Age | Commit message (Collapse) | Author |
|
s->close_first was on the wrong side of the inequality.
Caught by blender.
Thanks to AndrewR for reporting this.
|
|
util_fill_rect could not handle formats with more than 32 bits,
since the fill color was a uint32_t value. Fix this by using
a util_color union instead, and also expand the union so it
works with formats which have up to 256 bits (the max of any
format currently defined).
|
|
Before, we were discarding the compiled vertex program on each
vertex program change.
Now we compile the program as if there were 6 clip planes and
dynamically patch in an "end program" bit at the right place.
Also, nv30 should now work.
|
|
Signed-off-by: Brian Paul <brianp@vmware.com>
|
|
This is a typo fix of earlier commit 0f3b3751b8643352dcc242567b3696bd1505df1d.
|
|
dump_cpu is used only when DEBUG is defined.
Fixes the following GCC warning on builds without DEBUG defined.
util/u_cpu_detect.c:76: warning: 'debug_get_option_dump_cpu' defined but not used
|
|
|
|
Fixes build error with MSVC.
|
|
|
|
|
|
|
|
Thanks to Jose Fonseca for pointing out they were missing.
|
|
|
|
C++ doesn't accept function <-> void* conversions without a putting a
fight.
|
|
Unused now that check_os_katmai_support was removed.
|
|
|
|
This commit adds the ability to produce a log file containing all
reference count changes, and object creation/destruction, on Gallium
objects.
The data allows to answer these crucial questions:
1. This app is exhausting all my memory due to a resource leak: where
is the bug?
2. Which resources is this app using at a given moment? Which parts of
the code created them?
3. What kinds of resources does this app use?
4. How fast does this app create and destroy resources? Which parts of
the code create resources fast?
The output is compatible with the one produced by the similar facility
in Mozilla Firefox, allowing to use Mozilla's tools to analyze the data.
To get the log file:
export GALLIUM_REFCNT_LOG=<file>
To get function names and source lines in the log file:
tools/addr2line.sh <file>
To process the log file, see:
http://www.mozilla.org/performance/refcnt-balancer.html
|
|
|
|
Without this, any form of logging that dumps stack traces continuously
will spend a lot of time resolving symbol names.
|
|
|
|
|
|
|
|
|
|
Currently Gallium internals always use PIPE_TEXTURE_2D and normalized
coordinates to access textures.
However, PIPE_TEXTURE_2D is not always supported for NPOT textures,
and PIPE_TEXTURE_RECT requires unnormalized coordinates.
Hence, this change adds support for both kinds of normalization.
|
|
|
|
This seems to make sense, although I suspect the semantics of
TGSI_TEXTURE_RECT need to be closely reviewed.
|
|
Searched for them with:
git grep -E '[!=]=.*PIPE_TEXTURE_2D|PIPE_TEXTURE_2D.*[!=]=|case.*PIPE_TEXTURE_2D'
Behavior hasn't been changed.
|
|
|
|
check_os_katmai_support checks that the operating system running on a
SSE-capable processor supports SSE. This is necessary for unpatched
2.2.x and earlier kernels. 2.4.x and later kernels support SSE.
check_os_katmai_support will disable SSE capabilities for 32-bit x86
operating systems for which there is no code path. Currently, this
function handles Linux, Windows, and several BSDs. Mac OS, Cygwin, and
Solaris are several operating systems with no code paths.
Rather than add code for the unhandled operating systems, remove this
function altogether. This will fix SSE detection on all recent 32-bit
x86 operating systems. This completely breaks functionality on unpatched
2.2.x and earlier kernels, although there are likely no Gallium3D users
on such operating systems.
|
|
This commit silences the printing off most of the debug information
when running debug builds. The big culprits are: the tgsi sanity checker
that gets run on all shaders on debug; all the options; and
finaly the cpu caps printer.
|
|
This logic duplicates the one in p_config.h, so remove it and adjust
the only two places that were using it.
|
|
This reverts commit de4784e36505316c2a5ab34cc5b371d17f38d3c5.
|
|
No need to enable depth test for clear.
|
|
|
|
Fixes MSVC build.
|
|
util_framebuffer_copy was attempting to copy all elements of the
source framebuffer state.
However, this breaks if the user does not zero initialize the structure.
Instead, only copy the elements up to nr_cbufs, and clear elements up
to dst->nr_cbufs, if the destination was larger than the source.
|
|
|
|
Direct3D 10/11 has no concept of transfers. Applications instead
create resources with a STAGING or DYNAMIC usage, copy between them
and the real resource and use Map to map the STAGING/DYNAMIC resource.
This util module allows to implement Gallium drivers as a Direct3D
driver would be implemented: transfers allocate a resource with
PIPE_USAGE_STAGING, and copy the data between it and the real resource
with resource_copy_region.
|
|
Used to find out if a surface exists without creating one.
|
|
Using cso_hash directly is the right thing since util_hash_table
adds useless overhead and is harder to use for this application.
|
|
|
|
They are unimplemented, even though the framework makes it possible to
implement them well, and nv50 needs them.
|
|
Current practice is to start identifiers with "util_" instead of "u_".
|
|
This is a simple framework that handles splitting primitives in an
abstract way.
The user has to specify the primitive start, start index and count.
Then, it can ask the primitive splitter to "draw" a chunk of the
primitive, staying under a given vertex/index budget.
The primitive splitter will then call user-supplied functions to
emit a range of vertices/indices, as well as switch the edgeflag
on or off.
This is particularly useful for hardware that either has limits
on the vertex count field, or where vertices are pushed on a FIFO
or temporary buffer of limited size.
Note that unlike other splitters, it does not manipulate data in
any way, and merely asks a callback to do so, in vertex intervals.
|
|
To convert RGB -> SRGB format.
|
|
It's bound to be useful elsewhere.
|
|
|
|
|
|
This implements fast Z clear, Z compression, and HiZ support for r300->r500
GPUs.
It also allows cbzb clears when fast Z clears are being used for the ZB.
It requires a kernel with hyper-z support.
Thanks to Marek Olšák <maraeo@gmail.com>, who started this off, and Alex Deucher at AMD for providing lots of hints.
v2:
squashed zmask ram size fix]
squashed r300g/blitter: fix Z readback when compressed]
v3:
rebase around texture changes in master - .1 fix more bits
v4:
migrated to using u_mm in r300_texture to manage hiz/zmask rams consistently
disabled HiZ when using OQ
flush z-cache before turning hyper-z off
update hyper-z state on dsa state change
store depthclearvalue across cbzb clears and replace it afterwards.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
|