MrMonopoly/raylib
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix triangle cracking

Browse Source
This commit is contained in:
Bigfoot71 2025-10-25 04:32:21 +02:00
parent 39242dba23
commit 6fe42e32e5
1 changed files with 143 additions and 99 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

234
src/external/rlsw.h vendored
View File

@ -821,8 +821,8 @@ typedef struct {
float clearColor[4]; // Color used to clear the screen float clearColor[4]; // Color used to clear the screen
float clearDepth; // Depth value used to clear the screen float clearDepth; // Depth value used to clear the screen
int vpCenter[2]; // Viewport center float vpCenter[2]; // Viewport center
int vpHalfSize[2]; // Viewport half dimensions float vpHalf[2]; // Viewport half dimensions
int vpSize[2]; // Viewport dimensions (minus one) int vpSize[2]; // Viewport dimensions (minus one)
int vpMin[2]; // Viewport minimum renderable point (top-left) int vpMin[2]; // Viewport minimum renderable point (top-left)
int vpMax[2]; // Viewport maximum renderable point (bottom-right) int vpMax[2]; // Viewport maximum renderable point (bottom-right)
@ -1053,6 +1053,34 @@ static inline void sw_add_vertex_grad_PTCH(
out->homogeneous[3] += gradients->homogeneous[3]; out->homogeneous[3] += gradients->homogeneous[3];
} }
static inline void sw_add_vertex_grad_scaled_PTCH(
sw_vertex_t *SW_RESTRICT out,
const sw_vertex_t *SW_RESTRICT gradients,
float scale)
{
// Add gradients to Position
out->position[0] += gradients->position[0]*scale;
out->position[1] += gradients->position[1]*scale;
out->position[2] += gradients->position[2]*scale;
out->position[3] += gradients->position[3]*scale;
// Add gradients to Texture coordinates
out->texcoord[0] += gradients->texcoord[0]*scale;
out->texcoord[1] += gradients->texcoord[1]*scale;
// Add gradients to Color
out->color[0] += gradients->color[0]*scale;
out->color[1] += gradients->color[1]*scale;
out->color[2] += gradients->color[2]*scale;
out->color[3] += gradients->color[3]*scale;
// Add gradients to Homogeneous coordinates
out->homogeneous[0] += gradients->homogeneous[0]*scale;
out->homogeneous[1] += gradients->homogeneous[1]*scale;
out->homogeneous[2] += gradients->homogeneous[2]*scale;
out->homogeneous[3] += gradients->homogeneous[3]*scale;
}
static inline void sw_float_to_unorm8_simd(uint8_t dst[4], const float src[4]) static inline void sw_float_to_unorm8_simd(uint8_t dst[4], const float src[4])
{ {
#if defined(SW_HAS_NEON) #if defined(SW_HAS_NEON)
@ -1994,18 +2022,18 @@ static inline void sw_texture_sample_linear(float *color, const sw_texture_t *te
} }
} }
static inline void sw_texture_sample(float *color, const sw_texture_t *tex, float u, float v, float duDx, float duDy, float dvDx, float dvDy) static inline void sw_texture_sample(float *color, const sw_texture_t *tex, float u, float v, float dUdx, float dUdy, float dVdx, float dVdy)
{ {
// Previous method: There is no need to compute the square root // Previous method: There is no need to compute the square root
// because using the squared value, the comparison remains `L2 > 1.0f*1.0f` // because using the squared value, the comparison remains `L2 > 1.0f*1.0f`
//float du = sqrtf(duDx*duDx + duDy*duDy); //float du = sqrtf(dUdx*dUdx + dUdy*dUdy);
//float dv = sqrtf(dvDx*dvDx + dvDy*dvDy); //float dv = sqrtf(dVdx*dVdx + dVdy*dVdy);
//float L = (du > dv)? du : dv; //float L = (du > dv)? du : dv;
// Calculate the derivatives for each axis // Calculate the derivatives for each axis
float du2 = duDx*duDx + duDy*duDy; float dU2 = dUdx*dUdx + dUdy*dUdy;
float dv2 = dvDx*dvDx + dvDy*dvDy; float dV2 = dVdx*dVdx + dVdy*dVdy;
float L2 = (du2 > dv2)? du2 : dv2; float L2 = (dU2 > dV2)? dU2 : dV2;
SWfilter filter = (L2 > 1.0f)? tex->minFilter : tex->magFilter; SWfilter filter = (L2 > 1.0f)? tex->minFilter : tex->magFilter;
@ -2096,8 +2124,8 @@ static inline void sw_blend_colors(float *SW_RESTRICT dst/*[4]*/, const float *S
static inline void sw_project_ndc_to_screen(float screen[2], const float ndc[4]) static inline void sw_project_ndc_to_screen(float screen[2], const float ndc[4])
{ {
screen[0] = RLSW.vpCenter[0] + ndc[0]*RLSW.vpHalfSize[0]; screen[0] = RLSW.vpCenter[0] + ndc[0]*RLSW.vpHalf[0] + 0.5f;
screen[1] = RLSW.vpCenter[1] - ndc[1]*RLSW.vpHalfSize[1]; screen[1] = RLSW.vpCenter[1] - ndc[1]*RLSW.vpHalf[1] + 0.5f;
} }
// Polygon Clipping management // Polygon Clipping management
@ -2307,58 +2335,64 @@ static inline void sw_triangle_clip_and_project(void)
#define DEFINE_TRIANGLE_RASTER_SCANLINE(FUNC_NAME, ENABLE_TEXTURE, ENABLE_DEPTH_TEST, ENABLE_COLOR_BLEND) \ #define DEFINE_TRIANGLE_RASTER_SCANLINE(FUNC_NAME, ENABLE_TEXTURE, ENABLE_DEPTH_TEST, ENABLE_COLOR_BLEND) \
static inline void FUNC_NAME(const sw_texture_t *tex, const sw_vertex_t *start, \ static inline void FUNC_NAME(const sw_texture_t *tex, const sw_vertex_t *start, \
const sw_vertex_t *end, float duDy, float dvDy) \ const sw_vertex_t *end, float dUdy, float dVdy) \
{ \ { \
/* Convert and center the screen coordinates */ \ /* Gets the start and end coordinates */ \
int xStart = (int)(start->screen[0] + 0.5f); \ int xStart = (int)start->screen[0]; \
int xEnd = (int)(end->screen[0] + 0.5f); \ int xEnd = (int)end->screen[0]; \
int y = (int)start->screen[1]; \ \
/* Avoid empty lines */ \
if (xStart == xEnd) return; \
\
/* Compute the subpixel distance to traverse before the first pixel */ \
float xSubstep = 1.0f - sw_fract(start->screen[0]); \
\ \
/* Compute the inverse horizontal distance along the X axis */ \ /* Compute the inverse horizontal distance along the X axis */ \
float dx = end->screen[0] - start->screen[0]; \ float dxRcp = 1.0f/(end->screen[0] - start->screen[0]); \
if (fabsf(dx) < 1e-6f) return; \
float dxRcp = 1.0f/dx; \
\ \
/* Compute the interpolation steps along the X axis */ \ /* Compute the interpolation steps along the X axis */ \
float dzDx = (end->homogeneous[2] - start->homogeneous[2])*dxRcp; \ float dZdx = (end->homogeneous[2] - start->homogeneous[2])*dxRcp; \
float dwDx = (end->homogeneous[3] - start->homogeneous[3])*dxRcp; \ float dWdx = (end->homogeneous[3] - start->homogeneous[3])*dxRcp; \
\ \
float dcDx[4] = { 0 }; \ float dCdx[4] = { 0 }; \
dcDx[0] = (end->color[0] - start->color[0])*dxRcp; \ dCdx[0] = (end->color[0] - start->color[0])*dxRcp; \
dcDx[1] = (end->color[1] - start->color[1])*dxRcp; \ dCdx[1] = (end->color[1] - start->color[1])*dxRcp; \
dcDx[2] = (end->color[2] - start->color[2])*dxRcp; \ dCdx[2] = (end->color[2] - start->color[2])*dxRcp; \
dcDx[3] = (end->color[3] - start->color[3])*dxRcp; \ dCdx[3] = (end->color[3] - start->color[3])*dxRcp; \
\ \
float duDx = 0.0f, dvDx = 0.0f; \ float dUdx = 0.0f; \
float dVdx = 0.0f; \
if (ENABLE_TEXTURE) { \ if (ENABLE_TEXTURE) { \
duDx = (end->texcoord[0] - start->texcoord[0])*dxRcp; \ dUdx = (end->texcoord[0] - start->texcoord[0])*dxRcp; \
dvDx = (end->texcoord[1] - start->texcoord[1])*dxRcp; \ dVdx = (end->texcoord[1] - start->texcoord[1])*dxRcp; \
} \ } \
\ \
/* Initializing the interpolation starting values */ \ /* Initializing the interpolation starting values */ \
float z = start->homogeneous[2]; \ float z = start->homogeneous[2] + dZdx*xSubstep; \
float w = start->homogeneous[3]; \ float w = start->homogeneous[3] + dWdx*xSubstep; \
\ \
float color[4] = { 0 }; \ float color[4] = { \
color[0] = start->color[0]; \ start->color[0] + dCdx[0]*xSubstep, \
color[1] = start->color[1]; \ start->color[1] + dCdx[1]*xSubstep, \
color[2] = start->color[2]; \ start->color[2] + dCdx[2]*xSubstep, \
color[3] = start->color[3]; \ start->color[3] + dCdx[3]*xSubstep \
}; \
\ \
float u = 0.0f, v = 0.0f; \ float u = 0.0f; \
float v = 0.0f; \
if (ENABLE_TEXTURE) { \ if (ENABLE_TEXTURE) { \
u = start->texcoord[0]; \ u = start->texcoord[0] + dUdx*xSubstep; \
v = start->texcoord[1]; \ v = start->texcoord[1] + dVdx*xSubstep; \
} \ } \
\ \
/* Pre-calculate the starting pointers for the framebuffer row */ \ /* Pre-calculate the starting pointers for the framebuffer row */ \
int y = (int)start->screen[1]; \
void *cptr = GET_COLOR_PTR(RLSW.framebuffer.color, y*RLSW.framebuffer.width + xStart); \ void *cptr = GET_COLOR_PTR(RLSW.framebuffer.color, y*RLSW.framebuffer.width + xStart); \
void *dptr = GET_DEPTH_PTR(RLSW.framebuffer.depth, y*RLSW.framebuffer.width + xStart); \ void *dptr = GET_DEPTH_PTR(RLSW.framebuffer.depth, y*RLSW.framebuffer.width + xStart); \
\ \
/* Scanline rasterization */ \ /* Scanline rasterization */ \
for (int x = xStart; x < xEnd; x++) \ for (int x = xStart; x < xEnd; x++) \
{ \ { \
/* Pixel color computation */ \
float wRcp = 1.0f/w; \ float wRcp = 1.0f/w; \
float srcColor[4] = { \ float srcColor[4] = { \
color[0]*wRcp, \ color[0]*wRcp, \
@ -2367,7 +2401,6 @@ static inline void FUNC_NAME(const sw_texture_t *tex, const sw_vertex_t *start,
color[3]*wRcp \ color[3]*wRcp \
}; \ }; \
\ \
/* Test and write depth */ \
if (ENABLE_DEPTH_TEST) \ if (ENABLE_DEPTH_TEST) \
{ \ { \
/* TODO: Implement different depth funcs? */ \ /* TODO: Implement different depth funcs? */ \
@ -2382,7 +2415,7 @@ static inline void FUNC_NAME(const sw_texture_t *tex, const sw_vertex_t *start,
float texColor[4]; \ float texColor[4]; \
float s = u*wRcp; \ float s = u*wRcp; \
float t = v*wRcp; \ float t = v*wRcp; \
sw_texture_sample(texColor, tex, s, t, duDx, duDy, dvDx, dvDy); \ sw_texture_sample(texColor, tex, s, t, dUdx, dUdy, dVdx, dVdy); \
srcColor[0] *= texColor[0]; \ srcColor[0] *= texColor[0]; \
srcColor[1] *= texColor[1]; \ srcColor[1] *= texColor[1]; \
srcColor[2] *= texColor[2]; \ srcColor[2] *= texColor[2]; \
@ -2403,16 +2436,16 @@ static inline void FUNC_NAME(const sw_texture_t *tex, const sw_vertex_t *start,
\ \
/* Increment the interpolation parameter, UVs, and pointers */ \ /* Increment the interpolation parameter, UVs, and pointers */ \
discard: \ discard: \
z += dzDx; \ z += dZdx; \
w += dwDx; \ w += dWdx; \
color[0] += dcDx[0]; \ color[0] += dCdx[0]; \
color[1] += dcDx[1]; \ color[1] += dCdx[1]; \
color[2] += dcDx[2]; \ color[2] += dCdx[2]; \
color[3] += dcDx[3]; \ color[3] += dCdx[3]; \
if (ENABLE_TEXTURE) \ if (ENABLE_TEXTURE) \
{ \ { \
u += duDx; \ u += dUdx; \
v += dvDx; \ v += dVdx; \
} \ } \
\ \
INC_COLOR_PTR(cptr); \ INC_COLOR_PTR(cptr); \
@ -2442,60 +2475,71 @@ static inline void FUNC_NAME(const sw_vertex_t *v0, const sw_vertex_t *v1,
if (h02 < 1e-6f) return; \ if (h02 < 1e-6f) return; \
\ \
/* Precompute the inverse values without additional checks */ \ /* Precompute the inverse values without additional checks */ \
float invH02 = 1.0f/h02; \ float h02Rcp = 1.0f/h02; \
float invH01 = (h01 > 1e-6f)? 1.0f/h01 : 0.0f; \ float h01Rcp = (h01 > 1e-6f)? 1.0f/h01 : 0.0f; \
float invH12 = (h12 > 1e-6f)? 1.0f/h12 : 0.0f; \ float h12Rcp = (h12 > 1e-6f)? 1.0f/h12 : 0.0f; \
\ \
/* Pre-calculation of slopes */ \ /* Pre-calculation of slopes */ \
float dx02 = (x2 - x0)*invH02; \ float dXdy02 = (x2 - x0)*h02Rcp; \
float dx01 = (x1 - x0)*invH01; \ float dXdy01 = (x1 - x0)*h01Rcp; \
float dx12 = (x2 - x1)*invH12; \ float dXdy12 = (x2 - x1)*h12Rcp; \
\
/* Y subpixel correction */ \
float y0Substep = 1.0f - sw_fract(y0); \
float y1Substep = 1.0f - sw_fract(y1); \
\ \
/* Y bounds (vertical clipping) */ \ /* Y bounds (vertical clipping) */ \
int yTop = (int)(y0 + 0.5f); \ int yTop = (int)y0; \
int yMiddle = (int)(y1 + 0.5f); \ int yMid = (int)y1; \
int yBottom = (int)(y2 + 0.5f); \ int yBot = (int)y2; \
\ \
/* Compute gradients for each side of the triangle */ \ /* Compute gradients for each side of the triangle */ \
sw_vertex_t vDy02, vDy01, vDy12; \ sw_vertex_t dVXdy02, dVXdy01, dVXdy12; \
sw_get_vertex_grad_PTCH(&vDy02, v0, v2, invH02); \ sw_get_vertex_grad_PTCH(&dVXdy02, v0, v2, h02Rcp); \
sw_get_vertex_grad_PTCH(&vDy01, v0, v1, invH01); \ sw_get_vertex_grad_PTCH(&dVXdy01, v0, v1, h01Rcp); \
sw_get_vertex_grad_PTCH(&vDy12, v1, v2, invH12); \ sw_get_vertex_grad_PTCH(&dVXdy12, v1, v2, h12Rcp); \
\ \
/* Initializing scanline variables */ \ /* Get a copy of vertices for interpolation and apply substep correction */ \
sw_vertex_t vLeft = *v0; \ sw_vertex_t vLeft = *v0, vRight = *v0; \
vLeft.screen[0] = x0; \ sw_add_vertex_grad_scaled_PTCH(&vLeft, &dVXdy02, y0Substep); \
sw_vertex_t vRight = *v0; \ sw_add_vertex_grad_scaled_PTCH(&vRight, &dVXdy01, y0Substep); \
vRight.screen[0] = x0; \ \
vLeft.screen[0] += dXdy02*y0Substep; \
vRight.screen[0] += dXdy01*y0Substep; \
\ \
/* Scanline for the upper part of the triangle */ \ /* Scanline for the upper part of the triangle */ \
for (int y = yTop; y < yMiddle; y++) \ for (int y = yTop; y < yMid; y++) \
{ \ { \
vLeft.screen[1] = vRight.screen[1] = y; \ vLeft.screen[1] = vRight.screen[1] = y; \
\ \
if (vLeft.screen[0] < vRight.screen[0]) FUNC_SCANLINE(tex, &vLeft, &vRight, vDy02.texcoord[0], vDy02.texcoord[1]); \ if (vLeft.screen[0] < vRight.screen[0]) FUNC_SCANLINE(tex, &vLeft, &vRight, dVXdy02.texcoord[0], dVXdy02.texcoord[1]); \
else FUNC_SCANLINE(tex, &vRight, &vLeft, vDy02.texcoord[0], vDy02.texcoord[1]); \ else FUNC_SCANLINE(tex, &vRight, &vLeft, dVXdy02.texcoord[0], dVXdy02.texcoord[1]); \
\ \
sw_add_vertex_grad_PTCH(&vLeft, &vDy02); \ sw_add_vertex_grad_PTCH(&vLeft, &dVXdy02); \
vLeft.screen[0] += dx02; \ vLeft.screen[0] += dXdy02; \
sw_add_vertex_grad_PTCH(&vRight, &vDy01); \ \
vRight.screen[0] += dx01; \ sw_add_vertex_grad_PTCH(&vRight, &dVXdy01); \
vRight.screen[0] += dXdy01; \
} \ } \
\ \
/* Scanline for the lower part of the triangle */ \ /* Get a copy of next right for interpolation and apply substep correction */ \
vRight = *v1, vRight.screen[0] = x1; \ vRight = *v1; \
sw_add_vertex_grad_scaled_PTCH(&vRight, &dVXdy12, y1Substep); \
vRight.screen[0] += dXdy12*y1Substep; \
\ \
for (int y = yMiddle; y < yBottom; y++) \ /* Scanline for the lower part of the triangle */ \
for (int y = yMid; y < yBot; y++) \
{ \ { \
vLeft.screen[1] = vRight.screen[1] = y; \ vLeft.screen[1] = vRight.screen[1] = y; \
\ \
if (vLeft.screen[0] < vRight.screen[0]) FUNC_SCANLINE(tex, &vLeft, &vRight, vDy02.texcoord[0], vDy02.texcoord[1]); \ if (vLeft.screen[0] < vRight.screen[0]) FUNC_SCANLINE(tex, &vLeft, &vRight, dVXdy02.texcoord[0], dVXdy02.texcoord[1]); \
else FUNC_SCANLINE(tex, &vRight, &vLeft, vDy02.texcoord[0], vDy02.texcoord[1]); \ else FUNC_SCANLINE(tex, &vRight, &vLeft, dVXdy02.texcoord[0], dVXdy02.texcoord[1]); \
\ \
sw_add_vertex_grad_PTCH(&vLeft, &vDy02); \ sw_add_vertex_grad_PTCH(&vLeft, &dVXdy02); \
vLeft.screen[0] += dx02; \ vLeft.screen[0] += dXdy02; \
sw_add_vertex_grad_PTCH(&vRight, &vDy12); \ \
vRight.screen[0] += dx12; \ sw_add_vertex_grad_PTCH(&vRight, &dVXdy12); \
vRight.screen[0] += dXdy12; \
} \ } \
} }
@ -2727,10 +2771,10 @@ static inline void FUNC_NAME(void)
const sw_vertex_t *v3 = sortedVerts[3]; \ const sw_vertex_t *v3 = sortedVerts[3]; \
\ \
/* Screen bounds (axis-aligned) */ \ /* Screen bounds (axis-aligned) */ \
int xMin = (int)(v0->screen[0] + 0.5f); \ int xMin = (int)v0->screen[0]; \
int yMin = (int)(v0->screen[1] + 0.5f); \ int yMin = (int)v0->screen[1]; \
int xMax = (int)(v2->screen[0] + 0.5f); \ int xMax = (int)v2->screen[0]; \
int yMax = (int)(v2->screen[1] + 0.5f); \ int yMax = (int)v2->screen[1]; \
\ \
int width = xMax - xMin; \ int width = xMax - xMin; \
int height = yMax - yMin; \ int height = yMax - yMin; \
@ -3031,10 +3075,10 @@ static inline bool sw_line_clip_and_project(sw_vertex_t *v0, sw_vertex_t *v1)
#define DEFINE_LINE_RASTER(FUNC_NAME, ENABLE_DEPTH_TEST, ENABLE_COLOR_BLEND) \ #define DEFINE_LINE_RASTER(FUNC_NAME, ENABLE_DEPTH_TEST, ENABLE_COLOR_BLEND) \
static inline void FUNC_NAME(const sw_vertex_t *v0, const sw_vertex_t *v1) \ static inline void FUNC_NAME(const sw_vertex_t *v0, const sw_vertex_t *v1) \
{ \ { \
int x1 = (int)(v0->screen[0] + 0.5f); \ int x1 = (int)v0->screen[0]; \
int y1 = (int)(v0->screen[1] + 0.5f); \ int y1 = (int)v0->screen[1]; \
int x2 = (int)(v1->screen[0] + 0.5f); \ int x2 = (int)v1->screen[0]; \
int y2 = (int)(v1->screen[1] + 0.5f); \ int y2 = (int)v1->screen[1]; \
\ \
int dx = x2 - x1; \ int dx = x2 - x1; \
int dy = y2 - y1; \ int dy = y2 - y1; \
@ -3864,11 +3908,11 @@ void swViewport(int x, int y, int width, int height)
RLSW.vpSize[0] = width; RLSW.vpSize[0] = width;
RLSW.vpSize[1] = height; RLSW.vpSize[1] = height;
RLSW.vpHalfSize[0] = (int)(width/2.0f + 0.5f); RLSW.vpHalf[0] = width / 2.0f;
RLSW.vpHalfSize[1] = (int)(height/2.0f + 0.5f); RLSW.vpHalf[1] = height / 2.0f;
RLSW.vpCenter[0] = x + RLSW.vpHalfSize[0]; RLSW.vpCenter[0] = (float)x + RLSW.vpHalf[0];
RLSW.vpCenter[1] = y + RLSW.vpHalfSize[1]; RLSW.vpCenter[1] = (float)y + RLSW.vpHalf[1];
RLSW.vpMin[0] = sw_clampi(x, 0, RLSW.framebuffer.width - 1); RLSW.vpMin[0] = sw_clampi(x, 0, RLSW.framebuffer.width - 1);
RLSW.vpMin[1] = sw_clampi(y, 0, RLSW.framebuffer.height - 1); RLSW.vpMin[1] = sw_clampi(y, 0, RLSW.framebuffer.height - 1);