BitHeaven/picom
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Make buffer age not glx specific

Browse Source 
Buffer age is not a glx specific concept. xrender backend can have
buffer age too, if double buffering is used (required if we want to use
Present). So, make buffer age a generic concept in compton is required
for further backend improvements.

Moved buffer age based damage aggragation out of glx as well.

Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
This commit is contained in:
Yuxuan Shui
2019-01-08 23:15:00 +00:00
parent 377b18a00f
commit e80ff8530e
6 changed files with 180 additions and 283 deletions
Download Patch File Download Diff File Expand all files Collapse all files

257
src/render.c
View File

@ -59,6 +59,28 @@ static inline bool bkend_use_xrender(session_t *ps) {
return BKEND_XRENDER == ps->o.backend || BKEND_XR_GLX_HYBRID == ps->o.backend;
}
int maximum_buffer_age(session_t *ps) {
if (bkend_use_glx(ps) && ps->o.glx_swap_method == SWAPM_BUFFER_AGE) {
return CGLX_MAX_BUFFER_AGE;
}
return 1;
}
static int get_buffer_age(session_t *ps) {
#ifdef CONFIG_OPENGL
if (bkend_use_glx(ps)) {
if (ps->o.glx_swap_method == SWAPM_BUFFER_AGE) {
unsigned int val;
glXQueryDrawable(ps->dpy, get_tgt_window(ps), GLX_BACK_BUFFER_AGE_EXT, &val);
return (int)val ?: -1;
} else {
return -1;
}
}
#endif
return 1;
}
/**
* Reset filter on a <code>Picture</code>.
*/
@ -68,6 +90,7 @@ static inline void xrfilter_reset(session_t *ps, xcb_render_picture_t p) {
#undef FILTER
}
/// Set the input/output clip region of the target buffer (not the actual target!)
static inline void attr_nonnull(1, 2) set_tgt_clip(session_t *ps, region_t *reg) {
switch (ps->o.backend) {
case BKEND_XRENDER:
@ -113,8 +136,7 @@ void render(session_t *ps, int x, int y, int dx, int dy, int wid, int hei, doubl
int alpha_step = opacity * MAX_ALPHA;
xcb_render_picture_t alpha_pict = ps->alpha_picts[alpha_step];
if (alpha_step != 0) {
int op = ((!argb && !alpha_pict) ? XCB_RENDER_PICT_OP_SRC
: XCB_RENDER_PICT_OP_OVER);
int op = ((!argb && !alpha_pict) ? XCB_RENDER_PICT_OP_SRC : XCB_RENDER_PICT_OP_OVER);
xcb_render_composite(ps->c, op, pict, alpha_pict, ps->tgt_buffer.pict,
x, y, 0, 0, dx, dy, wid, hei);
}
@ -131,9 +153,8 @@ void render(session_t *ps, int x, int y, int dx, int dy, int wid, int hei, doubl
}
}
static inline void
paint_region(session_t *ps, win *w, int x, int y, int wid, int hei, double opacity,
const region_t *reg_paint, xcb_render_picture_t pict) {
static inline void paint_region(session_t *ps, win *w, int x, int y, int wid, int hei, double opacity,
const region_t *reg_paint, xcb_render_picture_t pict) {
const int dx = (w ? w->g.x : 0) + x;
const int dy = (w ? w->g.y : 0) + y;
const bool argb = (w && (win_has_alpha(w) || ps->o.force_win_blend));
@ -200,8 +221,7 @@ void paint_one(session_t *ps, win *w, const region_t *reg_paint) {
// Let glx_bind_pixmap() determine pixmap size, because if the user
// is resizing windows, the width and height we get may not be up-to-date,
// causing the jittering issue M4he reported in #7.
if (!paint_bind_tex(ps, &w->paint, 0, 0, 0,
(!ps->o.glx_no_rebind_pixmap && w->pixmap_damaged))) {
if (!paint_bind_tex(ps, &w->paint, 0, 0, 0, (!ps->o.glx_no_rebind_pixmap && w->pixmap_damaged))) {
log_error("Failed to bind texture for window %#010x.", w->id);
}
w->pixmap_damaged = false;
@ -229,15 +249,15 @@ void paint_one(session_t *ps, win *w, const region_t *reg_paint) {
pixman_region32_init(&reg);
pixman_region32_copy(&reg, (region_t *)reg_paint);
pixman_region32_translate(&reg, -x, -y);
// FIXME XFixesSetPictureClipRegion(ps->dpy, newpict, 0, 0, reg);
// FIXME XFixesSetPictureClipRegion(ps->dpy, newpict, 0,
// 0, reg);
pixman_region32_fini(&reg);
}
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC, pict, XCB_NONE,
newpict, 0, 0, 0, 0, 0, 0, wid, hei);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_DIFFERENCE,
ps->white_picture, XCB_NONE, newpict, 0, 0,
0, 0, 0, 0, wid, hei);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_DIFFERENCE, ps->white_picture,
XCB_NONE, newpict, 0, 0, 0, 0, 0, 0, wid, hei);
// We use an extra PictOpInReverse operation to get correct
// pixel alpha. There could be a better solution.
if (win_has_alpha(w))
@ -261,8 +281,7 @@ void paint_one(session_t *ps, win *w, const region_t *reg_paint) {
const int r = extents.right;
#define COMP_BDR(cx, cy, cwid, chei) \
paint_region(ps, w, (cx), (cy), (cwid), (chei), w->frame_opacity *dopacity, \
reg_paint, pict)
paint_region(ps, w, (cx), (cy), (cwid), (chei), w->frame_opacity *dopacity, reg_paint, pict)
// Sanitize the margins, in case some broken WM makes
// top_width + bottom_width > height in some cases.
@ -353,8 +372,7 @@ void paint_one(session_t *ps, win *w, const region_t *reg_paint) {
} break;
#ifdef CONFIG_OPENGL
case BKEND_GLX:
glx_dim_dst(ps, x, y, wid, hei, ps->psglx->z - 0.7, dim_opacity,
reg_paint);
glx_dim_dst(ps, x, y, wid, hei, ps->psglx->z - 0.7, dim_opacity, reg_paint);
break;
#endif
default: assert(false);
@ -380,9 +398,8 @@ static bool get_root_tile(session_t *ps) {
// Get the values of background attributes
for (int p = 0; background_props_str[p]; p++) {
winprop_t prop =
wid_get_prop(ps, ps->root, get_atom(ps, background_props_str[p]), 1L,
XCB_ATOM_PIXMAP, 32);
winprop_t prop = wid_get_prop(
ps, ps->root, get_atom(ps, background_props_str[p]), 1L, XCB_ATOM_PIXMAP, 32);
if (prop.nitems) {
pixmap = *prop.p32;
fill = false;
@ -465,15 +482,13 @@ static bool win_build_shadow(session_t *ps, win *w, double opacity) {
xcb_render_picture_t shadow_picture = XCB_NONE, shadow_picture_argb = XCB_NONE;
xcb_gcontext_t gc = XCB_NONE;
shadow_image =
make_shadow(ps->c, ps->gaussian_map, opacity, width, height);
shadow_image = make_shadow(ps->c, ps->gaussian_map, opacity, width, height);
if (!shadow_image) {
log_error("failed to make shadow");
return XCB_NONE;
}
shadow_pixmap =
x_create_pixmap(ps, 8, ps->root, shadow_image->width, shadow_image->height);
shadow_pixmap = x_create_pixmap(ps, 8, ps->root, shadow_image->width, shadow_image->height);
shadow_pixmap_argb =
x_create_pixmap(ps, 32, ps->root, shadow_image->width, shadow_image->height);
@ -570,9 +585,8 @@ static inline void normalize_conv_kern(int wid, int hei, xcb_render_fixed_t *ker
*
* @return true if successful, false otherwise
*/
static bool
xr_blur_dst(session_t *ps, xcb_render_picture_t tgt_buffer, int x, int y, int wid,
int hei, xcb_render_fixed_t **blur_kerns, const region_t *reg_clip) {
static bool xr_blur_dst(session_t *ps, xcb_render_picture_t tgt_buffer, int x, int y, int wid,
int hei, xcb_render_fixed_t **blur_kerns, const region_t *reg_clip) {
assert(blur_kerns[0]);
// Directly copying from tgt_buffer to it does not work, so we create a
@ -600,12 +614,10 @@ xr_blur_dst(session_t *ps, xcb_render_picture_t tgt_buffer, int x, int y, int wi
// be applied on source picture, to get the nearby pixels outside the
// window.
xcb_render_set_picture_filter(ps->c, src_pict, strlen(XRFILTER_CONVOLUTION),
XRFILTER_CONVOLUTION, kwid * khei + 2,
convolution_blur);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC, src_pict, XCB_NONE,
dst_pict, (rd_from_tgt ? x : 0),
(rd_from_tgt ? y : 0), 0, 0, (rd_from_tgt ? 0 : x),
(rd_from_tgt ? 0 : y), wid, hei);
XRFILTER_CONVOLUTION, kwid * khei + 2, convolution_blur);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC, src_pict, XCB_NONE, dst_pict,
(rd_from_tgt ? x : 0), (rd_from_tgt ? y : 0), 0, 0,
(rd_from_tgt ? 0 : x), (rd_from_tgt ? 0 : y), wid, hei);
xrfilter_reset(ps, src_pict);
{
@ -672,12 +684,10 @@ static inline void win_blur_background(session_t *ps, win *w, xcb_render_picture
}
// Modify the factor of the center pixel
kern_src[2 + (khei / 2) * kwid + kwid / 2] =
DOUBLE_TO_XFIXED(factor_center);
kern_src[2 + (khei / 2) * kwid + kwid / 2] = DOUBLE_TO_XFIXED(factor_center);
// Copy over
memcpy(kern_dst, kern_src,
(kwid * khei + 2) * sizeof(xcb_render_fixed_t));
memcpy(kern_dst, kern_src, (kwid * khei + 2) * sizeof(xcb_render_fixed_t));
normalize_conv_kern(kwid, khei, kern_dst + 2);
}
@ -708,10 +718,40 @@ static inline void win_blur_background(session_t *ps, win *w, xcb_render_picture
}
}
/**
* Resize a region.
*/
static inline void resize_region(region_t *region, short mod) {
if (!mod || !region)
return;
// Loop through all rectangles
int nrects;
int nnewrects = 0;
pixman_box32_t *rects = pixman_region32_rectangles(region, &nrects);
auto newrects = ccalloc(nrects, pixman_box32_t);
for (int i = 0; i < nrects; i++) {
int x1 = rects[i].x1 - mod;
int y1 = rects[i].y1 - mod;
int x2 = rects[i].x2 + mod;
int y2 = rects[i].y2 + mod;
int wid = x2 - x1;
int hei = y2 - y1;
if (wid <= 0 || hei <= 0)
continue;
newrects[nnewrects] = (pixman_box32_t){.x1 = x1, .x2 = x2, .y1 = y1, .y2 = y2};
++nnewrects;
}
pixman_region32_fini(region);
pixman_region32_init_rects(region, newrects, nnewrects);
free(newrects);
}
/// paint all windows
/// region = ??
/// region_real = the damage region
void paint_all(session_t *ps, region_t *region, const region_t *region_real, win *const t) {
void paint_all(session_t *ps, win *const t, bool ignore_damage) {
if (ps->o.xrender_sync_fence) {
if (!x_fence_sync(ps, ps->sync_fence)) {
log_error("x_fence_sync failed, xrender-sync-fence will be "
@ -722,24 +762,32 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
}
}
if (!region_real) {
region_real = region;
region_t region;
pixman_region32_init(&region);
int buffer_age = get_buffer_age(ps);
if (buffer_age == -1 || buffer_age > ps->ndamage || ignore_damage) {
pixman_region32_copy(&region, &ps->screen_reg);
} else {
for (int i = 0; i < get_buffer_age(ps); i++) {
const int curr = ((ps->damage - ps->damage_ring) + i) % ps->ndamage;
pixman_region32_union(&region, &region, &ps->damage_ring[curr]);
}
}
if (!pixman_region32_not_empty(&region)) {
return;
}
#ifdef DEBUG_REPAINT
static struct timespec last_paint = {0};
#endif
if (!region)
region_real = region = &ps->screen_reg;
else
// Remove the damaged area out of screen
pixman_region32_intersect(region, region, &ps->screen_reg);
if (ps->o.resize_damage > 0) {
resize_region(&region, ps->o.resize_damage);
}
#ifdef CONFIG_OPENGL
if (bkend_use_glx(ps))
glx_paint_pre(ps, region);
#endif
// Remove the damaged area out of screen
pixman_region32_intersect(&region, &region, &ps->screen_reg);
if (!paint_isvalid(ps, &ps->tgt_buffer)) {
if (!ps->tgt_buffer.pixmap) {
@ -759,26 +807,34 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
}
if (BKEND_XRENDER == ps->o.backend) {
x_set_picture_clip_region(ps, ps->tgt_picture, 0, 0, region_real);
x_set_picture_clip_region(ps, ps->tgt_picture, 0, 0, &region);
}
#ifdef CONFIG_OPENGL
if (bkend_use_glx(ps)) {
ps->psglx->z = 0.0;
}
#endif
region_t reg_tmp, *reg_paint;
pixman_region32_init(&reg_tmp);
if (t) {
// Calculate the region upon which the root window is to be painted
// based on the ignore region of the lowest window, if available
pixman_region32_subtract(&reg_tmp, region, t->reg_ignore);
// Calculate the region upon which the root window is to be
// painted based on the ignore region of the lowest window, if
// available
pixman_region32_subtract(&reg_tmp, &region, t->reg_ignore);
reg_paint = &reg_tmp;
} else {
reg_paint = region;
reg_paint = &region;
}
set_tgt_clip(ps, reg_paint);
paint_root(ps, reg_paint);
// Windows are sorted from bottom to top
// Each window has a reg_ignore, which is the region obscured by all the windows
// on top of that window. This is used to reduce the number of pixels painted.
// Each window has a reg_ignore, which is the region obscured by all the
// windows on top of that window. This is used to reduce the number of
// pixels painted.
//
// Whether this is beneficial is to be determined XXX
for (win *w = t; w; w = w->prev_trans) {
@ -790,39 +846,39 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
if (!win_build_shadow(ps, w, 1))
log_error("build shadow failed");
// Shadow doesn't need to be painted underneath the body of
// the window Because no one can see it
pixman_region32_subtract(&reg_tmp, region, w->reg_ignore);
// Shadow doesn't need to be painted underneath the body
// of the windows above. Because no one can see it
pixman_region32_subtract(&reg_tmp, &region, w->reg_ignore);
// Mask out the region we don't want shadow on
if (pixman_region32_not_empty(&ps->shadow_exclude_reg))
pixman_region32_subtract(&reg_tmp, &reg_tmp,
&ps->shadow_exclude_reg);
pixman_region32_subtract(&reg_tmp, &reg_tmp, &ps->shadow_exclude_reg);
// Might be worth while to crop the region to shadow border
pixman_region32_intersect_rect(
&reg_tmp, &reg_tmp, w->g.x + w->shadow_dx,
w->g.y + w->shadow_dy, w->shadow_width, w->shadow_height);
// Might be worth while to crop the region to shadow
// border
pixman_region32_intersect_rect(&reg_tmp, &reg_tmp, w->g.x + w->shadow_dx,
w->g.y + w->shadow_dy,
w->shadow_width, w->shadow_height);
// Mask out the body of the window from the shadow if needed
// Doing it here instead of in make_shadow() for saving GPU
// power and handling shaped windows (XXX unconfirmed)
// Mask out the body of the window from the shadow if
// needed Doing it here instead of in make_shadow() for
// saving GPU power and handling shaped windows (XXX
// unconfirmed)
if (!ps->o.wintype_option[w->window_type].full_shadow)
pixman_region32_subtract(&reg_tmp, &reg_tmp, &bshape);
#ifdef CONFIG_XINERAMA
if (ps->o.xinerama_shadow_crop && w->xinerama_scr >= 0 &&
w->xinerama_scr < ps->xinerama_nscrs)
// There can be a window where number of screens is
// updated, but the screen number attached to the
// windows have not.
// There can be a window where number of screens
// is updated, but the screen number attached to
// the windows have not.
//
// Window screen number will be updated eventually,
// so here we just check to make sure we don't access
// out of bounds.
// Window screen number will be updated
// eventually, so here we just check to make sure
// we don't access out of bounds.
pixman_region32_intersect(
&reg_tmp, &reg_tmp,
&ps->xinerama_scr_regs[w->xinerama_scr]);
&reg_tmp, &reg_tmp, &ps->xinerama_scr_regs[w->xinerama_scr]);
#endif
// Detect if the region is empty before painting
@ -832,10 +888,11 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
}
}
// Calculate the region based on the reg_ignore of the next (higher)
// window and the bounding region
// XXX XXX
pixman_region32_subtract(&reg_tmp, region, w->reg_ignore);
// Calculate the paint region based on the reg_ignore of the current
// window and its bounding region.
// Remeber, reg_ignore is the union of all windows above the current
// window.
pixman_region32_subtract(&reg_tmp, &region, w->reg_ignore);
pixman_region32_intersect(&reg_tmp, &reg_tmp, &bshape);
pixman_region32_fini(&bshape);
@ -855,6 +912,13 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
// Free up all temporary regions
pixman_region32_fini(&reg_tmp);
// Move the head of the damage ring
ps->damage = ps->damage - 1;
if (ps->damage < ps->damage_ring) {
ps->damage = ps->damage_ring + ps->ndamage - 1;
}
pixman_region32_clear(ps->damage);
// Do this as early as possible
set_tgt_clip(ps, &ps->screen_reg);
@ -886,14 +950,8 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
// the paint region. This is not very efficient, but since
// it's for debug only, we don't really care
// First, we clear tgt_buffer.pict's clip region, since we
// want to copy everything
x_set_picture_clip_region(ps, ps->tgt_buffer.pict, 0, 0,
&ps->screen_reg);
// Then we create a new picture, and copy content to it
xcb_render_pictforminfo_t *pictfmt =
x_get_pictform_for_visual(ps, ps->vis);
// First we create a new picture, and copy content from the buffer to it
xcb_render_pictforminfo_t *pictfmt = x_get_pictform_for_visual(ps, ps->vis);
xcb_render_picture_t new_pict = x_create_picture_with_pictfmt(
ps, ps->root_width, ps->root_height, pictfmt, 0, NULL);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC,
@ -901,22 +959,22 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
0, 0, 0, 0, ps->root_width, ps->root_height);
// Next, we set the region of paint and highlight it
x_set_picture_clip_region(ps, new_pict, 0, 0, region_real);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_OVER,
ps->white_picture,
x_set_picture_clip_region(ps, new_pict, 0, 0, &region);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_OVER, ps->white_picture,
ps->alpha_picts[MAX_ALPHA / 2], new_pict, 0, 0,
0, 0, 0, 0, ps->root_width, ps->root_height);
// Finally, clear clip region and put the whole thing on screen
// Finally, clear clip regions of new_pict and the screen, and put
// the whole thing on screen
x_set_picture_clip_region(ps, new_pict, 0, 0, &ps->screen_reg);
x_set_picture_clip_region(ps, ps->tgt_picture, 0, 0, &ps->screen_reg);
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC, new_pict,
XCB_NONE, ps->tgt_picture, 0, 0, 0, 0, 0, 0,
ps->root_width, ps->root_height);
xcb_render_free_picture(ps->c, new_pict);
} else
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC,
ps->tgt_buffer.pict, XCB_NONE,
ps->tgt_picture, 0, 0, 0, 0, 0, 0,
xcb_render_composite(ps->c, XCB_RENDER_PICT_OP_SRC, ps->tgt_buffer.pict,
XCB_NONE, ps->tgt_picture, 0, 0, 0, 0, 0, 0,
ps->root_width, ps->root_height);
break;
#ifdef CONFIG_OPENGL
@ -936,7 +994,7 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
glFlush();
glXWaitX();
glx_render(ps, ps->tgt_buffer.ptex, 0, 0, 0, 0, ps->root_width,
ps->root_height, 0, 1.0, false, false, region_real, NULL);
ps->root_height, 0, 1.0, false, false, &region, NULL);
// falls through
case BKEND_GLX: glXSwapBuffers(ps->dpy, get_tgt_window(ps)); break;
#endif
@ -967,6 +1025,9 @@ void paint_all(session_t *ps, region_t *region, const region_t *region_real, win
log_trace(" %#010lx", w->id);
#endif
// Free the paint region
pixman_region32_fini(&region);
// Check if fading is finished on all painted windows
{
win *pprev = NULL;
@ -1095,6 +1156,14 @@ bool init_render(session_t *ps) {
return false;
}
}
ps->ndamage = maximum_buffer_age(ps);
ps->damage_ring = ccalloc(ps->ndamage, region_t);
ps->damage = ps->damage_ring + ps->ndamage - 1;
for (int i = 0; i < ps->ndamage; i++) {
pixman_region32_init(&ps->damage_ring[i]);
}
return true;
}