#368 GBA pre-fill prev/next VertexLink's for non-clipped faces

XProger · May 26, 2022 · c820952 · c820952
1 parent cb681c2
commit c820952
Show file tree

Hide file tree

Showing 4 changed files with 179 additions and 217 deletions.
diff --git a/src/platform/gba/asm/common_asm.inc b/src/platform/gba/asm/common_asm.inc
@@ -44,18 +44,19 @@
 .equ CLIP_DISCARD,  (CLIP_LEFT + CLIP_RIGHT + CLIP_TOP + CLIP_BOTTOM + CLIP_FAR + CLIP_NEAR)
 .equ CLIP_MASK,     0xFF00
 
+.equ FACE_TYPE_SHIFT,   14
 .equ FACE_TEXTURE_BITS, 14
-.equ FACE_TEXTURE,      ((1 << FACE_TEXTURE_BITS) - 1)
+.equ FACE_TEXTURE,      ((1 << FACE_TYPE_SHIFT) - 1)
 .equ FACE_GOURAUD,      (2 << FACE_TYPE_SHIFT)
 .equ FACE_CLIPPED,      (1 << 18)
-.equ FACE_TRIANGLE,     (1 << 19)
+.equ FACE_TRIANGLE_BIT, 19
+.equ FACE_TRIANGLE,     (1 << FACE_TRIANGLE_BIT)
 
 .equ FACE_FLAGS,      0
 .equ FACE_NEXT,       4
 .equ FACE_INDICES,    8
 
-.equ FACE_TYPE_SHIFT, 14
-.equ FACE_TYPE_MASK,  (15 << FACE_TYPE_SHIFT)
+.equ FACE_TYPE_MASK,    (15 << FACE_TYPE_SHIFT)
 
 .equ FACE_TYPE_SHADOW,  (0 << FACE_TYPE_SHIFT)
 .equ FACE_TYPE_F,       (1 << FACE_TYPE_SHIFT)

diff --git a/src/platform/gba/asm/faceAddRoomQuads.s b/src/platform/gba/asm/faceAddRoomQuads.s
@@ -33,8 +33,6 @@ depth       .req vg0
 tmp         .req flags
 next        .req vp0
 
-SP_SIZE = 4
-
 .global faceAddRoomQuads_asm
 faceAddRoomQuads_asm:
     stmfd sp!, {r4-r7}

diff --git a/src/platform/gba/asm/flush.s b/src/platform/gba/asm/flush.s
@@ -1,16 +1,15 @@
 #include "common_asm.inc"
 
-flags        .req r0    // flags is always in r0 for rasterize & draw* calls
-vXY          .req r1
-vZG          .req r2
-tmp          .req r3
-
-OT           .req r4
+flags        .req r0    // flags must be in r0 for rasterize & draw* calls
+ptr          .req r1    // must be in r1
+vXY          .req r2
+vZG          .req r3
+tmp          .req r4
 list         .req r5
 face         .req r6
 VERTICES     .req r7
 TEXTURES     .req r8
-SPRITES      .req r9
+OT           .req r9
 TILE         .req r10
 MASK         .req r11
 
@@ -39,9 +38,20 @@ vZG0         .req vZG
 vXY1         .req index01
 vZG1         .req index23
 
-SP_SIZE = (16 * VERTEX_SIZEOF)
+vA           .req vXY
+vB           .req vZG
+
+Qs           .req ptr
+Qe           .req TILE
+Ts           .req MASK
+Te           .req index01
+PN           .req index23
+sprites      .req index01
 
-.extern rasterize_c, drawTriangle, drawQuad, drawPoly
+SP_SIZE = (7 * VERTEX_SIZEOF) + 4
+SP_SPRITES = SP_SIZE - 4
+
+.extern rasterize_c, drawPoly
 
 .global flush_asm
 flush_asm:
@@ -61,6 +71,25 @@ flush_asm:
 
     str faces, [tmp]
 
+    // fill VertexLink prev & next indices
+    sub sp, #SP_SIZE
+    add tmp, sp, #VERTEX_PREV
+    mov Qs, #255
+    add Qs, #4
+    mvn Qe, #512
+    sub Ts, Qs, #1
+    mvn Te, #256
+    mvn PN, #65024
+    // quad 
+    strh Qs, [tmp], #VERTEX_SIZEOF
+    strh PN, [tmp], #VERTEX_SIZEOF
+    strh PN, [tmp], #VERTEX_SIZEOF
+    strh Qe, [tmp], #VERTEX_SIZEOF
+    // triangle
+    strh Ts, [tmp], #VERTEX_SIZEOF
+    strh PN, [tmp], #VERTEX_SIZEOF
+    strh Te, [tmp], #VERTEX_SIZEOF
+
     ldr tmp, =level
     ldr TILE, =gTile
     ldr TEXTURES, [tmp, #LEVEL_TEXTURES]
@@ -71,7 +100,7 @@ flush_asm:
     mov MASK, #0xFF00
     orr MASK, MASK, MASK, lsl #16
 
-    sub sp, #SP_SIZE
+    str SPRITES, [sp, #SP_SPRITES]
 .loop_ot:
     ldr face, [list], #-4   // read the first face from the list and decrement
     cmp face, #0
@@ -84,37 +113,40 @@ flush_asm:
     ldmia face, {flags, face, index01, index23} // read face params and next face
 
     and type, flags, #FACE_TYPE_MASK
-    
+
 .draw_primitive: // shadows, triangles, quads and clipped polys
     cmp type, #FACE_TYPE_GTA
     bgt .draw_sprite
 
+    tst flags, #FACE_TRIANGLE
+    moveq ptr, sp                           // ptr to quad
+    addne ptr, sp, #(VERTEX_SIZEOF * 4)     // ptr to triangle
+
   .set_vertices:
     // 1st vertex
     mov vertex, index01, lsl #16
     add vertex, VERTICES, vertex, lsr #(16 - 3)
     ldmia vertex, {vXY, vZG}
-    stmia sp, {vXY, vZG}
+    stmia ptr, {vXY, vZG}
 
     // 2nd vertex
     add vertex, VERTICES, index01, lsr #(16 - 3) // assumption: vertex index will never exceed 8191
     ldmia vertex, {vXY, vZG}
-    str vXY, [sp, #(VERTEX_X + VERTEX_SIZEOF * 1)]
-    str vZG, [sp, #(VERTEX_Z + VERTEX_SIZEOF * 1)]
-    
+    str vXY, [ptr, #(VERTEX_X + VERTEX_SIZEOF * 1)]
+    str vZG, [ptr, #(VERTEX_Z + VERTEX_SIZEOF * 1)]
+
     // 3rd vertex
     mov vertex, index23, lsl #16
     add vertex, VERTICES, vertex, lsr #(16 - 3)
     ldmia vertex, {vXY, vZG}
-    str vXY, [sp, #(VERTEX_X + VERTEX_SIZEOF * 2)]
-    str vZG, [sp, #(VERTEX_Z + VERTEX_SIZEOF * 2)]
+    str vXY, [ptr, #(VERTEX_X + VERTEX_SIZEOF * 2)]
+    str vZG, [ptr, #(VERTEX_Z + VERTEX_SIZEOF * 2)]
 
     // 4th vertex (quads only)
-    tst flags, #FACE_TRIANGLE
     addeq vertex, VERTICES, index23, lsr #(16 - 3)
     ldmeqia vertex, {vXY, vZG}
-    streq vXY, [sp, #(VERTEX_X + VERTEX_SIZEOF * 3)]
-    streq vZG, [sp, #(VERTEX_Z + VERTEX_SIZEOF * 3)]
+    streq vXY, [ptr, #(VERTEX_X + VERTEX_SIZEOF * 3)]
+    streq vZG, [ptr, #(VERTEX_Z + VERTEX_SIZEOF * 3)]
 
     // skip texturing for FACE_TYPE_SHADOW and FACE_TYPE_F
     cmp type, #FACE_TYPE_F
@@ -133,35 +165,50 @@ flush_asm:
     str texTile, [TILE]
 
     and uv, MASK, uv01
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 0)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 0)]
     and uv, MASK, uv01, lsl #8
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 1)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 1)]
     and uv, MASK, uv23
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 2)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 2)]
     and uv, MASK, uv23, lsl #8
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 3)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 3)]
 
   .draw:
     // r0 = flags
-    mov r1, sp
+    // r1 = ptr
     adr lr, .next_face
 
     tst flags, #FACE_CLIPPED
     bne drawPoly
-    tst flags, #FACE_TRIANGLE
-    bne drawTriangle
-    beq drawQuad
+
+    // get top vertex for tri or quad rasterization
+    mov tmp, ptr
+    ldrsh vA, [tmp, #(VERTEX_Y + VERTEX_SIZEOF * 0)]
+    ldrsh vB, [tmp, #(VERTEX_Y + VERTEX_SIZEOF * 1)]
+    cmp vA, vB
+    addgt ptr, tmp, #(VERTEX_SIZEOF * 1)
+    movgt vA, vB
+    ldrsh vB, [tmp, #(VERTEX_Y + VERTEX_SIZEOF * 2)]
+    cmp vA, vB
+    addgt ptr, tmp, #(VERTEX_SIZEOF * 2)
+    movgt vA, vB
+    lsls vB, flags, #(31 - FACE_TRIANGLE_BIT) // check #FACE_TRIANGLE as sign bit for both pl and gt w/o branch
+    ldrplsh vB, [tmp, #(VERTEX_Y + VERTEX_SIZEOF * 3)]
+    cmppl vA, vB
+    addgt ptr, tmp, #(VERTEX_SIZEOF * 3)
+    b rasterize_asm
 
 .draw_sprite: // sprites and gui elements
+    mov ptr, sp
     mov vertex, index01, lsl #16
     add vertex, VERTICES, vertex, lsr #(16 - 3)
     ldmia vertex, {vXY0, vZG0, vXY1, vZG1}
-    stmia sp, {vXY0, vZG0}
-    str vXY1, [sp, #(VERTEX_X + VERTEX_SIZEOF * 1)]
-    str vZG1, [sp, #(VERTEX_Z + VERTEX_SIZEOF * 1)]
+    stmia ptr, {vXY0, vZG0}
+    str vXY1, [ptr, #(VERTEX_X + VERTEX_SIZEOF * 1)]
+    str vZG1, [ptr, #(VERTEX_Z + VERTEX_SIZEOF * 1)]
 
     // r0 = flags
-    mov r1, sp
+    // r1 = ptr
     adr lr, .next_face
 
     // gui
@@ -170,17 +217,18 @@ flush_asm:
 
     // sprite
     and sprIndex, flags, #0xFF
-    add sprite, SPRITES, sprIndex, lsl #4
+    ldr sprites, [sp, #SP_SPRITES]
+    add sprite, sprites, sprIndex, lsl #4
     ldmia sprite, {sprTile, uwvh}
     str sprTile, [TILE]
     and uv, uwvh, MASK
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 0)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 0)]
     bic uv, uwvh, MASK
-    str uv, [sp, #(VERTEX_T + VERTEX_SIZEOF * 1)]
+    str uv, [ptr, #(VERTEX_T + VERTEX_SIZEOF * 1)]
     b rasterize_asm
 
 .next_face:
-    cmp face, #0
+    tst face, face
     bne .loop_list
 
 .next_ot: