&0x1000)
mov r2, r9, lsl #21
add r2, r2, r10, lsl #17
mov r2, r2, lsr #17
eorcs r2, r2, #0x0e @ if (code&0x1000) addr^=0xe;
eorne r2, r2, #0x0e<<17 @ if (code&0x1000) addr^=0xe;
ldr r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
ldr r2, [lr, r2, lsr #16] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
.DrawStrip_samecode:
tst r9, #0x8000
@ tstne r10, #1<<20 @ !force[20]
@ -577,21 +574,19 @@ DrawLayer:
add r1, r1, #8
movs r2, r9, lsl #20 @ if (code&0x1000)
mov r2, r2, lsl #1
tst r9, #0x1000 @ if (code&0x1000)
mov r2, r9, lsl #21
add r2, r2, r10, lsl #17
mov r2, r2, lsr #17
eorcs r2, r2, #0x0e @ if (code&0x1000) addr^=0xe;
eorne r2, r2, #0x0e<<17 @ if (code&0x1000) addr^=0xe;
ldr r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
ldr r2, [lr, r2, lsr #16] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
mvn r7, #0
mov r4, r4, lsr #16-2 @ (dx&7)*4
tst r9, #0x0800
moveq r7, r7, lsl r4 @ mask = ~0 [shift] (dx&7)*4
movne r7, r7, lsr r4
mov r7, r7, ror #16
and r2, r2, r7 @ pack&mask
and r2, r2, r7, ror #16 @ pack&mask
bic r10,r10, #7<<16
b .DrawStrip_samecode @ one last time, with last tile now masked
@ -743,13 +738,12 @@ DrawLayer:
mov r9, r7 @ remember code
movs r2, r9, lsl #20 @ if (code&0x1000)
mov r2, r2, lsl #1
tst r9, #0x1000 @ if (code&0x1000)
mov r2, r9, lsl #21
add r2, r2, r10, lsl #17
mov r2, r2, lsr #17
eorcs r2, r2, #0x0e @ if (code&0x1000) addr^=0xe;
eorne r2, r2, #0x0e<<17 @ if (code&0x1000) addr^=0xe;
ldr r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(PicoMem.vram+addr); // Get 8 pixels
ldr r2, [lr, r2, lsr #16] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
.DrawStrip_vs_samecode:
tst r9, #0x8000
@ -1007,8 +1001,7 @@ DrawTilesFromCache:
tst r6, #0x0800 @ flipped?
moveq r12,r12, lsl r4
movne r12,r12, lsr r4
mov r12,r12, ror #16
and r2, r2, r12
and r2, r2, r12, ror #16
mov r12,#0xf
tst r8, #1
bne .dtfc_shadow
@ -1152,9 +1145,7 @@ DrawSpriteSHi:
cmp r0, #328
bge DrawSpriteSHi
mov r8, r8, lsl #17
mov r8, r8, lsr #17 @ tile&=0x7fff; // Clip tile address
bic r8, r8, #0xf8000 @ tile&=0x7fff; // Clip tile address
ldr r2, [lr, r8, lsl #1] @ pack=*(unsigned int *)(PicoMem.vram+addr); // Get 8 pixels
add r1, r11, r0 @ r1=pdest
tst r2, r2
@ -1346,9 +1337,7 @@ DrawSprite:
cmp r0, #328
bge DrawSprite
mov r8, r8, lsl #17
mov r8, r8, lsr #17 @ tile&=0x7fff; // Clip tile address
bic r8, r8, #0xf8000 @ tile&=0x7fff; // Clip tile address
ldr r2, [lr, r8, lsl #1] @ pack=*(unsigned int *)(PicoMem.vram+addr); // Get 8 pixels
add r1, r11, r0 @ r1=pdest
tst r2, r2
@ -1492,16 +1481,16 @@ DrawWindow:
mov r9, r7 @ remember code
movs r2, r9, lsl #20 @ if (code&0x1000)
mov r2, r2, lsl #1
add r2, r10, r2, lsr #17 @ r2=addr=(code&0x7ff)<<4; addr+=ty
eorcs r2, r2, #0x0e @ if (code&0x1000) addr^=0xe;
tst r9, #0x1000 @ if (code&0x1000)
mov r2, r9, lsl #21
add r2, r2, r10, lsl #17
eorne r2, r2, #0xe<<17 @ if (code&0x1000) addr^=0xe;
ldr r2, [lr, r2, lsr #16] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
and r3, r9, #0x6000
mov r3, r3, lsr #9 @ r3=pal=((code&0x6000)>>9);
ldr r2, [lr, r2, lsl #1] @ pack=*(unsigned int *)(Pico.vram+addr); // Get 8 pixels
.dw_samecode:
tst r6, #0x100
bne .dw_shadow