mrq/GLdc
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'clipping-rewrite-for-the-last-time-ffs' into 'master'

Browse Source 
Draft: Restructure clipping to be much MUCH faster in the visible case

See merge request simulant/GLdc!105
This commit is contained in:
Luke Benstead 2023-04-26 19:50:57 +00:00
commit cfc8b91e58
8 changed files with 970 additions and 315 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
CMakeLists.txt
View File

@ -17,6 +17,7 @@ string(TOUPPER ${BACKEND} BACKEND_UPPER)
add_definitions(-DBACKEND_${BACKEND_UPPER}) add_definitions(-DBACKEND_${BACKEND_UPPER})
set(CMAKE_C_STANDARD 99) set(CMAKE_C_STANDARD 99)
set(CMAKE_CXX_STANDARD 11)
include_directories(include) include_directories(include)
@ -178,6 +179,7 @@ gen_sample(scissor samples/scissor/main.c)
gen_sample(polymark samples/polymark/main.c) gen_sample(polymark samples/polymark/main.c)
gen_sample(cubes samples/cubes/main.cpp) gen_sample(cubes samples/cubes/main.cpp)
gen_sample(zclip_test tests/zclip/main.cpp)
if(PLATFORM_DREAMCAST) if(PLATFORM_DREAMCAST)
gen_sample(trimark samples/trimark/main.c) gen_sample(trimark samples/trimark/main.c)

590
GL/platforms/sh4.c
View File

@ -9,11 +9,7 @@
#define likely(x) __builtin_expect(!!(x), 1) #define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x), 0) #define unlikely(x) __builtin_expect(!!(x), 0)
#define SQ_BASE_ADDRESS (uint32_t *)(void *) \ #define SQ_BASE_ADDRESS (void*) 0xe0000000
(0xe0000000 | (((uint32_t)0x10000000) & 0x03ffffe0))
static volatile uint32_t* PVR_LMMODE0 = (uint32_t*) 0xA05F6884;
GL_FORCE_INLINE bool glIsVertex(const float flags) { GL_FORCE_INLINE bool glIsVertex(const float flags) {
@ -54,31 +50,28 @@ GL_FORCE_INLINE void _glPerspectiveDivideVertex(Vertex* vertex, const float h) {
const float f = _glFastInvert(vertex->w); const float f = _glFastInvert(vertex->w);
/* Convert to NDC and apply viewport */ /* Convert to NDC and apply viewport */
vertex->xyz[0] = __builtin_fmaf( vertex->xyz[0] = ((vertex->xyz[0] * f) * 320) + 320;
VIEWPORT.hwidth, vertex->xyz[0] * f, VIEWPORT.x_plus_hwidth vertex->xyz[1] = ((vertex->xyz[1] * f) * -240) + 240;
); vertex->xyz[2] = f;
vertex->xyz[1] = h - __builtin_fmaf(
VIEWPORT.hheight, vertex->xyz[1] * f, VIEWPORT.y_plus_hheight
);
/* Orthographic projections need to use invZ otherwise we lose /* Orthographic projections need to use invZ otherwise we lose
the depth information. As w == 1, and clip-space range is -w to +w the depth information. As w == 1, and clip-space range is -w to +w
we add 1.0 to the Z to bring it into range. We add a little extra to we add 1.0 to the Z to bring it into range. We add a little extra to
avoid a divide by zero. avoid a divide by zero.
*/ */
if(vertex->w == 1.0f) {
vertex->xyz[2] = (vertex->w == 1.0f) ? _glFastInvert(1.0001f + vertex->xyz[2]) : f; vertex->xyz[2] = _glFastInvert(1.0001f + vertex->xyz[2]);
}
} }
GL_FORCE_INLINE void _glSubmitHeaderOrVertex(uint32_t* d, const Vertex* v) { GL_FORCE_INLINE void _glSubmitHeaderOrVertex(volatile uint32_t* d, const Vertex* v) {
#ifndef NDEBUG #ifndef NDEBUG
gl_assert(!isnan(v->xyz[2])); gl_assert(!isnan(v->xyz[2]));
gl_assert(!isnan(v->w)); gl_assert(!isnan(v->w));
#endif #endif
#if CLIP_DEBUG #if CLIP_DEBUG
printf("Submitting: %x (%x)\n", v, v->flags); fprintf(stderr, "Submitting: %x (%x)\n", v, v->flags);
#endif #endif
uint32_t *s = (uint32_t*) v; uint32_t *s = (uint32_t*) v;
@ -94,336 +87,323 @@ GL_FORCE_INLINE void _glSubmitHeaderOrVertex(uint32_t* d, const Vertex* v) {
d += 8; d += 8;
} }
static struct __attribute__((aligned(32))) { static inline void _glClipEdge(const Vertex* const v1, const Vertex* const v2, Vertex* vout) {
Vertex* v; const static float o = 1.0f / 255.0f;
int visible;
} triangle[3];
static int tri_count = 0;
static int strip_count = 0;
static inline void interpolateColour(const uint32_t* a, const uint32_t* b, const float t, uint32_t* out) {
const static uint32_t MASK1 = 0x00FF00FF;
const static uint32_t MASK2 = 0xFF00FF00;
const uint32_t f2 = 256 * t;
const uint32_t f1 = 256 - f2;
*out = (((((*a & MASK1) * f1) + ((*b & MASK1) * f2)) >> 8) & MASK1) |
(((((*a & MASK2) * f1) + ((*b & MASK2) * f2)) >> 8) & MASK2);
}
static inline void _glClipEdge(const Vertex* v1, const Vertex* v2, Vertex* vout) {
/* Clipping time! */
const float d0 = v1->w + v1->xyz[2]; const float d0 = v1->w + v1->xyz[2];
const float d1 = v2->w + v2->xyz[2]; const float d1 = v2->w + v2->xyz[2];
const float sign = ((2.0f * (d1 < d0)) - 1.0f); const float t = (fabs(d0) * (1.0f / sqrtf((d1 - d0) * (d1 - d0)))) + 0.000001f;
const float epsilon = -0.00001f * sign; const float invt = 1.0f - t;
const float n = (d0 - d1);
const float r = (1.f / sqrtf(n * n)) * sign;
float t = fmaf(r, d0, epsilon);
vout->xyz[0] = fmaf(v2->xyz[0] - v1->xyz[0], t, v1->xyz[0]); vout->xyz[0] = invt * v1->xyz[0] + t * v2->xyz[0];
vout->xyz[1] = fmaf(v2->xyz[1] - v1->xyz[1], t, v1->xyz[1]); vout->xyz[1] = invt * v1->xyz[1] + t * v2->xyz[1];
vout->xyz[2] = fmaf(v2->xyz[2] - v1->xyz[2], t, v1->xyz[2]); vout->xyz[2] = invt * v1->xyz[2] + t * v2->xyz[2];
vout->w = fmaf(v2->w - v1->w, t, v1->w);
vout->uv[0] = fmaf(v2->uv[0] - v1->uv[0], t, v1->uv[0]); vout->uv[0] = invt * v1->uv[0] + t * v2->uv[0];
vout->uv[1] = fmaf(v2->uv[1] - v1->uv[1], t, v1->uv[1]); vout->uv[1] = invt * v1->uv[1] + t * v2->uv[1];
interpolateColour((uint32_t*) v1->bgra, (uint32_t*) v2->bgra, t, (uint32_t*) vout->bgra); vout->w = invt * v1->w + t * v2->w;
}
GL_FORCE_INLINE void ClearTriangle() { const float m = 255 * t;
tri_count = 0; const float n = 255 - m;
}
static inline void ShiftTriangle() { vout->bgra[0] = (v1->bgra[0] * n + v2->bgra[0] * m) * o;
if(!tri_count) { vout->bgra[1] = (v1->bgra[1] * n + v2->bgra[1] * m) * o;
return; vout->bgra[2] = (v1->bgra[2] * n + v2->bgra[2] * m) * o;
} vout->bgra[3] = (v1->bgra[3] * n + v2->bgra[3] * m) * o;
tri_count--;
triangle[0] = triangle[1];
triangle[1] = triangle[2];
#ifndef NDEBUG
triangle[2].v = NULL;
triangle[2].visible = false;
#endif
}
static inline void ShiftRotateTriangle() {
if(!tri_count) {
return;
}
if(triangle[0].v < triangle[1].v) {
triangle[0] = triangle[2];
} else {
triangle[1] = triangle[2];
}
tri_count--;
} }
#define SPAN_SORT_CFG 0x005F8030 #define SPAN_SORT_CFG 0x005F8030
static volatile uint32_t* PVR_LMMODE0 = (uint32_t*) 0xA05F6884;
static volatile uint32_t *PVR_LMMODE1 = (uint32_t*) 0xA05F6888;
static volatile uint32_t *QACR = (uint32_t*) 0xFF000038;
void SceneListSubmit(void* src, int n) { void SceneListSubmit(void* src, int n) {
/* You need at least a header, and 3 vertices to render anything */
if(n < 4) {
return;
}
const float h = GetVideoMode()->height; const float h = GetVideoMode()->height;
PVR_SET(SPAN_SORT_CFG, 0x0); PVR_SET(SPAN_SORT_CFG, 0x0);
//Set PVR DMA registers //Set PVR DMA registers
volatile int *pvrdmacfg = (int*)0xA05F6888; *PVR_LMMODE0 = 0;
pvrdmacfg[0] = 1; *PVR_LMMODE1 = 0;
pvrdmacfg[1] = 0;
//Set QACR registers //Set QACR registers
volatile int *qacr = (int*)0xFF000038; QACR[1] = QACR[0] = 0x11;
qacr[1] = qacr[0] = 0x11;
uint32_t *d = SQ_BASE_ADDRESS; volatile uint32_t *sq = SQ_BASE_ADDRESS;
Vertex __attribute__((aligned(32))) tmp; uint32_t clipping_disabled_mask = (_glNearZClippingEnabled()) ? 0 : 0x7;
/* Perform perspective divide on each vertex */ #if CLIP_DEBUG
Vertex* vertex = (Vertex*) src; for(int i = 0; i < n; ++i) {
fprintf(stderr, "{%f, %f, %f, %f}, // %x (%x)\n", vertex[i].xyz[0], vertex[i].xyz[1], vertex[i].xyz[2], vertex[i].w, vertex[i].flags, &vertex[i]);
}
if(!_glNearZClippingEnabled()) { fprintf(stderr, "----\n");
/* Prep store queues */ #endif
uint8_t counter = 0;
Vertex* v2 = (Vertex*) src;
while(n--) { while(n--) {
if(glIsVertex(vertex->flags)) { __builtin_prefetch(v2 + 1);
_glPerspectiveDivideVertex(vertex, h);
}
_glSubmitHeaderOrVertex(d, vertex); switch(v2->flags) {
++vertex; case GPU_CMD_VERTEX_EOL:
} case GPU_CMD_VERTEX:
if(++counter < 3) {
return; v2++;
}
tri_count = 0;
strip_count = 0;
#if CLIP_DEBUG
printf("----\n");
#endif
for(int i = 0; i < n; ++i, ++vertex) {
PREFETCH(vertex + 1);
PREFETCH(vertex + 2);
/* Wait until we fill the triangle */
if(tri_count < 3) {
if(glIsVertex(vertex->flags)) {
++strip_count;
triangle[tri_count].v = vertex;
triangle[tri_count].visible = vertex->xyz[2] >= -vertex->w;
if(++tri_count < 3) {
continue; continue;
} }
} else { break;
/* We hit a header */ default:
tri_count = 0; _glSubmitHeaderOrVertex(sq, v2++);
strip_count = 0; counter = 0;
_glSubmitHeaderOrVertex(d, vertex);
continue; continue;
} }
}
#if CLIP_DEBUG Vertex* const v0 = v2 - 2;
printf("SC: %d\n", strip_count); Vertex* const v1 = v2 - 1;
#endif
/* If we got here, then triangle contains 3 vertices */ const uint8_t visible_mask = (
int visible_mask = triangle[0].visible | (triangle[1].visible << 1) | (triangle[2].visible << 2); (v0->xyz[2] > -v0->w) << 0 |
(v1->xyz[2] > -v1->w) << 1 |
/* Clipping time! (v2->xyz[2] > -v2->w) << 2 |
((v2->flags == GPU_CMD_VERTEX_EOL) << 3) |
There are 6 distinct possibilities when clipping a triangle. 3 of them result clipping_disabled_mask // This forces everything to be marked visible if clipping is disabled
in another triangle, 3 of them result in a quadrilateral. );
Assuming you iterate the edges of the triangle in order, and create a new *visible*
vertex when you cross the plane, and discard vertices behind the plane, then the only
difference between the two cases is that the final two vertices that need submitting have
to be reversed.
Unfortunately we have to copy vertices here, because if we persp-divide a vertex it may
be used in a subsequent triangle in the strip and would end up being double divided.
*/
#define SUBMIT_QUEUED() \
if(strip_count > 3) { \
tmp = *(vertex - 2); \
/* If we had triangles ahead of this one, submit and finalize */ \
_glPerspectiveDivideVertex(&tmp, h); \
_glSubmitHeaderOrVertex(d, &tmp); \
tmp = *(vertex - 1); \
tmp.flags = GPU_CMD_VERTEX_EOL; \
_glPerspectiveDivideVertex(&tmp, h); \
_glSubmitHeaderOrVertex(d, &tmp); \
}
bool is_last_in_strip = glIsLastVertex(vertex->flags);
switch(visible_mask) { switch(visible_mask) {
case 1: {
SUBMIT_QUEUED();
/* 0, 0a, 2a */
tmp = *triangle[0].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[0].v, triangle[1].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[2].v, triangle[0].v, &tmp);
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 2: {
SUBMIT_QUEUED();
/* 0a, 1, 1a */
_glClipEdge(triangle[0].v, triangle[1].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[1].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[1].v, triangle[2].v, &tmp);
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 3: {
SUBMIT_QUEUED();
/* 0, 1, 2a, 1a */
tmp = *triangle[0].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[1].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[2].v, triangle[0].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[1].v, triangle[2].v, &tmp);
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 4: {
SUBMIT_QUEUED();
/* 1a, 2, 2a */
_glClipEdge(triangle[1].v, triangle[2].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[2].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[2].v, triangle[0].v, &tmp);
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 5: {
SUBMIT_QUEUED();
/* 0, 0a, 2, 1a */
tmp = *triangle[0].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[0].v, triangle[1].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[2].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[1].v, triangle[2].v, &tmp);
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 6: {
SUBMIT_QUEUED();
/* 0a, 1, 2a, 2 */
_glClipEdge(triangle[0].v, triangle[1].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[1].v;
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
_glClipEdge(triangle[2].v, triangle[0].v, &tmp);
tmp.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
tmp = *triangle[2].v;
tmp.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(&tmp, h);
_glSubmitHeaderOrVertex(d, &tmp);
} break;
case 7: {
/* All the vertices are visible! We divide and submit v0, then shift */
_glPerspectiveDivideVertex(vertex - 2, h);
_glSubmitHeaderOrVertex(d, vertex - 2);
if(is_last_in_strip) {
_glPerspectiveDivideVertex(vertex - 1, h);
_glSubmitHeaderOrVertex(d, vertex - 1);
_glPerspectiveDivideVertex(vertex, h);
_glSubmitHeaderOrVertex(d, vertex);
tri_count = 0;
strip_count = 0;
}
ShiftRotateTriangle();
continue;
} break;
case 0: case 0:
break;
case 15: /* All visible, but final vertex in strip */
{
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
Vertex __attribute__((aligned(32))) a = *v1;
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
a = *v2;
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
}
break;
case 7:
/* All visible, push the first vertex and move on */
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
break;
case 9:
/* First vertex was visible, last in strip */
{
Vertex __attribute__((aligned(32))) a, b;
_glClipEdge(v0, v1, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v2, v0, &b);
b.flags = GPU_CMD_VERTEX_EOL;
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
}
break;
case 1:
/* First vertex was visible, but not last in strip */
{
Vertex __attribute__((aligned(32))) a, b;
_glClipEdge(v0, v1, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v2, v0, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
_glSubmitHeaderOrVertex(sq, &b);
}
break;
case 10:
case 2:
/* Second vertex was visible. In self case we need to create a triangle and produce
two new vertices: 1-2, and 2-3. */
{
Vertex __attribute__((aligned(32))) a;
Vertex __attribute__((aligned(32))) c = *v1;
_glClipEdge(v0, &c, &a);
a.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glClipEdge(&c, v2, &a);
a.flags = v2->flags;
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
}
break;
case 11:
case 3: /* First and second vertex were visible */
{
Vertex __attribute__((aligned(32))) a, b;
Vertex __attribute__((aligned(32))) c = *v1;
_glClipEdge(v2, v0, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
_glClipEdge(v1, v2, &a);
a.flags = v2->flags;
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &c);
_glSubmitHeaderOrVertex(sq, &a);
}
break;
case 12:
case 4:
/* Third vertex was visible. */
{
Vertex __attribute__((aligned(32))) a, b;
Vertex __attribute__((aligned(32))) c = *v2;
_glClipEdge(v2, v0, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v1, v2, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &a);
_glSubmitHeaderOrVertex(sq, &b);
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
}
break;
case 13:
{
Vertex __attribute__((aligned(32))) a, b;
Vertex __attribute__((aligned(32))) c = *v2;
c.flags = GPU_CMD_VERTEX;
_glClipEdge(v0, v1, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v1, v2, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
c.flags = GPU_CMD_VERTEX_EOL;
_glSubmitHeaderOrVertex(sq, &c);
}
break;
case 5: /* First and third vertex were visible */
{
Vertex __attribute__((aligned(32))) a, b;
Vertex __attribute__((aligned(32))) c = *v2;
c.flags = GPU_CMD_VERTEX;
_glClipEdge(v0, v1, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v1, v2, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glSubmitHeaderOrVertex(sq, v0);
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
_glSubmitHeaderOrVertex(sq, &c);
}
break;
case 14:
case 6: /* Second and third vertex were visible */
{
Vertex __attribute__((aligned(32))) a, b;
Vertex __attribute__((aligned(32))) c = *v1;
_glClipEdge(v0, v1, &a);
a.flags = GPU_CMD_VERTEX;
_glClipEdge(v2, v0, &b);
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&a, h);
_glSubmitHeaderOrVertex(sq, &a);
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(sq, &b);
_glSubmitHeaderOrVertex(sq, &c);
c = *v2;
_glPerspectiveDivideVertex(&c, h);
_glSubmitHeaderOrVertex(sq, &c);
}
break;
default: default:
break; break;
} }
/* If this was the last in the strip, we don't need to if(v2->flags == GPU_CMD_VERTEX_EOL) {
submit anything else, we just wipe the tri_count */ counter = 0;
if(is_last_in_strip) {
tri_count = 0;
strip_count = 0;
} else { } else {
ShiftRotateTriangle(); --counter;
strip_count = 2;
} }
v2++;
} }
} }

36
GL/private.h
View File

@ -233,11 +233,41 @@ GL_FORCE_INLINE float clamp(float d, float min, float max) {
return (d < min) ? min : (d > max) ? max : d; return (d < min) ? min : (d > max) ? max : d;
} }
GL_FORCE_INLINE void memcpy_vertex(Vertex *dest, const Vertex *src) {
#ifdef __DREAMCAST__
_Complex float double_scratch;
asm volatile (
"fschg\n\t"
"clrs\n\t"
".align 2\n\t"
"fmov.d @%[in]+, %[scratch]\n\t"
"fmov.d %[scratch], @%[out]\n\t"
"fmov.d @%[in]+, %[scratch]\n\t"
"add #8, %[out]\n\t"
"fmov.d %[scratch], @%[out]\n\t"
"fmov.d @%[in]+, %[scratch]\n\t"
"add #8, %[out]\n\t"
"fmov.d %[scratch], @%[out]\n\t"
"fmov.d @%[in], %[scratch]\n\t"
"add #8, %[out]\n\t"
"fmov.d %[scratch], @%[out]\n\t"
"fschg\n"
: [in] "+&r" ((uint32_t) src), [scratch] "=&d" (double_scratch), [out] "+&r" ((uint32_t) dest)
:
: "t", "memory" // clobbers
);
#else
*dest = *src;
#endif
}
#define swapVertex(a, b) \ #define swapVertex(a, b) \
do { \ do { \
Vertex c = *a; \ Vertex __attribute__((aligned(32))) c; \
*a = *b; \ memcpy_vertex(&c, a); \
*b = c; \ memcpy_vertex(a, b); \
memcpy_vertex(b, &c); \
} while(0) } while(0)
/* ClipVertex doesn't have room for these, so we need to parse them /* ClipVertex doesn't have room for these, so we need to parse them

1
containers/aligned_vector.h
View File

@ -12,6 +12,7 @@ extern "C" {
#if defined(__APPLE__) || defined(__WIN32__) #if defined(__APPLE__) || defined(__WIN32__)
/* Linux + Kos define this, OSX does not, so just use malloc there */ /* Linux + Kos define this, OSX does not, so just use malloc there */
static inline void* memalign(size_t alignment, size_t size) { static inline void* memalign(size_t alignment, size_t size) {
(void) alignment;
return malloc(size); return malloc(size);
} }
#else #else

8
samples/loadbmp.c
View File

@ -23,7 +23,11 @@ int ImageLoad(char *filename, Image *image) {
} }
// seek through the bmp header, up to the width/height: // seek through the bmp header, up to the width/height:
fseek(file, 18, SEEK_CUR); fseek(file, 10, SEEK_CUR);
uint32_t offset;
fread(&offset, 4, 1, file);
fseek(file, 4, SEEK_CUR);
// read the width // read the width
if ((i = fread(&sizeX, 4, 1, file)) != 1) { if ((i = fread(&sizeX, 4, 1, file)) != 1) {
@ -65,7 +69,7 @@ int ImageLoad(char *filename, Image *image) {
} }
// seek past the rest of the bitmap header. // seek past the rest of the bitmap header.
fseek(file, 24, SEEK_CUR); fseek(file, offset, SEEK_SET);
// read the data. // read the data.
image->data = (char *) malloc(size); image->data = (char *) malloc(size);

3
samples/zclip_triangle/main.c
View File

@ -86,12 +86,13 @@ void DrawGLScene()
rotation = (rotation > 360.0f) ? rotation - 360.0f : rotation; rotation = (rotation > 360.0f) ? rotation - 360.0f : rotation;
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT); // Clear The Screen And The Depth Buffer glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT); // Clear The Screen And The Depth Buffer
glClearColor(0.5f, 0.5f, 0.5f, 0.5f);
glLoadIdentity(); // Reset The View glLoadIdentity(); // Reset The View
glDisable(GL_CULL_FACE); glDisable(GL_CULL_FACE);
glPushMatrix(); glPushMatrix();
glTranslatef(0.0f, -1.0f, movement); glTranslatef(0.0f, -1.0f, -movement);
glRotatef(rotation, 0.0f, 1.0f, 0.0f); glRotatef(rotation, 0.0f, 1.0f, 0.0f);
glBegin(GL_TRIANGLES); glBegin(GL_TRIANGLES);

637
tests/zclip/main.cpp Normal file
View File

@ -0,0 +1,637 @@
#include <cstdint>
#include <vector>
#include <cstdio>
#include <cmath>
#include <stdexcept>
#include <cassert>
#define SQ_BASE_ADDRESS 0
#define SPAN_SORT_CFG 0
#define PVR_SET(x, y) (void)(x); (void)(y)
struct Vertex {
uint32_t flags;
float xyz[3];
float uv[2];
float w;
uint8_t bgra[4];
};
struct {
float hwidth;
float x_plus_hwidth;
float hheight;
float y_plus_hheight;
} VIEWPORT = {320, 320, 240, 240};
struct VideoMode {
float height;
};
static VideoMode* GetVideoMode() {
static VideoMode mode = {320.0f};
return &mode;
}
enum GPUCommand {
GPU_CMD_POLYHDR = 0x80840000,
GPU_CMD_VERTEX = 0xe0000000,
GPU_CMD_VERTEX_EOL = 0xf0000000,
GPU_CMD_USERCLIP = 0x20000000,
GPU_CMD_MODIFIER = 0x80000000,
GPU_CMD_SPRITE = 0xA0000000
};
static std::vector<Vertex> sent;
static inline void interpolateColour(const uint32_t* a, const uint32_t* b, const float t, uint32_t* out) {
const static uint32_t MASK1 = 0x00FF00FF;
const static uint32_t MASK2 = 0xFF00FF00;
const uint32_t f2 = 256 * t;
const uint32_t f1 = 256 - f2;
*out = (((((*a & MASK1) * f1) + ((*b & MASK1) * f2)) >> 8) & MASK1) |
(((((*a & MASK2) * f1) + ((*b & MASK2) * f2)) >> 8) & MASK2);
}
static inline void _glClipEdge(const Vertex* v1, const Vertex* v2, Vertex* vout) {
/* Clipping time! */
const float d0 = v1->w + v1->xyz[2];
const float d1 = v2->w + v2->xyz[2];
const float sign = ((2.0f * (d1 < d0)) - 1.0f);
const float epsilon = -0.00001f * sign;
const float n = (d0 - d1);
const float r = (1.f / sqrtf(n * n)) * sign;
float t = fmaf(r, d0, epsilon);
vout->xyz[0] = fmaf(v2->xyz[0] - v1->xyz[0], t, v1->xyz[0]);
vout->xyz[1] = fmaf(v2->xyz[1] - v1->xyz[1], t, v1->xyz[1]);
vout->xyz[2] = fmaf(v2->xyz[2] - v1->xyz[2], t, v1->xyz[2]);
vout->w = fmaf(v2->w - v1->w, t, v1->w);
vout->uv[0] = fmaf(v2->uv[0] - v1->uv[0], t, v1->uv[0]);
vout->uv[1] = fmaf(v2->uv[1] - v1->uv[1], t, v1->uv[1]);
interpolateColour((uint32_t*) v1->bgra, (uint32_t*) v2->bgra, t, (uint32_t*) vout->bgra);
}
bool glIsVertex(const uint32_t flags) {
return flags == GPU_CMD_VERTEX_EOL || flags == GPU_CMD_VERTEX;
}
bool glIsLastVertex(const uint32_t flags) {
return flags == GPU_CMD_VERTEX_EOL;
}
void _glSubmitHeaderOrVertex(volatile uint32_t*, Vertex* vtx) {
sent.push_back(*vtx);
}
float _glFastInvert(float x) {
return (1.f / __builtin_sqrtf(x * x));
}
void _glPerspectiveDivideVertex(Vertex* vertex, const float h) {
const float f = _glFastInvert(vertex->w);
/* Convert to NDC and apply viewport */
vertex->xyz[0] = __builtin_fmaf(
VIEWPORT.hwidth, vertex->xyz[0] * f, VIEWPORT.x_plus_hwidth
);
vertex->xyz[1] = h - __builtin_fmaf(
VIEWPORT.hheight, vertex->xyz[1] * f, VIEWPORT.y_plus_hheight
);
/* Orthographic projections need to use invZ otherwise we lose
the depth information. As w == 1, and clip-space range is -w to +w
we add 1.0 to the Z to bring it into range. We add a little extra to
avoid a divide by zero.
*/
vertex->xyz[2] = (vertex->w == 1.0f) ? _glFastInvert(1.0001f + vertex->xyz[2]) : f;
}
void memcpy_vertex(Vertex* dst, Vertex* src) {
*dst = *src;
}
/* Zclipping is so difficult to get right, that self sample tests all the cases of clipping and makes sure that things work as expected */
#ifdef __DREAMCAST__
static volatile int *pvrdmacfg = (int*)0xA05F6888;
static volatile int *qacr = (int*)0xFF000038;
#else
static int pvrdmacfg[2];
static int qacr[2];
#endif
void SceneListSubmit(void* src, int n) {
/* You need at least a header, and 3 vertices to render anything */
if(n < 4) {
return;
}
const float h = GetVideoMode()->height;
PVR_SET(SPAN_SORT_CFG, 0x0);
//Set PVR DMA registers
pvrdmacfg[0] = 1;
pvrdmacfg[1] = 1;
//Set QACR registers
qacr[1] = qacr[0] = 0x11;
volatile uint32_t *d = SQ_BASE_ADDRESS;
int8_t queue_head = 0;
int8_t queue_tail = 0;
/* The most vertices ever in the queue is 5 (as some clipping operations
* produce and additional couple of vertice, but we add one more so the ring buffer doesn't
* trip over itself (e.g. if tail == head we can guarantee it's empty, not full) */
Vertex __attribute__((aligned(32))) queue[4];
const int queue_capacity = sizeof(queue) / sizeof(Vertex);
Vertex* vertex = (Vertex*) src;
uint32_t visible_mask = 0;
#if CLIP_DEBUG
for(int i = 0; i < n; ++i) {
fprintf(stderr, "{%f, %f, %f, %f}, // %x (%x)\n", vertex[i].xyz[0], vertex[i].xyz[1], vertex[i].xyz[2], vertex[i].w, vertex[i].flags, &vertex[i]);
}
fprintf(stderr, "----\n");
#endif
while(n--) {
bool last_vertex = false;
memcpy_vertex(queue + queue_tail, vertex);
++vertex;
switch(queue[queue_tail].flags) {
case GPU_CMD_POLYHDR:
_glSubmitHeaderOrVertex(d, &queue[queue_tail]);
break;
case GPU_CMD_VERTEX_EOL:
last_vertex = true;
case GPU_CMD_VERTEX:
visible_mask = (visible_mask >> 1) | (queue[queue_tail].xyz[2] >= -queue[queue_tail].w) << 2;
assert(visible_mask < 15);
queue_tail = (queue_tail + 1) % queue_capacity;
default:
break;
}
int counter = (queue_tail - queue_head + queue_capacity) % queue_capacity;
if(counter < 3) {
continue;
}
#if CLIP_DEBUG
fprintf(stderr, "%d\n", visible_mask);
#endif
Vertex __attribute__((aligned(32))) a, b; // Scratch vertices
switch(visible_mask) {
case 0:
break;
case 7:
/* All visible, push the first vertex and move on */
_glPerspectiveDivideVertex(&queue[queue_head], h);
_glSubmitHeaderOrVertex(d, &queue[queue_head]);
if(last_vertex) {
/* If this was the last vertex in the strip, we need to flush the queue and then
restart it again */
int v1 = (queue_head + 1) % queue_capacity;
int v2 = (queue_head + 2) % queue_capacity;
_glPerspectiveDivideVertex(&queue[v1], h);
_glSubmitHeaderOrVertex(d, &queue[v1]);
_glPerspectiveDivideVertex(&queue[v2], h);
_glSubmitHeaderOrVertex(d, &queue[v2]);
}
break;
case 1:
/* First vertex was visible */
{
Vertex* v0 = &queue[queue_head];
Vertex* v1 = &queue[(queue_head + 1) % queue_capacity];
Vertex* v2 = &queue[(queue_head + 2) % queue_capacity];
_glClipEdge(v0, v1, &a);
_glClipEdge(v2, v0, &b);
a.flags = GPU_CMD_VERTEX;
/* If v2 was the last in the strip, then b should be. If it wasn't
we'll create a degenerate triangle by adding b twice in a row so that the
strip processing will continue correctly after crossing the plane so it can
cross back*/
b.flags = v2->flags;
_glPerspectiveDivideVertex(v0, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, v0);
_glSubmitHeaderOrVertex(d, &a);
_glSubmitHeaderOrVertex(d, &b);
_glSubmitHeaderOrVertex(d, &b);
}
break;
case 2:
/* Second vertex was visible. In self case we need to create a triangle and produce
two new vertices: 1-2, and 2-3. */
{
Vertex* v0 = &queue[queue_head];
const Vertex* v1 = &queue[(queue_head + 1) % queue_capacity];
const Vertex* v2 = &queue[(queue_head + 2) % queue_capacity];
_glClipEdge(v0, v1, &a);
_glClipEdge(v1, v2, &b);
a.flags = GPU_CMD_VERTEX;
b.flags = v2->flags;
_glPerspectiveDivideVertex(v0, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, &a);
_glSubmitHeaderOrVertex(d, v0);
_glSubmitHeaderOrVertex(d, &b);
}
break;
case 3: /* First and second vertex were visible */
{
Vertex* v0 = &queue[queue_head];
Vertex __attribute__((aligned(32))) v1 = queue[(queue_head + 1) % queue_capacity];
Vertex* v2 = &queue[(queue_head + 2) % queue_capacity];
_glClipEdge(&v1, v2, &a);
_glClipEdge(v2, v0, &b);
a.flags = v2->flags;
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glPerspectiveDivideVertex(&v1, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, v0);
_glSubmitHeaderOrVertex(d, &v1);
_glSubmitHeaderOrVertex(d, &b);
_glSubmitHeaderOrVertex(d, &v1);
_glSubmitHeaderOrVertex(d, &a);
}
break;
case 4:
/* Third vertex was visible. */
{
Vertex* v0 = &queue[queue_head];
Vertex* v1 = &queue[(queue_head + 1) % queue_capacity];
Vertex __attribute__((aligned(32))) v2 = queue[(queue_head + 2) % queue_capacity];
_glClipEdge(&v2, v0, &a);
_glClipEdge(v1, &v2, &b);
a.flags = GPU_CMD_VERTEX;
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&v2, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, &a);
_glSubmitHeaderOrVertex(d, &a);
_glSubmitHeaderOrVertex(d, &b);
_glSubmitHeaderOrVertex(d, &v2);
}
break;
case 5: /* First and third vertex were visible */
{
Vertex* v0 = &queue[queue_head];
Vertex* v1 = &queue[(queue_head + 1) % queue_capacity];
Vertex __attribute__((aligned(32))) v2 = queue[(queue_head + 2) % queue_capacity];
_glClipEdge(v0, v1, &a);
_glClipEdge(v1, &v2, &b);
a.flags = GPU_CMD_VERTEX;
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(v0, h);
_glPerspectiveDivideVertex(&v2, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, v0);
_glSubmitHeaderOrVertex(d, &a);
uint32_t v2_flags = v2.flags;
v2.flags = GPU_CMD_VERTEX;
_glSubmitHeaderOrVertex(d, &v2);
v2.flags = v2_flags;
_glSubmitHeaderOrVertex(d, &b);
_glSubmitHeaderOrVertex(d, &v2);
}
break;
case 6: /* Second and third vertex were visible */
{
Vertex* v0 = &queue[queue_head];
Vertex __attribute__((aligned(32))) v1 = queue[(queue_head + 1) % queue_capacity];
Vertex __attribute__((aligned(32))) v2 = queue[(queue_head + 2) % queue_capacity];
_glClipEdge(v0, &v1, &a);
_glClipEdge(&v2, v0, &b);
a.flags = GPU_CMD_VERTEX;
b.flags = GPU_CMD_VERTEX;
_glPerspectiveDivideVertex(&v1, h);
_glPerspectiveDivideVertex(&v2, h);
_glPerspectiveDivideVertex(&a, h);
_glPerspectiveDivideVertex(&b, h);
_glSubmitHeaderOrVertex(d, &a);
_glSubmitHeaderOrVertex(d, &v1);
_glSubmitHeaderOrVertex(d, &b);
_glSubmitHeaderOrVertex(d, &v1);
_glSubmitHeaderOrVertex(d, &v2);
}
break;
default:
break;
}
if(last_vertex) {
visible_mask = queue_head = queue_tail = 0;
} else {
queue_head = (queue_head + 1) % queue_capacity;
}
}
}
struct VertexTmpl {
VertexTmpl(float x, float y, float z, float w):
x(x), y(y), z(z), w(w) {}
float x, y, z, w;
};
std::vector<Vertex> make_vertices(const std::vector<VertexTmpl>& verts) {
std::vector<Vertex> result;
Vertex r;
r.flags = GPU_CMD_POLYHDR;
result.push_back(r);
for(auto& v: verts) {
r.flags = GPU_CMD_VERTEX;
r.xyz[0] = v.x;
r.xyz[1] = v.y;
r.xyz[2] = v.z;
r.uv[0] = 0.0f;
r.uv[1] = 0.0f;
r.w = v.w;
result.push_back(r);
}
result.back().flags = GPU_CMD_VERTEX_EOL;
return result;
}
template<typename T, typename U>
void check_equal(const T& lhs, const U& rhs) {
if(lhs != rhs) {
throw std::runtime_error("Assertion failed");
}
}
template<>
void check_equal(const Vertex& lhs, const Vertex& rhs) {
if(lhs.xyz[0] != rhs.xyz[0] ||
lhs.xyz[1] != rhs.xyz[1] ||
lhs.xyz[2] != rhs.xyz[2] ||
lhs.w != rhs.w) {
throw std::runtime_error("Assertion failed");
}
}
bool test_clip_case_001() {
/* The first vertex is visible only */
sent.clear();
auto data = make_vertices({
{0.000000, -2.414213, 3.080808, 5.000000},
{-4.526650, -2.414213, -7.121212, -5.000000},
{4.526650, -2.414213, -7.121212, -5.000000}
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 5);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
// Because we're sending a single triangle, we end up sending a
// degenerate final vert. But if we were sending more than one triangle
// this would be GPU_CMD_VERTEX twice
check_equal(sent[3].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[4].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[3], sent[4]);
return true;
}
bool test_clip_case_010() {
/* The third vertex is visible only */
sent.clear();
auto data = make_vertices({
{-4.526650, -2.414213, -7.121212, -5.000000},
{0.000000, -2.414213, 3.080808, 5.000000},
{4.526650, -2.414213, -7.121212, -5.000000}
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 4);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
check_equal(sent[3].flags, GPU_CMD_VERTEX_EOL);
return true;
}
bool test_clip_case_100() {
/* The third vertex is visible only */
sent.clear();
auto data = make_vertices({
{-4.526650, -2.414213, -7.121212, -5.000000},
{4.526650, -2.414213, -7.121212, -5.000000},
{0.000000, -2.414213, 3.080808, 5.000000}
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 5);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
// Because we're sending a single triangle, we end up sending a
// degenerate final vert. But if we were sending more than one triangle
// this would be GPU_CMD_VERTEX twice
check_equal(sent[3].flags, GPU_CMD_VERTEX);
check_equal(sent[4].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[1], sent[2]);
return true;
}
bool test_clip_case_110() {
/* 2nd and 3rd visible */
sent.clear();
auto data = make_vertices({
{0.0, -2.414213, -7.121212, -5.000000},
{-4.526650, -2.414213, 3.080808, 5.000000},
{4.526650, -2.414213, 3.080808, 5.000000}
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 6);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
check_equal(sent[3].flags, GPU_CMD_VERTEX);
check_equal(sent[4].flags, GPU_CMD_VERTEX);
check_equal(sent[5].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[2], sent[4]);
return true;
}
bool test_clip_case_011() {
/* 1st and 2nd visible */
sent.clear();
auto data = make_vertices({
{-4.526650, -2.414213, 3.080808, 5.000000},
{4.526650, -2.414213, 3.080808, 5.000000},
{0.0, -2.414213, -7.121212, -5.000000}
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 6);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
check_equal(sent[3].flags, GPU_CMD_VERTEX);
check_equal(sent[4].flags, GPU_CMD_VERTEX);
check_equal(sent[5].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[2], sent[4]);
return true;
}
bool test_clip_case_101() {
/* 1st and 3rd visible */
sent.clear();
auto data = make_vertices({
{-4.526650, -2.414213, 3.080808, 5.000000},
{0.0, -2.414213, -7.121212, -5.000000},
{4.526650, -2.414213, 3.080808, 5.000000},
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 6);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
check_equal(sent[3].flags, GPU_CMD_VERTEX);
check_equal(sent[4].flags, GPU_CMD_VERTEX);
check_equal(sent[5].flags, GPU_CMD_VERTEX_EOL);
check_equal(sent[3], sent[5]);
return true;
}
bool test_clip_case_111() {
/* 1st and 3rd visible */
sent.clear();
auto data = make_vertices({
{-4.526650, -2.414213, 3.080808, 5.000000},
{0.0, -2.414213, -7.121212, 8.000000},
{4.526650, -2.414213, 3.080808, 5.000000},
});
SceneListSubmit(&data[0], data.size());
check_equal(sent.size(), 4);
check_equal(sent[0].flags, GPU_CMD_POLYHDR);
check_equal(sent[1].flags, GPU_CMD_VERTEX);
check_equal(sent[2].flags, GPU_CMD_VERTEX);
check_equal(sent[3].flags, GPU_CMD_VERTEX_EOL);
return true;
}
bool test_start_behind() {
/* Triangle behind the plane, but the strip continues in front */
sent.clear();
auto data = make_vertices({
{-3.021717, -2.414213, -10.155344, -9.935254},
{5.915236, -2.414213, -9.354721, -9.136231},
{-5.915236, -2.414213, -0.264096, -0.063767},
{3.021717, -2.414213, 0.536527, 0.735255},
{-7.361995, -2.414213, 4.681529, 4.871976},
{1.574958, -2.414213, 5.482152, 5.670999},
});
SceneListSubmit(&data[0], data.size());
return true;
}
bool test_longer_strip() {
sent.clear();
auto data = make_vertices({
{-4.384623, -2.414213, -5.699644, -5.488456},
{4.667572, -2.414213, -5.621354, -5.410322},
{-4.667572, -2.414213, 4.319152, 4.510323},
{4.384623, -2.414213, 4.397442, 4.588456},
{-4.809045, -2.414213, 9.328549, 9.509711},
{4.243149, -2.414213, 9.406840, 9.587846},
});
SceneListSubmit(&data[0], data.size());
return true;
}
int main(int argc, char* argv[]) {
// test_clip_case_000();
test_clip_case_001();
test_clip_case_010();
test_clip_case_100();
test_clip_case_110();
test_clip_case_011();
test_clip_case_101();
test_clip_case_111();
test_start_behind();
test_longer_strip();
return 0;
}