Looking for memory leaks that were causing crashes in my environment
in a situation where valgrind would not work, I ended up improving
the i965 debug traces so I could better see where the memory was
being allocated and where it was going, in the regions and miptrees
code, and in the state caches. These traces were specific enough
that external scripts could determine what elements were not being
released, and where the memory leaks were.
I also ended up creating my own backtrace code in intel_regions.c,
to determine exactly where regions were being allocated and for what,
since valgrind wasn't working. Because it was useful, I left it in,
but disabled and compiled out. It can be activated by changing a flag
at the top of the file.
A temporary change to the intelMakeCurrent() function to make
it work with frame buffer objects causes the static regions
associated with the context (the front_region, back_region,
and depth_region) to take on an additional reference, with
no corresponding release. This causes a memory leak if a
program repeatedly creates and destroys contexts.
The fix is the corresponding hack, to unreference these
regions when the context is deleted, but only if the
framebuffer objects are still present and the same
regions are still referenced within.
Both sets of code have comment blocks referring to each
other.
When out of memory (in at least one case, triggered by a longrunning
memory leak), this code will segfault and crash. By checking for the
out-of-memory condition, the system can continue, and will report
the out-of-memory error later, a much preferable outcome.
The current texture for any particular texture unit is given an additional
reference in update_texture_state(); but if the context is closed before
that texture can be released (which is quite frequent in normal use, unless
a program unbinds and deletes the texture and renders without it to force
a call to update_texture_state(), the memory is lost.
This affects general Mesa; but the i965 is particularly affected because
it allocates a considerable amount of additional memory for each allocated
texture.
In addition to being HW accelerated, it avoids the incorrect
(black) rendering of the mipmaps that SW was doing in fbo-generatemipmap.
Improves the performance of the mipmap generation and drawing in
fbo-generatemipmap by 30%.
Make this be its own function for setup/teardown of the binding of these
texcoords. No performance difference in the engine demo (I just felt dirty
not using a VBO for this), and I think it should be more resilient to
interference from current GL state.
They seem to be used for something else and using them for shader temps
seems to lead to GPU lock-ups.
Call _mesa_warning() when we run out of temps.
Also, clean up some debug code.
The src/dest x,y, and w,h arguments of the pipe->surface_copy
function are unsigned and the drivers aren't expecting negative
(or extremly-large unsigned) values as inputs. Trim the requests
at the state-tracker level before passing down.
Add a new flag mvp_with_dp4 in the context, and use that to switch
both ffvertex.c and programopt.c vertex transformation code to
either DP4 or MUL/MAD implementations.
This is a quick fix for z fighting in quake4 caused by the mismatch
between vertex transformation here and in the position_invarient code.
Full fix would be to make this driver-tunable and adjust both
position_invarient and ffvertex_prog.c code to respect driver
preferences.
