Currently we only unroll loops with conditional breaks at the end, which is
the form that lower_jumps generates.
However, if breaks are not lowered, they tend to appear at the beginning, so
add support for a conditional break anywhere.
Signed-off-by: Luca Barbieri <luca@luca-barbieri.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Found this bug by code inspection. Based off the comments just before
this code, the intent is to find whether the break exists in the "then"
branch or the "else" branch. However, the code actually looked at the
last instruction in the "then" branch twice.
If you used a constant array index to access the matrix, we'd flag a
bunch of wrong inputs/outputs as being used because the index was
multiplied by matrix columns and the actual used index was left out.
Fixes glsl-mat-attribute.
Fixes this GCC warning.
brw_fs.cpp: In function 'brw_reg brw_reg_from_fs_reg(fs_reg*)':
brw_fs.cpp:3255: warning: 'brw_reg' may be used uninitialized in this function
Allow important performance increase by doing hw specific implementation
of the upload manager helper. Drop the range flushing that is not hit with
this code (and wasn't with previous neither). Performance improvement are
mostly visible on slow CPU.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
r600g is up to a point where all small CPU cycle matter and pb* turn
high on profile. It's mostly because pb try to be generic and thus
trigger unecessary check for r600g driver. To avoid having too much
abstraction & too much depth in the call embedded everythings into
r600_bo. Make code simpler & faster. The performance win highly depend
on the CPU & application considered being more important on slower CPU
and marginal/unoticeable on faster one.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
This eases the gen6 implementation, which can only handle up to 32
registers of constants, while likely not penalizing real apps using
reladdr since all of those I've seen also end up hitting the pull
constant buffer. On gen6, the constant map means that simple NV VPs
fit under the 32-reg limit and now succeed. Fixes around 10 testcases.
Raise error if a sampler array is indexed with a non-constant expression.
From section 4.1.7 of the GLSL 1.30 spec:
"Samplers aggregated into arrays within a shader (using square
brackets [ ]) can only be indexed with integral constant
expressions [...]."
Attribute 0 has no special meaning in GLES2. Add VertexAttrib4f_nopos
for that purpose and make _es_VertexAttrib* call the new function.
Rename _vbo_* to _es_* to avoid confusion. These functions are only
used by GLES, and now some of them (_es_VertexAttrib*) even behave
differently than vbo_VertexAttrib*.
CMP may now use two less temps, other non-native instructions may end up
using one less temp, except for SIN/COS/SCS, which I am leaving unchanged
for now.
This may reduce register pressure inside loops, because the register
allocator doesn't do a very good job there.
That's what I get for not running piglit before pushing.
Don't try to patch types of unsized arrays when linking fails.
Don't try to patch types of unsized arrays that are shared between
shader stages.
Fixes piglit test case glsl-vec-array (bugzilla #31908).
NOTE: This bug does not affect 7.9, but I think this patch is a
candiate for the 7.9 branch anyway.
Types of declared variables and their initializer must match excatly
except for unsized arrays. Previously the type inherritance for
unsized arrays happened implicitly in the emitted assignment.
However, this assignment is never emitted for uniforms. Now that type
is explicitly copied unconditionally.
Fixes piglit test array-compare-04.vert (bugzilla #32035) and
glsl-array-uniform-length (bugzilla #31985).
NOTE: This is a candidate for the 7.9 branch.
With the change of extended math from having the arguments moved into
mrfs and handed off through message passing to being directly hooked
up to the EU, it looks like the piece for doing source modifiers
(negate and abs) was left out.
Fixes:
fog-modes
glean/fp1-ARB_fog_exp test
glean/fp1-ARB_fog_exp2 test
glean/fp1-Computed fog exp test
glean/fp1-Computed fog exp2 test
ext_fog_coord-modes
Previously the code only handled scalars and vectors. This new code
is modeled somewhat after similar code in ir_to_mesa.
Reviewed-by: Eric Anholt <eric@anholt.net>
R6XX GPU doesn't like to have two partial flush writting
back to memory in row without a prior flush of the pipeline.
Add PS_PARTIAL_FLUSH to flush all work between the CP and
the ES, GS, VS, PS shaders.
Thanks a lot to Alban Browaeys (prahal on irc) for investigating
this issue.
Signed-off-by: Alban Browaeys <prahal@yahoo.com>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
gen6 builtin RSQ apparently clamps negative values to 0 instead of
returning the RSQ of the absolute value like ARB_fragment_program
desires and pre-gen6 apparently does.
Fixes:
glean/fp1-RSQ test 2 (reciprocal square root of negative value)
glean/vp1-RSQ test 2 (reciprocal square root of negative value)
Fixes this GCC warning.
r200_maos_arrays.c: In function 'r200EmitArrays':
r200_maos_arrays.c:113: warning: 'emitsize' may be used uninitialized in this function
It's not always possible to preprocess the content of 3D_LOAD_VBPNTR
in a command buffer, because the offset to all vertex buffers (which
the packet depends on) is derived from the "start" parameter of draw_arrays
and the "indexBias" parameter of draw_elements, but we can at least lazily
make a command buffer for the case when offset == 0, which should occur
most of the time.
The original lazy built-in importing patch did not add the newly created
function to the symbol table, nor actually emit it into the IR stream.
Adding it to the symbol table is non-trivial since importing occurs when
generating some ir_call in a nested scope. A new add_global_function
method, backed by new symbol_table code in the previous patch, handles
this.
Fixes bug #32030.
Avoid rebuilding constant shader state at each draw call,
factor out spi update that might change at each draw call.
Best would be to update spi only when revealent states
change (likely only flat shading & sprite point).
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Vertex elements change are less frequent than draw call, those to
avoid rebuilding fetch shader to often build the fetch shader along
vertex elements. This also allow to move vertex buffer setup out
of draw path and make update to it less frequent.
Shader update can still be improved to only update SPI regs (based
on some rasterizer state like flat shading or point sprite ...).
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
More than 1023 temporaries were being used for a Cinebench shader before
doing temporary optimization, causing the index value to wrap around to
-1024.
In st_finalize_texture() we were looking at the st_texture_object::
lastLevel field instead of the pipe_resource::last_level field to
determine which resource to store the mipmap in.
Then, in st_generate_mipmap() we need to call st_finalize_texture() to
make sure the destination resource is properly allocated.
These changes fix the broken piglit fbo-generatemipmap-formats test.
Use --cppflags instead of --cflags so that we get the -I and -D flags
we want, but not compiler options like -O3.
A similar change should probably be made for autoconf.
When X or Y or Z is close to W the outcome of the floating point clip
test comparision may be different between the C and x86 asm paths.
That's OK; don't report an error.
See fd.o bug 32093
It was only used for gen6 fragment programs (not GLSL shaders) at this
point, and it was clearly unsuited to the task -- missing opcodes,
corrupted texturing, and assertion failures hit various applications
of all sorts. It was easier to patch up the non-glsl for remaining
gen6 changes than to make brw_wm_glsl.c complete.
Bug #30530
Since the 8-wide first-quarter and 16-wide first-half have the same
bit encoding, we now need to track "do you want instruction
compression" in the compile state.
The FS backend is fine with register level granularity. But for the
brw_wm_emit.c backend, it expects pairs of regs to be used for the
constants, because the whole world is pairs of regs. If an odd number
got used, we went looking for interpolation in the wrong place.
In the SF and brw_fs.cpp fixes to set up interpolation sanely on gen6,
the setup for 16-wide interpolation was left behind. This brings
relative sanity to that path too.
Payload reg setup on gen6 depends more on the dispatch width as well
as the uses_depth, computes_depth, and other flags. That's something
we want to decide at compile time, not at cache lookup. As a bonus,
the fragment shader program cache lookup should be cheaper now that
there's less to compute for the hash key.
The preprocessor magic in mapi was nothing but obfuscation. Rewrite
mapi_abi.py to generate real C code.
This commit removes the hack added in
43121f2086.
Need to check the required primitive type for GS on Sandybridge,
and when GS is disabled, the new state has to be issued too, instead
of only updating URB state with no GS entry, that caused hang on Sandybridge.
This fixes hang issue during conformance suite testing.
This fixes endless vertex shader recompilations in find_translated_vp
if the shader contains an edge flag output.
NOTE: This is a candidate for the 7.9 branch.
Signed-off-by Brian Paul <brianp@vmware.com>
Finished up by Marek Olšák.
We can set the constant space to use a different area per-call to the shader,
we can avoid flushing the PVS as often as we do by spreading out the constants
across the whole constant space.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
It appears to be a constant buffer index (in case there are more constant
buffers explicitly used by a shader), i.e. something that Gallium currently
does not use. We treated it incorrectly as the offset to a constant buffer.
Sometimes I'm on the train and want to just read what's generated
under INTEL_DEBUG=vs,wm for some code on another generation. Or, for
the next gen enablement we'll want to dump aub files before we have
the actual hardware. This will let us do that.
rgb_src_factor and rgb_dst_factor should be PIPE_BLENDFACTOR_ONE for
VG_BLEND_SRC_IN and VG_BLEND_DST_IN respectively. VG_BLEND_SRC_OVER can
be supported only when the fb has no alpha channel. VG_BLEND_DST_OVER
and VG_BLEND_ADDITIVE have to be supported with a shader.
Note that Porter-Duff blending rules assume premultiplied alpha.
Convert color values to and back from premultiplied form for blending.
Finally the rendering result of the blend demo looks much closer to that
of the reference implementation.
Drawing an image in VG_DRAW_IMAGE_STENCIL mode produces per-channel
alpha for use in blending. Add a new shader stage to produce and save
it in TEMP[1].
For other modes that do not need per-channel alpha, the stage does
MOV TEMP[1], TEMP[0].wwww
Add a helper function, blend_generic, that supports all blend modes and
per-channel alpha. Make other blend generators a wrapper to it.
Both the old and new code expects premultiplied colors, yet the input is
non-premultiplied. Per-channel alpha is also not used for stencil
image. They still need to be fixed.
In find_value() check if we've hit the 0th/invalid entry before checking
if the pname matches.
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=31987
NOTE: This is a candidate for the 7.9 branch.
If querying the default/window-system FBO's attachment type, return
GL_FRAMEBUFFER_DEFAULT (per the GL_ARB_framebuffer_object spec).
See http://bugs.freedesktop.org/show_bug.cgi?id=31947
NOTE: This is a candidate for the 7.9 branch.
Return 0 instead of generating an error.
See http://bugs.freedesktop.org/show_bug.cgi?id=30993
Note that piglit fbo-getframebufferattachmentparameter-01 still does
not pass. But Mesa behaves the same as the NVIDIA driver in this case.
Perhaps the test is incorrect.
NOTE: This is a candidate for the 7.9 branch.
It was done in path-to-polygon conversion. That meant that the
results were invalidated when the transformation was modified, and CPU
had to recreate the vertex buffer with new vertices. It could be a
performance hit for apps that animate.
gl_FragCoord.y needs to be flipped upside down if a FBO is bound.
This fixes:
- piglit/fbo-fragcoord
- https://bugs.freedesktop.org/show_bug.cgi?id=29420
Here I add a new program state STATE_FB_WPOS_Y_TRANSFORM, which is set based
on whether a FBO is bound. The state contains a pair of transformations.
It can be either (XY=identity, ZW=transformY) if a FBO is bound,
or (XY=transformY, ZW=identity) otherwise, where identity = (1, 0),
transformY = (-1, height-1).
A classic driver (or st/mesa) may, based on some other state, choose whether
to use XY or ZW, thus negate the conditional "if (is a FBO bound) ...".
The reason for this is that a Gallium driver is allowed to only support WPOS
relative to either the lower left or the upper left corner, so we must flip
the Y axis accordingly again. (the "invert" parameter in emit_wpos_inversion)
NOTE: This is a candidate for the 7.9 branch.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Brian Paul <brianp@vmware.com>
Fixes nasty bug where some parts of the code didn't define WIN32_THREADS
and were using the integer mutex implementation, causing even confusion
to the debuggers.
And there is little interest of other thread implemenation on Win32
besides Win32 threads.
This allows 16K x 16K 2D textures, for example, but we don't want to
allow that for 3D textures. The new gl_constants::MaxTextureMBytes
field is used to prevent allocating too large of texture image.
This allows a 16K x 32 x 32 3D texture, for example, but prevents 16K^3.
Drivers can override this limit. The default is currently 1GB.
Apps should use the proxy texture mechanism to determine the actual
max texture size.
resources have a array_size parameter now.
get_tex_surface and tex_surface_destroy have been renamed to create_surface
and surface_destroy and moved to context, similar to sampler views (and
create_surface now uses a template just like create_sampler_view). Surfaces
now really should only be used for rendering. In particular they shouldn't be
used as some kind of 2d abstraction for sharing a texture. offset/layout fields
don't make sense any longer and have been removed, width/height should go too.
surfaces and sampler views now specify a layer range (for texture resources),
layer is either array slice, depth slice or cube face.
pipe_subresource is gone array slices (or cube faces) are now treated the same
as depth slices in transfers etc. (that is, they use the z coord of the
respective functions).
Squashed commit of the following:
commit a45bd509014743d21a532194d7b658a1aeb00cb7
Merge: 1aeca28 32e1e59
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Dec 2 04:32:06 2010 +0100
Merge remote branch 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/drivers/i915/i915_resource_texture.c
src/gallium/drivers/i915/i915_state_emit.c
src/gallium/drivers/i915/i915_surface.c
commit 1aeca287a827f29206078fa1204715a477072c08
Merge: 912f042 6f7c8c3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Dec 2 00:37:11 2010 +0100
Merge remote branch 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/state_trackers/vega/api_filters.c
src/gallium/state_trackers/vega/api_images.c
src/gallium/state_trackers/vega/mask.c
src/gallium/state_trackers/vega/paint.c
src/gallium/state_trackers/vega/renderer.c
src/gallium/state_trackers/vega/st_inlines.h
src/gallium/state_trackers/vega/vg_context.c
src/gallium/state_trackers/vega/vg_manager.c
commit 912f042e1d439de17b36be9a740358c876fcd144
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Dec 1 03:01:55 2010 +0100
gallium: even more compile fixes after merge
commit 6fc95a58866d2a291def333608ba9c10c3f07e82
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Dec 1 00:22:26 2010 +0100
gallium: some fixes after merge
commit a8d5ffaeb5397ffaa12fb422e4e7efdf0494c3e2
Merge: f7a202f 2da02e7
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Nov 30 23:41:26 2010 +0100
Merge remote branch 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/drivers/i915/i915_state_emit.c
src/gallium/state_trackers/vega/api_images.c
src/gallium/state_trackers/vega/vg_context.c
commit f7a202fde2aea2ec78ef58830f945a5e214e56ab
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Nov 24 19:19:32 2010 +0100
gallium: even more fixes/cleanups after merge
commit 6895a7f969ed7f9fa8ceb788810df8dbcf04c4c9
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Nov 24 03:07:36 2010 +0100
gallium: more compile fixes after merge
commit af0501a5103b9756bc4d79167bd81051ad6e8670
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Nov 23 19:24:45 2010 +0100
gallium: lots of compile fixes after merge
commit 0332003c2feb60f2a20e9a40368180c4ecd33e6b
Merge: 26c6346 b6b91fa
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Nov 23 17:02:26 2010 +0100
Merge remote branch 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/auxiliary/gallivm/lp_bld_sample.c
src/gallium/auxiliary/util/u_blit.c
src/gallium/auxiliary/util/u_blitter.c
src/gallium/auxiliary/util/u_inlines.h
src/gallium/auxiliary/util/u_surface.c
src/gallium/auxiliary/util/u_surfaces.c
src/gallium/docs/source/context.rst
src/gallium/drivers/llvmpipe/lp_rast.c
src/gallium/drivers/nv50/nv50_state_validate.c
src/gallium/drivers/nvfx/nv04_surface_2d.c
src/gallium/drivers/nvfx/nv04_surface_2d.h
src/gallium/drivers/nvfx/nvfx_buffer.c
src/gallium/drivers/nvfx/nvfx_miptree.c
src/gallium/drivers/nvfx/nvfx_resource.c
src/gallium/drivers/nvfx/nvfx_resource.h
src/gallium/drivers/nvfx/nvfx_state_fb.c
src/gallium/drivers/nvfx/nvfx_surface.c
src/gallium/drivers/nvfx/nvfx_transfer.c
src/gallium/drivers/r300/r300_state_derived.c
src/gallium/drivers/r300/r300_texture.c
src/gallium/drivers/r600/r600_blit.c
src/gallium/drivers/r600/r600_buffer.c
src/gallium/drivers/r600/r600_context.h
src/gallium/drivers/r600/r600_screen.c
src/gallium/drivers/r600/r600_screen.h
src/gallium/drivers/r600/r600_state.c
src/gallium/drivers/r600/r600_texture.c
src/gallium/include/pipe/p_defines.h
src/gallium/state_trackers/egl/common/egl_g3d_api.c
src/gallium/state_trackers/glx/xlib/xm_st.c
src/gallium/targets/libgl-gdi/gdi_softpipe_winsys.c
src/gallium/targets/libgl-gdi/libgl_gdi.c
src/gallium/tests/graw/tri.c
src/mesa/state_tracker/st_cb_blit.c
src/mesa/state_tracker/st_cb_readpixels.c
commit 26c6346b385929fba94775f33838d0cceaaf1127
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Aug 2 19:37:21 2010 +0200
fix more merge breakage
commit b30d87c6025eefe7f6979ffa8e369bbe755d5c1d
Merge: 9461bf3 1f1928d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Aug 2 19:15:38 2010 +0200
Merge remote branch 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/drivers/llvmpipe/lp_rast.c
src/gallium/drivers/llvmpipe/lp_rast_priv.h
src/gallium/drivers/r300/r300_blit.c
src/gallium/drivers/r300/r300_screen_buffer.c
src/gallium/drivers/r300/r300_state_derived.c
src/gallium/drivers/r300/r300_texture.c
src/gallium/drivers/r300/r300_texture.h
src/gallium/drivers/r300/r300_transfer.c
src/gallium/drivers/r600/r600_screen.c
src/gallium/drivers/r600/r600_state.c
src/gallium/drivers/r600/r600_texture.c
src/gallium/drivers/r600/r600_texture.h
src/gallium/state_trackers/dri/common/dri1_helper.c
src/gallium/state_trackers/dri/sw/drisw.c
src/gallium/state_trackers/xorg/xorg_exa.c
commit 9461bf3cfb647d2301364ae29fc3084fff52862a
Merge: 17492d7 0eaccb3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jul 15 20:13:45 2010 +0200
Merge commit 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/auxiliary/util/u_blitter.c
src/gallium/drivers/llvmpipe/lp_rast.c
src/gallium/drivers/llvmpipe/lp_surface.c
src/gallium/drivers/r300/r300_render.c
src/gallium/drivers/r300/r300_state.c
src/gallium/drivers/r300/r300_texture.c
src/gallium/drivers/r300/r300_transfer.c
src/gallium/tests/trivial/quad-tex.c
commit 17492d705e7b7f607b71db045c3bf344cb6842b3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Jun 18 10:58:08 2010 +0100
gallium: rename element_offset/width fields in views to first/last_element
This is much more consistent with the other fields used there
(first/last level, first/last layer).
Actually thinking about removing the ugly union/structs again and
rename first/last_layer to something even more generic which could also
be used for buffers (like first/last_member) without inducing headaches.
commit 1b717a289299f942de834dcccafbab91361e20ab
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 17 14:46:09 2010 +0100
gallium: remove PIPE_SURFACE_LAYOUT_LINEAR definition
This was only used by the layout field of pipe_surface, but this
driver internal stuff is gone so there's no need for this driver independent
layout definition neither.
commit 10cb644b31b3ef47e6c7b55e514ad24bb891fac4
Merge: 5691db9 c85971d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 17 12:20:41 2010 +0100
Merge commit 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/docs/source/glossary.rst
src/gallium/tests/graw/fs-test.c
src/gallium/tests/graw/gs-test.c
commit 5691db960ca3d525ce7d6c32d9c7a28f5e907f3b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 17 11:29:03 2010 +0100
st/wgl: fix interface changes bugs
commit 2303ec32143d363b46e59e4b7c91b0ebd34a16b2
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 16 19:42:32 2010 +0100
gallium: adapt code to interface changes...
commit dcae4f586f0d0885b72674a355e5d56d47afe77d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 16 19:42:05 2010 +0100
gallium: separate depth0 and array_size in the resource itself.
These fields are still mutually exclusive (since no 3d array textures exist)
but it ultimately seemed to error-prone to adapt all code accept the new
meaning of depth0 (drivers stick that into hardware regs, calculate mipmap
sizes etc.). And it isn't really cleaner anyway.
So, array textures will have depth0 of 1, but instead use array_size,
3D textures will continue to use depth0 (and have array_size of 1). Cube
maps also will use array_size to indicate their 6 faces, but since all drivers
should just be fine by inferring this themselves from the fact it's a cube map
as they always used to nothing should break.
commit 621737a638d187d208712250fc19a91978fdea6b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 16 17:47:38 2010 +0100
gallium: adapt code to interface changes
There are still usages of pipe_surface where pipe_resource should be used,
which should eventually be fixed.
commit 2d17f5efe166b2c3d51957c76294165ab30b8ae2
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 16 17:46:14 2010 +0100
gallium: more interface changes
In particular to enable usage of buffers in views, and ability to use a
different pipe_format in pipe_surface.
Get rid of layout and offset parameter in pipe_surface - the former was
not used in any (public) code anyway, and the latter should either be computed
on-demand or driver can use subclass of pipe_surface.
Also make create_surface() use a template to be more consistent with
other functions.
commit 71f885ee16aa5cf2742c44bfaf0dc5b8734b9901
Merge: 3232d11 8ad410d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 14 14:19:51 2010 +0100
Merge commit 'origin/master' into gallium-array-textures
Conflicts:
src/gallium/auxiliary/util/u_box.h
src/gallium/drivers/nv50/nv50_surface.c
src/gallium/drivers/nvfx/nvfx_surface.c
src/gallium/drivers/r300/r300_blit.c
src/gallium/drivers/r300/r300_texture.c
src/gallium/drivers/r300/r300_transfer.c
src/gallium/drivers/r600/r600_blit.c
src/gallium/drivers/r600/r600_screen.h
src/gallium/include/pipe/p_state.h
commit 3232d11fe3ebf7686286013c357b404714853984
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 14 11:40:04 2010 +0100
mesa/st: adapt to interface changes
still need to fix pipe_surface sharing
(as that is now per-context).
Also broken is depth0 handling - half the code assumes
this is also used for array textures (and hence by extension
of that cube maps would have depth 6), half the code does not...
commit f433b7f7f552720e5eade0b4078db94590ee85e1
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 14 11:35:52 2010 +0100
gallium: fix a couple of bugs in interface chnage fixes
commit 818366b28ea18f514dc791646248ce6f08d9bbcf
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:42:11 2010 +0200
targets: adapt to interface changes
Yes even that needs adjustments...
commit 66c511ab1682c9918e0200902039247793acb41e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:41:13 2010 +0200
tests: adapt to interface changes
Everything needs to be fixed :-(.
commit 6b494635d9dbdaa7605bc87b1ebf682b138c5808
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:39:50 2010 +0200
st: adapt non-rendering state trackers to interface changes
might not be quite right in all places, but they really don't want
to use pipe_surface.
commit 00c4289a35d86e4fe85919ec32aa9f5ffe69d16d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:38:48 2010 +0200
winsys: adapt to interface changes
commit 39d858554dc9ed5dbc795626fec3ef9deae552a0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:26:54 2010 +0200
st/python: adapt to interface changes
don't think that will work, sorry.
commit 6e9336bc49b32139cec4e683857d0958000e15e3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:26:07 2010 +0200
st/vega: adapt to interface changes
commit e07f2ae9aaf8842757d5d50865f76f8276245e11
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:25:56 2010 +0200
st/xorg: adapt to interface changes
commit 05531c10a74a4358103e30d3b38a5eceb25c947f
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:24:53 2010 +0200
nv50: adapt to interface changes
commit 97704f388d7042121c6d496ba8c003afa3ea2bf3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:24:45 2010 +0200
nvfx: adapt to interface changes
commit a8a9c93d703af6e8f5c12e1cea9ec665add1abe0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:24:01 2010 +0200
i965g: adapt to interface changes
commit 0dde209589872d20cc34ed0b237e3ed7ae0e2de3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:22:38 2010 +0200
i915g: adapt to interface changes
commit 5cac9beede69d12f5807ee1a247a4c864652799e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:20:58 2010 +0200
svga: adapt to interface changes
resource_copy_region still looking fishy.
Was not very suited to unified zslice/face approach...
commit 08b5a6af4b963a3e4c75fc336bf6c0772dce5150
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:20:01 2010 +0200
rbug: adapt to interface changes
Not sure if that won't need changes elsewhere?
commit c9fd24b1f586bcef2e0a6e76b68e40fca3408964
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:19:31 2010 +0200
trace: adapt to interface changes
commit ed84e010afc5635a1a47390b32247a266f65b8d1
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:19:21 2010 +0200
failover: adapt to interface changes
commit a1d4b4a293da933276908e3393435ec4b43cf201
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:19:12 2010 +0200
identity: adapt to interface changes
commit a8dd73e2c56c7d95ffcf174408f38f4f35fd2f4c
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:18:55 2010 +0200
softpipe: adapt to interface changes
commit a886085893e461e8473978e8206ec2312b7077ff
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:18:44 2010 +0200
llvmpipe: adapt to interface changes
commit 70523f6d567d8b7cfda682157556370fd3c43460
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:18:14 2010 +0200
r600g: adapt to interface changes
commit 3f4bc72bd80994865eb9f6b8dfd11e2b97060d19
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:18:05 2010 +0200
r300g: adapt to interface changes
commit 5d353b55ee14db0ac0515b5a3cf9389430832c19
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:17:37 2010 +0200
cell: adapt to interface changes
not even compile tested
commit cf5d03601322c2dcb12d7a9c2f1745e2b2a35eb4
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:14:59 2010 +0200
util: adapt to interface changes
amazing how much code changes just due to some subtle interface changes?
commit dc98d713c6937c0e177fc2caf23020402cc7ea7b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 12 02:12:40 2010 +0200
gallium: more interface fail, docs
this also changes flush_frontbuffer to use a pipe_resource instead of
a pipe_surface - pipe_surface is not meant to be (or at least no longer)
an abstraction for standalone 2d images which get passed around.
(This has also implications for the non-rendering state-trackers.)
commit 08436d27ddd59857c22827c609b692aa0c407b7b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 10 17:42:52 2010 +0200
gallium: fix array texture interface changes bugs, docs
commit 4a4d927609b62b4d7fb9dffa35158afe282f277b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 3 22:02:44 2010 +0200
gallium: interface changes for array textures and related cleanups
This patch introduces array textures to gallium (note they are not immediately
usable without the associated changes to the shader side).
Also, this abandons pipe_subresource in favor of using level and layer
parameters since the distinction between several faces (which was part of
pipe_subresource for cube textures) and several z slices (which were not part
of pipe_subresource but instead part of pipe_box where appropriate for 3d
textures) is gone at the resource level.
Textures, be it array, cube, or 3d, now use a "unified" set of parameters,
there is no distinction between array members, cube faces, or 3d zslices.
This is unlike d3d10, whose subresource index includes layer information for
array textures, but which considers all z slices of a 3d texture to be part
of the same subresource.
In contrast to d3d10, OpenGL though reuses old 2d and 3d function entry points
for 1d and 2d array textures, respectively, which also implies that for instance
it is possible to specify all layers of a 2d array texture at once (note that
this is not possible for cube maps, which use the 2d entry points, although
it is possible for cube map arrays, which aren't supported yet in gallium).
This should possibly make drivers a bit simpler, and also get rid of mutually
exclusive parameters in some functions (as z and face were exclusive), one
potential downside would be that 3d array textures could not easily be supported
without reverting this, but those are nowhere to be seen.
Also along with adjusting to new parameters, rename get_tex_surface /
tex_surface_destroy to create_surface / surface_destroy and move them from
screen to context, which reflects much better what those do (they are analogous
to create_sampler_view / sampler_view_destroy).
PIPE_CAP_ARRAY_TEXTURES is used to indicate if a driver supports all of this
functionality (that is, both sampling from array texture as well as use a range
of layers as a render target, with selecting the layer from the geometry shader).
Malloc likes to reuse old address as soon as possible this would cause the
new vbo buffer to get the same address as the old. So make sure we set it to
NULL when we allocate a new one. This fixes ipers which will fill up a couple
of VBO buffers per frame.
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Byte offsets simply don't work with tiled render targets when using
tiling bits. Luckily we can cox the hw into doing the right thing
with the DRAWING_RECT command by disabling the drawing rect offset
for the depth buffer.
Minor fixes by Jakob Bornecrantz.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Tiling is rather fragile in general and results in pure blackness when
unlucky. Hence add a new option to disable tiling.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
libdrm-intel can refuse to tile buffers for various reasons. For
potentially tiled buffers the stride is therefore only known after
the iws->buffer_create_tiled call. Unconditionally rounding up to whatever
tiling requires wastes space, so rework the code to not use tex->stride
in the layout code.
Luckily only the mimap/face offset calculation uses it which can easily
be solved by storing an (x, y) coordinate pair. Furthermore this will
be usefull later for properly supporting rendering into the different
levels of tiled mipmap textures.
v2: switch to nblocks(x|y): More in line with gallium and better
suited for rendering into mipmap textures.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
This is needed to properly implement tiling flags. And the gem
implemention fo buffer_from_handle already calls get_tiling, so
it's for free.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
This way relaxed fencing is handled by libdrm. And buffers _can't_
ever change their tiling.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Different kernels have different restrictions for tiled buffers.
Hence use the libdrm abstraction to calculate the necessary
stride and height alignment requirements.
Not yet used.
v2: Incorporate review comments from Jakob Bornecrantz
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
It's unnecessary. The kernel gem ignores it totally and we can't
run on the old userspace fake bo manager due to lack of dri2.
Also drop the redundant name string from the sw winsys as suggested
by Jakob Bornecrantz
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
By not doing so, the uniform contents of
glsl-uniform-non-uniform-array-compare.shader_test was getting thrown
out since nobody was recorded as dereferencing the array.
The new lower_discard and opt_discard_simplification passes should
handle all the necessary transformations, so lower_jumps doesn't need to
support it.
Also, lower_jumps incorrectly handled conditional discards - it would
unconditionally truncate all code after the discard. Rather than fixing
the bug, simply remove the code.
NOTE: This is a candidate for the 7.9 branch.
This should allow lower_if_to_cond_assign to work in the presence of
discards, fixing bug #31690 and likely #31983.
NOTE: This is a candidate for the 7.9 branch.
There were a few places where we were using the wrong opcodes
on evergreen. arl still needs to be fixed on evergreen; see
r600g for reference.
NOTE: This is a candidate for the 7.9 branch.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
As the blend texture, a drawing surface mask is used when masking is
enabled. It should be created as needed.
s/alpha_mask/surface_mask/ to follow OpenVG 1.1 naming.
Depending on whether vgDrawPath(mode), vgDrawImage, or vgDrawGlyph[s] is
called, different paint-to-user and user-to-surface matrices should be
used to derive the sample points for the paint.
This fixes "paint" demo.
Divide bits of VegaShaderType into 6 groups: paint, image, mask, fill,
premultiply, and bw. Each group represents a stage. At most one shader
from each group will be selected when constructing the final fragment
shader.
Optional features such as auth-hinting are not implemented. There is no
anti-aliasing, and no effort is done to keep the glyph origin integral.
So the text quality is poor.
With this commit, the pipe states are entirely managed by the renderer.
The rest of the code interfaces with the renderer instead of
manipulating the states directly.
vg_manager_validate_framebuffer should mark the fb dirty and have
vg_validate_state call cso_set_framebuffer. Rename VIEWPORT_DIRTY to
FRAMEBUFFER_DIRTY.
The states are designated for polygon filling. Polygon filling is a
two-pass process utilizing the stencil buffer. polygon_fill and
polygon_array_fill functions are updated to make use of the state.
This state provides the ability to clear rectangles of the framebuffer
to the specified color, honoring scissoring. vegaClear is updated to
make use of the state.
This state provides glDrawTex-like function. It can be used for
vgSetPixels.
Rather than modifying every user of the renderer, this commit instead
modifies renderer_copy_surface to use DRAWTEX or COPY state internally
depending on whether the destination is the framebuffer.
Renderer states are high-level states to perform specific tasks. The
renderer is initially in INIT state. In that state, the renderer is
used for OpenVG pipeline.
This commit adds a new COPY state to the renderer. The state is used
for copying between two pipe resources using textured drawing. It can
be used for vgCopyImage, for example.
Rather than modifying every user of the renderer, this commit instead
modifies renderer_copy_texture to use the COPY state internally.
This branch defines a gallivm_state structure which contains the
LLVMBuilderRef, LLVMContextRef, etc. All data structures built with
this object can be periodically freed during a "garbage collection"
operation.
The gallivm_state object has to be passed to most of the builder
functions where LLVMBuilderRef used to be used.
Conflicts:
src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
src/gallium/drivers/llvmpipe/lp_state_setup.c
I made the texwrap test be more thorough and realized that this driver code
had not been quite right. This commit fixes the border color for depth
textures, compressed textures, and 16-bits-per-channel textures
with up to 2 channels (R16, RG16).
NOTE: This is a candidate for the 7.9 branch.
Previously, IR for a linked shader was allocated directly out of the
gl_shader object - meaning all of it lived as long as the shader.
Now, IR is allocated out of a temporary context, and any -live- IR is
reparented/stolen to (effectively) the gl_shader. Any remaining IR can
be freed.
NOTE: This is a candidate for the 7.9 branch.
This makes a very simple 1.30 shader go from 196k of memory to 9k.
NOTE: This -may- be a candidate for the 7.9 branch, as the benefit is
substantial. However, it's not a simple change, so it may be wiser to
wait for 7.10.
We were trying to emit a single ir_expression to compare the whole
thing. The backends (ir_to_mesa.cpp and brw_fs.cpp so far) expected
ir_binop_any_nequal or ir_binop_all_equal to apply to at most a vector
(with matrices broken down by the lowering pass). Break them down to
a bunch of ORed or ANDed any_nequals/all_equals.
Fixes:
glsl-array-compare
glsl-array-compare-02
glsl-fs-struct-equal
glsl-fs-struct-notequal
Bug #31909
This doesn't cover all expressions or all operand types, but it will
complain if you overreach and it allows for much greater slack on the
programmer's part.
This code was for the old GLcore build of the software rasteriser. The
X server switched to a DRI driver for software indirect GLX long ago.
Signed-off-by: Adam Jackson <ajax@redhat.com>
There are also some u_simple_shaders changes.
On r300, the TGSI_SEMANTIC_COLOR varying is a fixed-point number clamped
to the range [0,1] and limited to 12 bits of precision. Therefore we can't
use it for passing through a clear color in order to clear high precision
texture formats.
This also makes u_blitter use only one vertex shader instead of two.
Since the viewport is not updated on RandR12 mode switches anymore,
clipping to viewport may incorrectly clip away the video.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
This may help paint the colorkey before overlay updates in some
situations where the app paints the color key (mainly xine).
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
The compiler seriously needs a cleanup as far as the arrangement of functions
is concerned. It's hard to know whether some function was implemented or not
because there are so many places to search in and it can be anywhere and
named anyhow.
If nvfx_framebuffer prepare and validate were called successively with
fb->zsbuf not NULL and then NULL, nvfx->hw_zeta would contain garbage and
this would cause failures in nvfx_framebuffer_relocate/OUT_RELOC(hw_zeta).
This was triggered by piglit/texwrap 2D GL_DEPTH_COMPONENT24 and caused
first a 'write to user buffer!!' error in libdrm and then worse things.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Currently, the standalone compiler tries to do function inlining before
linking shaders (including linking against the built-in functions).
This resulted in the built-in function _prototypes_ being inlined rather
than the actual function definition.
This is only known to fix a bug in the standalone compiler; most
programs should be unaffected. Still, it seems like a good idea.
NOTE: This is a candidate for the 7.9 branch.
Fixes this GCC warning with linux-x86 build.
radeon_pair_regalloc.c: In function ‘compute_live_intervals’:
radeon_pair_regalloc.c:222: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning with linux-x86 build.
radeon_pair_regalloc.c: In function ‘compute_live_intervals’:
radeon_pair_regalloc.c:221: warning: ISO C90 forbids mixed declarations and code
The alpha mask is addressed with unnormalized coordinates in the
fragment shader. It should have the same orientation as the surface
does.
This fixes "mask" OpenVG demo.
Previously, the get table listed all three as having custom locations,
yet find_custom_value did not have cases to handle them.
MAX_VARYING_VECTORS does not need a custom location since MaxVaryings is
already stored as float[4] (or vec4). MaxUniformComponents is stored as
the number of floats, however, so a custom implementation that divides
by 4 is necessary.
Fixes bugs.freedesktop.org #31495.
This fixes incorrect behaviour when the stencil clear value exceeds
the size of the stencil buffer, for example, when set with:
glClearStencil (~1); /* Set a bit pattern of 111...11111110 */
glClear (GL_STENCIL_BUFFER_BIT);
The clear value needs to be masked by the value 2^m - 1, where m is the
number of bits in the stencil buffer. Previously, we passed the value
masked with 0x7fffffff to _mesa_StencilFuncSeparate which then clamps,
NOT masks the value to the range 0 to 2^m - 1.
The result would be clearing the stencil buffer to 0xff, rather than 0xfe.
Signed-off-by: Peter Clifton <pcjc2@cam.ac.uk>
Signed-off-by: Brian Paul <brianp@vmware.com>
Make sure regions are properly updated and that the colorkey painting is
flushed before we update the HW overlay.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
This is needed to properly sync with host side rendering. For example,
make sure we flush colorkey painting before updating the overlay.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
When a context was made current to another surface, the old code did
this
dri2_dpy->core->bindContext(cctx, ddraw, rdraw);
dri2_dpy->core->unbindContext(old_cctx);
and there will be no current context due to the second line.
unbindContext should be called only when bindContext is not. This fixes
a regression since d19afc57. Thanks to Neil Roberts for noticing the
issue and creating a test case.
(Marek: added the initializion of "vec" in the default statement)
NOTE: This is a candidate for the 7.9 branch.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
This fixes piglit/glsl-vs-main-return and glsl-fs-main-return for the drivers
which don't support RET (i915g, r300g, r600g, svga).
ir_to_mesa does not currently generate subroutines, but it's a matter of time
till it's added. It would then break all the drivers which don't implement
them, so this CAP makes sense.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
When the result of the alpha instruction is being replicated to the RGB
destination register, we do not need to use alpha's destination register.
This fixes an invalid "Too many hardware temporaries used" error in
the case where a transcendent operation writes to a temporary register
greater than max_temp_regs.
NOTE: This is a candidate for the 7.9 branch.
This fixes an invalid "Too many hardware temporaries used" error in the
case where a source reads from a temporary register with an index greater
than max_temp_regs and then the source is marked as unused before the
register allocation pass.
NOTE: This is a candidate for the 7.9 branch.
Reads of registers that where not written to within the same block were
not being tracked. So in a situations like this:
0: IF
1: ADD t0, t1, t2
2: MOV t2, t1
Instruction 2 didn't know that instruction 1 read from t2, so
in some cases instruction 2 was being scheduled before instruction 1.
NOTE: This is a candidate for the 7.9 branch.
Gallium drivers pass all piglit tests for the two (there are 12 tests
for separate_shader_objects and 5 tests for explicit_attrib_location),
and I was told the extensions don't need any driver-specific code.
I made them dependent on PIPE_CAP_GLSL.
Signed-off-by: Brian Paul <brianp@vmware.com>
The drm winsys only ever handles one gem memory manager. Rip out
the unnecessary complication.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Not using the gtt is considered harmful for performance. And for
partial uploads there's always drm_intel_bo_subdata.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
It looks like this was meant to facilitate unfenced access to textures/
color/renderbuffers. It's totally incomplete and fundamentally broken
on a few levels:
- broken: The kernel needs to about every tiled bo to fix up bit17
swizzling on swap-in.
- unflexible: fenced/unfenced relocs from execbuffer2 do the same, much
simpler.
- unneeded: with relaxed fencing tiled gem bos are as memory-efficient
as this trick.
Hence kill it.
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Fix a crash when the subrectangle is not inside the fb. Fix wrong
pipe transfer when sx > 0 or sy + height != fb->height.
This fixes "readpixels" demo.
These two samplers use non-normalized texture coordinates. wrap_r
cannot be PIPE_TEX_WRAP_REPEAT (the default).
This fixes
sp_tex_sample.c:1790:get_linear_unorm_wrap: Assertion `0' failed
assertion failure.
Fix OpenVG "filter" demo
Program received signal SIGSEGV, Segmentation fault.
0xb7153dc9 in str_match_no_case (pcur=0xbfffe564, str=0x0) at
tgsi/tgsi_text.c:86
86 while (*str != '\0' && *str == uprcase( *cur )) {
This is quite common for multitexture sampling, and not only cuts down
on the second and later set of MOVs, but typically also allows
compute-to-MRF on the first set.
No statistically siginficant performance difference in nexuiz (n=3),
but it reduces instruction count in one of its shaders and seems like
a good idea.
We were skipping it if the instruction producing the value we were
going to compute-to-mrf used its result reg as a source reg. This
meant that the typical "write interpolated color to fragment color" or
"texture from interpolated texcoord" shader didn't compute-to-MRF.
Just don't check for the interference cases until after we've checked
if this is the instruction we wanted to compute-to-MRF.
Improves nexuiz high-settings performance on my laptop 0.48% +- 0.08%
(n=3).
On pre-gen6, this turns 4 instructions into 1. We could still do
better by folding the saturate into the instruction generating the
value if nobody else uses it, but that should be a separate pass.
Hardware pretty commonly has saturate modifiers on instructions, and
this can be used in codegen to produce those, without everyone else
needing to understand clamping other than min and max.
This should save on the overhead of tree-walking and provide a
convenient place to add more instruction lowering in the future.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
The vector operator collects 2, 3, or 4 scalar components into a
vector. Doing this has several advantages. First, it will make
ud-chain tracking for components of vectors much easier. Second, a
later optimization pass could collect scalars into vectors to allow
generation of SWZ instructions (or similar as operands to other
instructions on R200 and i915). It also enables an easy way to
generate IR for SWZ instructions in the ARB_vertex_program assembler.
The operate just like ir_unop_sin and ir_unop_cos except that they
expect their inputs to be limited to the range [-pi, pi]. Several
GPUs require this limited range for their sine and cosine
instructions, so having these as operations (along with a to-be-written
lowering pass) helps this architectures.
These new operations also matche the semantics of the
GL_ARB_fragment_program SCS instruction. Having these as operations
helps in generating GLSL IR directly from assembly fragment programs.
Use fetch shader instead of having fetch instruction in the vertex
shader. Allow to restrict shader update to a smaller part when
vertex buffer input layout changes.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Occlusion query on evergreen need the event index field to be
set otherwise we endup locking up the GPU.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Condiation moves with a condition of (a < 0), (a > 0), (a <= 0), or (a
>= 0) can be generated with "a" directly as an operand of the CMP
instruction. This doesn't help much now, but it will help with
assembly shaders that use the CMP instruction.
This should prevent the field going unset in the future. See bug
http://bugs.freedesktop.org/show_bug.cgi?id=31544 for background.
Also remove unneeded calls to clear_teximage_fields().
Finally, call _mesa_set_fetch_functions() from the
_mesa_init_teximage_fields() function so callers have one less
thing to worry about.
Fix this GCC warning.
ir.cpp: In static member function
'static unsigned int ir_expression::get_num_operands(ir_expression_operation)':
ir.cpp:199: warning: control reaches end of non-void function
If an instruction writes reg but nothing later uses it, then we don't
need to bother doing it. Before, we were just killing code that was
never read after it was ever written.
This removes many interpolation instructions for attributes with only
a few comopnents used. Improves nexuiz high-settings performance .46%
+/- .12% (n=3) on my Ironlake.
It's annoying to use test suites under a Mesa debug build because
pretty output is cluttered with stderr's continuous reports that
you're still using the debug driver.
The new usage message lists possible command line options. (Newcomers to Mesa
currently have to trawl through the source to find the command line options,
and we should save them from that trouble.)
Example Output
--------------
usage: ./glsl_compiler [options] <file.vert | file.geom | file.frag>
Possible options are:
--glsl-es
--dump-ast
--dump-hir
--dump-lir
--link
This adds sentinel values to the ir_expression_operation enum type:
ir_last_unop, ir_last_binop, and ir_last_opcode. They are set to the
previous one so they don't trigger "unhandled case in switch statement"
warnings, but should never be handled directly.
This allows us to remove the huge array of 1s and 2s in
ir_expression::get_num_operands().
We are not aware of any GPU that actually implements the cross product
as a single instruction. Hence, there's no need for it to be an opcode.
Future commits will remove it entirely.
This is really supposed to be defined only if the driver supports highp
in the fragment shader - but all of our current ES2 implementations do.
So, just define it. In the future, we'll need to add a flag to
gl_context and only define the macro if the flag is set.
"Fixes" freedesktop.org bug #31673.
ir_binop_less, ir_binop_greater, ir_binop_lequal, and ir_binop_gequal
are defined to work on vectors as well as scalars, as long as the two
operands have the same type.
This is evident from both ir_validate.cpp and our use of these opcodes
in the GLSL lessThan, greaterThan, lessThanEqual, greaterThanEqual
built-in functions.
Found by code inspection. Not known to fix any bugs. Presumably, our
tests for the built-in comparison functions must pass because C.E.
handling is done on the ir_call of "greaterThan" rather than the inlined
opcode. The C.E. handling of the built-in function calls is correct.
NOTE: This is a candidate for the 7.9 branch.
Default group bytes to 512 on evergreen. Don't query
tiling config yet for evergreen, the current info returned is not
adequate for evergreen (no way to get bank info).
RET was interpreted as END, which was wrong. Instead, if a shader contains RET
in the main function, it will fail to compile with an error message
from now on.
The hack is from early days.
When drawing GL_DEPTH_COMPONENT the usual fragment pipeline steps apply
so don't override the depth state.
When drawing GL_STENCIL_INDEX (or GL_DEPTH_STENCIL) the fragment pipeline
does not apply (only the stencil and Z writemasks apply) so disable writes
to the color buffers.
Fixes some regressions from commit ef8bb7ada9
This consolidates the TOKEN_OR_IDENTIFIER and RESERVED_WORD macros into
a single KEYWORD macro.
The old TOKEN_OR_IDENTIFIER macros handled the case of a word going from
an identifier to a keyword; the RESERVED_WORD macro handled a word going
from a reserved word to a language keyword. However, neither could
properly handle samplerBuffer (for example), which is an identifier in
1.10 and 1.20, a reserved word in 1.30, and a keyword in 1.40 and on.
Furthermore, the existing macros didn't properly handle reserved words
in GLSL ES 1.00. The best they could do was return a token (rather than
an identifier), resulting in an obtuse parser error, rather than a
user-friendly "you used a reserved word" error message.
Functions are not first class objects in GLSL, so there is never a value
of function type. No code actually used this except for one function
which asserted it shouldn't occur. One comment mentioned it, but was
incorrect. So we may as well remove it entirely.
This driver is a fake swdri driver that perform no operations
beside allocation gallium structure and buffer for upper layer
usage.
It's purpose is to help profiling core mesa/gallium without
having pipe driver overhead hidding hot spot of core code.
scons file are likely inadequate i am unfamiliar with this
build system.
To use it simply rename is to swrast_dri.so and properly set
LIBGL_DRIVERS_PATH env variable.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
src/mesa/drivers/dri/*/*/*.[chS] is a superset of
src/mesa/drivers/dri/*/server/*.[ch] and
src/mesa/drivers/dri/common/xmlpool/*.[ch].
include/GL/internal/glcore.h is already in MAIN_FILES, no need for it in
DRI_FILES too. src/glx/Makefile was listed twice.
Signed-off-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Brian Paul <brianp@vmware.com>
This showed up as cairo-gl gradients being inverted on everyone but
Intel, where I'd apparently tweaked the transformation to work around
the bug. Fixes piglit fbo-fragcoord.
Required because ATI and NVIDIA DX9 GPUs do not support indirect addressing
of temps, inputs, outputs, and consts (FS-only) or the hw support is so
limited that we cannot use it.
This should make r300g and possibly nvfx more feature complete.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
Since this was talloced off of NULL instead of the compile state, it
was a real leak over the course of the program. Noticed with
valgrind --leak-check=full --show-reachable=yes. We should really
change these passes to generally get the compile context as an argument
so simple mistakes like this stop mattering.
This fixes a regression (failed assertion) from commit
c552f273f5 which was hit if glDeleteBuffers()
was called on a buffer that was never bound.
NOTE: this is a candidate for the 7.9 branch.
Currently r600_resource_copy_region() will turn these copies into
transfers + memcpys, so to avoid recursion we must not turn those
transfers back into blits.
Previously queries of MAX_SAMPLES were only allowed with
ARB_framebuffer_object, but EXT_framebuffer_multisample also enables
this query. This seems to only effect the i915. All other drivers
support both extensions or neither extension.
This patch is based on a patch that Kenneth sent along with the report.
NOTE: this is a candidate for the 7.9 branch.
Reported-by: Kenneth Waters <kwaters@chromium.org>
For driver performance analysis it usefull to be able to
disable as much as possible the GPU interaction so that
one can profile the userspace only.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
IF statements were getting flattened while they were broken. With
Zhenyu's last fix for ENDIF's type, everything appears to have lined
up to actually work.
This regresses two tests:
glsl1-! (not) operator (1, fail)
glsl1-! (not) operator (1, pass)
but fixes tests that couldn't work before because the IFs couldn't be
flattened:
glsl-fs-discard-01
occlusion-query-discard
(and, naturally, this should be a performance improvement for apps
that actually use IF statements to avoid executing a bunch of code).
Sometimes we swizzled in a different channel it looked like, and
sometimes we swizzled in zero. Or something.
Having looked at the output of another code generator for this chip,
this is approximately what they do, too: use align1 math on
temporaries, and then move the results into place.
Fixes:
glean/vp1-EX2 test
glean/vp1-EXP test
glean/vp1-LG2 test
glean/vp1-RCP test (reciprocal)
glean/vp1-RSQ test 1 (reciprocal square root)
shaders/glsl-cos
shaders/glsl-sin
shaders/glsl-vs-masked-cos
shaders/vpfp-generic/vp-exp-alias
Instead of messing with the callers simply copy our inputs into a
alloca array at the beginning of the function and then use it.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
DD_POINT_SIZE was broken for quite some time, and the only driver (r200) relying
on this no longer needs it.
Both DD_POINT_SIZE and DD_LINE_WIDTH have no users left outside of debugging
output, hence instead of fixing DD_POINT_SIZE setting just drop both of them -
there was a plan to remove tricaps flags entirely at some point.
DD_POINT_SIZE got never set for some time now (as it was set only in ifdefed
out code), which caused the r200 driver to use the point primitive mistakenly
in some cases which can only do size 1 instead of point sprite. Since the
logic to use point instead of point sprite prim is flaky at best anyway (can't
work correctly for per-vertex point size), just drop this and always emit point
sprites (except for AA points) - reasons why the driver tried to use points for
size 1.0 are unknown though it is possible they are faster or more conformant.
Note that we can't emit point sprites without point sprite cntl as that might
result in undefined point sizes, hence need drm version check (which was
unnecessary before as it should always have selected points). An
alternative would be to rely on the RE point size clamp controls which could
clamp the size to 1.0 min/max even if the SE point size is undefined, but currently
always use 0 for min clamp. (As a side note, this also means the driver does
not honor the gl spec which mandates points, but not point sprites, with zero size
to be rendered as size 1.)
This should fix recent reports of https://bugs.freedesktop.org/show_bug.cgi?id=702.
This is a candidate for the mesa 7.9 branch.
Fixes assertion failure with texture swizzling in the GLSL path when
it's triggered (such as gen6 FF or ARB_fp shadow comparisons).
Fixes:
texdepth
texSwizzle
fp1-DST test
fp1-LIT test 3
Also fixup code comment to reflect that the GPU requires DWORD
alignment, but in this case does not actually pass the value "in
DWORDs" as I previously stated.
This reverts commit 76360d6abc. On
second thought, it turned out that sync objects also used the
wait_rendering API like this, and would need the same treatment, and
so wait_rendering itself is fixed in libdrm now.
We were asking for a wait to GTT read (all GPU rendering to it
complete), instead of asking for all GPU reading from it to be
complete. Prevents swapbuffers-based apps from running away with
rendering, and produces a better input experience.
For some reason I though we needed the _DISCARD flag to avoid
readbacks, which isn't true at all. Now write operations should
pipeline properly, gives a good speedup to demos/tunnel.
When linked with certain builds of libstdc++, it appears like powf is resolved
by a symbol in that library. Other builds of libstdc++ doesn't contain that
symbol resulting in a linker / loader error. Optionally
resolve that symbol and replace it with calls to logf and expf.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Fixes this GCC warning.
graw.h:93: warning: 'struct pipe_surface' declared inside parameter list
graw.h:93: warning: its scope is only this definition or declaration,
which is probably not what you want
A call to radeon_prepare_render() at the beginning of draw
operations was placed too deep in the call chain,
inside r300RunRenderPrimitive(), instead of
r300DrawPrims() where it belongs. This leads to
emission of stale target color renderbuffer into the cs if
bufferswaps via page-flipping are used, and thereby causes
massive rendering corruption due to unsynchronized
rendering into the active frontbuffer.
This patch fixes such problems for use with the
upcoming radeon page-flipping patches.
Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de>
The width of the 2D blits used to copy the data is defined as a 16-bit
signed integer, but the pitch must be DWORD aligned. Limit to an integral
number of DWORDs, (1 << 15 - 4) rather than (1 << 15 -1).
Fixes corruption to data uploaded with glBufferSubData.
Signed-off-by: Peter Clifton <pcjc2@cam.ac.uk>
Allows applications to dump surfaces to file without
referencing gallium/auxiliary entry points statically.
Existing test apps have been modified such that
they save the contents of the fronbuffer only
when the `-o' option's specified.
These ES1 features were only tested for in the vertex array code.
Checking the ctx->API field at runtime is cleaner than the #ifdef
stuff and supports choosing the API at runtime.
Flushing the vertices after having already updated the state doesn't
do any good. Fixes useshaderprogram-flushverts-1. As a side effect,
by moving it to the right place we end up skipping no-op state changes
for traditional glUseProgram.
Rebuilding the vertex format from scratch every time we see a new
vertex attribute is rather costly, new attributes can be appended at
the end avoiding a copy to current and then back again, and the full
attr pointer recalculation.
In the not so likely case of an already existing attribute having its
size increased the old behavior is preserved, this could be optimized
more, not sure if it's worth it.
It's a modest improvement in FlightGear (that game punishes the VBO
module pretty hard in general, framerate goes from some 46 FPS to 50
FPS with the nouveau classic driver).
Signed-off-by: Brian Paul <brianp@vmware.com>
Silences this GCC warning.
brw_wm_fp.c: In function 'brw_wm_pass_fp':
brw_wm_fp.c:966: warning: 'last_inst' may be used uninitialized in this function
brw_wm_fp.c:966: note: 'last_inst' was declared here
Silences this GCC warning.
brw_wm_fp.c: In function 'precalc_tex':
brw_wm_fp.c:666: warning: 'tmpcoord.Index' may be used uninitialized in this function
Fixes this GCC warning with linux-x86 build.
radeon_dataflow.c: In function 'get_readers_normal_read_callback':
radeon_dataflow.c:472: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning with linux-x86 build.
radeon_pair_schedule.c: In function 'merge_presub_sources':
radeon_pair_schedule.c:312: warning: ISO C90 forbids mixed declarations and code
Commit 8dfafbf086 forgot to update r300g.
There is a buf == NULL check, but buf is used before for var init.
Tested-by: Guillermo S. Romero <gsromero@infernal-iceberg.com>
Fixes this GCC warning.
nouveau_vbo_t.c: In function 'nv10_vbo_render_prims':
nouveau_render_t.c:161: warning: 'max_out' may be used uninitialized in this function
nouveau_render_t.c:161: note: 'max_out' was declared here
We really only want to print spaces -between- elements, not after each
element. This cleans up error messages from IR reader, making them
(mildly) easier to read.
In particular, calling the abs function is silly, since there's already
an expression opcode for that. Also, assigning to temporaries then
assigning those to the final location is rather redundant.
Trivial change that avoids a segmentation fault when the blitter state
happens to be bound when the context is destroyed.
The free calls should probably removed altogether in the future -- the
responsibility to destroy the state atoms lies with whoever created it,
and the safest thing for the pipe driver is to not touch any bound state
in its destructor.
In testing on Ironlake, the histogram of clocks/pixel results for the
system memcpy and magic unaligned memcpy show no noticeable difference
(and no statistically significant difference with the 5510 samples
taken, though the stddev is large due to what looks like the cache
effects from the different texture sizes used).
This provides the optimizer with hints about code hotness, which we're
quite certain about for debug printouts (or, rather, while we
developers often hit the checks for debug printouts, we don't care
about performance while doing so).
Discard fractional bits from linewidth. This matches the nvidia
closed drivers, my reading of the OpenGL SI and current llvmpipe
behaviour.
It looks a lot nicer & avoids ugliness where lines alternate between n
and n+1 pixels in width along their length.
Also fix up r600g to match.
These were previously being left in the default (D3D) mode. This mean
that triangles were drawn slightly incorrectly, but also because this
state is relied on by the u_blitter code, all blits were half a pixel
off.
Generalize the existing tiled_buffer path in texture transfers for use
in some non-tiled up and downloads.
Use a staging buffer, which the winsys will restrict to GTT memory.
GTT buffers have the major advantage when they are mapped, they are
cachable, which is a very nice property for downloads, usually the CPU
will want to do look at the data it downloaded.
This opens the question of what interface the winsys layer should
really have for talking about these concepts.
For now I'm using the existing gallium resource usage concept, but
there is no reason not use terms closer to what the hardware
understands - eg. the domains themselves.
The callback presents the given attachment to the native engine. It
allows the swap behavior and interval to be controlled. It will replace
native_surface::flush_frontbuffer and native_surface::swap_buffers
shortly.
Silences warning such as:
main/texobj.c:442:40: warning: ISO C99 requires rest arguments to be used
main/texobj.c:498:58: warning: ISO C99 requires rest arguments to be used
This ensures that we increase bo->map_count when radeon_bo_map_internal()
returns successfully, which in turn makes sure we don't decrement
bo->map_count below zero later.
Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
The call to draw_bind_fragment_shader() was using the old fragment
shader. This bug would have really only effected the draw module's
use of the fragment shader in the wide point stage.
Some C++ header files were included in an extern "C" block. When building with
Clang, this caused the build to fail due to namespace errors. (GCC did not
report any errors.)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Important as more constant buffers per shader start to get used.
Fix up r600 (tested) and nv50 (untested) to cope with this. Drivers
previously didn't see unbinds of constant buffers often or ever, so
this isn't always dealt with cleanly.
For r600 just return and keep the reference. Will try to do better in
a followup change.
Don't trim triangle bounding box to scissor/draw-region until after
the logic for emitting tri_16. Don't generate tri_16 commands for
triangles with untrimmed bounding boxes outside the current tile.
This is important as the tri-16 itself can extend past tile bounds and
we don't want to add code to it to check against tile bounds (slow) or
restrict it to locations within a tile (pessimistic).
This effectively redoes 1741ddb747 in a
way that allows contexts of different APIs to coexist.
First, the changes to the remap table are reverted. The remap table
(driDispatchRemapTable) is always initialized in the same way regardless
of the context API.
es_generator.py is updated to use a local remap table, whose sole
purpose is to help initialize its dispatch table. The local remap table
and the global one are always different, as they use different
glapidispatch.h. But the dispatch tables initialized by both remap
tables are always compatible with glapi (libGL.so).
Finally, the semantics of one_time_init are changed to per-api one-time
initialization.
Core mesa should query glapi for the positions of the functions in
_glapi_table when multiple APIs are supported. It does not know which
glapitable.h glapi used.
Use scons target and dependency system instead of ad-hoc options.
Now is simply a matter of naming what to build. For example:
scons libgl-xlib
scons libgl-gdi
scons graw-progs
scons llvmpipe
and so on. And there is also the possibility of scepcified subdirs, e.g.
scons src/gallium/drivers
If nothing is specified then everything will be build.
There might be some rough corners over the next days. Please bare with me.
Without this, update_textures may not pick up the new pipe_resource.
It is actually update_textures that should check
stObj->sampler_view->texture != stObj->pt, but let's follow st_TexImage
and others for now.
Make autoconf decide the client APIs enabled first. Then when OpenGL
and OpenGL ES are disabled, there is no need to build src/mesa/; when
OpenGL is disabled, no $mesa_driver should be built. Finally, add
--enable-openvg to enable OpenVG.
With these changes, an OpenVG only build can be configured with
$ ./configure --disable-opengl --enable-openvg
src/mesa, src/glsl, and src/glx will be skipped, which saves a great
deal of compilation time.
And an OpenGL ES only build can be configured with
$ ./configure --disable-opengl --enable-gles-overlay
Fixes this GCC warning.
state_tracker/st_program.c: In function 'st_print_shaders':
state_tracker/st_program.c:735: warning: 'sh' may be used uninitialized in this function
This fixes some insanity that would otherwise be required for GLSL
1.30 bit ops or gen6 integer uniform operations in general, at the
cost of upload-time pain. Given that we only have that pain because
mesa's mangling our integer uniforms to be floats, this something that
should be fixed outside of the shader codegen.
First, it changes autoconf to use a "python2" binary when available,
rather than plain "python" (which is ambiguous). Secondly, it changes
the Makefiles to use $(PYTHON) $(PYTHON_FLAGS) rather than calling
python directly.
Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Matthew William Cox <matt@mattcox.ca>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Fixes this GCC warning.
r300_state_derived.c: In function 'r300_update_derived_state':
r300_state_derived.c:593: warning: 'r' may be used uninitialized in this function
r300_state_derived.c:593: note: 'r' was declared here
radeon_bo_destroy() will want to read the list field. Without this patch,
we'd end up evaluating the list pointers before they have been properly
set up when we destroyed the newly created bo if it cannot be mapped.
Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
With 07b85457d9, glapitable.h is included
by core mesa only to know the size of _glapi_table. It is not necessary
as the same info is given by _gloffset_COUNT.
This change makes _glapi_table opaque to core mesa. All operations on
it are supposed to go through one of the SET/GET/CALL macros.
Move defines in glapioffsets.h to glapidispatch.h. Rename
_gloffset_FIRST_DYNAMIC to _gloffset_COUNT, which is equal to the number
of entries in _glapi_table.
Consistently use SET_by_offset, GET_by_offset, CALL_by_offset, and
_gloffset_* to recursively define all SET/GET/CALL macros.
glapioffsets.h exists for the same reason as glapidispatch.h does. It
is of no use to glapi. This commit also drops the use of glapioffsets.h
in glx as glx is considered an extension to glapi when it comes to
defining public GL entries.
glapidispatch.h exists so that core mesa (libmesa.a) can be built for
DRI drivers or for non-DRI drivers as a compile time decision (whether
IN_DRI_DRIVER is defined). It is of no use to glapi. This commit also
drops the use of glapidispatch.h in glx and libgl-xlib as they are
considered extensions to glapi when it comes to defining public GL
entries.
This lets us simplify and consolidate some state checking code.
This implements the GL_INVALID_OPERATION check for all drawing commands
required by GL_EXT_texture_integer.
It's a little more painful than before because we don't have the handy
mask register any more, and have to make do with cooking up a value
out of the flag register.
This is apparently required, as the thread will be initiated while it
still has dependencies, and this is what waits for those to be
resolved before writing color.
Function ast_declarator_list::hir(), when processing keywords added by
extension ARB_fragment_coord_conventions, made the mistake of checking only if
the extension was __supported by the driver__. The correct behavior is to check
if the extensi0n is __enabled in the parse state__.
NOTE: this is a candidate for the 7.9 branch.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Ensure -L$(TOP)/$(LIB_DIR) (the staging dir for build products), appears
in the link line before any -L in $LDFLAGS, so that we link driver we are
building with libEGL we have just built, and not an installed version
[olv: make a similar change to targets/egl]
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
This call sequence
eglMakeCurrent(dpy, surf, surf, ctx1);
eglMakeCurrent(dpy, surf, surf, ctx2);
should be valid if ctx1 and ctx2 have the same client API and are not
current in another thread.
Internally a mode belongs to a screen. But functions like
eglGetModeAttribMESA treat a mode as a display resource: a mode can be
looked up without a screen. Considering how KMS works, it is better to
stick to the current implementation.
To properly support looking up a mode without a screen, this commit
assigns each mode (of all screens) a unique ID.
The opaque nature of EGLImage implies that extensions almost always
define their own attributes. Move attributes in _EGLImage to
_EGLImageAttribs and add a helper function to parse attribute lists.
Fixes glsl-fs-convolution-2, which was blowing up due to the array
access insanity getting at the uniform values within the loop. Each
temporary was considered live across the whole loop.
One, it was allocating increments of 1kb, but per thread scratch space
is a power of two. Two, the new FS wasn't getting total_scratch set
at all, so everyone thought they had 1kb and writes beyond 1kb would
go stomping on a neighbor thread.
With this plus the previous register spilling for the new FS,
glsl-fs-convolution-1 passes.
g0 is used by others, and is expected to be left exactly as it was
dispatched to us. So manually move g0 into our message reg when
spilling/unspilling and update the offset in the MRF. Fixes failures
in texture sampling after having spilled a register.
It's amazing this code worked. Basically, we would get lucky in
register allocation and the tests using frontfacing would happen to
allocate gl_FrontFacing storage and the instructions generating
gl_FrontFacing but pointing at another register to the same hardware
register. Noticed during register spilling debug, when suddenly they
didn't get allocatd the same storage.
Since this is just generated by python, it's questionable whether this
should continue to live in the repository - Mesa already has other
things generated from python as part of the build process.
This works around MSVC's 65535 byte limit, unfortunately at the expense
of any semblance of readability and much larger file size. Hopefully I
can implement a better solution later, but for now this fixes the build.
Fixes this GCC warning.
gallivm/lp_bld_tgsi_aos.c: In function 'lp_build_tgsi_aos':
gallivm/lp_bld_tgsi_aos.c:516: warning: 'dst0' may be used uninitialized in this function
gallivm/lp_bld_tgsi_aos.c:516: note: 'dst0' was declared here
Fixes these GCC warnings.
gallivm/lp_bld_sample_aos.c: In function 'lp_build_sample_image_nearest':
gallivm/lp_bld_sample_aos.c:271: warning: 't_ipart' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:271: warning: 'r_ipart' may be used uninitialized in this function
Fixes these GCC warnings.
gallivm/lp_bld_sample_aos.c: In function 'lp_build_sample_image_linear':
gallivm/lp_bld_sample_aos.c:439: warning: 'r_ipart' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:438: warning: 't_ipart' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:438: warning: 't_fpart' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:439: warning: 'r_fpart' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:438: warning: 't_fpart_hi' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:438: warning: 't_fpart_lo' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:439: warning: 'r_fpart_hi' may be used uninitialized in this function
gallivm/lp_bld_sample_aos.c:439: warning: 'r_fpart_lo' may be used uninitialized in this function
At the moment you need kernel patches to have texture tiling work
with the kernel CS checker, so once they are upstream and the drm version
is bumped we can make this enable flip the other way most likely.
Since the hw transitions from 2D->1D sampling below the 2D macrotile
size we need to keep track of the array mode per level so we can
render to it using the CB.
Silences this GCC warning.
ast_to_hir.cpp: In function 'void apply_type_qualifier_to_variable(const
ast_type_qualifier*, ir_variable*, _mesa_glsl_parse_state*, YYLTYPE*)'
ast_to_hir.cpp:1768: warning: enumeration value 'ir_shader' not handled
in switch
We were working around an LLVM 2.5 bug but we're using LLVM 2.6 or later now.
This basically reverts commit baddcbc522.
This fixes the piglit bug/tri-tex-crash.c failure.
Silences these GCC warnings.
r600_shader.c: In function 'tgsi_exp':
r600_shader.c:2339: warning: 'r600_src[0].rel' is used uninitialized in this function
r600_shader.c:2339: warning: 'r600_src[0].abs' is used uninitialized in this function
r600_shader.c:2339: warning: 'r600_src[0].neg' is used uninitialized in this function
r600_shader.c:2339: warning: 'r600_src[0].chan' is used uninitialized in this function
r600_shader.c:2339: warning: 'r600_src[0].sel' is used uninitialized in this function
Previously some shader input or outputs that hadn't received location
assignments could slip through. This could happen when a shader
contained user-defined varyings and was used with either
fixed-function or assembly shaders.
See the piglit tests glsl-[fv]s-user-varying-ff and
sso-user-varying-0[12].
NOTE: this is a candidate for the 7.9 branch.
This function type checks the operands of and returns the result type of
bit-logic operations. It replaces the type checking performed in the
following cases of ast_expression::hir() :
- ast_bit_and
- ast_bit_or
- ast_bit_xor
This function type checks the operands of and returns the result type of
bit-shift operations. It replaces the type checking performed in the following
cases of ast_expression::hir() :
- ast_lshift
- ast_rshift
This fixes hangs in some Z-writes-in-shaders tests, though other
pieces don't come out correctly.
Bug #30392: hang in fbo-fblit-d24s8. (still fails with bad color drawn
to some targets)
rc_get_readers_normal() supplies a list of readers for a given
instruction. This function is now being used by the copy propagate
optimization and will eventually be used by most other optimization
passes as well.
Existing code relies on IR being generated (possibly with error type)
rather than returning NULL. So, don't break - go ahead and generate the
operation. As long as an error is flagged, things will work out.
Fixes fd.o bug #30914.
Thanks to Alex Deucher for pointing out the FLT to int conversion is necessary
and writing an initial patch, this brings about 20 piglits, and I think this
is the last piece to make evergreen and r600 equal in terms of features.
We need to keep track of three different fragment shaders: Z-only, stencil-
only, and Z+stencil. Before, we were only keeping track of the first one
we encountered.
We often use reg_null as the destination when setting up the flag
regs. However, on gen6 there aren't general implicit conversions to
destination types from src types, so the comparison to produce the
flag regs would be done on the integer result interpreted as a float.
Hilarity ensued.
Fixes 20 piglit cases.
In ir_validate::visit_leave(), the cases for
- ir_binop_bit_and
- ir_binop_bit_xor
- ir_binop_bit_or
were incorrect. It was incorrectly asserted that both operands must be the
same type, when in fact one may be scalar and the other a vector. It was also
incorrectly asserted that the resultant type was the type of the left operand,
which in fact does not hold when the left operand is a scalar and the right
operand is a vector.
Implement by adding the following cases to ast_expression::hir():
- ast_lshift
- ast_rshift
Also, implement ir validation for the new operators by adding the following
cases to ir_validate::visit_leave():
- ir_binop_lshift
- ir_binop_rshift
On evergreen, interpolation has moved into the fragment shader,
with the interpolation parmaters being passed via GPRs and LDS entries.
This works out the number of interps required and reserves GPR/LDS
storage for them, it also correctly routes face/position values which
aren't interpolated from the vertex shader.
Also if we noticed nothing is to be interpolated we always setup perspective
interpolation for one value otherwise the GPU appears to lockup.
This fixes about 15 piglit tests on evergreen.
Previously _LinkedShaders was a compact array of the linked shaders
for each shader stage. Now it is arranged such that each slot,
indexed by the MESA_SHADER_* defines, refers to a specific shader
stage. As a result, some slots will be NULL. This makes things a
little more complex in the linker, but it simplifies things in other
places.
As a side effect _NumLinkedShaders is removed.
NOTE: This may be a candidate for the 7.9 branch. If there are other
patches that get backported to 7.9 that use _LinkedShader, this patch
should be cherry picked also.
This implements round() via the ir_unop_round_even opcode, rather than
adding a new opcode. We may wish to add one in the future, since it
might enable a small performance increase on some hardware, but for now,
this should suffice.
Tested with demos/pixeltest - line rasterization doesn't seem to be
set up for GL conventions yet, but at least width is respected now.
Signed-off-by: Dave Airlie <airlied@redhat.com>
It is now to the point where we have no regressing piglit tests. It
also fixes Yo Frankie! and Humus DynamicBranching, probably due to the
piglit bias tests that work that didn't on the Mesa IR backend.
As a downside, performance takes about a 5-10% performance hit at the
moment (e.g. nexuiz 19.8fps -> 18.8fps), which I plan to resolve by
reintroducing 16-wide fragment shaders where possible. It is a win,
though, for fragment shaders using flow control.
Simply using RNDU, RNDZ, or RNDE does not produce the desired result.
Rather, the RND* instructions place a value in the destination register
that may be 1 less than the correct answer. They can also set per-channel
"increment bits" in a flag register, which, if set, mean dest needs to
be incremented by 1. A second instruction - a predicated add -
completes the job.
Notably, RNDD always produces the correct answer in a single
instruction.
Fixes piglit test glsl-fs-trunc.
This cuts usually 2 out of 3 instructions for flag reg generation (if
statements, conditional assignment) by producing the conditional mod
in the expression representing the boolean value.
Fixes glsl-fs-vec4-indexing-temp-dst-in-nested-loop-combined (register
allocation no longer fails for the conditional generation
proliferation)
This will be a place to peephole comparisions directly to the flag
regs, and for now avoids using MOV with conditional mod on gen6, which
is now illegal.
GLES1 and GLES2 install their own exec pointers and don't need the
Save table. Also, the SET_* macros use different indices for the different
APIs so the offsets used in vtxfmt.c are actually wrong for the ES APIs.
Just always check for FLUSH_UPDATE_CURRENT and call Driver.BeginVertices
when necessary. By using the unlikely() macros, this ends up as
a 10% performance improvement (for isosurf, anyway) over the old,
complicated function pointer swapping.
Don't use r0 for FF_SYNC dest reg on Sandybridge, which would
smash FFID field in GS payload, that cause later URB write fail.
Also not use r0 in any URB write requiring allocate.
Actually validate that the implementation supports the particular
shader target as well. Previously if a driver only supported vertex
shaders, for example, glCreateShaderObjectARB would gladly create a
fragment shader.
NOTE: this is a candidate for the 7.9 branch.
This really amounts to just using the return value from
link_function_calls. All the work was being done, but the result was
being ignored.
Fixes piglit test link-unresolved-funciton.
NOTE: this is a candidate for the 7.9 branch.
Together with the previous commit, this generalize the benefits of
d2cf757f44 to all depth formats, in
particular:
- simpler float -> 24unorm conversion
- avoid unsigned comparisons (not directly supported on SSE) by aligning
to the least significant bit
- avoid unecessary/repeated mask ANDing
Verified with trivial/tri-z that the exact same assembly is produced for
X8Z24.
Z32_FLOAT uses <4 x float> as intermediate/destination type,
instead of <4 x i32>.
The necessary bitcasts got removed with commit
5b7eb868fd
Also use depth/stencil type and build contexts consistently, and
make the depth pointer argument a ordinary <i8 *>, to catch this
sort of issues in the future (and also to pave way for Z16 and
Z32_FLOAT_S8_X24 support).
__GLcontextModes is always only used as an implementation internal struct
at this point and we shouldn't install glcore.h anymore. Anything that
needs __GLcontextModes should just include the struct in its headers files
directly.
This is relying on lp_build_pack2 using the sse2 pack intrinsics which
handle clamping.
(Alternatively could have make it use lp_build_packs2 but it might
not even produce more efficient code than not using the fastpath
in the first place.)
Fixes this GCC warning.
r300_state.c: In function 'r300InvalidateState':
r300_state.c:2247: warning: 'hw_format' may be used uninitialized in this function
r300_state.c:2247: note: 'hw_format' was declared here
This adds proper support for the GL_ARB_shader_stencil_export extension
to the GLSL compiler. Thanks to Ian for pointing out where I need to add things.
If the pipe driver has shader stencil export we can accelerate DrawPixels
using it. It tries to pick an S8 texture and works its way to X24S8 and S8X24
if that isn't supported.
We need a texture to put the drawpixels stuff into, an S8 texture is less
memory/bandwidth than the 32-bit X24S8, but we might not be able to render
directly to an S8, so this lets us specify we won't be rendering to this
texture.
this improves mesa texstore for 8/24 so it can create S24X8/X24S8 variants
by keeping the depth bits static.
it also adds a texstore for S8 so we can write out an S8 texture to use
in the sampler for accel draw pixels to save memory bw.
The logic seems sound here, I've worked it out a few times on paper, though
it would be good to have some review.
Signed-off-by: Dave Airlie <airlied@redhat.com>
There was a check to only do the rebase if we didn't have everything
in VBOs, but nexuiz apparently hands us a mix of VBOs and arrays,
resulting in blocking on the GPU to do a rebase.
Improves nexuiz 800x600, high-settings performance on my Ironlake 41%
(+/- 1.3%), from 14.0fps to 19.7fps.
The format selection of the CopyTexSubImage is pretty bogus still, but
this at least avoids software fallbacks in nexuiz, bringing
performance from 7.5fps to 12.8fps on my machine.
This assertion was added in commit f1c1ee11, but it did not notice
that the array is accessed with 'size-1' instead of 'size'. As a
result, the assertion was off by one. This caused failures in at
least glsl-orangebook-ch06-bump.
If an GLSL shader is used that does not provide all stages and
assembly shaders are provided for the missing stages, validate the
assembly shaders.
Fixes bugzilla #30787 and piglit tests glsl-invalid-asm0[12].
NOTE: this is a candidate for the 7.9 branch.
There was actually a large quantity of scalar code in these functions
previously. This tries to move more into intrinsics.
Introduce an sse2 mm_mullo_epi32 replacement to avoid sse4 dependency
in the new rasterization code.
It's now much more correct for gen6 than the old backend, with just 2
regressions I've found (one of which is common with pre-gen6 and will
be fixed by an array splitting IR pass).
This does leave the old Mesa IR backend getting used still when we
don't have GLSL IR, but the plan is to get GLSL IR input to the driver
for the ARB programs and fixed function by the next release.
Pre-gen6, you could mix int and float just fine. Now, you get goofy
results.
Fixes:
glsl-arb-fragment-coord-conventions
glsl-fs-fragcoord
glsl-fs-if-greater
glsl-fs-if-greater-equal
glsl-fs-if-less
glsl-fs-if-less-equal
This is a hw requirement in math args. This also is inefficient, as
we're calculating the same result 8 times, but then we've been doing
that on pre-gen6 as well. If we're doing math on uniforms, though,
we'd probably be better served by having some sort of mechanism for
precalculating those results into another uniform value to use.
Fixes 7 piglit math tests.
Add ability to set the GLSL version used by the GLcontext by setting the
environment variable INTEL_GLSL_VERSION. For example,
env INTEL_GLSL_VERSION=130 prog args
If the environment variable is missing, the GLSL versions defaults to 120.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
By calling radeon_draw_buffers (which sets the necessary flags
in radeon->NewGLState) and revalidating if NewGLState is non-zero
in r200TclPrimitive. This fixes an assert in libdrm (the color-/
depthbuffer was changed but not yet validated) and and stops the
kernel cs checker from complaining about them (when they're too
small).
Thanks to Mario Kleiner for the hint to call radeon_draw_buffer
(instead of my half-broken hack).
v2: Also fix the swtcl r200 path.
Cc: Mario Kleiner <mario.kleiner@tuebingen.mpg.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This didn't produce a statistically significant performance difference
in my demo (n=4) or nexuiz (n=3), but it still seems like a good idea
and is recommended by the HW team.
Having the single opcode write then read the reg meant that single
instruction opcodes had to consider their source regs to interfere
with their dest regs.
SSE support for 32bit and 16bit unsigned arithmetic is not complete, and
can easily result in inefficient code.
In most cases signed/unsigned doesn't make a difference, such as for
integer texture coordinates.
So remove uint_coord_type and uint_coord_bld to avoid inefficient
operations to sneak in the future.
We need to move the texture sampler resources out of the range of the vertex attribs.
We could probably improve this using an allocator but this is the simple answer for now.
makes mesa-demos/src/glsl/vert-tex work.
We can't patch true-block at end-if time, as there is no guarantee that
the block at the beginning of the true stanza is the same at the end of
the true stanza -- other control flow elements may have been emitted half
way the true stanza.
Although this bug surfaced recently with the commit to skip mip filtering
when lod is an integer the bug was always there, although probably it
was avoided until now: e.g., cubemap selection nests if-then-else on the
else stanza, which does not suffer from the same problem.
We've been using these in the linear path for a while now. Based on
Chris's SSSE3 code, but using only sse2 opcodes. Speed seems to be
identical, but code is simpler & removes dependency on SSE3.
Should be easier to extend to other rgba8 formats.
Specifically, can do early-depth-test even when alpahtest or
kill-pixel are active, providing we defer the actual z write until the
final mask is avaialable.
Improves demos/fire.c especially in the case where you get close to
the trees.
The current interpolation schemes causes precision loss.
Changing the operation order helps, but does not completely avoid the
problem.
The only short term solution is to clamp z to 1.0.
This is unfortunate, but probably unavoidable until interpolation is
improved.
Operate simultanouesly on <width, height, depth> vector as much as possible,
instead of doing the operations on vectors with broadcasted scalars.
Also do the 24.8 fixed point scalar with integer shift of the texture size,
for unnormalized coordinates.
AoS path only for now -- the same thing can be done for SoA.
Fixes these GCC warnings.
brw_wm_fp.c: In function 'search_or_add_const4f':
brw_wm_fp.c:92: warning: 'reg.Index2' is used uninitialized in this function
brw_wm_fp.c:84: note: 'reg.Index2' was declared here
brw_wm_fp.c:92: warning: 'reg.RelAddr2' is used uninitialized in this function
brw_wm_fp.c:84: note: 'reg.RelAddr2' was declared here
Clamp against 0 instead of -0.5, which simplifies things.
The former version would have resulted in both int coords being zero
(in case of coord being smaller than 0) and some "unused" weight value,
whereas now the int coords will be 0 and 1, but weight will be 0, hence the
lerp should produce the same value.
Still not happy about differences between normalized and non-normalized...
Haven't looked at what code this exactly generates but URem can't be fast.
Instead of using two URem only use one and replace the second one with
select/add (this is what the corresponding aos code already does).
Rearrange order of operations a bit to make some clamps easier.
All calculations should be equivalent.
Note there seems to be some inconsistency in the clamp to edge case
wrt normalized/non-normalized coords, could potentially simplify this too.
Sometimes coords are clamped to positive numbers before doing conversion
to int, or clamped to 0 afterwards, in this case can use itrunc
instead of ifloor which is easier. This is only the case for nearest
calculations unfortunately, except linear MIRROR_CLAMP_TO_EDGE which
for the same reason can use a unsigned float build context so the
ifloor_fract helper can reduce this to itrunc in the ifloor helper itself.
sse2 supports round to nearest directly (or rather, assuming default nearest
rounding mode in MXCSR). Use intrinsic to use this rather than round (sse41)
or bit manipulation whenever possible.
The constness of the function parameter gets inlined with the rest of
the function. However, there is also an assignment to the parameter.
If this occurs inside a loop the loop analysis code will get confused
by the assignment to a read-only variable.
Fixes bugzilla #30552.
NOTE: this is a candidate for the 7.9 branch.
Improves performance of my GLSL demo 14.3% (+/- 4%, n=4) by
eliminating the moves used in ir_assignment and ir_swizzle handling.
Still 16.5% to go to catch up to the Mesa IR backend, presumably
because instructions are almost perfectly mis-scheduled now.
We were trying to remap a fully-filled array down to only handing the
WM the components it uses. This is called attribute swizzling, and if
you don't enable it you just get 1:1 mappings of inputs to outputs.
This almost fixes glsl-routing, except for the highest gl_TexCoord[]
indices.
We would compute a new buffer, but never point the hardware at the new
buffer. This partially fixes glsl-routing, as now it get the updated
uniform for which attribute to draw.
Avoid multiplying fixed-point values. Calculate triangle area in
floating point use that for culling.
Lift area calculations up a level as we are already doing this in the
triangle_both() case.
Would like to share the calculated area with attribute interpolation,
but the way the code is structured makes this difficult.
interp data is stored in gpr0 so first interp overwrote it
and subsequent ones got wrong values
reserve register 0 so it's not used for attribs.
alternative is to interpolate attrib0 last (reverse, as r600c does)
We sensibly only provide it if the FS asks for it. We could actually
skip WPOS unless the FS needed WPOS.zw, but that's something for
later.
Fixes: glsl-texture2d and probably many others.
- Handle PIPE_FORMAT_Z32_FLOAT packing correctly.
- In the integer version z shouldn't be passed as as double.
- Make it clear that the integer versions should only be used for masks.
- Make integer type sizes explicit (uint32_t for now, although
uint64_t will be necessary later to encode f32_s8_x24).
The jump delta is now in the part of the instruction where the
destination fields used to be, and the src args are ignored (or not,
for the new non-predicated IF that we don't use yet).
Once a fragment is generated with LP_INTERP_PERSPECTIVE set for an input,
it will do a divide by w for that input. Therefore it's not OK to treat LP_INTERP_PERSPECTIVE as
LP_INTERP_LINEAR or vice-versa, even if the attribute is known to not
vary.
A better strategy would be to take the primitive in consideration when
generating the fragment shader key, and therefore avoid the per-fragment
perspective divide.
Got a speed up by tracking the dirty blocks in a seperate list instead of looping through all blocks. This version should work with block that get their dirty state disabled again and I added a dirty check during the flush as some blocks were already dirty.
Flush read cache before writting register. Track flushing inside
of a same cs and avoid reflushing same bo if not necessary. Allmost
properly force flush if bo rendered too and then use as a texture
in same cs (missing pipeline flush dunno if it's needed or not).
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
When we go to do a lot of bos in one draw like constant bufs we need
to avoid bouncing off the busy ioctl, this mitigates by backing off
on busy bos for a short amount of times.
These texture formats (like R16G16B16A16_UNORM) were untested until now
because st/mesa doesn't use them. I am testing this with a hacked st/mesa
here.
I am trying to be exhaustive, but still I might have missed tons of other
changes to Gallium.
(cherry picked from commit 968a9ec76e)
Conflicts:
docs/relnotes-7.9.html
The glsl core should be handling most dead code issues for us, but we
generate some things in codegen that may not get used, like the 1/w
value or pixel deltas. It seems a lot easier this way than trying to
work out up front whether we're going to use those values or not.
this code was memcmp'ing two structs, but refcounting one of them afterwards,
so any subsequent memcmp was never going to work.
again this stops unnecessary uploads of vertex program,
The brw_wm_surface_state.c handling of GL_DEPTH_TEXTURE_MODE doesn't
apply to shadow compares, which always return an intensity value. The
texture swizzles can do the job for us.
Fixes:
glsl1-shadow2D(): 1
glsl1-shadow2D(): 3
The hw swizzles have been obtained by a brute force approach,
and only C0 and C2 are stored in UV88, the other channels are
ignored.
R16G16 is going to be a lot trickier.
Luckily, one of them would result in failing out register allocation
when the other bugs were encountered. Applies to
glsl-fs-vec4-indexing-temp-dst-in-nested-loop-combined, which still
fails register allocation, but now legitimately.
Fixes several problems:
The half-float, float, and integer internal formats depend on
ARB_texture_rg and other extensions.
RG_INTEGER is not a valid internal format.
Generic compressed formats depend on ARB_texture_rg, not
EXT_texture_compression_rgtc.
Use GL_RED instead of GL_R.
We should fix the SF to actually give us just the data we need, but
this fixes regressions in the new FS until then.
Fixes:
glsl-kwin-blur
glsl-routing
Trying to track the insanity of the different argument layouts for
normal/shadow crossed with normal/lod/bias one generation at a time is
enough.
Fixes: glsl1-texture2D() with bias.
(first test passing in this code that doesn't pass without it!)
We should end up with the same code, but anyone else with this issue
could share the handling (which I got wrong for shadow comparisons in
the driver before).
I've screwed this up enough times that I don't think it's worth it.
This time, it was that I was doing it once per top-level body
instruction instead of just once at the end of the loop body.
The interference is totally bogus (maximal), so this is equivalent to
our trivial register assignment before. As in, passes the same set of
piglit tests.
Now, virtual GRFs are consecutive integers, rather than offsetting the
next one by the size. We need the size information to still be around
for real register allocation, anyway.
Fixes this GCC warning on linux-x86 build.
r3xx_vertprog.c: In function ‘ei_if’:
r3xx_vertprog.c:396: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
r500_fragprog_emit.c: In function ‘emit_paired’:
r500_fragprog_emit.c:237: warning: ISO C90 forbids mixed declarations and code
r500_fragprog_emit.c: In function ‘emit_tex’:
r500_fragprog_emit.c:367: warning: ISO C90 forbids mixed declarations and code
r500_fragprog_emit.c: In function ‘emit_flowcontrol’:
r500_fragprog_emit.c:415: warning: ISO C90 forbids mixed declarations and code
r500_fragprog_emit.c: In function ‘r500BuildFragmentProgramHwCode’:
r500_fragprog_emit.c:633: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
r500_fragprog.c: In function ‘r500_transform_IF’:
r500_fragprog.c:45: warning: ISO C90 forbids mixed declarations and code
r500_fragprog.c: In function ‘r500FragmentProgramDump’:
r500_fragprog.c:256: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
r300_fragprog_emit.c: In function ‘emit_alu’:
r300_fragprog_emit.c:143: warning: ISO C90 forbids mixed declarations and code
r300_fragprog_emit.c:156: warning: ISO C90 forbids mixed declarations and code
r300_fragprog_emit.c: In function ‘finish_node’:
r300_fragprog_emit.c:271: warning: ISO C90 forbids mixed declarations and code
r300_fragprog_emit.c: In function ‘emit_tex’:
r300_fragprog_emit.c:344: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
r300_fragprog_swizzle.c: In function ‘r300_swizzle_is_native’:
r300_fragprog_swizzle.c:120: warning: ISO C90 forbids mixed declarations and code
r300_fragprog_swizzle.c: In function ‘r300_swizzle_split’:
r300_fragprog_swizzle.c:159: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning on linux-x86 build.
radeon_rename_regs.c: In function ‘rc_rename_regs’:
radeon_rename_regs.c:112: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning on linux-x86 build.
radeon_remove_constants.c: In function ‘rc_remove_unused_constants’:
radeon_remove_constants.c💯 warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warning on linux-x86 build.
radeon_optimize.c: In function ‘constant_folding’:
radeon_optimize.c:419: warning: ISO C90 forbids mixed declarations and code
radeon_optimize.c:425: warning: ISO C90 forbids mixed declarations and code
radeon_optimize.c:432: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
radeon_dataflow_deadcode.c: In function ‘push_branch’:
radeon_dataflow_deadcode.c:112: warning: ISO C90 forbids mixed declarations and code
radeon_dataflow_deadcode.c: In function ‘update_instruction’:
radeon_dataflow_deadcode.c:183: warning: ISO C90 forbids mixed declarations and code
radeon_dataflow_deadcode.c: In function ‘rc_dataflow_deadcode’:
radeon_dataflow_deadcode.c:352: warning: ISO C90 forbids mixed declarations and code
radeon_dataflow_deadcode.c:379: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning on linux-x86 build.
radeon_pair_regalloc.c: In function ‘rc_pair_regalloc_inputs_only’:
radeon_pair_regalloc.c:330: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
radeon_pair_schedule.c: In function ‘emit_all_tex’:
radeon_pair_schedule.c:244: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c: In function ‘destructive_merge_instructions’:
radeon_pair_schedule.c:291: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c:438: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c: In function ‘scan_read’:
radeon_pair_schedule.c:619: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c: In function ‘scan_write’:
radeon_pair_schedule.c:645: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c: In function ‘schedule_block’:
radeon_pair_schedule.c:673: warning: ISO C90 forbids mixed declarations and code
radeon_pair_schedule.c: In function ‘rc_pair_schedule’:
radeon_pair_schedule.c:730: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
radeon_pair_translate.c: In function ‘set_pair_instruction’:
radeon_pair_translate.c:153: warning: ISO C90 forbids mixed declarations and code
radeon_pair_translate.c:170: warning: ISO C90 forbids mixed declarations and code
radeon_pair_translate.c: In function ‘rc_pair_translate’:
radeon_pair_translate.c:336: warning: ISO C90 forbids mixed declarations and code
radeon_pair_translate.c:341: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings on linux-x86 build.
radeon_program_alu.c: In function ‘r300_transform_trig_simple’:
radeon_program_alu.c:882: warning: ISO C90 forbids mixed declarations and code
radeon_program_alu.c:932: warning: ISO C90 forbids mixed declarations and code
radeon_program_alu.c: In function ‘radeonTransformTrigScale’:
radeon_program_alu.c:996: warning: ISO C90 forbids mixed declarations and code
radeon_program_alu.c: In function ‘r300_transform_trig_scale_vertex’:
radeon_program_alu.c:1033: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning on linux-x86 build.
radeon_emulate_loops.c: In function ‘rc_emulate_loops’:
radeon_emulate_loops.c:517: warning: ISO C90 forbids mixed declarations and code
Fixes these GCC warnings with linux-x86 build.
radeon_emulate_branches.c: In function ‘handle_if’:
radeon_emulate_branches.c:65: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c:71: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c: In function ‘handle_else’:
radeon_emulate_branches.c:94: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c: In function ‘handle_endif’:
radeon_emulate_branches.c:201: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c: In function ‘fix_output_writes’:
radeon_emulate_branches.c:267: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c:284: warning: ISO C90 forbids mixed declarations and code
radeon_emulate_branches.c: In function ‘rc_emulate_branches’:
radeon_emulate_branches.c:307: warning: ISO C90 forbids mixed declarations and code
Fixes this GCC warning.
math/m_debug_xform.c: In function '_math_test_all_transform_functions':
math/m_debug_xform.c:320: warning: format not a string literal and no format arguments
Fixes this GCC warning.
math/m_debug_norm.c: In function '_math_test_all_normal_transform_functions':
math/m_debug_norm.c:365: warning: format not a string literal and no format arguments
Fixes this GCC warning.
math/m_debug_clip.c: In function '_math_test_all_cliptest_functions':
math/m_debug_clip.c:363: warning: format not a string literal and no format arguments
Instead of creating group of register use a hash table
to lookup into which block each register belongs. This
simplify code a bit.
Signed-off-by: Jerome Glisse <jglisse@redhat.com
Fixes glean pbo crash.
It would be possible to avoid crashing without decoupling, but given
that state trackers give no guarantee that number of views is consistent,
that would likely cause too many state updates (or miss some).
Corrections in store_clip to store clip coordinates in AoS form.
Viewport & cliptest flag options based on variant key.
Put back draw_pt_post_vs and now 2 paths based on whether clipping
occurs or not.
Cliptesting now done at the end of vs in draw_llvm instead of
draw_pt_post_vs.
Added viewport mapping transformation and further cliptesting to
vertex shader in draw_llvm.c
Alternative path where vertex header setup, clip coordinates store,
cliptesting and viewport mapping are done earlier in the vertex
shader.
Still need to hook this up properly according to the return value of
"draw_llvm_shader" function.
I'm still not pleased with how builtin uniforms are handled, but as
long as we're relying on the prog_statevar stuff this seems about as
good as it'll get.
Otherwise, we'll often end up with gl_TexCoord being 0 length, for
example. With ir_to_mesa, things ended up working out anyway, as long
as multiple implicitly-sized arrays weren't involved.
Fixes these tests using gl_FragData or just gl_FragDepth:
glsl1-Preprocessor test (extension test 1)
glsl1-Preprocessor test (extension test 2)
glsl-bug-22603
We don't set the type on the array virtual reg as a whole, so here's
the right place.
Fixes:
glsl1-GLSL 1.20 arrays
glsl1-temp array with constant indexing, fragment shader
glsl1-temp array with swizzled variable indexing
We need to re-run channel expressions afterwards as it generates new
vector expressions, and we need to successfully support conditional
assignment (brw_CMP takes 2 operands, not 1).
While much of this we will want to support natively, this should make
the task of reaching the Mesa IR backend's quality easier.
Fixes:
glsl-fs-main-return.
New design seems to be on parity according to piglit,
make it default to get more exposure and see if there
is any show stopper in the coming days.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Sandybridge's PS constant buffer payload size is decided from
push const buffer command, incorrect size would cause wrong data
in payload for position and vertex attributes. This fixes coefficients
for tex2d/tex3d.
The driver actually creates a 3D texture aligned to POT and does all
the magic with texture coordinates in the fragment shader. It first
emulates REPEAT and MIRRORED wrap modes in the fragment shader to get
the coordinates into the range [0, 1]. (already done for 2D NPOT)
Then it scales them to get the coordinates of the NPOT subtexture.
NPOT textures are now less of a lie and we can at least display
something meaningful even for the 3D ones.
Supported wrap modes:
- REPEAT
- MIRRORED_REPEAT
- CLAMP_TO_EDGE (NEAREST filtering only)
- MIRROR_CLAMP_TO_EDGE (NEAREST filtering only)
- The behavior of other CLAMP modes is undefined on borders, but they usually
give results very close to CLAMP_TO_EDGE with mirroring working perfectly.
This fixes:
- piglit/fbo-3d
- piglit/tex3d-npot
Taking the W component from coords directly ignores swizzling. Instead,
take the component which is mapped to W in the TEX instruction parameter.
The same for Z.
NOTE: This is a candidate for the 7.9 branch.
Some random stuff I had here.
1) Fixed some misleading comments.
2) Removed fake_npot, since it's redundant.
3) lower_texture_rect -> scale_texcoords
4) Reordered and reindented some TEX transform code.
It's trying to get an int smeared across all channels, not trying to
get a 1:1 mapping of a subset of a vector's channels. This usually
ended up not mattering with ir_to_mesa, since it just smears floats
into every chan of a vec4.
Fixes:
glsl1-temp array with swizzled variable indexing
The pipe_sampler_view's swizzle terms also apply to the texture border
color. Simply move the apply_sampler_swizzle() call after we fetch
the border color.
Fixes many piglit texwrap failures.
Changes in v2:
- Invalidate last_tile_addr on any change, fixing regressions
- Correct coding style
Currently softpipe ends up allocating more than 200 MB of memory
for each context due to the tile caches.
Even worse, this memory is all explicitly cleared, which means that the
kernel must actually back it with physical RAM right away.
This change allocates tile memory on demand.
Signed-off-by: Brian Paul <brianp@vmware.com>
Build packet header once and allow to add fake register support so
we can handle things like indexed set of register (evergreen sampler
border registers for instance.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Commit 80ee3a440c added a PROGS_DEPS
definition, but no uses, even though it seems clearly intended
to be a set of additional dependencies for $(PROGS).
Correct this.
Currently makedepend is used by the Mesa Makefile-based build system,
but not required.
Unfortunately, not having it makes dependency resolution non-existent,
which is a source of subtle bugs, and is a rarely tested
configuration, since all Mesa developers likely have it installed.
Furthermore some idioms require dependency resolution to work at all,
such as making headers depend on generated files.
Fixes these GCC warnings.
r600_shader.c: In function 'tgsi_tex':
r600_shader.c:1611: warning: 'src2_chan' may be used uninitialized in this function
r600_shader.c:1611: warning: 'src_chan' may be used uninitialized in this function
When occlusion query are running we want to have accurate
fragment count thus disable any early culling optimization
GPU has.
Based on work from Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Fixes this GCC warning.
radeon_state.c: In function 'radeon_state_fini':
radeon_state.c:140: warning: 'return' with a value, in function returning void
Move GB_ENABLE to derived rs state, and find sprite coord for the correct
generic and enable the tex coord for that generic.
Signed-off-by: Dave Airlie <airlied@redhat.com>
1. We can't turn an instruction into a presubtract operation if it
writes to one of the registers it reads from.
2. If we turn an instruction into a presubtract operation, we can't
remove that intruction unless all readers can use the presubtract
operation.
This fixes fdo bug 30337.
This is a candidate for the 7.9 branch.
The variables are used only in currently disabled code.
Fixes this GCC warning.
r600_context.c: In function 'r600_flush':
r600_context.c:76: warning: unused variable 'dname'
r600_context.c:75: warning: unused variable 'dc'
this is more a proof to show vector shifts on x86 with per-element shift count
are evil. Since we can avoid the shift with a single compare/select, use that
instead. Replaces more than 20 instructions (and slow ones at that) with about 3,
and cuts compiled shader size with mesa's yuvsqure demo by over 10%
(no performance measurements done - but selection is blazing fast).
Might want to revisit that for future cpus - unfortunately AVX won't have vector
shifts neither, but AMD's XOP will, but even in that case using selection here
is probably not slower.
While it's true that llvm can and will indeed replace this with bit
arithmetic (since block height/width is POT), it does so (llvm 2.7) by element
and hence extracts/shifts/reinserts each element individually.
This costs about 16 instructions (and extract is not really fast) vs. 1...
Fixes this GCC warning.
r600_hw_states.c: In function 'r600_translate_fill':
r600_state_inlines.h:136: warning: control reaches end of non-void function
The variables are only used in currently disabled code.
Fixes this GCC warning.
r600_state2.c: In function 'r600_flush2':
r600_state2.c:613: warning: unused variable 'dname'
r600_state2.c:612: warning: unused variable 'dc'
Fix this GCC warning.
intel_pixel_bitmap.c: In function 'intelBitmap':
intel_pixel_bitmap.c:343: warning: implicit declaration of function '_mesa_meta_Bitmap'
Silence this GCC warning.
r300_state_derived.c: In function 'r300_update_derived_state':
r300_state_derived.c:578: warning: 'r' may be used uninitialized in this function
r300_state_derived.c:578: note: 'r' was declared here
Up to 2010-09-19:
r600g: fix tiling support for ddx supplied buffers
9b146eae25
user buffer seems to be broken... new to fix that.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Before, changing any of these sampler values triggered generation
of new JIT code. Added a new flag for the special case of
min_lod == max_lod which is hit during auto mipmap generation.
TX_BORDER_COLOR should be formatted according to the texture format.
Also the interaction with ARB_texture_swizzle should be fixed too.
NOTE: This is a candidate for the 7.9 branch.
This commit fixes an infinite loop in foreach_s if remove_from_list is used
more than once on the same item with other list operations in between.
NOTE: This is a candidate for the 7.9 branch because the commit
"r300g: fixup long-lived BO maps being incorrectly unmapped when flushing"
depends on it.
This is already covered by radeon_mipmap_tree.c, and my convolution
cleanups broke in the presence of this code. Thanks to Marek Olšák
for tracking down the relevant miptree code for me.
Make libr300compiler.a a PHONY target so that this library will always be
built. This fixes the problem of libr300compiler.a not being updated
when r300g is being built and r300c is not.
This is a candidate for the Mesa 7.9 branch.
Many of the EXT_ extensions in the subset have significant code
overhead with no users. It is not a required part of GL -- though
text describing the extension is part of the core spec since 1.2, it
is always conditional on the ARB_imaging extension.
Some pathological triangles cause a theoritically impossible number of
clipped vertices.
The clipper will still assert, but at least release builds will not
crash, while this problem is further investigated.
use the blitter + custom stage to avoid doing a whole lot of state
setup by hand. This makes life a lot easier for doing this on evergreen
it also keeps all the state setup in one place.
We setup a custom context state at the start with a flag to denote
its for the flush, when it gets generated we generate the correct state
for the flush and no longer have to do it all by hand.
this should also make adding texture *to* depth easier.
It turns out that most people new to this IR are surprised when an
assignment to (say) 3 components on the LHS takes 4 components on the
RHS. It also makes for quite strange IR output:
(assign (constant bool (1)) (x) (var_ref color) (swiz x (var_ref v) ))
(assign (constant bool (1)) (y) (var_ref color) (swiz yy (var_ref v) ))
(assign (constant bool (1)) (z) (var_ref color) (swiz zzz (var_ref v) ))
But even worse, even we get it wrong, as shown by this line of our
current step(float, vec4):
(assign (constant bool (1)) (w)
(var_ref t)
(expression float b2f (expression bool >=
(swiz w (var_ref x))(var_ref edge))))
where we try to assign a float to the writemasked-out x channel and
don't supply anything for the actual w channel we're writing. Drivers
right now just get lucky since ir_to_mesa spams the float value across
all the source channels of a vec4.
Instead, the RHS will now have a number of components equal to the
number of components actually being written. Hopefully this confuses
everyone less, and it also makes codegen for a scalar target simpler.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We can't expect to have a context when this is called, and we don't need one
so just require a __DRIscreen instead.
Reported by Yu Dai <yu.dai@intel.com>
Shader rebuild should be more clever, we should store along each
shader all the value that change shader program rather than using
flags in context (ie change sequence like : change vs buffer, draw,
change vs buffer, switch shader will trigger useless shader rebuild).
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Normally the Mesa state tracker uses TXP instructions for texturing.
But if a fragment shader uses texture2D() that's a TEX instruction.
In that case we were incorrectly computing the texcoord coefficients
in the point sprite setup code. Some new comments in the code explain
things.
Fix commit e7087175f8. Move the reference to
GL_VERSION_2_1_functions to intel_extensions.c where it's available,
don't try to enable a non-existing extension and advertise 1.20 for all
intel chipsets, not just GEN4 and up.
progs requires winsys, which hasn't yet been built by the time we
go into state_trackers.
It may be a good idea to also move it into tests.
After a normal build, run make in src/gallium/state_trackers/d3d1x/progs
to build them.
The Gallium EGL state tracker reuses dri2.c but not the GLX code.
Currently there is a bit of code in dri2.c that is incorrectly tied
to GLX: instead, make it call an helper that both GLX and Gallium EGL
implement, like dri2InvalidateBuffers.
This avoids a link error complaining that dri2GetGlxDrawableFromXDrawableId
is undefined.
Note that we might want to move the whole event translation elsewhere,
and probably stop using non-XCB DRI2 altogether, but this seems to be
the minimal fix.
This may just be hiding some other bug though, since the types are supposed
to be the same (and it compiles for me).
Anyway, this interface will likely need to changed, since it seems Wine needs
a more powerful one capable of expressing window subregions and called at
every Present.
add cb/db flush states to the blit code.
add support for the rv6xx that need special treatment.
according to R6xx_7xx_3D.pdf
set r700 CB_SHADER_CONTROL reg in blit code
docs say dual export should be disabled for DB->CB
Instead of using the invalid GL_ARB_shading_language_120 extension to
determine the GLSL version, use a new ctx->Const.GLSLVersion field.
Updated the intel and r600 drivers, but untested.
See fd.o bug 29910
NOTE: This is a candidate for the 7.9 branch (but let's wait and see if
there's any regressions).
The old code didn't really make sense. We only need to compare the
X channel of the texture (depth) against the texcoord.
For (bi)linear sampling we should move the calls to this function
and compute the final result as (s1+s2+s3+s4) * 0.25. Someday.
This fixes the glean glsl1 shadow2D() tests. See fd.o bug 29307.
There was some libstdc++-specific code that would only build with GCC 4.5
Now it should be much more compatible, at the price of reimplementing
the generic hash function.
Looks like the problem was we weren't passing the depth to the render
target as expected, so the chip would wedge. Fixes GPU hang in
occlusion-query-discard.
Bug #30097
Recent Wine versions provide a d3d11shader.h, which is however empty
and was getting used instead of our non-empty one.
Correct the include path order to fix this.
This is a new implementation of the Direct3D 11 COM API for Gallium.
Direct3D 10 and 10.1 implementations are also provided, which are
automatically generated with s/D3D11/D3D10/g plus a bunch of #ifs.
While this is an initial version, most of the code is there (limited
to what Gallium can express), and tri, gears and texturing demos
are working.
The primary goal is to realize Gallium's promise of multiple API
support, and provide an API that can be easily implemented with just
a very thin wrapper over Gallium, instead of the enormous amount of
complex code needed for OpenGL.
The secondary goal is to run Windows Direct3D 10/11 games on Linux
using Wine.
Wine dlls are currently not provided, but adding them should be
quite easy.
Fglrx and nvidia drivers can also be supported by writing a Gallium
driver that talks to them using OpenGL, which is a relatively easy
task.
Thanks to the great design of Direct3D 10/11 and closeness to Gallium,
this approach should not result in detectable overhead, and is the
most maintainable way to do it, providing a path to switch to the
open Gallium drivers once they are on par with the proprietary ones.
Currently Wine has a very limited Direct3D 10 implementation, and
completely lacks a Direct3D 11 implementation.
Note that Direct3D 10/11 are completely different from Direct3D 9
and earlier, and thus warrant a fully separate implementation.
The third goal is to provide a superior alternative to OpenGL for
graphics programming on non-Windows systems, particularly Linux
and other free and open systems.
Thanks to a very clean and well-though design done from scratch,
the Direct3D 10/11 APIs are vastly better than OpenGL and can be
supported with orders of magnitude less code and development time,
as you can see by comparing the lines of code of this commit and
those in the existing Mesa OpenGL implementation.
This would have been true for the Longs Peak proposal as well, but
unfortunately it was abandoned by Khronos, leaving the OpenGL
ecosystem without a graphics API with a modern design.
A binding of Direct3D 10/11 to EGL would solve this issue in the
most economical way possible, and this would be great to provide
in Mesa, since DXGI, the API used to bind Direct3D 10/11 to Windows,
is a bit suboptimal, especially on non-Windows platforms.
Finally, a mature Direct3D 10/11 implementation is intrinsically going
to be faster and more reliable than an OpenGL implementation, thanks
to the dramatically smaller API and the segregation of all nontrivial
work to object creation that the application must perform ahead of
time.
Currently, this commit contains:
- Independently created headers for Direct3D 10, 10.1, 11 and DXGI 1.1,
partially based on the existing Wine headers for D3D10 and DXGI 1.0
- A parser for Direct3D 10/11 DXBC and TokenizedProgramFormat (TPF)
- A shader translator from TokenizedProgramFormat to TGSI
- Implementation of the Direct3D 11 core interfaces
- Automatically generated implementation of Direct3D 10 and 10.1
- Implementation of DXGI using the "native" framework of the EGL st
- Demos, usable either on Windows or on this implementation
- d3d11tri, a clone of tri
- d3d11tex, a (multi)texturing demo
- d3d11gears, an improved version of glxgears
- d3d11spikysphere, a D3D11 tessellation demo (currently Windows-only)
- A downloader for the Microsoft HLSL compiler, needed to recompile
the shaders (compiled shader bytecode is also included)
To compile this, configure at least with these options:
--with-state-trackers=egl,d3d1x --with-egl-platforms=x11
plus some gallium drivers (such as softpipe with --enable-gallium-swrast)
The Wine headers (usually from a wine-dev or wine-devel package) must
be installed.
Only x86-32 has been tested.
You may need to run "make" in the subdirectories of src/gallium/winsys/sw
and you may need to manually run "sudo make install" in
src/gallium/targets/egl
To test it, run the demos in the "progs" directory.
Windows binaries are included to find out how demos should work, and to
test Wine integration when it will be done.
Enjoy, and let me know if you manage to compile and run this, or
which issues you are facing if not.
Using softpipe is recommended for now, and your mileage with hardware
drivers may vary.
However, getting this to work on hardware drivers is also obviously very
important.
Note that currently llvmpipe is buggy and causes all 3 gears to be
drawn with the same color.
Use export GALLIUM_DRIVER=softpipe to avoid this.
Thanks to all the Gallium contributors and especially the VMware
team, whose work made it possible to implement Direct3D 10/11 much
more easily than it would have been otherwise.
Use rc_pair_ prefix for all pair instruction structs
Create a named struct for pair instruction args
Replace structs radeon_pair_instruction_{rgb,alpha} with struct
radeon_pair_sub_instruction. These two structs were nearly identical
and were creating a lot of cut and paste code. These changes are the
first step towards removing some of that code.
This allow to share code path btw old & new, also
remove check on reference this might make things
a little slower but new design doesn't use reference
stuff.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
If we are not going to write to the X or Y components of the destination
vector we also don't need to prepare to compute SIN or COS.
Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
When ir_binop_all_equal and ir_binop_any_nequal were introduced, the
meaning of these two opcodes changed to return vectors rather than a
single scalar, but the constant expression handling code was incorrectly
written and only worked for scalars. As a result, only the first
component of the returned vector would be properly initialized.
This reverts commit c32bac57ed
and silences the warning differently.
The _mesa_reference_texobj() function is called quite a bit and
we don't want to call valid_texture_object() all the time in non-
debug builds.
Count int and float constants independently. Since there are only
few i# constants available and hundreds of c# constants, it would
be too easy to end up with an i# declaration out of its range.
Pixel shaders do not have address registers a#, only one
loop register aL. Our only hope is to assume the address
register is in fact a loop counter and replace it with aL.
Do not translate ARL instruction for pixel shaders -- MOVA
instruction is only valid for vertex saders.
Make it more explicit relative addressing of inputs is only valid
for pixel shaders and constants for vertex shaders.
we were leaking buffers since the flush code was added, it wasn't dropping references.
move setting up flush to the set_framebuffer_state.
clean up the flush state object.
make more space in the BOs array for flushing.
This reverts commit a1d9a58b82.
Flushing the upload buffers on draw is wrong, uploads aren't supposed to
cause flushes in the first place. The real issue was
radeon_bo_pb_map_internal() not respecting PB_USAGE_UNSYNCHRONIZED.
Depth-only and stencil-only clears should mask out depth/stencil from the
output, mask out stencil/input from input, and OR or ADD them together.
However, due to a typo they were being ANDed, resulting in zeroing the buffer.
If a upload buffer is used by a previous draw that's still in the CS,
accessing it would need a context flush. However, doing a context flush when
mapping the upload buffer would then flush/destroy the same buffer we're trying
to map there. Flushing the upload buffers before a draw avoids both the CS
flush and the upload buffer going away while it's being used. Note that
u_upload_data() could e.g. use a pool of buffers instead of allocating new
ones all the time if that turns out to be a significant issue.
Silences this GCC warning.
nvfx_vertprog.c: In function 'nvfx_vertprog_translate':
nvfx_vertprog.c:998: warning: assignment discards qualifiers from pointer target type
Those have the callee field set to the null pointer, so
calling the public constructor will segfault.
Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
Fixes this GCC warning.
lower_variable_index_to_cond_assign.cpp:
In member function
'bool variable_index_to_cond_assign_visitor::needs_lowering(ir_dereference_array*) const':
lower_variable_index_to_cond_assign.cpp:261:
warning: control reaches end of non-void function
This diagram shows the rendering pipeline with an emphasis on
the inputs/outputs for each stage. Some stages emit new vertex
attributes and others consume some attributes.
Implement the pipe_rasterizer_state::sprite_coord_enable field
in the draw module (and softpipe) according to what's specified
in the documentation.
The draw module can now add any number of extra vertex attributes
to a post-transformed vertex and generate texcoords for those
attributes per sprite_coord_enable. Auto-generated texcoords
for sprites only worked for one texcoord unit before.
The frag shader gl_PointCoord input is now implemented like any
other generic/texcoord attribute.
The draw module now needs to be informed about fragment shaders
since we need to look at the fragment shader's inputs to know
which ones need auto-generated texcoords.
Only softpipe has been updated so far.
Fixes this GCC warning.
r600_state2.c: In function 'r600_context_flush':
r600_state2.c:946: error: implicit declaration of function 'drmCommandWriteRead'
Winsys context build a list of register block a register block is
a set of consecutive register that will be emited together in the
same pm4 packet (the various r600_block* are there to provide basic
grouping that try to take advantage of states that are linked together)
Some consecutive register are emited each in a different block,
for instance the various cb[0-7]_base. At winsys context creation,
the list of block is created & an index into the list of block. So
to find into which block a register is in you simply use the register
offset and lookup the block index. Block are grouped together into
group which are the various pkt3 group of config, context, resource,
Pipe state build a list of register each state want to modify,
beside register value it also give a register mask so only subpart
of a register can be updated by a given pipe state (the oring is
in the winsys) There is no prebuild register list or define for
each pipe state. Once pipe state are built they are bound to
the winsys context.
Each of this functions will go through the list of register and
will find into which block each reg falls and will update the
value of the block with proper masking (vs/ps resource/constant
are specialized variant with somewhat limited capabilities).
Each block modified by r600_context_pipe_state_set* is marked as
dirty and we update a count of dwords needed to emit all dirty
state so far.
r600_context_pipe_state_set* should be call only when pipe context
change some of the state (thus when pipe bind state or set state)
Then to draw primitive you make a call to r600_context_draw
void r600_context_draw(struct r600_context *ctx, struct r600_draw *draw)
It will check if there is enough dwords in current cs buffer and
if not will flush. Once there is enough room it will copy packet
from dirty block and then add the draw packet3 to initiate the draw.
The flush will send the current cs, reset the count of dwords to
0 and remark all states that are enabled as dirty and recompute
the number of dwords needed to send the current context.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Currenly GLSL happily generates indirect addressing of any kind of
arrays.
Unfortunately DirectX 9 GPUs are not guaranteed to support any of them in
general.
This pass fixes that by lowering such constructs to a binary search on the
values, followed at the end by vectorized generation of equality masks, and
4 conditional assignments for each mask generation.
Note that this requires the ir_binop_equal change so that we can emit SEQ
to generate the boolean masks.
Unfortunately, ir_structure_splitting is too dumb to turn the resulting
constant array references to individual variables, so this will need to
be added too before this pass can actually be effective for temps.
Several patches in the glsl2-lower-variable-indexing were squashed
into this commit. These patches fix bugs in Luca's original
implementation, and the individual patches can be seen in that branch.
This was done to aid bisecting in the future.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
this adds the bo caching layer and uses it for vertex/index/constant bos.
ctx needs to take references on hw bos so the flushing works okay, also
needs to flush the maps.
This function is executed outside _mesa_meta_begin/end(), that means
that e.g. _mesa_meta_Bitmap() clobbers the texturing state because it
changes the currently active texture object.
There's no need to bind the new texture when it's created, it's done
again later anyway (from setup_drawpix/copypix_texture()).
Signed-off-by: Brian Paul <brianp@vmware.com>
Allows disabling various operations (mainly texture-related, but
will grow) to try & identify bottlenecks.
Unlike LP_DEBUG, this is active even in release builds - which is
necessary for performance investigation.
The current context should be notified when the the front/back buffers
of the current drawable are swapped. The notification was skipped when
xmesa_strict_invalidate is false (the default).
This fixes fdo bug #29774.
Enable some extensions now that the needed tokens are defined in
GLES/glext.h and GLES2/glext.h. Update the prototype of MultiDrawArrays
now that the prototype of _mesa_MultiDrawArraysEXT has been updated.
The shadow versions of the texture targets use an extra component
(Z) to express distance from light source to the fragment.
Fixes the shadowtex demo with llvmpipe.
Execute before folding loads, because we don't check if it's legal
in lower_mods.
Ensure that a value's insn pointer is updated when transferring it
to a different instruction.
Fix the following GCC warning.
loop_controls.cpp: In function 'int calculate_iterations(ir_rvalue*, ir_rvalue*, ir_rvalue*, ir_expression_operation)':
loop_controls.cpp:88: warning: format not a string literal and no format arguments
This fixes a DRM deadlock in the cubestorm xscreensaver, because somehow
there must not be 2 different BOs relocated in one CS if both BOs back
the same handle. I was told it is impossible to happen, but apparently
it is not, or there is something else wrong.
Since atm our OPs aren't typed but instead values are, we need to
take care if they're used as different types (e.g. a load makes a
value u32 by default).
Maybe this should be changed (also to match TGSI), but it should
work as well if done properly.
At some point we'll want to support real subroutines instead of
just inlining them into the main shader.
Since recursive calls are forbidden, we can just save all used
registers to a fixed local memory region and restore them on a
return, no need for a stack pointer.
We only uploaded up to the highest offset a program would use,
and if the constant buffer isn't changed when a new program is
used, the new program is missing the rest of them.
Might want to introduce a "fill state" for user mem constbufs.
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=22622">Bug 22622</a> - [GM965 GLSL] noise*() cause GPU lockup</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=23743">Bug 23743</a> - For loop from 0 to 0 not optimized out</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=24553">Bug 24553</a> - shader compilation times explode when using more () pairs</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25664">Bug 25664</a> - [GLSL] re-declaring an empty array fails to compile</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25769">Bug 25769</a> - [GLSL] "float" can be implicitly converted to "int"</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25808">Bug 25808</a> - [GLSL] const variable is modified successfully</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25826">Bug 25826</a> - [GLSL] declaring an unsized array then re-declaring with a size fails</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25827">Bug 25827</a> - [GLSL] vector constructor accepts too many arguments successfully</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25829">Bug 25829</a> - [GLSL] allowing non-void function without returning value</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25830">Bug 25830</a> - [GLSL] allowing non-constant-expression as const declaration initializer</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25877">Bug 25877</a> - [GLSL 1.10] implicit conversion from "int" to "float" should not be allowed</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25878">Bug 25878</a> - [GLSL] sampler is converted to int successfully</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25994">Bug 25994</a> - [GM45][GLSL] 'return' statement in vertex shader unsupported</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=25999">Bug 25999</a> - [GLSL] embedded structure constructor fails to compile</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=26000">Bug 26000</a> - [GLSL] allowing different parameter qualifier between the function definition and declaration</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=26001">Bug 26001</a> - [GLSL 1.10] constructing matrix from matrix succeeds</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=26224">Bug 26224</a> - [GLSL] Cannot get location of a uniform struct member</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=26990">Bug 26990</a> - [GLSL] variable declaration in "while" fails to compile</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=27060">Bug 27060</a> - [965] piglit glsl-fs-raytrace failure due to lack of function calls.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=27216">Bug 27216</a> - Assignment with a function call in an if statement causes an assertion failure</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=27261">Bug 27261</a> - GLSL Compiler fails on the following vertex shader</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=27265">Bug 27265</a> - GLSL Compiler doesnt link the attached vertex shader</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=28834">Bug 28834</a> - Add support for system fpclassify to GL_OES_query_matrix function for OpenBSD / NetBSD</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=28837">Bug 28837</a> - varying vec4 index support</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=28845">Bug 28845</a> - The GLU tesselator code has some warnings</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=28889">Bug 28889</a> - [regression] wine game crash</li>
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
