brl/mutter
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

cogl: Implements a software only read-pixel fast-path

Browse Source 
This adds a transparent optimization to cogl_read_pixels for when a
single pixel is being read back and it happens that all the geometry of
the current frame is still available in the framebuffer's associated
journal.

The intention is to indirectly optimize Clutter's render based picking
mechanism in such a way that the 99% of cases where scenes are comprised
of trivial quad primitives that can easily be intersected we can avoid
the latency of kicking a GPU render and blocking for the result when we
know we can calculate the result manually on the CPU probably faster
than we could even kick a render.

A nice property of this solution is that it maintains all the
flexibility of the render based picking provided by Clutter and it can
gracefully fall back to GPU rendering if actors are drawn using anything
more complex than a quad for their geometry.

It seems worth noting that there is a limitation to the extensibility of
this approach in that it can only optimize picking a against geometry
that passes through Cogl's journal which isn't something Clutter
directly controls.  For now though this really doesn't matter since
basically all apps should end up hitting this fast-path. The current
idea to address this longer term would be a pick2 vfunc for ClutterActor
that can support geometry and render based input regions of actors and
move this optimization up into Clutter instead.

Note: currently we don't have a primitive count threshold to consider
that there could be scenes with enough geometry for us to compensate for
the cost of kicking a render and determine a result more efficiently by
utilizing the GPU. We don't currently expect this to be common though.

Note: in the future it could still be interesting to revive something
like the wip/async-pbo-picking branch to provide an asynchronous
read-pixels based optimization for Clutter picking in cases where more
complex input regions that necessitate rendering are in use or if we do
add a threshold for rendering as mentioned above.
This commit is contained in:
Robert Bragg
2011-01-12 22:12:41 +00:00
parent 3e42af2a00
commit a8d6c3f686
6 changed files with 800 additions and 180 deletions
Download Patch File Download Diff File Expand all files Collapse all files

196
cogl/cogl-framebuffer.c
View File

@ -160,6 +160,13 @@ _cogl_framebuffer_init (CoglFramebuffer *framebuffer,
framebuffer->journal = _cogl_journal_new ();
/* Ensure we know the framebuffer->clear_color* members can't be
* referenced for our fast-path read-pixel optimization (see
* _cogl_journal_try_read_pixel()) until some region of the
* framebuffer is initialized.
*/
framebuffer->clear_clip_dirty = TRUE;
/* XXX: We have to maintain a central list of all framebuffers
* because at times we need to be able to flush all known journals.
*
@ -248,6 +255,12 @@ _cogl_clear4f (unsigned long buffers,
GE (glClear (gl_buffers));
}
void
_cogl_framebuffer_dirty (CoglFramebuffer *framebuffer)
{
framebuffer->clear_clip_dirty = TRUE;
}
void
_cogl_framebuffer_clear4f (CoglFramebuffer *framebuffer,
unsigned long buffers,
@ -256,11 +269,96 @@ _cogl_framebuffer_clear4f (CoglFramebuffer *framebuffer,
float blue,
float alpha)
{
CoglClipStack *clip_stack = _cogl_framebuffer_get_clip_stack (framebuffer);
int scissor_x0;
int scissor_y0;
int scissor_x1;
int scissor_y1;
_cogl_clip_stack_get_bounds (clip_stack,
&scissor_x0, &scissor_y0,
&scissor_x1, &scissor_y1);
/* NB: the previous clear could have had an arbitrary clip.
* NB: everything for the last frame might still be in the journal
* but we can't assume anything about how each entry was
* clipped.
* NB: Clutter will scissor its pick renders which would mean all
* journal entries have a common ClipStack entry, but without
* a layering violation Cogl has to explicitly walk the journal
* entries to determine if this is the case.
* NB: We have a software only read-pixel optimization in the
* journal that determines the color at a given framebuffer
* coordinate for simple scenes without rendering with the GPU.
* When Clutter is hitting this fast-path we can expect to
* receive calls to clear the framebuffer with an un-flushed
* journal.
* NB: To fully support software based picking for Clutter we
* need to be able to reliably detect when the contents of a
* journal can be discarded and when we can skip the call to
* glClear because it matches the previous clear request.
*/
/* Note: we don't check for the stencil buffer being cleared here
* since there isn't any public cogl api to manipulate the stencil
* buffer.
*
* Note: we check for an exact clip match here because
* 1) a smaller clip could mean existing journal entries may
* need to contribute to regions outside the new clear-clip
* 2) a larger clip would mean we need to issue a real
* glClear and we only care about cases avoiding a
* glClear.
*
* Note: Comparing without an epsilon is considered
* appropriate here.
*/
if (buffers & COGL_BUFFER_BIT_COLOR &&
buffers & COGL_BUFFER_BIT_DEPTH &&
!framebuffer->clear_clip_dirty &&
framebuffer->clear_color_red == red &&
framebuffer->clear_color_green == green &&
framebuffer->clear_color_blue == blue &&
framebuffer->clear_color_alpha == alpha &&
scissor_x0 == framebuffer->clear_clip_x0 &&
scissor_y0 == framebuffer->clear_clip_y0 &&
scissor_x1 == framebuffer->clear_clip_x1 &&
scissor_y1 == framebuffer->clear_clip_y1)
{
/* NB: We only have to consider the clip state of journal
* entries if the current clear is clipped since otherwise we
* know every pixel of the framebuffer is affected by the clear
* and so all journal entries become redundant and can simply be
* discarded.
*/
if (clip_stack)
{
/*
* Note: the function for checking the journal entries is
* quite strict. It avoids detailed checking of all entry
* clip_stacks by only checking the details of the first
* entry and then it only verifies that the remaining
* entries share the same clip_stack ancestry. This means
* it's possible for some false negatives here but that will
* just result in us falling back to a real clear.
*/
if (_cogl_journal_all_entries_within_bounds (framebuffer->journal,
scissor_x0, scissor_y0,
scissor_x1, scissor_y1))
{
_cogl_journal_discard (framebuffer->journal);
goto cleared;
}
}
else
{
_cogl_journal_discard (framebuffer->journal);
goto cleared;
}
}
COGL_NOTE (DRAW, "Clear begin");
/* XXX: in the case where it's the color buffer being cleared and
* the current clip-stack is empty we could instead discard the
* journal here instead of flushing it. */
_cogl_framebuffer_flush_journal (framebuffer);
/* NB: _cogl_framebuffer_flush_state may disrupt various state (such
@ -282,6 +380,38 @@ _cogl_framebuffer_clear4f (CoglFramebuffer *framebuffer,
}
COGL_NOTE (DRAW, "Clear end");
cleared:
if (buffers & COGL_BUFFER_BIT_COLOR && buffers & COGL_BUFFER_BIT_DEPTH)
{
/* For our fast-path for reading back a single pixel of simple
* scenes where the whole frame is in the journal we need to
* track the cleared color of the framebuffer in case the point
* read doesn't intersect any of the journal rectangles. */
framebuffer->clear_clip_dirty = FALSE;
framebuffer->clear_color_red = red;
framebuffer->clear_color_green = green;
framebuffer->clear_color_blue = blue;
framebuffer->clear_color_alpha = alpha;
/* NB: A clear may be scissored so we need to track the extents
* that the clear is applicable too... */
if (clip_stack)
{
_cogl_clip_stack_get_bounds (clip_stack,
&framebuffer->clear_clip_x0,
&framebuffer->clear_clip_y0,
&framebuffer->clear_clip_x1,
&framebuffer->clear_clip_y1);
}
else
{
/* FIXME: set degenerate clip */
}
}
else
_cogl_framebuffer_dirty (framebuffer);
}
/* XXX: We'll need to consider if this API is a good approach for the
@ -1097,3 +1227,63 @@ _cogl_framebuffer_get_alpha_bits (CoglFramebuffer *framebuffer)
return framebuffer->alpha_bits;
}
gboolean
_cogl_framebuffer_try_fast_read_pixel (CoglFramebuffer *framebuffer,
int x,
int y,
CoglReadPixelsFlags source,
CoglPixelFormat format,
guint8 *pixel)
{
gboolean found_intersection;
if (source != COGL_READ_PIXELS_COLOR_BUFFER)
return FALSE;
if (format != COGL_PIXEL_FORMAT_RGBA_8888_PRE &&
format != COGL_PIXEL_FORMAT_RGBA_8888)
return FALSE;
if (!_cogl_journal_try_read_pixel (framebuffer->journal,
x, y, format, pixel,
&found_intersection))
return FALSE;
/* If we can't determine the color from the primitives in the
* journal then see if we can use the last recorded clear color
*/
/* If _cogl_journal_try_read_pixel() failed even though there was an
* intersection of the given point with a primitive in the journal
* then we can't fallback to the framebuffer's last clear color...
* */
if (found_intersection)
return TRUE;
/* If the framebuffer has been rendered too since it was last
* cleared then we can't return the last known clear color. */
if (framebuffer->clear_clip_dirty)
return FALSE;
if (x >= framebuffer->clear_clip_x0 &&
x < framebuffer->clear_clip_x1 &&
y >= framebuffer->clear_clip_y0 &&
y < framebuffer->clear_clip_y1)
{
/* we currently only care about cases where the premultiplied or
* unpremultipled colors are equivalent... */
if (framebuffer->clear_color_alpha != 1.0)
return FALSE;
pixel[0] = framebuffer->clear_color_red * 255.0;
pixel[1] = framebuffer->clear_color_green * 255.0;
pixel[2] = framebuffer->clear_color_blue * 255.0;
pixel[3] = framebuffer->clear_color_alpha * 255.0;
return TRUE;
}
return FALSE;
}