Brian Paul [Wed, 21 Jul 2010 00:48:53 +0000 (18:48 -0600)]
gallivm: fix indirect addressing of constant buffer
The previous code assumed that all elements of the address register
were the same. But it can vary from pixel to pixel or vertex to
vertex so we must use a gather operation when dynamically indexing
the constant buffer.
Still need to fix this for the temporary register file...
Brian Paul [Wed, 21 Jul 2010 00:45:45 +0000 (18:45 -0600)]
gallivm: added lp_build_const_int32() helper
Vinson Lee [Tue, 20 Jul 2010 21:18:42 +0000 (14:18 -0700)]
radeon: Remove unnecessary header.
Jerome Glisse [Tue, 20 Jul 2010 19:04:55 +0000 (15:04 -0400)]
r600g: split trans unit computation
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Marek Olšák [Sat, 5 Jun 2010 17:32:29 +0000 (19:32 +0200)]
st/mesa: implement and advertise GL_ARB_draw_elements_base_vertex
Signed-off-by: Brian Paul <brianp@vmware.com>
Brian Paul [Tue, 20 Jul 2010 17:46:43 +0000 (11:46 -0600)]
draw: correctly handle max_index=0xffffffff case in vcache code
If max_index=0xffffffff and elt_bias > 0 the test for
elt_bias + max_index >= DRAW_PIPE_MAX_VERTICES
was wrong. Check earlier if max_index=0xffffffff and do the
"fail" case.
This fixes the piglit draw-elements-base-vertex test (and probably
some other things).
Marek Olšák [Mon, 19 Jul 2010 18:46:15 +0000 (20:46 +0200)]
st/mesa: implement depth texture modes
Brian Paul [Tue, 20 Jul 2010 14:40:19 +0000 (08:40 -0600)]
mesa: call ctx->Driver.ChooseTextureFormat() only when necessary.
When defining mipmap level 'L' and level L-1 exists and the new level's
internalFormat matches level L-1's internalFormat, then use the same hw
format. Otherwise, do the regular ctx->Driver.ChooseTextureFormat() call.
This avoids a problem where we end up choosing different hw formats for
different mipmap levels depending on how the levels are defined (glTexImage
vs. glCopyTexImage vs. glGenerateMipmap, etc).
The root problem is the ChooseTextureFormat() implementation in some
drivers uses the user's glTexImage format/type parameters in the choosing
heuristic. Later mipmap levels might be generated with different calls
(ex: glCopyTexImage()) so we don't always have format/type info and the
driver may choose a different format.
For more background info see the July 2010 mesa-dev thread "Bug in
_mesa_meta_GenerateMipmap"
Brian Paul [Tue, 20 Jul 2010 14:39:30 +0000 (08:39 -0600)]
graw/tests: pass -e option to test draw_elements_instanced()
Michal Krol [Tue, 20 Jul 2010 13:00:28 +0000 (15:00 +0200)]
softpipe: Support non-depth-stencil formats in sp_tile_cache_flush_clear().
Kristian Høgsberg [Tue, 20 Jul 2010 11:43:50 +0000 (07:43 -0400)]
glx: #ifdef DRI specific prototype
https://bugs.freedesktop.org/show_bug.cgi?id=29162
Michel Dänzer [Tue, 20 Jul 2010 10:21:17 +0000 (12:21 +0200)]
glx/dri2: Fix dri2CopySubBuffer() again.
Only refresh the fake front buffer if there is one, and only destroy the region
once.
Fixes X11 protocol errors reported by 'mcgreg' on IRC.
Marek Olšák [Mon, 19 Jul 2010 21:26:17 +0000 (23:26 +0200)]
util: mempool: initialize last block's magic number in a page
Chia-I Wu [Tue, 20 Jul 2010 03:49:59 +0000 (11:49 +0800)]
st/egl: Fixes for recent GLX cleanup.
Mainly, the type of __GLXdisplayPrivateRec::screenConfigs has changed
from "__GLXscreenConfigs *" to "__GLXscreenConfigs **".
Chia-I Wu [Tue, 20 Jul 2010 04:35:47 +0000 (12:35 +0800)]
st/dri: Remove driFrameTrackingExtension.
The extension has been removed in
22266c391fbe17603b15a83d4ccf5fa9455ccf8d.
Kristian Høgsberg [Tue, 20 Jul 2010 02:40:22 +0000 (22:40 -0400)]
glx: Move DRI CopySubBuffer extension to DRI1 code
We do this in the X server for DRI2.
Kristian Høgsberg [Tue, 20 Jul 2010 02:12:22 +0000 (22:12 -0400)]
glx: Move driver_configs to DRI screen privates
Kristian Høgsberg [Tue, 20 Jul 2010 01:15:50 +0000 (21:15 -0400)]
glx: Move drawHash to display private
The XIDs are display wide so the natural location of the hash is here.
This way we don't have to lookup in each of the screen hashes.
Kristian Høgsberg [Tue, 20 Jul 2010 01:00:09 +0000 (21:00 -0400)]
glx: Don't use __glXInitialize() when we might be holding __glXLock()
Kristian Høgsberg [Tue, 20 Jul 2010 00:01:59 +0000 (20:01 -0400)]
glx: Workaround mismatch in signedness between extensions and protocol
The DRI2 protocol for ust, msc and sbc are unsigned but the extensions
talk about int64_t. Do a little dance to make the compiler shut up.
Kristian Høgsberg [Mon, 19 Jul 2010 22:11:03 +0000 (18:11 -0400)]
glx: Move __DRIdrawable pointers to DRI drawable privates
Kristian Høgsberg [Mon, 19 Jul 2010 21:15:42 +0000 (17:15 -0400)]
glx: Remove support for MESA_swap_frame_usage
The extension never worked, the implementation returns GLX_BAD_CONTEXT
when enabling the frame tracking.
Kristian Høgsberg [Mon, 19 Jul 2010 20:45:23 +0000 (16:45 -0400)]
glx: Move __driScreen into the dri screen privates
Kristian Høgsberg [Mon, 19 Jul 2010 20:39:53 +0000 (16:39 -0400)]
glx: Move DRI1 specific extensions and code to DRI1 screen private
Kristian Høgsberg [Mon, 19 Jul 2010 19:35:48 +0000 (15:35 -0400)]
glx: Move DRI2 extensions to DRI2 screen private
Kristian Høgsberg [Mon, 19 Jul 2010 19:15:15 +0000 (15:15 -0400)]
glx: Drop support for GLX_MESA_allocate_memory
Only r200 implemented it.
Kristian Høgsberg [Mon, 19 Jul 2010 18:57:59 +0000 (14:57 -0400)]
glx: Add screen privates for dri drivers and moved some fields there
GLXscreenConfigs is badly named and a dumping ground for a lot of stuff.
This patch creates private screen structs for the dri drivers and moves
some of their fields over there.
Kristian Høgsberg [Mon, 19 Jul 2010 14:51:09 +0000 (10:51 -0400)]
glx: Rename various DRI structs away from obnoxious __GLXfooRec convention
Enough is enough.
Kristian Høgsberg [Mon, 19 Jul 2010 14:32:02 +0000 (10:32 -0400)]
glx: Factor out common code from dri2WaitGL() and dri2WaitX()
Kristian Høgsberg [Mon, 19 Jul 2010 14:19:22 +0000 (10:19 -0400)]
glx: Dont use dri2WaitX() to update fake front
This saves a superfluous flush and a create/destryo region.
Eric Anholt [Tue, 20 Jul 2010 00:22:20 +0000 (17:22 -0700)]
i965: Mostly fix glsl-max-varyings.
There was confusion on both the size of message we can send, and on
what the URB destination offset means.
The remaining problems appear to be due to spilling of regs in the
fragment shader being broken.
Eric Anholt [Tue, 20 Jul 2010 00:02:52 +0000 (17:02 -0700)]
i965: Clean up message register setup in emit_vertex_write().
Eric Anholt [Mon, 19 Jul 2010 23:44:38 +0000 (16:44 -0700)]
i965: Reduce repeated calculation of the attribute-offset-in-VUE.
This cleans up some chipset dependency sprinkled around, and fixes a
potential overflow of the attribute offset array for many vertex
results.
Eric Anholt [Mon, 19 Jul 2010 23:11:19 +0000 (16:11 -0700)]
i965: Clarify the nr_regs calculation in brw_clip.c
Eric Anholt [Mon, 19 Jul 2010 22:54:15 +0000 (15:54 -0700)]
i965: Don't set up VUE space for the disabled user clip distances on gen6.
Brian Paul [Tue, 20 Jul 2010 00:29:12 +0000 (18:29 -0600)]
draw: fix incorrect instancing divisor in LLVM code
Brian Paul [Mon, 19 Jul 2010 21:28:53 +0000 (15:28 -0600)]
mesa: remove restart.c from build
Marek Olšák [Mon, 19 Jul 2010 18:47:30 +0000 (20:47 +0200)]
r300g: fix possible crash in destroy_context
The problem is destroy_context is almost NEVER called.
The only test for destroy_context I know is compiz.
Reported by Vinson Lee.
FDO bug #29150.
Marek Olšák [Mon, 19 Jul 2010 16:08:53 +0000 (18:08 +0200)]
util: remove the dummy field in mempool
It should allocate less memory now.
Vinson Lee [Mon, 19 Jul 2010 18:31:44 +0000 (11:31 -0700)]
scons: Fix Mac OS X SCons build on 32-bit CPUs.
The Mac OS X SCons build failed on 32-bit CPUs starting with commit
2f6d47a7c8d6e69e5154de44115aab9ba35a41d2 during linking of graw-null.
The build succeeds though on a 64-bit CPU. See FDO bug 29117.
This was the compiler error.
scons: building associated VariantDir targets: build/darwin-x86-debug
Linking build/darwin-x86-debug/gallium/targets/graw-null/libgraw.dylib ...
Undefined symbols:
"_lp_swizzled_cbuf", referenced from:
_lp_swizzled_cbuf$non_lazy_ptr in libllvmpipe.a(lp_rast.os)
_lp_swizzled_cbuf$non_lazy_ptr in libllvmpipe.a(lp_rast_tri.os)
(maybe you meant: _lp_swizzled_cbuf$non_lazy_ptr)
"_lp_dummy_tile", referenced from:
_lp_dummy_tile$non_lazy_ptr in libllvmpipe.a(lp_rast.os)
_lp_dummy_tile$non_lazy_ptr in libllvmpipe.a(lp_rast_tri.os)
_lp_dummy_tile$non_lazy_ptr in libllvmpipe.a(lp_setup.os)
(maybe you meant: _lp_dummy_tile$non_lazy_ptr)
The patch adds -fno-common to all Mac OS X builds to work around this issue.
Brian Paul [Mon, 19 Jul 2010 16:50:08 +0000 (10:50 -0600)]
util: add dummy field to empty structure types
Empty structure types aren't allowed with MSVC.
I haven't tested this change. Hope I haven't broken it...
Marek Olšák [Mon, 19 Jul 2010 15:33:40 +0000 (17:33 +0200)]
r300g: fix typo
Marek Olšák [Mon, 19 Jul 2010 12:31:25 +0000 (14:31 +0200)]
r300g: use memory pools for buffer_create and get_transfer
The improvement in Tremulous: 68.9 fps -> 71.1 fps.
Marek Olšák [Mon, 19 Jul 2010 09:04:30 +0000 (11:04 +0200)]
util: add a memory pool for equally sized memory allocations
malloc/free are in O(1).
Marek Olšák [Sat, 17 Jul 2010 13:55:24 +0000 (15:55 +0200)]
st/mesa: fix FRAMEBUFFER_UNSUPPORTED with the D24S8 format
Fixes FDO bug #29116.
NOTE: this is a candidate for the 7.8 branch
Kristian Høgsberg [Mon, 19 Jul 2010 13:37:07 +0000 (09:37 -0400)]
glx: Fix drawable lookup in DRI2 event handler
DRI2 events are sent to the X drawable ID used to create the DRI2 drawable,
not the GLX drawable ID. So when an event comes in, we need to look up
the __GLXDRIdrawable by its X drawable ID, which needs a new hash table.
Dave Airlie [Sun, 18 Jul 2010 08:47:08 +0000 (18:47 +1000)]
r300g: u_upload optimisation
fix vb/ib uploads
Sven Arvidsson [Sun, 18 Jul 2010 09:17:14 +0000 (19:17 +1000)]
gallium/st/dri2: add dri2 vblank query extension support
from bugzilla:
https://bugs.freedesktop.org/show_bug.cgi?id=28771
Signed-off-by: Dave Airlie <airlied@redhat.com>
Vinson Lee [Sun, 18 Jul 2010 08:14:36 +0000 (01:14 -0700)]
llvmpipe: Remove dead initialization.
Marek Olšák [Sat, 17 Jul 2010 13:35:03 +0000 (15:35 +0200)]
r300g: final fix for r3xx constant buffer emission
Marek Olšák [Sat, 17 Jul 2010 12:40:41 +0000 (14:40 +0200)]
r300g: fix typo in r3xx constant buffer emission
Ooops.
Marek Olšák [Sat, 17 Jul 2010 11:37:14 +0000 (13:37 +0200)]
r300g: fix constant buffer emission on r3xx
FDO bug #29128.
Vinson Lee [Sat, 17 Jul 2010 07:35:10 +0000 (00:35 -0700)]
llvmpipe: Remove unused variable in lp_test_sincos.
Vinson Lee [Sat, 17 Jul 2010 01:41:32 +0000 (18:41 -0700)]
nv50: s/__func__/__FUNCTION__/
Vinson Lee [Sat, 17 Jul 2010 01:14:11 +0000 (18:14 -0700)]
nouveau: s/__func__/__FUNCTION__/
Vinson Lee [Sat, 17 Jul 2010 01:03:03 +0000 (18:03 -0700)]
nouveau: s/snprintf/util_snprintf/
Vinson Lee [Sat, 17 Jul 2010 00:55:57 +0000 (17:55 -0700)]
r300g: Remove unnecessary header.
Vinson Lee [Sat, 17 Jul 2010 00:49:18 +0000 (17:49 -0700)]
nouveau: s/inline/INLINE/
José Fonseca [Fri, 16 Jul 2010 20:48:36 +0000 (21:48 +0100)]
glut: Remove duplicate symbol definition.
José Fonseca [Fri, 16 Jul 2010 20:26:43 +0000 (21:26 +0100)]
scons: Make PIPE_ALIGN_VAR() of static/global vars work on MinGW.
Workaround http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37216
Vinson Lee [Fri, 16 Jul 2010 19:34:22 +0000 (12:34 -0700)]
scons: Fix Cygwin build.
The Cygwin SCons build needed several file names to be fully qualified.
Marek Olšák [Fri, 16 Jul 2010 10:54:11 +0000 (12:54 +0200)]
r300g: inline winsys_buffer_destroy
Marek Olšák [Fri, 16 Jul 2010 10:53:40 +0000 (12:53 +0200)]
r300g: do not make copies of constant buffers, emit them directly
Brian Paul [Fri, 16 Jul 2010 17:10:25 +0000 (11:10 -0600)]
draw: updated debug/dump code
Brian Paul [Fri, 16 Jul 2010 17:09:17 +0000 (11:09 -0600)]
draw: added array element debug / bounds checking code (disabled)
Keith Whitwell [Fri, 16 Jul 2010 13:40:30 +0000 (14:40 +0100)]
llvmpipe: use single swizzled tile
Use a single swizzled tile per colorbuf (and per thread) to avoid
accumulating large amounts of cached swizzled data.
Now that the SSE3 code has been merged to master, the performance delta
of this change is minimal, the main benefit is reduced memory usage
due to no longer keeping swizzled copies of render targets.
It's clear from the performance of the in-place version of this code
that there is still quite a bit of time being spent swizzling &
unswizzling, but it's not clear exactly how to reduce that.
José Fonseca [Fri, 16 Jul 2010 12:16:57 +0000 (13:16 +0100)]
llvmpipe: Describe _mm_shuffle_epi8() with gcc extended inline assembly when -mssse3 is not supported/enabled.
José Fonseca [Thu, 15 Jul 2010 17:22:39 +0000 (18:22 +0100)]
llvmpipe: Only use -mssse3 on gcc 4.3+
Zack Rusin [Fri, 16 Jul 2010 16:19:26 +0000 (12:19 -0400)]
draw/llvm: adjust the instance id at run time
fixes instancing in draw llvm
Zack Rusin [Fri, 16 Jul 2010 15:58:20 +0000 (11:58 -0400)]
draw: use the instance id when fetching vertex data
Brian Paul [Fri, 16 Jul 2010 13:36:19 +0000 (07:36 -0600)]
mesa: return retval in _mesa_RenderObjectUnpurgeable()
Found by Vinson with static analysis.
NOTE: This is a candidate for the 7.8 branch.
Chia-I Wu [Fri, 16 Jul 2010 12:09:29 +0000 (20:09 +0800)]
st/egl: Fix build on FreeBSD.
There is no libdl on FreeBSD. Based on patch from Thinker
<thinker@branda.to>, which is against 7.8.
This fixes fdo bug #29093.
Chia-I Wu [Fri, 16 Jul 2010 11:48:52 +0000 (19:48 +0800)]
egl: Build egl_dri2 only when DRI drivers are built.
That is, build egl_dri2 only when --with-driver=dri is given (the
default).
Chia-I Wu [Fri, 16 Jul 2010 11:39:58 +0000 (19:39 +0800)]
egl: Build egl_dri2 only when xcb-dri2 is available.
The driver does not build when xcb-dri2 is not available.
Vinson Lee [Fri, 16 Jul 2010 08:11:03 +0000 (01:11 -0700)]
glu/sgi: Remove dead initialization in extract565rev.
Vinson Lee [Fri, 16 Jul 2010 08:07:13 +0000 (01:07 -0700)]
i965g: Remove dead initialization in precalc_tex.
Marek Olšák [Tue, 13 Jul 2010 23:59:57 +0000 (01:59 +0200)]
r300g: rebuild winsys and command submission to support multiple contexts
Vinson Lee [Fri, 16 Jul 2010 06:45:25 +0000 (23:45 -0700)]
mesa: Add error path in compressed_texture_error_check.
Add error path for unhandled dimensions in
compressed_texture_error_check.
Vinson Lee [Fri, 16 Jul 2010 01:45:20 +0000 (18:45 -0700)]
glsl/apps: Handle ftell errors in non-debug builds.
Brian Paul [Thu, 15 Jul 2010 21:47:03 +0000 (15:47 -0600)]
llvmpipe: implement instanced drawing functions
And express all the other drawing functions in terms of
llvmpipe_draw_range_elements_instanced().
Brian Paul [Thu, 15 Jul 2010 21:41:48 +0000 (15:41 -0600)]
draw: update comments for drawing functions
Brian Paul [Thu, 15 Jul 2010 21:41:15 +0000 (15:41 -0600)]
draw: move prototype, update comment
Brian Paul [Thu, 15 Jul 2010 21:40:52 +0000 (15:40 -0600)]
softpipe: re-order drawing functions to get rid of prototype
Brian Paul [Thu, 15 Jul 2010 21:39:56 +0000 (15:39 -0600)]
graw: new tri-instanced.c program to test instanced drawing
Alex Deucher [Thu, 15 Jul 2010 21:03:58 +0000 (17:03 -0400)]
r600: fix typo in r700 assembler
Noticed by Henri Verbeet on IRC.
NOTE: This is a candidate for the 7.8 branch.
Henri Verbeet [Thu, 15 Jul 2010 18:53:16 +0000 (14:53 -0400)]
radeon: Also flush if it's not the current context that's being destroyed.
This avoids calling radeonFlush() during context destruction, when
ctx->DrawBuffer would be NULL.
NOTE: This is a candidate for the 7.8 branch.
Alex Deucher [Thu, 15 Jul 2010 18:47:06 +0000 (14:47 -0400)]
radeon: allow driconf vblank settings with dri2
fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=28771
NOTE: This is a candidate for the 7.8 branch.
Brian Paul [Thu, 15 Jul 2010 16:46:33 +0000 (10:46 -0600)]
st/mesa: fix quad strip trimming bug
The translate_prim() function tries to convert quad strips into
tri strips. This is normally OK but we have to check for an odd
number of vertices so that we don't accidentally draw an extra
triangle. The mesa-demos/src/samples/prim.c demo exercises that.
With this fix the stray yellow triangle is no longer drawn.
Use the u_trim_pipe_prim() function to make sure that prims have
the right number of vertices and avoid calling gallium drawing
functions when the prim has a degenerate number of vertices.
Plus add comments, clean-up formatting, etc.
NOTE: This is a candidate for the 7.8 branch.
José Fonseca [Thu, 15 Jul 2010 09:59:03 +0000 (10:59 +0100)]
llvmpipe: Remove redundant statement.
Thanks to Vinson for spotting this.
Vinson Lee [Thu, 15 Jul 2010 08:42:02 +0000 (01:42 -0700)]
glu/sgi: Remove dead initialization in extract565.
Vinson Lee [Thu, 15 Jul 2010 08:26:09 +0000 (01:26 -0700)]
tgsi: Remove dead assignment in uprcase function.
Vinson Lee [Thu, 15 Jul 2010 07:53:07 +0000 (00:53 -0700)]
mesa: Fix potential out-of-bounds access by _vbo_Materialf.
_vbo_Materialf calls _vbo_Materialfv, which uses the params argument as
an array.
Vinson Lee [Thu, 15 Jul 2010 07:20:41 +0000 (00:20 -0700)]
glx: Move dereference and initialization to after NULL check.
Chia-I Wu [Wed, 14 Jul 2010 16:31:39 +0000 (00:31 +0800)]
egl: Return the correct array size in _eglFlattenArray.
The function is used by _eglGetConfigs and _eglGetScreens. The array
size should not be limited by the buffer size when the buffer is NULL.
This fixes fdo bug #29052.
Brian Paul [Wed, 14 Jul 2010 21:11:32 +0000 (15:11 -0600)]
mesa: fix _mesa_Texture/Render/BufferObjectUnpurgeable() return values
Fixes piglit object_purgeable-api-pbo, object_purgeable-api-vbo
and object_purgeable-api-texture failures with swrast.
NOTE: This is a candidate for the 7.8 branch.
Brian Paul [Wed, 14 Jul 2010 20:50:29 +0000 (14:50 -0600)]
mesa: update assertions and fix refcounting in depth/stencil renderbuffer code
Brian Paul [Wed, 14 Jul 2010 20:36:25 +0000 (14:36 -0600)]
mesa: silence a printf warning
Brian Paul [Wed, 14 Jul 2010 20:35:21 +0000 (14:35 -0600)]
llvmpipe: delete lp_test_*.o files with make clean
Brian Paul [Wed, 14 Jul 2010 20:34:45 +0000 (14:34 -0600)]
gallium: added CLEAN_EXTRA var for make clean target
Nick Bowler [Wed, 14 Jul 2010 16:01:49 +0000 (12:01 -0400)]
dri2: Track event mask in client code.
When direct rendering is being used, DRI2 BufferSwapComplete events are
sent unconditionally to clients, even if they haven't been requested.
This causes error messages to be printed by every freeglut application
of the form
freeglut (./gears): Unknown X event type: 104
and might confuse other clients.
This is a fixed up version of the patch by Jesse Barnes, which drops
BufferSwapComplete events if they are not requested by clients.
Fixes fdo bug 27962.
Signed-off-by: Nick Bowler <nbowler@draconx.ca>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>