* \author Brian Paul
*/
-#include "sp_setup.h"
-
#include "sp_context.h"
-#include "sp_headers.h"
+#include "sp_prim_setup.h"
#include "sp_quad.h"
+#include "sp_quad_pipe.h"
+#include "sp_setup.h"
#include "sp_state.h"
-#include "sp_prim_setup.h"
+#include "draw/draw_context.h"
#include "draw/draw_private.h"
#include "draw/draw_vertex.h"
-#include "pipe/p_util.h"
#include "pipe/p_shader_tokens.h"
+#include "pipe/p_thread.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
#define DEBUG_VERTS 0
#define DEBUG_FRAGS 0
int lines; /**< number of lines on this edge */
};
+#if SP_NUM_QUAD_THREADS > 1
+
+/* Set to 1 if you want other threads to be instantly
+ * notified of pending jobs.
+ */
+#define INSTANT_NOTEMPTY_NOTIFY 0
+
+struct thread_info
+{
+ struct setup_context *setup;
+ uint id;
+ pipe_thread handle;
+};
+
+struct quad_job;
+
+typedef void (* quad_job_routine)( struct setup_context *setup, uint thread, struct quad_job *job );
+
+struct quad_job
+{
+ struct quad_header_input input;
+ struct quad_header_inout inout;
+ quad_job_routine routine;
+};
+
+#define NUM_QUAD_JOBS 64
+
+struct quad_job_que
+{
+ struct quad_job jobs[NUM_QUAD_JOBS];
+ uint first;
+ uint last;
+ pipe_mutex que_mutex;
+ pipe_condvar que_notfull_condvar;
+ pipe_condvar que_notempty_condvar;
+ uint jobs_added;
+ uint jobs_done;
+ pipe_condvar que_done_condvar;
+};
+
+static void
+add_quad_job( struct quad_job_que *que, struct quad_header *quad, quad_job_routine routine )
+{
+#if INSTANT_NOTEMPTY_NOTIFY
+ boolean empty;
+#endif
+
+ /* Wait for empty slot, see if the que is empty.
+ */
+ pipe_mutex_lock( que->que_mutex );
+ while ((que->last + 1) % NUM_QUAD_JOBS == que->first) {
+#if !INSTANT_NOTEMPTY_NOTIFY
+ pipe_condvar_broadcast( que->que_notempty_condvar );
+#endif
+ pipe_condvar_wait( que->que_notfull_condvar, que->que_mutex );
+ }
+#if INSTANT_NOTEMPTY_NOTIFY
+ empty = que->last == que->first;
+#endif
+ que->jobs_added++;
+ pipe_mutex_unlock( que->que_mutex );
+
+ /* Submit new job.
+ */
+ que->jobs[que->last].input = quad->input;
+ que->jobs[que->last].inout = quad->inout;
+ que->jobs[que->last].routine = routine;
+ que->last = (que->last + 1) % NUM_QUAD_JOBS;
+
+#if INSTANT_NOTEMPTY_NOTIFY
+ /* If the que was empty, notify consumers there's a job to be done.
+ */
+ if (empty) {
+ pipe_mutex_lock( que->que_mutex );
+ pipe_condvar_broadcast( que->que_notempty_condvar );
+ pipe_mutex_unlock( que->que_mutex );
+ }
+#endif
+}
+
+#endif
/**
* Triangle setup info (derived from draw_stage).
struct tgsi_interp_coef posCoef; /* For Z, W */
struct quad_header quad;
+#if SP_NUM_QUAD_THREADS > 1
+ struct quad_job_que que;
+ struct thread_info threads[SP_NUM_QUAD_THREADS];
+#endif
+
struct {
int left[2]; /**< [0] = row0, [1] = row1 */
int right[2];
unsigned winding; /* which winding to cull */
};
+#if SP_NUM_QUAD_THREADS > 1
+
+static PIPE_THREAD_ROUTINE( quad_thread, param )
+{
+ struct thread_info *info = (struct thread_info *) param;
+ struct quad_job_que *que = &info->setup->que;
+ for (;;) {
+ struct quad_job job;
+ boolean full;
+ /* Wait for an available job.
+ */
+ pipe_mutex_lock( que->que_mutex );
+ while (que->last == que->first)
+ pipe_condvar_wait( que->que_notempty_condvar, que->que_mutex );
+
+ /* See if the que is full.
+ */
+ full = (que->last + 1) % NUM_QUAD_JOBS == que->first;
+ /* Take a job and remove it from que.
+ */
+ job = que->jobs[que->first];
+ que->first = (que->first + 1) % NUM_QUAD_JOBS;
+
+ /* Notify the producer if the que is not full.
+ */
+ if (full)
+ pipe_condvar_signal( que->que_notfull_condvar );
+ pipe_mutex_unlock( que->que_mutex );
+
+ job.routine( info->setup, info->id, &job );
+
+ /* Notify the producer if that's the last finished job.
+ */
+ pipe_mutex_lock( que->que_mutex );
+ que->jobs_done++;
+ if (que->jobs_added == que->jobs_done)
+ pipe_condvar_signal( que->que_done_condvar );
+ pipe_mutex_unlock( que->que_mutex );
+ }
+
+ return NULL;
+}
+
+#define WAIT_FOR_COMPLETION(setup) \
+ do {\
+ pipe_mutex_lock( setup->que.que_mutex );\
+ if (!INSTANT_NOTEMPTY_NOTIFY)\
+ pipe_condvar_broadcast( setup->que.que_notempty_condvar );\
+ while (setup->que.jobs_added != setup->que.jobs_done)\
+ pipe_condvar_wait( setup->que.que_done_condvar, setup->que.que_mutex );\
+ pipe_mutex_unlock( setup->que.que_mutex );\
+ } while (0)
+
+#else
+
+#define WAIT_FOR_COMPLETION(setup) ((void) 0)
+
+#endif
-static boolean cull_tri( struct setup_context *setup,
- float det )
+
+
+/**
+ * Do triangle cull test using tri determinant (sign indicates orientation)
+ * \return true if triangle is to be culled.
+ */
+static INLINE boolean
+cull_tri(const struct setup_context *setup, float det)
{
- if (det != 0)
- {
+ if (det != 0) {
/* if (det < 0 then Z points toward camera and triangle is
* counter-clockwise winding.
*/
unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW;
-
- if ((winding & setup->winding) == 0)
+
+ if ((winding & setup->winding) == 0)
return FALSE;
}
* Clip setup->quad against the scissor/surface bounds.
*/
static INLINE void
-quad_clip(struct setup_context *setup)
+quad_clip( struct setup_context *setup, struct quad_header *quad )
{
const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect;
const int minx = (int) cliprect->minx;
const int miny = (int) cliprect->miny;
const int maxy = (int) cliprect->maxy;
- if (setup->quad.x0 >= maxx ||
- setup->quad.y0 >= maxy ||
- setup->quad.x0 + 1 < minx ||
- setup->quad.y0 + 1 < miny) {
+ if (quad->input.x0 >= maxx ||
+ quad->input.y0 >= maxy ||
+ quad->input.x0 + 1 < minx ||
+ quad->input.y0 + 1 < miny) {
/* totally clipped */
- setup->quad.mask = 0x0;
+ quad->inout.mask = 0x0;
return;
}
- if (setup->quad.x0 < minx)
- setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
- if (setup->quad.y0 < miny)
- setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
- if (setup->quad.x0 == maxx - 1)
- setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
- if (setup->quad.y0 == maxy - 1)
- setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+ if (quad->input.x0 < minx)
+ quad->inout.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+ if (quad->input.y0 < miny)
+ quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+ if (quad->input.x0 == maxx - 1)
+ quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+ if (quad->input.y0 == maxy - 1)
+ quad->inout.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
}
* Emit a quad (pass to next stage) with clipping.
*/
static INLINE void
-clip_emit_quad(struct setup_context *setup)
+clip_emit_quad( struct setup_context *setup, struct quad_header *quad, uint thread )
{
- quad_clip(setup);
- if (setup->quad.mask) {
+ quad_clip( setup, quad );
+ if (quad->inout.mask) {
struct softpipe_context *sp = setup->softpipe;
- sp->quad.first->run(sp->quad.first, &setup->quad);
+
+ sp->quad[thread].first->run( sp->quad[thread].first, quad );
}
}
+#if SP_NUM_QUAD_THREADS > 1
+
+static void
+clip_emit_quad_job( struct setup_context *setup, uint thread, struct quad_job *job )
+{
+ struct quad_header quad;
+
+ quad.input = job->input;
+ quad.inout = job->inout;
+ quad.coef = setup->quad.coef;
+ quad.posCoef = setup->quad.posCoef;
+ quad.nr_attrs = setup->quad.nr_attrs;
+ clip_emit_quad( setup, &quad, thread );
+}
+
+#define CLIP_EMIT_QUAD(setup) add_quad_job( &setup->que, &setup->quad, clip_emit_quad_job )
+
+#else
+
+#define CLIP_EMIT_QUAD(setup) clip_emit_quad( setup, &setup->quad, 0 )
+
+#endif
/**
* Emit a quad (pass to next stage). No clipping is done.
*/
static INLINE void
-emit_quad( struct setup_context *setup, int x, int y, unsigned mask )
+emit_quad( struct setup_context *setup, struct quad_header *quad, uint thread )
{
struct softpipe_context *sp = setup->softpipe;
- setup->quad.x0 = x;
- setup->quad.y0 = y;
- setup->quad.mask = mask;
+#if DEBUG_FRAGS
+ uint mask = quad->inout.mask;
+#endif
+
#if DEBUG_FRAGS
if (mask & 1) setup->numFragsEmitted++;
if (mask & 2) setup->numFragsEmitted++;
if (mask & 4) setup->numFragsEmitted++;
if (mask & 8) setup->numFragsEmitted++;
#endif
- sp->quad.first->run(sp->quad.first, &setup->quad);
+ sp->quad[thread].first->run( sp->quad[thread].first, quad );
#if DEBUG_FRAGS
- mask = setup->quad.mask;
+ mask = quad->inout.mask;
if (mask & 1) setup->numFragsWritten++;
if (mask & 2) setup->numFragsWritten++;
if (mask & 4) setup->numFragsWritten++;
#endif
}
+#if SP_NUM_QUAD_THREADS > 1
+
+static void
+emit_quad_job( struct setup_context *setup, uint thread, struct quad_job *job )
+{
+ struct quad_header quad;
+
+ quad.input = job->input;
+ quad.inout = job->inout;
+ quad.coef = setup->quad.coef;
+ quad.posCoef = setup->quad.posCoef;
+ quad.nr_attrs = setup->quad.nr_attrs;
+ emit_quad( setup, &quad, thread );
+}
+
+#define EMIT_QUAD(setup,x,y,mask) do {\
+ setup->quad.input.x0 = x;\
+ setup->quad.input.y0 = y;\
+ setup->quad.inout.mask = mask;\
+ add_quad_job( &setup->que, &setup->quad, emit_quad_job );\
+ } while (0)
+
+#else
+
+#define EMIT_QUAD(setup,x,y,mask) do {\
+ setup->quad.input.x0 = x;\
+ setup->quad.input.y0 = y;\
+ setup->quad.inout.mask = mask;\
+ emit_quad( setup, &setup->quad, 0 );\
+ } while (0)
+
+#endif
/**
* Given an X or Y coordinate, return the block/quad coordinate that it
mask |= MASK_TOP_RIGHT;
if (x+1 >= xleft1 && x+1 < xright1)
mask |= MASK_BOTTOM_RIGHT;
- emit_quad( setup, x, setup->span.y, mask );
+ EMIT_QUAD( setup, x, setup->span.y, mask );
}
break;
mask |= MASK_TOP_LEFT;
if (x+1 >= xleft0 && x+1 < xright0)
mask |= MASK_TOP_RIGHT;
- emit_quad( setup, x, setup->span.y, mask );
+ EMIT_QUAD( setup, x, setup->span.y, mask );
}
break;
mask |= MASK_BOTTOM_LEFT;
if (x+1 >= xleft1 && x+1 < xright1)
mask |= MASK_BOTTOM_RIGHT;
- emit_quad( setup, x, setup->span.y, mask );
+ EMIT_QUAD( setup, x, setup->span.y, mask );
}
break;
const float (*v)[4])
{
int i;
- debug_printf("Vertex: (%p)\n", v);
+ debug_printf(" Vertex: (%p)\n", v);
for (i = 0; i < setup->quad.nr_attrs; i++) {
- debug_printf(" %d: %f %f %f %f\n", i,
+ debug_printf(" %d: %f %f %f %f\n", i,
v[i][0], v[i][1], v[i][2], v[i][3]);
+ if (util_is_inf_or_nan(v[i][0])) {
+ debug_printf(" NaN!\n");
+ }
}
}
#endif
+/**
+ * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
+ */
static boolean setup_sort_vertices( struct setup_context *setup,
float det,
const float (*v0)[4],
setup->ebot.dx * setup->emaj.dy);
setup->oneoverarea = 1.0f / area;
+
/*
debug_printf("%s one-over-area %f area %f det %f\n",
__FUNCTION__, setup->oneoverarea, area, det );
*/
+ if (util_is_inf_or_nan(setup->oneoverarea))
+ return FALSE;
}
/* We need to know if this is a front or back-facing triangle for:
* - the GLSL gl_FrontFacing fragment attribute (bool)
* - two-sided stencil test
*/
- setup->quad.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);
+ setup->quad.input.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);
return TRUE;
}
setup->coef[slot].dadx[0] = 1.0;
setup->coef[slot].dady[0] = 0.0;
/*Y*/
- if (setup->softpipe->rasterizer->origin_lower_left) {
- /* y=0=bottom */
- const int winHeight = setup->softpipe->framebuffer.height;
- setup->coef[slot].a0[1] = (float) (winHeight - 1);
- setup->coef[slot].dady[1] = -1.0;
- }
- else {
- /* y=0=top */
- setup->coef[slot].a0[1] = 0.0;
- setup->coef[slot].dady[1] = 1.0;
- }
+ setup->coef[slot].a0[1] = 0.0;
setup->coef[slot].dadx[1] = 0.0;
+ setup->coef[slot].dady[1] = 1.0;
/*Z*/
setup->coef[slot].a0[2] = setup->posCoef.a0[2];
setup->coef[slot].dadx[2] = setup->posCoef.dadx[2];
/* setup interpolation for all the remaining attributes:
*/
for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->src_index[fragSlot];
+ const uint vertSlot = vinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->interp_mode[fragSlot]) {
+ switch (vinfo->attrib[fragSlot].interp_mode) {
case INTERP_CONSTANT:
for (j = 0; j < NUM_CHANNELS; j++)
const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
/* FOG.y = front/back facing XXX fix this */
- setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
setup->coef[fragSlot].dadx[1] = 0.0;
setup->coef[fragSlot].dady[1] = 0.0;
}
float vmid_y = setup->vmid[0][1] - 0.5f;
float vmax_y = setup->vmax[0][1] - 0.5f;
- setup->emaj.sy = CEILF(vmin_y);
- setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy);
+ setup->emaj.sy = ceilf(vmin_y);
+ setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy);
setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy;
setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy;
- setup->etop.sy = CEILF(vmid_y);
- setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy);
+ setup->etop.sy = ceilf(vmid_y);
+ setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy);
setup->etop.dxdy = setup->etop.dx / setup->etop.dy;
setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy;
- setup->ebot.sy = CEILF(vmin_y);
- setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy);
+ setup->ebot.sy = ceilf(vmin_y);
+ setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy);
setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy;
setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy;
}
if (cull_tri( setup, det ))
return;
- setup_sort_vertices( setup, det, v0, v1, v2 );
+ if (!setup_sort_vertices( setup, det, v0, v1, v2 ))
+ return;
setup_tri_coefficients( setup );
setup_tri_edges( setup );
- setup->quad.prim = PRIM_TRI;
+ setup->quad.input.prim = QUAD_PRIM_TRI;
setup->span.y = 0;
setup->span.y_flags = 0;
flush_spans( setup );
+ WAIT_FOR_COMPLETION(setup);
+
#if DEBUG_FRAGS
printf("Tri: %u frags emitted, %u written\n",
setup->numFragsEmitted,
* for a line.
*/
static void
-line_linear_coeff(struct setup_context *setup,
+line_linear_coeff(const struct setup_context *setup,
struct tgsi_interp_coef *coef,
uint vertSlot, uint i)
{
* for a line.
*/
static void
-line_persp_coeff(struct setup_context *setup,
- struct tgsi_interp_coef *coef,
- uint vertSlot, uint i)
+line_persp_coeff(const struct setup_context *setup,
+ struct tgsi_interp_coef *coef,
+ uint vertSlot, uint i)
{
/* XXX double-check/verify this arithmetic */
const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3];
* Compute the setup->coef[] array dadx, dady, a0 values.
* Must be called after setup->vmin,vmax are initialized.
*/
-static INLINE void
+static INLINE boolean
setup_line_coefficients(struct setup_context *setup,
const float (*v0)[4],
const float (*v1)[4])
const struct sp_fragment_shader *spfs = softpipe->fs;
const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
uint fragSlot;
+ float area;
/* use setup->vmin, vmax to point to vertices */
setup->vprovoke = v1;
setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
- /* NOTE: this is not really 1/area */
- setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx +
- setup->emaj.dy * setup->emaj.dy);
+
+ /* NOTE: this is not really area but something proportional to it */
+ area = setup->emaj.dx * setup->emaj.dx + setup->emaj.dy * setup->emaj.dy;
+ if (area == 0.0f || util_is_inf_or_nan(area))
+ return FALSE;
+ setup->oneoverarea = 1.0f / area;
/* z and w are done by linear interpolation:
*/
/* setup interpolation for all the remaining attributes:
*/
for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->src_index[fragSlot];
+ const uint vertSlot = vinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->interp_mode[fragSlot]) {
+ switch (vinfo->attrib[fragSlot].interp_mode) {
case INTERP_CONSTANT:
for (j = 0; j < NUM_CHANNELS; j++)
const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
/* FOG.y = front/back facing XXX fix this */
- setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
setup->coef[fragSlot].dadx[1] = 0.0;
setup->coef[fragSlot].dady[1] = 0.0;
}
}
+ return TRUE;
}
const int quadY = y - iy;
const int mask = (1 << ix) << (2 * iy);
- if (quadX != setup->quad.x0 ||
- quadY != setup->quad.y0)
+ if (quadX != setup->quad.input.x0 ||
+ quadY != setup->quad.input.y0)
{
/* flush prev quad, start new quad */
- if (setup->quad.x0 != -1)
- clip_emit_quad(setup);
+ if (setup->quad.input.x0 != -1)
+ CLIP_EMIT_QUAD(setup);
- setup->quad.x0 = quadX;
- setup->quad.y0 = quadY;
- setup->quad.mask = 0x0;
+ setup->quad.input.x0 = quadX;
+ setup->quad.input.y0 = quadY;
+ setup->quad.inout.mask = 0x0;
}
- setup->quad.mask |= mask;
+ setup->quad.inout.mask |= mask;
}
if (dx == 0 && dy == 0)
return;
- setup_line_coefficients(setup, v0, v1);
+ if (!setup_line_coefficients(setup, v0, v1))
+ return;
+
+ assert(v0[0][0] < 1.0e9);
+ assert(v0[0][1] < 1.0e9);
+ assert(v1[0][0] < 1.0e9);
+ assert(v1[0][1] < 1.0e9);
if (dx < 0) {
dx = -dx; /* make positive */
assert(dx >= 0);
assert(dy >= 0);
- setup->quad.x0 = setup->quad.y0 = -1;
- setup->quad.mask = 0x0;
- setup->quad.prim = PRIM_LINE;
+ setup->quad.input.x0 = setup->quad.input.y0 = -1;
+ setup->quad.inout.mask = 0x0;
+ setup->quad.input.prim = QUAD_PRIM_LINE;
/* XXX temporary: set coverage to 1.0 so the line appears
* if AA mode happens to be enabled.
*/
- setup->quad.coverage[0] =
- setup->quad.coverage[1] =
- setup->quad.coverage[2] =
- setup->quad.coverage[3] = 1.0;
+ setup->quad.input.coverage[0] =
+ setup->quad.input.coverage[1] =
+ setup->quad.input.coverage[2] =
+ setup->quad.input.coverage[3] = 1.0;
if (dx > dy) {
/*** X-major line ***/
}
/* draw final quad */
- if (setup->quad.mask) {
- clip_emit_quad(setup);
+ if (setup->quad.inout.mask) {
+ CLIP_EMIT_QUAD(setup);
}
+
+ WAIT_FOR_COMPLETION(setup);
}
static void
-point_persp_coeff(struct setup_context *setup,
+point_persp_coeff(const struct setup_context *setup,
const float (*vert)[4],
struct tgsi_interp_coef *coef,
uint vertSlot, uint i)
const_coeff(setup, &setup->posCoef, 0, 3);
for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
- const uint vertSlot = vinfo->src_index[fragSlot];
+ const uint vertSlot = vinfo->attrib[fragSlot].src_index;
uint j;
- switch (vinfo->interp_mode[fragSlot]) {
+ switch (vinfo->attrib[fragSlot].interp_mode) {
case INTERP_CONSTANT:
/* fall-through */
case INTERP_LINEAR:
if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
/* FOG.y = front/back facing XXX fix this */
- setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
setup->coef[fragSlot].dadx[1] = 0.0;
setup->coef[fragSlot].dady[1] = 0.0;
}
}
- setup->quad.prim = PRIM_POINT;
+ setup->quad.input.prim = QUAD_PRIM_POINT;
if (halfSize <= 0.5 && !round) {
/* special case for 1-pixel points */
const int ix = ((int) x) & 1;
const int iy = ((int) y) & 1;
- setup->quad.x0 = (int) x - ix;
- setup->quad.y0 = (int) y - iy;
- setup->quad.mask = (1 << ix) << (2 * iy);
- clip_emit_quad(setup);
+ setup->quad.input.x0 = (int) x - ix;
+ setup->quad.input.y0 = (int) y - iy;
+ setup->quad.inout.mask = (1 << ix) << (2 * iy);
+ CLIP_EMIT_QUAD(setup);
}
else {
if (round) {
for (ix = ixmin; ix <= ixmax; ix += 2) {
float dx, dy, dist2, cover;
- setup->quad.mask = 0x0;
+ setup->quad.inout.mask = 0x0;
dx = (ix + 0.5f) - x;
dy = (iy + 0.5f) - y;
dist2 = dx * dx + dy * dy;
if (dist2 <= rmax2) {
cover = 1.0F - (dist2 - rmin2) * cscale;
- setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
- setup->quad.mask |= MASK_TOP_LEFT;
+ setup->quad.input.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
+ setup->quad.inout.mask |= MASK_TOP_LEFT;
}
dx = (ix + 1.5f) - x;
dist2 = dx * dx + dy * dy;
if (dist2 <= rmax2) {
cover = 1.0F - (dist2 - rmin2) * cscale;
- setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
- setup->quad.mask |= MASK_TOP_RIGHT;
+ setup->quad.input.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
+ setup->quad.inout.mask |= MASK_TOP_RIGHT;
}
dx = (ix + 0.5f) - x;
dist2 = dx * dx + dy * dy;
if (dist2 <= rmax2) {
cover = 1.0F - (dist2 - rmin2) * cscale;
- setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
- setup->quad.mask |= MASK_BOTTOM_LEFT;
+ setup->quad.input.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
+ setup->quad.inout.mask |= MASK_BOTTOM_LEFT;
}
dx = (ix + 1.5f) - x;
dist2 = dx * dx + dy * dy;
if (dist2 <= rmax2) {
cover = 1.0F - (dist2 - rmin2) * cscale;
- setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
- setup->quad.mask |= MASK_BOTTOM_RIGHT;
+ setup->quad.input.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
+ setup->quad.inout.mask |= MASK_BOTTOM_RIGHT;
}
- if (setup->quad.mask) {
- setup->quad.x0 = ix;
- setup->quad.y0 = iy;
- clip_emit_quad(setup);
+ if (setup->quad.inout.mask) {
+ setup->quad.input.x0 = ix;
+ setup->quad.input.y0 = iy;
+ CLIP_EMIT_QUAD(setup);
}
}
}
mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
}
- setup->quad.mask = mask;
- setup->quad.x0 = ix;
- setup->quad.y0 = iy;
- clip_emit_quad(setup);
+ setup->quad.inout.mask = mask;
+ setup->quad.input.x0 = ix;
+ setup->quad.input.y0 = iy;
+ CLIP_EMIT_QUAD(setup);
}
}
}
}
+
+ WAIT_FOR_COMPLETION(setup);
}
void setup_prepare( struct setup_context *setup )
softpipe_update_derived(sp);
}
- /* Mark surfaces as defined now */
- for (i = 0; i < sp->framebuffer.num_cbufs; i++){
- if (sp->framebuffer.cbufs[i]) {
- sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED;
- }
- }
- if (sp->framebuffer.zsbuf) {
- sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED;
- }
+ /* Note: nr_attrs is only used for debugging (vertex printing) */
+ setup->quad.nr_attrs = draw_num_vs_outputs(sp->draw);
- {
- const struct sp_fragment_shader *fs = setup->softpipe->fs;
- setup->quad.nr_attrs = fs->info.num_inputs;
- sp->quad.first->begin(sp->quad.first);
+ for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+ sp->quad[i].first->begin( sp->quad[i].first );
}
if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES &&
struct setup_context *setup_create_context( struct softpipe_context *softpipe )
{
struct setup_context *setup = CALLOC_STRUCT(setup_context);
+#if SP_NUM_QUAD_THREADS > 1
+ uint i;
+#endif
setup->softpipe = softpipe;
setup->quad.coef = setup->coef;
setup->quad.posCoef = &setup->posCoef;
+#if SP_NUM_QUAD_THREADS > 1
+ setup->que.first = 0;
+ setup->que.last = 0;
+ pipe_mutex_init( setup->que.que_mutex );
+ pipe_condvar_init( setup->que.que_notfull_condvar );
+ pipe_condvar_init( setup->que.que_notempty_condvar );
+ setup->que.jobs_added = 0;
+ setup->que.jobs_done = 0;
+ pipe_condvar_init( setup->que.que_done_condvar );
+ for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+ setup->threads[i].setup = setup;
+ setup->threads[i].id = i;
+ setup->threads[i].handle = pipe_thread_create( quad_thread, &setup->threads[i] );
+ }
+#endif
+
return setup;
}
+
projects / mesa.git / blobdiff