This repository has been archived on 2023-10-09. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
blender-archive/source/blender/render/intern/source/envmap.c
Sergey Sharybin 5c6f6301b0 Image thread safe improvements
This commit makes BKE_image_acquire_ibuf referencing result, which means once
some area requested for image buffer, it'll be guaranteed this buffer wouldn't
be freed by image signal.

To de-reference buffer BKE_image_release_ibuf should now always be used.

To make referencing working correct we can not rely on result of
image_get_ibuf_threadsafe called outside from thread lock. This is so because
we need to guarantee getting image buffer from list of loaded buffers and it's
referencing happens atomic. Without lock here it is possible that between call
of image_get_ibuf_threadsafe and referencing the buffer IMA_SIGNAL_FREE would
be called. Image signal handling too is blocking now to prevent such a
situation.

Threads are locking by spinlock, which are faster than mutexes. There were some
slowdown reports in the past about render slowdown when using OSX on Xeon CPU.
It shouldn't happen with spin locks, but more tests on different hardware would
be really welcome. So far can not see speed regressions on own computers.

This commit also removes BKE_image_get_ibuf, because it was not so intuitive
when get_ibuf and acquire_ibuf should be used.

Thanks to Ton and Brecht for discussion/review :)
2012-11-15 15:59:58 +00:00

774 lines
19 KiB
C

/*
* ***** BEGIN GPL LICENSE BLOCK *****
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*
* The Original Code is Copyright (C) 2001-2002 by NaN Holding BV.
* All rights reserved.
*
* Contributors: 2004/2005/2006 Blender Foundation, full recode
*
* ***** END GPL LICENSE BLOCK *****
*/
/** \file blender/render/intern/source/envmap.c
* \ingroup render
*/
#include <math.h>
#include <string.h>
/* external modules: */
#include "MEM_guardedalloc.h"
#include "BLI_math.h"
#include "BLI_blenlib.h"
#include "BLI_threads.h"
#include "BLI_utildefines.h"
#include "IMB_imbuf_types.h"
#include "IMB_imbuf.h" /* for rectcpy */
#include "DNA_group_types.h"
#include "DNA_image_types.h"
#include "DNA_object_types.h"
#include "DNA_scene_types.h"
#include "DNA_texture_types.h"
#include "BKE_library.h"
#include "BKE_main.h"
#include "BKE_image.h" /* BKE_imbuf_write */
#include "BKE_texture.h"
/* this module */
#include "render_types.h"
#include "renderpipeline.h"
#include "envmap.h"
#include "rendercore.h"
#include "renderdatabase.h"
#include "texture.h"
#include "zbuf.h"
#include "initrender.h"
/* ------------------------------------------------------------------------- */
static void envmap_split_ima(EnvMap *env, ImBuf *ibuf)
{
int dx, part;
/* after lock we test cube[1], if set the other thread has done it fine */
BLI_lock_thread(LOCK_IMAGE);
if (env->cube[1] == NULL) {
BKE_free_envmapdata(env);
dx = ibuf->y;
dx /= 2;
if (3 * dx == ibuf->x) {
env->type = ENV_CUBE;
env->ok = ENV_OSA;
}
else if (ibuf->x == ibuf->y) {
env->type = ENV_PLANE;
env->ok = ENV_OSA;
}
else {
printf("Incorrect envmap size\n");
env->ok = 0;
env->ima->ok = 0;
}
if (env->ok) {
if (env->type == ENV_CUBE) {
for (part = 0; part < 6; part++) {
env->cube[part] = IMB_allocImBuf(dx, dx, 24, IB_rect | IB_rectfloat);
}
IMB_float_from_rect(ibuf);
IMB_rectcpy(env->cube[0], ibuf,
0, 0, 0, 0, dx, dx);
IMB_rectcpy(env->cube[1], ibuf,
0, 0, dx, 0, dx, dx);
IMB_rectcpy(env->cube[2], ibuf,
0, 0, 2 * dx, 0, dx, dx);
IMB_rectcpy(env->cube[3], ibuf,
0, 0, 0, dx, dx, dx);
IMB_rectcpy(env->cube[4], ibuf,
0, 0, dx, dx, dx, dx);
IMB_rectcpy(env->cube[5], ibuf,
0, 0, 2 * dx, dx, dx, dx);
}
else { /* ENV_PLANE */
env->cube[1] = IMB_dupImBuf(ibuf);
IMB_float_from_rect(env->cube[1]);
}
}
}
BLI_unlock_thread(LOCK_IMAGE);
}
/* ------------------------------------------------------------------------- */
/* ****************** RENDER ********************** */
/* copy current render */
static Render *envmap_render_copy(Render *re, EnvMap *env)
{
Render *envre;
float viewscale;
int cuberes;
envre = RE_NewRender("Envmap");
env->lastsize = re->r.size;
cuberes = (env->cuberes * re->r.size) / 100;
cuberes &= 0xFFFC;
/* this flag has R_ZTRA in it for example */
envre->flag = re->flag;
/* set up renderdata */
envre->r = re->r;
envre->r.mode &= ~(R_BORDER | R_PANORAMA | R_ORTHO | R_MBLUR);
envre->r.layers.first = envre->r.layers.last = NULL;
envre->r.filtertype = 0;
envre->r.tilex = envre->r.xsch / 2;
envre->r.tiley = envre->r.ysch / 2;
envre->r.size = 100;
envre->r.yasp = envre->r.xasp = 1;
RE_InitState(envre, NULL, &envre->r, NULL, cuberes, cuberes, NULL);
envre->scene = re->scene; /* unsure about this... */
envre->scene_color_manage = re->scene_color_manage;
envre->lay = re->lay;
/* view stuff in env render */
viewscale = (env->type == ENV_PLANE) ? env->viewscale : 1.0f;
RE_SetEnvmapCamera(envre, env->object, viewscale, env->clipsta, env->clipend);
/* callbacks */
envre->display_draw = re->display_draw;
envre->ddh = re->ddh;
envre->test_break = re->test_break;
envre->tbh = re->tbh;
/* and for the evil stuff; copy the database... */
envre->totvlak = re->totvlak;
envre->totvert = re->totvert;
envre->tothalo = re->tothalo;
envre->totstrand = re->totstrand;
envre->totlamp = re->totlamp;
envre->sortedhalos = re->sortedhalos;
envre->lights = re->lights;
envre->objecttable = re->objecttable;
envre->customdata_names = re->customdata_names;
envre->raytree = re->raytree;
envre->totinstance = re->totinstance;
envre->instancetable = re->instancetable;
envre->objectinstance = re->objectinstance;
envre->qmcsamplers = re->qmcsamplers;
return envre;
}
static void envmap_free_render_copy(Render *envre)
{
envre->totvlak = 0;
envre->totvert = 0;
envre->tothalo = 0;
envre->totstrand = 0;
envre->totlamp = 0;
envre->totinstance = 0;
envre->sortedhalos = NULL;
envre->lights.first = envre->lights.last = NULL;
envre->objecttable.first = envre->objecttable.last = NULL;
envre->customdata_names.first = envre->customdata_names.last = NULL;
envre->raytree = NULL;
envre->instancetable.first = envre->instancetable.last = NULL;
envre->objectinstance = NULL;
envre->qmcsamplers = NULL;
RE_FreeRender(envre);
}
/* ------------------------------------------------------------------------- */
static void envmap_transmatrix(float mat[][4], int part)
{
float tmat[4][4], eul[3], rotmat[4][4];
eul[0] = eul[1] = eul[2] = 0.0;
if (part == 0) { /* neg z */
/* pass */
}
else if (part == 1) { /* pos z */
eul[0] = M_PI;
}
else if (part == 2) { /* pos y */
eul[0] = M_PI / 2.0;
}
else if (part == 3) { /* neg x */
eul[0] = M_PI / 2.0;
eul[2] = M_PI / 2.0;
}
else if (part == 4) { /* neg y */
eul[0] = M_PI / 2.0;
eul[2] = M_PI;
}
else { /* pos x */
eul[0] = M_PI / 2.0;
eul[2] = -M_PI / 2.0;
}
copy_m4_m4(tmat, mat);
eul_to_mat4(rotmat, eul);
mul_serie_m4(mat, tmat, rotmat,
NULL, NULL, NULL,
NULL, NULL, NULL);
}
/* ------------------------------------------------------------------------- */
static void env_rotate_scene(Render *re, float mat[][4], int mode)
{
GroupObject *go;
ObjectRen *obr;
ObjectInstanceRen *obi;
LampRen *lar = NULL;
HaloRen *har = NULL;
float imat[3][3], pmat[4][4], smat[4][4], tmat[4][4], cmat[3][3], tmpmat[4][4];
int a;
if (mode == 0) {
invert_m4_m4(tmat, mat);
copy_m3_m4(imat, tmat);
}
else {
copy_m4_m4(tmat, mat);
copy_m3_m4(imat, mat);
}
for (obi = re->instancetable.first; obi; obi = obi->next) {
/* append or set matrix depending on dupli */
if (obi->flag & R_DUPLI_TRANSFORMED) {
copy_m4_m4(tmpmat, obi->mat);
mult_m4_m4m4(obi->mat, tmat, tmpmat);
}
else if (mode == 1)
copy_m4_m4(obi->mat, tmat);
else
unit_m4(obi->mat);
copy_m3_m4(cmat, obi->mat);
invert_m3_m3(obi->nmat, cmat);
transpose_m3(obi->nmat);
/* indicate the renderer has to use transform matrices */
if (mode == 0)
obi->flag &= ~R_ENV_TRANSFORMED;
else
obi->flag |= R_ENV_TRANSFORMED;
}
for (obr = re->objecttable.first; obr; obr = obr->next) {
for (a = 0; a < obr->tothalo; a++) {
if ((a & 255) == 0) har = obr->bloha[a >> 8];
else har++;
mul_m4_v3(tmat, har->co);
}
}
for (go = re->lights.first; go; go = go->next) {
lar = go->lampren;
/* removed here some horrible code of someone in NaN who tried to fix
* prototypes... just solved by introducing a correct cmat[3][3] instead
* of using smat. this works, check square spots in reflections (ton) */
copy_m3_m3(cmat, lar->imat);
mul_m3_m3m3(lar->imat, cmat, imat);
mul_m3_v3(imat, lar->vec);
mul_m4_v3(tmat, lar->co);
lar->sh_invcampos[0] = -lar->co[0];
lar->sh_invcampos[1] = -lar->co[1];
lar->sh_invcampos[2] = -lar->co[2];
mul_m3_v3(lar->imat, lar->sh_invcampos);
lar->sh_invcampos[2] *= lar->sh_zfac;
if (lar->shb) {
if (mode == 1) {
invert_m4_m4(pmat, mat);
mult_m4_m4m4(smat, lar->shb->viewmat, pmat);
mult_m4_m4m4(lar->shb->persmat, lar->shb->winmat, smat);
}
else mult_m4_m4m4(lar->shb->persmat, lar->shb->winmat, lar->shb->viewmat);
}
}
}
/* ------------------------------------------------------------------------- */
static void env_layerflags(Render *re, unsigned int notlay)
{
ObjectRen *obr;
VlakRen *vlr = NULL;
int a;
/* invert notlay, so if face is in multiple layers it will still be visible,
* unless all 'notlay' bits match the face bits.
* face: 0110
* not: 0100
* ~not: 1011
* now (face & ~not) is true
*/
notlay = ~notlay;
for (obr = re->objecttable.first; obr; obr = obr->next) {
if ((obr->lay & notlay) == 0) {
for (a = 0; a < obr->totvlak; a++) {
if ((a & 255) == 0) vlr = obr->vlaknodes[a >> 8].vlak;
else vlr++;
vlr->flag |= R_HIDDEN;
}
}
}
}
static void env_hideobject(Render *re, Object *ob)
{
ObjectRen *obr;
VlakRen *vlr = NULL;
int a;
for (obr = re->objecttable.first; obr; obr = obr->next) {
for (a = 0; a < obr->totvlak; a++) {
if ((a & 255) == 0) vlr = obr->vlaknodes[a >> 8].vlak;
else vlr++;
if (obr->ob == ob)
vlr->flag |= R_HIDDEN;
}
}
}
static void env_showobjects(Render *re)
{
ObjectRen *obr;
VlakRen *vlr = NULL;
int a;
for (obr = re->objecttable.first; obr; obr = obr->next) {
for (a = 0; a < obr->totvlak; a++) {
if ((a & 255) == 0) vlr = obr->vlaknodes[a >> 8].vlak;
else vlr++;
vlr->flag &= ~R_HIDDEN;
}
}
}
/* ------------------------------------------------------------------------- */
static void env_set_imats(Render *re)
{
Base *base;
float mat[4][4];
base = re->scene->base.first;
while (base) {
mult_m4_m4m4(mat, re->viewmat, base->object->obmat);
invert_m4_m4(base->object->imat, mat);
base = base->next;
}
}
/* ------------------------------------------------------------------------- */
static void render_envmap(Render *re, EnvMap *env)
{
/* only the cubemap and planar map is implemented */
Render *envre;
ImBuf *ibuf;
float orthmat[4][4];
float oldviewinv[4][4], mat[4][4], tmat[4][4];
short part;
/* need a recalc: ortho-render has no correct viewinv */
invert_m4_m4(oldviewinv, re->viewmat);
envre = envmap_render_copy(re, env);
/* precalc orthmat for object */
copy_m4_m4(orthmat, env->object->obmat);
normalize_m4(orthmat);
/* need imat later for texture imat */
mult_m4_m4m4(mat, re->viewmat, orthmat);
invert_m4_m4(tmat, mat);
copy_m3_m4(env->obimat, tmat);
for (part = 0; part < 6; part++) {
if (env->type == ENV_PLANE && part != 1)
continue;
re->display_clear(re->dch, envre->result);
copy_m4_m4(tmat, orthmat);
envmap_transmatrix(tmat, part);
invert_m4_m4(mat, tmat);
/* mat now is the camera 'viewmat' */
copy_m4_m4(envre->viewmat, mat);
copy_m4_m4(envre->viewinv, tmat);
/* we have to correct for the already rotated vertexcoords */
mult_m4_m4m4(tmat, envre->viewmat, oldviewinv);
invert_m4_m4(env->imat, tmat);
env_rotate_scene(envre, tmat, 1);
init_render_world(envre);
project_renderdata(envre, projectverto, 0, 0, 1);
env_layerflags(envre, env->notlay);
env_hideobject(envre, env->object);
env_set_imats(envre);
if (re->test_break(re->tbh) == 0) {
RE_TileProcessor(envre);
}
/* rotate back */
env_showobjects(envre);
env_rotate_scene(envre, tmat, 0);
if (re->test_break(re->tbh) == 0) {
RenderLayer *rl = envre->result->layers.first;
int y;
float *alpha;
ibuf = IMB_allocImBuf(envre->rectx, envre->recty, 24, IB_rect | IB_rectfloat);
memcpy(ibuf->rect_float, rl->rectf, ibuf->channels * ibuf->x * ibuf->y * sizeof(float));
/* envmap renders without alpha */
alpha = ibuf->rect_float + 3;
for (y = ibuf->x * ibuf->y - 1; y >= 0; y--, alpha += 4)
*alpha = 1.0;
env->cube[part] = ibuf;
}
if (re->test_break(re->tbh)) break;
}
if (re->test_break(re->tbh)) BKE_free_envmapdata(env);
else {
if (envre->r.mode & R_OSA) env->ok = ENV_OSA;
else env->ok = ENV_NORMAL;
env->lastframe = re->scene->r.cfra;
}
/* restore */
envmap_free_render_copy(envre);
env_set_imats(re);
}
/* ------------------------------------------------------------------------- */
void make_envmaps(Render *re)
{
Tex *tex;
int do_init = FALSE, depth = 0, trace;
if (!(re->r.mode & R_ENVMAP)) return;
/* we don't raytrace, disabling the flag will cause ray_transp render solid */
trace = (re->r.mode & R_RAYTRACE);
re->r.mode &= ~R_RAYTRACE;
re->i.infostr = "Creating Environment maps";
re->stats_draw(re->sdh, &re->i);
/* 5 = hardcoded max recursion level */
while (depth < 5) {
tex = re->main->tex.first;
while (tex) {
if (tex->id.us && tex->type == TEX_ENVMAP) {
if (tex->env && tex->env->object) {
EnvMap *env = tex->env;
if (env->object->lay & re->lay) {
if (env->stype == ENV_LOAD) {
float orthmat[4][4], mat[4][4], tmat[4][4];
/* precalc orthmat for object */
copy_m4_m4(orthmat, env->object->obmat);
normalize_m4(orthmat);
/* need imat later for texture imat */
mult_m4_m4m4(mat, re->viewmat, orthmat);
invert_m4_m4(tmat, mat);
copy_m3_m4(env->obimat, tmat);
}
else {
/* decide if to render an envmap (again) */
if (env->depth >= depth) {
/* set 'recalc' to make sure it does an entire loop of recalcs */
if (env->ok) {
/* free when OSA, and old one isn't OSA */
if ((re->r.mode & R_OSA) && env->ok == ENV_NORMAL)
BKE_free_envmapdata(env);
/* free when size larger */
else if (env->lastsize < re->r.size)
BKE_free_envmapdata(env);
/* free when env is in recalcmode */
else if (env->recalc)
BKE_free_envmapdata(env);
}
if (env->ok == 0 && depth == 0) env->recalc = 1;
if (env->ok == 0) {
do_init = TRUE;
render_envmap(re, env);
if (depth == env->depth) env->recalc = 0;
}
}
}
}
}
}
tex = tex->id.next;
}
depth++;
}
if (do_init) {
re->display_init(re->dih, re->result);
re->display_clear(re->dch, re->result);
// re->flag |= R_REDRAW_PRV;
}
/* restore */
re->r.mode |= trace;
}
/* ------------------------------------------------------------------------- */
static int envcube_isect(EnvMap *env, const float vec[3], float answ[2])
{
float labda;
int face;
if (env->type == ENV_PLANE) {
face = 1;
labda = 1.0f / vec[2];
answ[0] = env->viewscale * labda * vec[0];
answ[1] = -env->viewscale * labda * vec[1];
}
else {
/* which face */
if (vec[2] <= -fabsf(vec[0]) && vec[2] <= -fabsf(vec[1]) ) {
face = 0;
labda = -1.0f / vec[2];
answ[0] = labda * vec[0];
answ[1] = labda * vec[1];
}
else if (vec[2] >= fabsf(vec[0]) && vec[2] >= fabsf(vec[1])) {
face = 1;
labda = 1.0f / vec[2];
answ[0] = labda * vec[0];
answ[1] = -labda * vec[1];
}
else if (vec[1] >= fabsf(vec[0])) {
face = 2;
labda = 1.0f / vec[1];
answ[0] = labda * vec[0];
answ[1] = labda * vec[2];
}
else if (vec[0] <= -fabsf(vec[1])) {
face = 3;
labda = -1.0f / vec[0];
answ[0] = labda * vec[1];
answ[1] = labda * vec[2];
}
else if (vec[1] <= -fabsf(vec[0])) {
face = 4;
labda = -1.0f / vec[1];
answ[0] = -labda * vec[0];
answ[1] = labda * vec[2];
}
else {
face = 5;
labda = 1.0f / vec[0];
answ[0] = -labda * vec[1];
answ[1] = labda * vec[2];
}
}
answ[0] = 0.5f + 0.5f * answ[0];
answ[1] = 0.5f + 0.5f * answ[1];
return face;
}
/* ------------------------------------------------------------------------- */
static void set_dxtdyt(float r_dxt[3], float r_dyt[3], const float dxt[3], const float dyt[3], int face)
{
if (face == 2 || face == 4) {
r_dxt[0] = dxt[0];
r_dyt[0] = dyt[0];
r_dxt[1] = dxt[2];
r_dyt[1] = dyt[2];
}
else if (face == 3 || face == 5) {
r_dxt[0] = dxt[1];
r_dxt[1] = dxt[2];
r_dyt[0] = dyt[1];
r_dyt[1] = dyt[2];
}
else {
r_dxt[0] = dxt[0];
r_dyt[0] = dyt[0];
r_dxt[1] = dxt[1];
r_dyt[1] = dyt[1];
}
}
/* ------------------------------------------------------------------------- */
int envmaptex(Tex *tex, const float texvec[3], float dxt[3], float dyt[3], int osatex, TexResult *texres)
{
extern Render R; /* only in this call */
/* texvec should be the already reflected normal */
EnvMap *env;
ImBuf *ibuf;
float fac, vec[3], sco[3], dxts[3], dyts[3];
int face, face1;
env = tex->env;
if (env == NULL || (env->stype != ENV_LOAD && env->object == NULL)) {
texres->tin = 0.0;
return 0;
}
if (env->stype == ENV_LOAD) {
env->ima = tex->ima;
if (env->ima && env->ima->ok) {
if (env->cube[1] == NULL) {
ImBuf *ibuf_ima = BKE_image_acquire_ibuf(env->ima, NULL, NULL);
if (ibuf_ima)
envmap_split_ima(env, ibuf_ima);
else
env->ok = 0;
BKE_image_release_ibuf(env->ima, ibuf_ima, NULL);
}
}
}
if (env->ok == 0) {
texres->tin = 0.0;
return 0;
}
/* rotate to envmap space, if object is set */
copy_v3_v3(vec, texvec);
if (env->object) mul_m3_v3(env->obimat, vec);
else mul_mat3_m4_v3(R.viewinv, vec);
face = envcube_isect(env, vec, sco);
ibuf = env->cube[face];
if (osatex) {
if (env->object) {
mul_m3_v3(env->obimat, dxt);
mul_m3_v3(env->obimat, dyt);
}
else {
mul_mat3_m4_v3(R.viewinv, dxt);
mul_mat3_m4_v3(R.viewinv, dyt);
}
set_dxtdyt(dxts, dyts, dxt, dyt, face);
imagewraposa(tex, NULL, ibuf, sco, dxts, dyts, texres);
/* edges? */
if (texres->ta < 1.0f) {
TexResult texr1, texr2;
texr1.nor = texr2.nor = NULL;
texr1.talpha = texr2.talpha = texres->talpha; /* boxclip expects this initialized */
add_v3_v3(vec, dxt);
face1 = envcube_isect(env, vec, sco);
sub_v3_v3(vec, dxt);
if (face != face1) {
ibuf = env->cube[face1];
set_dxtdyt(dxts, dyts, dxt, dyt, face1);
imagewraposa(tex, NULL, ibuf, sco, dxts, dyts, &texr1);
}
else texr1.tr = texr1.tg = texr1.tb = texr1.ta = 0.0;
/* here was the nasty bug! results were not zero-ed. FPE! */
add_v3_v3(vec, dyt);
face1 = envcube_isect(env, vec, sco);
sub_v3_v3(vec, dyt);
if (face != face1) {
ibuf = env->cube[face1];
set_dxtdyt(dxts, dyts, dxt, dyt, face1);
imagewraposa(tex, NULL, ibuf, sco, dxts, dyts, &texr2);
}
else texr2.tr = texr2.tg = texr2.tb = texr2.ta = 0.0;
fac = (texres->ta + texr1.ta + texr2.ta);
if (fac != 0.0f) {
fac = 1.0f / fac;
texres->tr = fac * (texres->ta * texres->tr + texr1.ta * texr1.tr + texr2.ta * texr2.tr);
texres->tg = fac * (texres->ta * texres->tg + texr1.ta * texr1.tg + texr2.ta * texr2.tg);
texres->tb = fac * (texres->ta * texres->tb + texr1.ta * texr1.tb + texr2.ta * texr2.tb);
}
texres->ta = 1.0;
}
}
else {
imagewrap(tex, NULL, ibuf, sco, texres);
}
return 1;
}