/**************************************************************************
*
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
#include "lp_setup_context.h"
#include "lp_rast.h"
#include "lp_state_fs.h"
+#include "lp_state_setup.h"
+#include "lp_context.h"
+#include "draw/draw_context.h"
#define NUM_CHANNELS 4
+struct lp_line_info {
-static const int step_scissor_minx[16] = {
- 0, 1, 0, 1,
- 2, 3, 2, 3,
- 0, 1, 0, 1,
- 2, 3, 2, 3
-};
-
-static const int step_scissor_maxx[16] = {
- 0, -1, 0, -1,
- -2, -3, -2, -3,
- 0, -1, 0, -1,
- -2, -3, -2, -3
-};
+ float dx;
+ float dy;
+ float oneoverarea;
+ boolean frontfacing;
-static const int step_scissor_miny[16] = {
- 0, 0, 1, 1,
- 0, 0, 1, 1,
- 2, 2, 3, 3,
- 2, 2, 3, 3
-};
+ const float (*v1)[4];
+ const float (*v2)[4];
-static const int step_scissor_maxy[16] = {
- 0, 0, -1, -1,
- 0, 0, -1, -1,
- -2, -2, -3, -3,
- -2, -2, -3, -3
+ float (*a0)[4];
+ float (*dadx)[4];
+ float (*dady)[4];
};
-
/**
* Compute a0 for a constant-valued coefficient (GL_FLAT shading).
*/
static void constant_coef( struct lp_setup_context *setup,
- struct lp_rast_triangle *tri,
+ struct lp_line_info *info,
unsigned slot,
const float value,
unsigned i )
{
- tri->inputs.a0[slot][i] = value;
- tri->inputs.dadx[slot][i] = 0.0f;
- tri->inputs.dady[slot][i] = 0.0f;
+ info->a0[slot][i] = value;
+ info->dadx[slot][i] = 0.0f;
+ info->dady[slot][i] = 0.0f;
}
* for a triangle.
*/
static void linear_coef( struct lp_setup_context *setup,
- struct lp_rast_triangle *tri,
- float oneoverarea,
+ struct lp_line_info *info,
unsigned slot,
- const float (*v1)[4],
- const float (*v2)[4],
unsigned vert_attr,
unsigned i)
{
- float a1 = v1[vert_attr][i];
- float a2 = v2[vert_attr][i];
+ float a1 = info->v1[vert_attr][i];
+ float a2 = info->v2[vert_attr][i];
float da21 = a1 - a2;
- float dadx = da21 * tri->dx * oneoverarea;
- float dady = da21 * tri->dy * oneoverarea;
+ float dadx = da21 * info->dx * info->oneoverarea;
+ float dady = da21 * info->dy * info->oneoverarea;
- tri->inputs.dadx[slot][i] = dadx;
- tri->inputs.dady[slot][i] = dady;
+ info->dadx[slot][i] = dadx;
+ info->dady[slot][i] = dady;
- tri->inputs.a0[slot][i] = (a1 -
- (dadx * (v1[0][0] - setup->pixel_offset) +
- dady * (v1[0][1] - setup->pixel_offset)));
+ info->a0[slot][i] = (a1 -
+ (dadx * (info->v1[0][0] - setup->pixel_offset) +
+ dady * (info->v1[0][1] - setup->pixel_offset)));
}
* divide the interpolated value by the interpolated W at that fragment.
*/
static void perspective_coef( struct lp_setup_context *setup,
- struct lp_rast_triangle *tri,
- float oneoverarea,
+ struct lp_line_info *info,
unsigned slot,
- const float (*v1)[4],
- const float (*v2)[4],
unsigned vert_attr,
unsigned i)
{
/* premultiply by 1/w (v[0][3] is always 1/w):
*/
- float a1 = v1[vert_attr][i] * v1[0][3];
- float a2 = v2[vert_attr][i] * v2[0][3];
+ float a1 = info->v1[vert_attr][i] * info->v1[0][3];
+ float a2 = info->v2[vert_attr][i] * info->v2[0][3];
float da21 = a1 - a2;
- float dadx = da21 * tri->dx * oneoverarea;
- float dady = da21 * tri->dy * oneoverarea;
+ float dadx = da21 * info->dx * info->oneoverarea;
+ float dady = da21 * info->dy * info->oneoverarea;
- tri->inputs.dadx[slot][i] = dadx;
- tri->inputs.dady[slot][i] = dady;
+ info->dadx[slot][i] = dadx;
+ info->dady[slot][i] = dady;
- tri->inputs.a0[slot][i] = (a1 -
- (dadx * (v1[0][0] - setup->pixel_offset) +
- dady * (v1[0][1] - setup->pixel_offset)));
+ info->a0[slot][i] = (a1 -
+ (dadx * (info->v1[0][0] - setup->pixel_offset) +
+ dady * (info->v1[0][1] - setup->pixel_offset)));
+}
+
+static void
+setup_fragcoord_coef( struct lp_setup_context *setup,
+ struct lp_line_info *info,
+ unsigned slot,
+ unsigned usage_mask)
+{
+ /*X*/
+ if (usage_mask & TGSI_WRITEMASK_X) {
+ info->a0[slot][0] = 0.0;
+ info->dadx[slot][0] = 1.0;
+ info->dady[slot][0] = 0.0;
+ }
+
+ /*Y*/
+ if (usage_mask & TGSI_WRITEMASK_Y) {
+ info->a0[slot][1] = 0.0;
+ info->dadx[slot][1] = 0.0;
+ info->dady[slot][1] = 1.0;
+ }
+
+ /*Z*/
+ if (usage_mask & TGSI_WRITEMASK_Z) {
+ linear_coef(setup, info, slot, 0, 2);
+ }
+
+ /*W*/
+ if (usage_mask & TGSI_WRITEMASK_W) {
+ linear_coef(setup, info, slot, 0, 3);
+ }
}
/**
* Compute the tri->coef[] array dadx, dady, a0 values.
*/
static void setup_line_coefficients( struct lp_setup_context *setup,
- struct lp_rast_triangle *tri,
- float oneoverarea,
- const float (*v1)[4],
- const float (*v2)[4])
+ struct lp_line_info *info)
{
+ const struct lp_setup_variant_key *key = &setup->setup.variant->key;
unsigned fragcoord_usage_mask = TGSI_WRITEMASK_XYZ;
unsigned slot;
/* setup interpolation for all the remaining attributes:
*/
- for (slot = 0; slot < setup->fs.nr_inputs; slot++) {
- unsigned vert_attr = setup->fs.input[slot].src_index;
- unsigned usage_mask = setup->fs.input[slot].usage_mask;
+ for (slot = 0; slot < key->num_inputs; slot++) {
+ unsigned vert_attr = key->inputs[slot].src_index;
+ unsigned usage_mask = key->inputs[slot].usage_mask;
unsigned i;
- switch (setup->fs.input[slot].interp) {
+ switch (key->inputs[slot].interp) {
case LP_INTERP_CONSTANT:
- if (setup->flatshade_first) {
+ if (key->flatshade_first) {
for (i = 0; i < NUM_CHANNELS; i++)
if (usage_mask & (1 << i))
- constant_coef(setup, tri, slot+1, v1[vert_attr][i], i);
+ constant_coef(setup, info, slot+1, info->v1[vert_attr][i], i);
}
else {
for (i = 0; i < NUM_CHANNELS; i++)
if (usage_mask & (1 << i))
- constant_coef(setup, tri, slot+1, v2[vert_attr][i], i);
+ constant_coef(setup, info, slot+1, info->v2[vert_attr][i], i);
}
break;
case LP_INTERP_LINEAR:
for (i = 0; i < NUM_CHANNELS; i++)
if (usage_mask & (1 << i))
- linear_coef(setup, tri, oneoverarea, slot+1, v1, v2, vert_attr, i);
+ linear_coef(setup, info, slot+1, vert_attr, i);
break;
case LP_INTERP_PERSPECTIVE:
for (i = 0; i < NUM_CHANNELS; i++)
if (usage_mask & (1 << i))
- perspective_coef(setup, tri, oneoverarea, slot+1, v1, v2, vert_attr, i);
+ perspective_coef(setup, info, slot+1, vert_attr, i);
fragcoord_usage_mask |= TGSI_WRITEMASK_W;
break;
fragcoord_usage_mask |= usage_mask;
break;
+ case LP_INTERP_FACING:
+ for (i = 0; i < NUM_CHANNELS; i++)
+ if (usage_mask & (1 << i))
+ constant_coef(setup, info, slot+1,
+ info->frontfacing ? 1.0f : -1.0f, i);
+ break;
+
default:
assert(0);
}
/* The internal position input is in slot zero:
*/
- lp_setup_fragcoord_coef(setup, tri, oneoverarea, 0, v1, v2, v2,
- fragcoord_usage_mask);
+ setup_fragcoord_coef(setup, info, 0,
+ fragcoord_usage_mask);
}
const float (*v1)[4],
const float (*v2)[4])
{
+ const struct lp_setup_variant_key *key = &setup->setup.variant->key;
uint i;
debug_printf("llvmpipe line\n");
- for (i = 0; i < 1 + setup->fs.nr_inputs; i++) {
+ for (i = 0; i < 1 + key->num_inputs; i++) {
debug_printf(" v1[%d]: %f %f %f %f\n", i,
v1[i][0], v1[i][1], v1[i][2], v1[i][3]);
}
- for (i = 0; i < 1 + setup->fs.nr_inputs; i++) {
+ for (i = 0; i < 1 + key->num_inputs; i++) {
debug_printf(" v2[%d]: %f %f %f %f\n", i,
v2[i][0], v2[i][1], v2[i][2], v2[i][3]);
}
}
-static void
-lp_setup_line( struct lp_setup_context *setup,
+static INLINE boolean sign(float x){
+ return x >= 0;
+}
+
+
+/* Used on positive floats only:
+ */
+static INLINE float fracf(float f)
+{
+ return f - floorf(f);
+}
+
+
+
+static boolean
+try_setup_line( struct lp_setup_context *setup,
const float (*v1)[4],
const float (*v2)[4])
{
- struct lp_scene *scene = lp_setup_get_current_scene(setup);
+ struct llvmpipe_context *lp_context = (struct llvmpipe_context *)setup->pipe;
+ struct lp_scene *scene = setup->scene;
+ const struct lp_setup_variant_key *key = &setup->setup.variant->key;
struct lp_rast_triangle *line;
- float oneoverarea;
- float half_width = setup->line_width / 2;
- int minx, maxx, miny, maxy;
- int ix0, ix1, iy0, iy1;
+ struct lp_rast_plane *plane;
+ struct lp_line_info info;
+ float width = MAX2(1.0, setup->line_width);
+ struct u_rect bbox;
unsigned tri_bytes;
int x[4];
int y[4];
int i;
int nr_planes = 4;
- boolean opaque;
-
+ unsigned viewport_index = 0;
+ unsigned layer = 0;
+
+ /* linewidth should be interpreted as integer */
+ int fixed_width = util_iround(width) * FIXED_ONE;
+
+ float x_offset=0;
+ float y_offset=0;
+ float x_offset_end=0;
+ float y_offset_end=0;
+
+ float x1diff;
+ float y1diff;
+ float x2diff;
+ float y2diff;
+ float dx, dy;
+ float area;
+
+ boolean draw_start;
+ boolean draw_end;
+ boolean will_draw_start;
+ boolean will_draw_end;
+
if (0)
print_line(setup, v1, v2);
if (setup->scissor_test) {
nr_planes = 8;
+ if (setup->viewport_index_slot > 0) {
+ unsigned *udata = (unsigned*)v1[setup->viewport_index_slot];
+ viewport_index = lp_clamp_viewport_idx(*udata);
+ }
}
else {
nr_planes = 4;
}
- line = lp_setup_alloc_triangle(scene,
- setup->fs.nr_inputs,
- nr_planes,
- &tri_bytes);
- if (!line)
- return;
+ if (setup->layer_slot > 0) {
+ layer = *(unsigned*)v1[setup->layer_slot];
+ layer = MIN2(layer, scene->fb_max_layer);
+ }
-#ifndef DEBUG
- line->v[0][0] = v1[0][0];
- line->v[1][0] = v2[0][0];
- line->v[0][1] = v1[0][1];
- line->v[1][1] = v2[0][1];
-#endif
+ dx = v1[0][0] - v2[0][0];
+ dy = v1[0][1] - v2[0][1];
+ area = (dx * dx + dy * dy);
+ if (area == 0) {
+ LP_COUNT(nr_culled_tris);
+ return TRUE;
+ }
- /* pre-calculation(based on given vertices) to determine if line is
- * more horizontal or more vertical
- */
- line->dx = v1[0][0] - v2[0][0];
- line->dy = v1[0][1] - v2[0][1];
-
- /* x-major line */
- if (fabsf(line->dx) >= fabsf(line->dy)) {
- if (line->dx < 0) {
+ info.oneoverarea = 1.0f / area;
+ info.dx = dx;
+ info.dy = dy;
+ info.v1 = v1;
+ info.v2 = v2;
+
+
+ /* X-MAJOR LINE */
+ if (fabsf(dx) >= fabsf(dy)) {
+ float dydx = dy / dx;
+
+ x1diff = v1[0][0] - (float) floor(v1[0][0]) - 0.5;
+ y1diff = v1[0][1] - (float) floor(v1[0][1]) - 0.5;
+ x2diff = v2[0][0] - (float) floor(v2[0][0]) - 0.5;
+ y2diff = v2[0][1] - (float) floor(v2[0][1]) - 0.5;
+
+ if (y2diff==-0.5 && dy<0){
+ y2diff = 0.5;
+ }
+
+ /*
+ * Diamond exit rule test for starting point
+ */
+ if (fabsf(x1diff) + fabsf(y1diff) < 0.5) {
+ draw_start = TRUE;
+ }
+ else if (sign(x1diff) == sign(-dx)) {
+ draw_start = FALSE;
+ }
+ else if (sign(-y1diff) != sign(dy)) {
+ draw_start = TRUE;
+ }
+ else {
+ /* do intersection test */
+ float yintersect = fracf(v1[0][1]) + x1diff * dydx;
+ draw_start = (yintersect < 1.0 && yintersect > 0.0);
+ }
+
+
+ /*
+ * Diamond exit rule test for ending point
+ */
+ if (fabsf(x2diff) + fabsf(y2diff) < 0.5) {
+ draw_end = FALSE;
+ }
+ else if (sign(x2diff) != sign(-dx)) {
+ draw_end = FALSE;
+ }
+ else if (sign(-y2diff) == sign(dy)) {
+ draw_end = TRUE;
+ }
+ else {
+ /* do intersection test */
+ float yintersect = fracf(v2[0][1]) + x2diff * dydx;
+ draw_end = (yintersect < 1.0 && yintersect > 0.0);
+ }
+
+ /* Are we already drawing start/end?
+ */
+ will_draw_start = sign(-x1diff) != sign(dx);
+ will_draw_end = (sign(x2diff) == sign(-dx)) || x2diff==0;
+
+ if (dx < 0) {
/* if v2 is to the right of v1, swap pointers */
const float (*temp)[4] = v1;
v1 = v2;
v2 = temp;
- line->dx = -line->dx;
- line->dy = -line->dy;
+ dx = -dx;
+ dy = -dy;
+ /* Otherwise shift planes appropriately */
+ if (will_draw_start != draw_start) {
+ x_offset_end = - x1diff - 0.5;
+ y_offset_end = x_offset_end * dydx;
+
+ }
+ if (will_draw_end != draw_end) {
+ x_offset = - x2diff - 0.5;
+ y_offset = x_offset * dydx;
+ }
+
}
-
+ else{
+ /* Otherwise shift planes appropriately */
+ if (will_draw_start != draw_start) {
+ x_offset = - x1diff + 0.5;
+ y_offset = x_offset * dydx;
+ }
+ if (will_draw_end != draw_end) {
+ x_offset_end = - x2diff + 0.5;
+ y_offset_end = x_offset_end * dydx;
+ }
+ }
+
/* x/y positions in fixed point */
- x[0] = subpixel_snap(v1[0][0] - setup->pixel_offset);
- x[1] = subpixel_snap(v2[0][0] - setup->pixel_offset);
- x[2] = subpixel_snap(v2[0][0] - setup->pixel_offset);
- x[3] = subpixel_snap(v1[0][0] - setup->pixel_offset);
+ x[0] = subpixel_snap(v1[0][0] + x_offset - setup->pixel_offset);
+ x[1] = subpixel_snap(v2[0][0] + x_offset_end - setup->pixel_offset);
+ x[2] = subpixel_snap(v2[0][0] + x_offset_end - setup->pixel_offset);
+ x[3] = subpixel_snap(v1[0][0] + x_offset - setup->pixel_offset);
+
+ y[0] = subpixel_snap(v1[0][1] + y_offset - setup->pixel_offset) - fixed_width/2;
+ y[1] = subpixel_snap(v2[0][1] + y_offset_end - setup->pixel_offset) - fixed_width/2;
+ y[2] = subpixel_snap(v2[0][1] + y_offset_end - setup->pixel_offset) + fixed_width/2;
+ y[3] = subpixel_snap(v1[0][1] + y_offset - setup->pixel_offset) + fixed_width/2;
- y[0] = subpixel_snap(v1[0][1] - half_width - setup->pixel_offset);
- y[1] = subpixel_snap(v2[0][1] - half_width - setup->pixel_offset);
- y[2] = subpixel_snap(v2[0][1] + half_width - setup->pixel_offset);
- y[3] = subpixel_snap(v1[0][1] + half_width - setup->pixel_offset);
}
- else{
- /* y-major line */
- if (line->dy > 0) {
+ else {
+ const float dxdy = dx / dy;
+
+ /* Y-MAJOR LINE */
+ x1diff = v1[0][0] - (float) floor(v1[0][0]) - 0.5;
+ y1diff = v1[0][1] - (float) floor(v1[0][1]) - 0.5;
+ x2diff = v2[0][0] - (float) floor(v2[0][0]) - 0.5;
+ y2diff = v2[0][1] - (float) floor(v2[0][1]) - 0.5;
+
+ if (x2diff==-0.5 && dx<0) {
+ x2diff = 0.5;
+ }
+
+ /*
+ * Diamond exit rule test for starting point
+ */
+ if (fabsf(x1diff) + fabsf(y1diff) < 0.5) {
+ draw_start = TRUE;
+ }
+ else if (sign(-y1diff) == sign(dy)) {
+ draw_start = FALSE;
+ }
+ else if (sign(x1diff) != sign(-dx)) {
+ draw_start = TRUE;
+ }
+ else {
+ /* do intersection test */
+ float xintersect = fracf(v1[0][0]) + y1diff * dxdy;
+ draw_start = (xintersect < 1.0 && xintersect > 0.0);
+ }
+
+ /*
+ * Diamond exit rule test for ending point
+ */
+ if (fabsf(x2diff) + fabsf(y2diff) < 0.5) {
+ draw_end = FALSE;
+ }
+ else if (sign(-y2diff) != sign(dy) ) {
+ draw_end = FALSE;
+ }
+ else if (sign(x2diff) == sign(-dx) ) {
+ draw_end = TRUE;
+ }
+ else {
+ /* do intersection test */
+ float xintersect = fracf(v2[0][0]) + y2diff * dxdy;
+ draw_end = (xintersect < 1.0 && xintersect >= 0.0);
+ }
+
+ /* Are we already drawing start/end?
+ */
+ will_draw_start = sign(y1diff) == sign(dy);
+ will_draw_end = (sign(-y2diff) == sign(dy)) || y2diff==0;
+
+ if (dy > 0) {
/* if v2 is on top of v1, swap pointers */
const float (*temp)[4] = v1;
v1 = v2;
v2 = temp;
- line->dx = -line->dx;
- line->dy = -line->dy;
+ dx = -dx;
+ dy = -dy;
+
+ /* Otherwise shift planes appropriately */
+ if (will_draw_start != draw_start) {
+ y_offset_end = - y1diff + 0.5;
+ x_offset_end = y_offset_end * dxdy;
+ }
+ if (will_draw_end != draw_end) {
+ y_offset = - y2diff + 0.5;
+ x_offset = y_offset * dxdy;
+ }
}
-
- x[0] = subpixel_snap(v1[0][0] - half_width - setup->pixel_offset);
- x[1] = subpixel_snap(v2[0][0] - half_width - setup->pixel_offset);
- x[2] = subpixel_snap(v2[0][0] + half_width - setup->pixel_offset);
- x[3] = subpixel_snap(v1[0][0] + half_width - setup->pixel_offset);
+ else {
+ /* Otherwise shift planes appropriately */
+ if (will_draw_start != draw_start) {
+ y_offset = - y1diff - 0.5;
+ x_offset = y_offset * dxdy;
+
+ }
+ if (will_draw_end != draw_end) {
+ y_offset_end = - y2diff - 0.5;
+ x_offset_end = y_offset_end * dxdy;
+ }
+ }
+
+ /* x/y positions in fixed point */
+ x[0] = subpixel_snap(v1[0][0] + x_offset - setup->pixel_offset) - fixed_width/2;
+ x[1] = subpixel_snap(v2[0][0] + x_offset_end - setup->pixel_offset) - fixed_width/2;
+ x[2] = subpixel_snap(v2[0][0] + x_offset_end - setup->pixel_offset) + fixed_width/2;
+ x[3] = subpixel_snap(v1[0][0] + x_offset - setup->pixel_offset) + fixed_width/2;
- y[0] = subpixel_snap(v1[0][1] - setup->pixel_offset);
- y[1] = subpixel_snap(v2[0][1] - setup->pixel_offset);
- y[2] = subpixel_snap(v2[0][1] - setup->pixel_offset);
- y[3] = subpixel_snap(v1[0][1] - setup->pixel_offset);
+ y[0] = subpixel_snap(v1[0][1] + y_offset - setup->pixel_offset);
+ y[1] = subpixel_snap(v2[0][1] + y_offset_end - setup->pixel_offset);
+ y[2] = subpixel_snap(v2[0][1] + y_offset_end - setup->pixel_offset);
+ y[3] = subpixel_snap(v1[0][1] + y_offset - setup->pixel_offset);
}
- /* calculate the deltas */
- line->plane[0].dcdy = x[0] - x[1];
- line->plane[1].dcdy = x[1] - x[2];
- line->plane[2].dcdy = x[2] - x[3];
- line->plane[3].dcdy = x[3] - x[0];
-
- line->plane[0].dcdx = y[0] - y[1];
- line->plane[1].dcdx = y[1] - y[2];
- line->plane[2].dcdx = y[2] - y[3];
- line->plane[3].dcdx = y[3] - y[0];
-
-
- LP_COUNT(nr_tris);
-
-
/* Bounding rectangle (in pixels) */
{
/* Yes this is necessary to accurately calculate bounding boxes
* up needing a bottom-left fill convention, which requires
* slightly different rounding.
*/
- int adj = (setup->pixel_offset != 0) ? 1 : 0;
+ int adj = (setup->bottom_edge_rule != 0) ? 1 : 0;
+
+ bbox.x0 = (MIN4(x[0], x[1], x[2], x[3]) + (FIXED_ONE-1)) >> FIXED_ORDER;
+ bbox.x1 = (MAX4(x[0], x[1], x[2], x[3]) + (FIXED_ONE-1)) >> FIXED_ORDER;
+ bbox.y0 = (MIN4(y[0], y[1], y[2], y[3]) + (FIXED_ONE-1) + adj) >> FIXED_ORDER;
+ bbox.y1 = (MAX4(y[0], y[1], y[2], y[3]) + (FIXED_ONE-1) + adj) >> FIXED_ORDER;
- minx = (MIN4(x[0], x[1], x[2], x[3]) + (FIXED_ONE-1)) >> FIXED_ORDER;
- maxx = (MAX4(x[0], x[1], x[2], x[3]) + (FIXED_ONE-1)) >> FIXED_ORDER;
- miny = (MIN4(y[0], y[1], y[3], y[3]) + (FIXED_ONE-1) + adj) >> FIXED_ORDER;
- maxy = (MAX4(y[0], y[1], y[3], y[3]) + (FIXED_ONE-1) + adj) >> FIXED_ORDER;
+ /* Inclusive coordinates:
+ */
+ bbox.x1--;
+ bbox.y1--;
}
- if (setup->scissor_test) {
- minx = MAX2(minx, setup->scissor.current.minx);
- maxx = MIN2(maxx, setup->scissor.current.maxx);
- miny = MAX2(miny, setup->scissor.current.miny);
- maxy = MIN2(maxy, setup->scissor.current.maxy);
+ if (bbox.x1 < bbox.x0 ||
+ bbox.y1 < bbox.y0) {
+ if (0) debug_printf("empty bounding box\n");
+ LP_COUNT(nr_culled_tris);
+ return TRUE;
}
- else {
- minx = MAX2(minx, 0);
- miny = MAX2(miny, 0);
- maxx = MIN2(maxx, scene->fb.width);
- maxy = MIN2(maxy, scene->fb.height);
+
+ if (!u_rect_test_intersection(&setup->draw_regions[viewport_index], &bbox)) {
+ if (0) debug_printf("offscreen\n");
+ LP_COUNT(nr_culled_tris);
+ return TRUE;
}
+ /* Can safely discard negative regions:
+ */
+ bbox.x0 = MAX2(bbox.x0, 0);
+ bbox.y0 = MAX2(bbox.y0, 0);
+
+ line = lp_setup_alloc_triangle(scene,
+ key->num_inputs,
+ nr_planes,
+ &tri_bytes);
+ if (!line)
+ return FALSE;
+
+#ifdef DEBUG
+ line->v[0][0] = v1[0][0];
+ line->v[1][0] = v2[0][0];
+ line->v[0][1] = v1[0][1];
+ line->v[1][1] = v2[0][1];
+#endif
+
+ LP_COUNT(nr_tris);
- if (miny >= maxy || minx >= maxx) {
- lp_scene_putback_data( scene, tri_bytes );
- return;
+ if (lp_context->active_statistics_queries &&
+ !llvmpipe_rasterization_disabled(lp_context)) {
+ lp_context->pipeline_statistics.c_primitives++;
}
- oneoverarea = 1.0f / (line->dx * line->dx + line->dy * line->dy);
+ /* calculate the deltas */
+ plane = GET_PLANES(line);
+ plane[0].dcdy = x[0] - x[1];
+ plane[1].dcdy = x[1] - x[2];
+ plane[2].dcdy = x[2] - x[3];
+ plane[3].dcdy = x[3] - x[0];
+
+ plane[0].dcdx = y[0] - y[1];
+ plane[1].dcdx = y[1] - y[2];
+ plane[2].dcdx = y[2] - y[3];
+ plane[3].dcdx = y[3] - y[0];
+
+ if (draw_will_inject_frontface(lp_context->draw) &&
+ setup->face_slot > 0) {
+ line->inputs.frontfacing = v1[setup->face_slot][0];
+ } else {
+ line->inputs.frontfacing = TRUE;
+ }
/* Setup parameter interpolants:
*/
- setup_line_coefficients( setup, line, oneoverarea, v1, v2);
+ info.a0 = GET_A0(&line->inputs);
+ info.dadx = GET_DADX(&line->inputs);
+ info.dady = GET_DADY(&line->inputs);
+ info.frontfacing = line->inputs.frontfacing;
+ setup_line_coefficients(setup, &info);
+
+ line->inputs.disable = FALSE;
+ line->inputs.opaque = FALSE;
+ line->inputs.layer = layer;
+ line->inputs.viewport_index = viewport_index;
for (i = 0; i < 4; i++) {
- struct lp_rast_plane *plane = &line->plane[i];
/* half-edge constants, will be interated over the whole render
* target.
*/
- plane->c = plane->dcdx * x[i] - plane->dcdy * y[i];
+ plane[i].c = IMUL64(plane[i].dcdx, x[i]) - IMUL64(plane[i].dcdy, y[i]);
/* correct for top-left vs. bottom-left fill convention.
- *
- * note that we're overloading gl_rasterization_rules to mean
- * both (0.5,0.5) pixel centers *and* bottom-left filling
- * convention.
- *
- * GL actually has a top-left filling convention, but GL's
- * notion of "top" differs from gallium's...
- *
- * Also, sometimes (in FBO cases) GL will render upside down
- * to its usual method, in which case it will probably want
- * to use the opposite, top-left convention.
*/
- if (plane->dcdx < 0) {
+ if (plane[i].dcdx < 0) {
/* both fill conventions want this - adjust for left edges */
- plane->c++;
+ plane[i].c++;
}
- else if (plane->dcdx == 0) {
+ else if (plane[i].dcdx == 0) {
if (setup->pixel_offset == 0) {
/* correct for top-left fill convention:
*/
- if (plane->dcdy > 0) plane->c++;
+ if (plane[i].dcdy > 0) plane[i].c++;
}
else {
/* correct for bottom-left fill convention:
*/
- if (plane->dcdy < 0) plane->c++;
+ if (plane[i].dcdy < 0) plane[i].c++;
}
}
- plane->dcdx *= FIXED_ONE;
- plane->dcdy *= FIXED_ONE;
+ plane[i].dcdx *= FIXED_ONE;
+ plane[i].dcdy *= FIXED_ONE;
/* find trivial reject offsets for each edge for a single-pixel
* sized block. These will be scaled up at each recursive level to
* match the active blocksize. Scaling in this way works best if
* the blocks are square.
*/
- plane->eo = 0;
- if (plane->dcdx < 0) plane->eo -= plane->dcdx;
- if (plane->dcdy > 0) plane->eo += plane->dcdy;
-
- /* Calculate trivial accept offsets from the above.
- */
- plane->ei = plane->dcdy - plane->dcdx - plane->eo;
-
- plane->step = line->step[i];
-
- /* Fill in the inputs.step[][] arrays.
- * We've manually unrolled some loops here.
- */
-#define SETUP_STEP(j, x, y) \
- line->step[i][j] = y * plane->dcdy - x * plane->dcdx
-
- SETUP_STEP(0, 0, 0);
- SETUP_STEP(1, 1, 0);
- SETUP_STEP(2, 0, 1);
- SETUP_STEP(3, 1, 1);
-
- SETUP_STEP(4, 2, 0);
- SETUP_STEP(5, 3, 0);
- SETUP_STEP(6, 2, 1);
- SETUP_STEP(7, 3, 1);
-
- SETUP_STEP(8, 0, 2);
- SETUP_STEP(9, 1, 2);
- SETUP_STEP(10, 0, 3);
- SETUP_STEP(11, 1, 3);
-
- SETUP_STEP(12, 2, 2);
- SETUP_STEP(13, 3, 2);
- SETUP_STEP(14, 2, 3);
- SETUP_STEP(15, 3, 3);
-#undef STEP
+ plane[i].eo = 0;
+ if (plane[i].dcdx < 0) plane[i].eo -= plane[i].dcdx;
+ if (plane[i].dcdy > 0) plane[i].eo += plane[i].dcdy;
}
* these planes elsewhere.
*/
if (nr_planes == 8) {
- line->plane[4].step = step_scissor_maxx;
- line->plane[4].dcdx = 1;
- line->plane[4].dcdy = 0;
- line->plane[4].c = maxx;
- line->plane[4].ei = -1;
- line->plane[4].eo = 0;
-
- line->plane[5].step = step_scissor_miny;
- line->plane[5].dcdx = 0;
- line->plane[5].dcdy = 1;
- line->plane[5].c = 1-miny;
- line->plane[5].ei = 0;
- line->plane[5].eo = 1;
-
- line->plane[6].step = step_scissor_maxy;
- line->plane[6].dcdx = 0;
- line->plane[6].dcdy = -1;
- line->plane[6].c = maxy;
- line->plane[6].ei = -1;
- line->plane[6].eo = 0;
-
- line->plane[7].step = step_scissor_minx;
- line->plane[7].dcdx = -1;
- line->plane[7].dcdy = 0;
- line->plane[7].c = 1-minx;
- line->plane[7].ei = 0;
- line->plane[7].eo = 1;
+ const struct u_rect *scissor =
+ &setup->scissors[viewport_index];
+
+ plane[4].dcdx = -1;
+ plane[4].dcdy = 0;
+ plane[4].c = 1-scissor->x0;
+ plane[4].eo = 1;
+
+ plane[5].dcdx = 1;
+ plane[5].dcdy = 0;
+ plane[5].c = scissor->x1+1;
+ plane[5].eo = 0;
+
+ plane[6].dcdx = 0;
+ plane[6].dcdy = 1;
+ plane[6].c = 1-scissor->y0;
+ plane[6].eo = 1;
+
+ plane[7].dcdx = 0;
+ plane[7].dcdy = -1;
+ plane[7].c = scissor->y1+1;
+ plane[7].eo = 0;
}
+ return lp_setup_bin_triangle(setup, line, &bbox, nr_planes, viewport_index);
+}
- /*
- * All fields of 'tri' are now set. The remaining code here is
- * concerned with binning.
- */
-
- /* Convert to tile coordinates, and inclusive ranges:
- */
- ix0 = minx / TILE_SIZE;
- iy0 = miny / TILE_SIZE;
- ix1 = (maxx-1) / TILE_SIZE;
- iy1 = (maxy-1) / TILE_SIZE;
- /*
- * Clamp to framebuffer size
- */
- assert(ix0 == MAX2(ix0, 0));
- assert(iy0 == MAX2(iy0, 0));
- assert(ix1 == MIN2(ix1, scene->tiles_x - 1));
- assert(iy1 == MIN2(iy1, scene->tiles_y - 1));
-
- /* Determine which tile(s) intersect the triangle's bounding box
- */
- if (iy0 == iy1 && ix0 == ix1)
- {
- /* Triangle is contained in a single tile:
- */
- lp_scene_bin_command( scene, ix0, iy0,
- lp_rast_tri_tab[nr_planes],
- lp_rast_arg_triangle(line, (1<<nr_planes)-1) );
- }
- else
+static void lp_setup_line( struct lp_setup_context *setup,
+ const float (*v0)[4],
+ const float (*v1)[4] )
+{
+ if (!try_setup_line( setup, v0, v1 ))
{
- int c[8];
- int ei[8];
- int eo[8];
- int xstep[8];
- int ystep[8];
- int x, y;
- int is_blit = -1; /* undetermined */
-
- for (i = 0; i < nr_planes; i++) {
- c[i] = (line->plane[i].c +
- line->plane[i].dcdy * iy0 * TILE_SIZE -
- line->plane[i].dcdx * ix0 * TILE_SIZE);
-
- ei[i] = line->plane[i].ei << TILE_ORDER;
- eo[i] = line->plane[i].eo << TILE_ORDER;
- xstep[i] = -(line->plane[i].dcdx << TILE_ORDER);
- ystep[i] = line->plane[i].dcdy << TILE_ORDER;
- }
-
-
+ if (!lp_setup_flush_and_restart(setup))
+ return;
- /* Test tile-sized blocks against the triangle.
- * Discard blocks fully outside the tri. If the block is fully
- * contained inside the tri, bin an lp_rast_shade_tile command.
- * Else, bin a lp_rast_triangle command.
- */
- for (y = iy0; y <= iy1; y++)
- {
- boolean in = FALSE; /* are we inside the triangle? */
- int cx[8];
-
- for (i = 0; i < nr_planes; i++)
- cx[i] = c[i];
-
- for (x = ix0; x <= ix1; x++)
- {
- int out = 0;
- int partial = 0;
-
- for (i = 0; i < nr_planes; i++) {
- int planeout = cx[i] + eo[i];
- int planepartial = cx[i] + ei[i] - 1;
- out |= (planeout >> 31);
- partial |= (planepartial >> 31) & (1<<i);
- }
- if (out) {
- /* do nothing */
- if (in)
- break; /* exiting triangle, all done with this row */
- LP_COUNT(nr_empty_64);
- }
- else if (partial) {
- /* Not trivially accepted by at least one plane -
- * rasterize/shade partial tile
- */
- int count = util_bitcount(partial);
- in = TRUE;
- lp_scene_bin_command( scene, x, y,
- lp_rast_tri_tab[count],
- lp_rast_arg_triangle(line, partial) );
-
- LP_COUNT(nr_partially_covered_64);
- }
- else {
- /* triangle covers the whole tile- shade whole tile */
- LP_COUNT(nr_fully_covered_64);
- in = TRUE;
- /* leverages on existing code in lp_setup_tri.c */
- do_triangle_ccw_whole_tile(setup, scene, line, x, y,
- opaque, &is_blit);
- }
-
- /* Iterate cx values across the region:
- */
- for (i = 0; i < nr_planes; i++)
- cx[i] += xstep[i];
- }
-
- /* Iterate c values down the region:
- */
- for (i = 0; i < nr_planes; i++)
- c[i] += ystep[i];
- }
+ if (!try_setup_line( setup, v0, v1 ))
+ return;
}
}
projects / mesa.git / blobdiff