MCS buffers use 32 bits per pixel in 8x MSAA, and 8 bits per pixel in
4x MSAA. This patch adjusts the format we use to allocate the buffer
so that enough memory is set aside for 8x MSAA.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The code to emit 3DSTATE_SAMPLE_MASK was already correct for 8x
MSAA--this patch just removes an assertion that would have prevented
it from being used for 8x MSAA.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch updates the blorp functions encode_msaa() and decode_msaa()
to properly handle the encoding of IMS MSAA buffers when
num_samples=8.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
When operating in persample dispatch mode, the blorp engine would
previously assume that subspan N always represented sample N (this is
correct assuming 4x MSAA and a 16-wide dispatch). In order to support
8x MSAA, we must compute which sample is associated with each subspan,
using the "Starting Sample Pair Index" field in the thread payload.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When rendering to an IMS MSAA surface on Gen7, blorp sets up the
rendering pipeline as though it were rendering to a single-sampled
surface; accordingly it must adjust the size of the primitive it sends
down the pipeline to account for the interleaving of samples in an IMS
surface.
This patch modifies the size adjustment code to properly handle 8x
MSAA, which makes room for the extra samples by using an interleaving
pattern that is twice as wide as 4x MSAA.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch adds a num_samples argument to the blorp function
manual_blend(), allowing it to be told how many samples need to be
blended together. Previously it assumed 4x MSAA, since that was all
we supported.
We also bump up LOG2_MAX_BLEND_SAMPLES from 2 to 3, so that
manual_blend() will be able to handle 8x MSAA.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When the client program uses glDrawBuffer() or glDrawBuffers() to
select more than one color buffer for drawing into, and then performs
a blit, we need to blit into every single enabled draw buffer.
+2 oglconforms.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50407
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
This patch rearranges the order of steps performed by a blorp blit
from this:
- Sync up state of window system buffers.
- Find buffers.
- Find miptrees.
- Make sure buffer formats match.
- Handle mirroring.
- Make sure width and height match.
- Handle clipping/scissoring.
- Account for window system origin conventions.
- Do depth resolves, if applicable.
- Do the blit.
- Record the need for a future HiZ resolve, if applicable.
To this:
- Sync up state of window system buffers.
- Handle mirroring.
- Make sure width and height match.
- Handle clipping/scissoring.
- Account for window system origin conventions.
- Find buffers.
- Make sure buffer formats match.
- Find miptrees.
- Do depth resolves, if applicable.
- Do the blit.
- Record the need for a future HiZ resolve, if applicable.
The steps are the same, but they are now performed in an order that
will make it possible to implement correct DrawBuffers support. Note
that the last four steps are now in a separate function
(do_blorp_blit), since they will need to be executed repeatedly when
DrawBuffers support is added.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Previously, the blorp engine would fall back to swrast if the source
or destination of a blit had no associated miptree. This was
unnecessary, since _mesa_BlitFramebufferEXT() already takes care of
making the blit silently succeed if there are no buffers bound, so the
fallback paths could never actually happen in practice.
Removing these fallback paths will simplify the implementation of
correct DrawBuffers support in blorp.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
This patch modifies the order of operations in the blorp engine so
that clipping and scissoring are performed before adjusting the
coordinates to account for the difference in origin convention between
window system buffers and framebuffer objects. Previously, we would
do clipping and scissoring after adjusting for origin conventions, so
we would get scissoring wrong in window system buffers.
Fixes Piglit test "fbo-scissor-blit window".
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
When checking that the source and destination dimensions match, we
don't need to store the width and height in variables; doing so just
risks confusion since right after the check, we do clipping and
scissoring, which may alter the width and height.
No functional change.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
On Gen6, multisampled null render targets don't seem to work
properly--they cause the GPU to hang. So, as a workaround, we render
into a dummy color buffer.
Fortunately this situation (multisampled rendering without a color
buffer) is rare, and we don't have to waste too much memory, because
we can give the workaround buffer a very small pitch.
Fixes piglit test "EXT_framebuffer_multisample/no-color {2,4}
depth-computed *" on Gen6.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
The HW docs say that the width and height of null render targets need
to match the width and height of the corresponding depth and/or
stencil buffers, and that they need to be marked as Y-tiled. Although
leaving these values at 0 doesn't seem to cause any ill effects, it
seems wise to follow the documented requirements.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Previously, we used the number of samples in draw buffer 0 to
determine whether to set up the 3D pipeline for multisampling. Using
the visual is cleaner, and has the benefit of working properly when
there is no color buffer.
Fixes all piglit tests "EXT_framebuffer_multisample/no-color" on Gen7.
On Gen6, the "depth-computed" variants of these tests still fail; this
will be addresed in a later patch.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
This patch ensures that Visual.samples and Visual.sampleBuffers are
set correctly even in the case where there is no color buffer.
Previously, these values would retain their default value of 0 in this
circumstance, even if the depth or stencil buffer was multisampled.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Mesa misses a few checks when compiling on a uclibc system
which cause it to fall back on glibc-ism. This patch
addresses those issues.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
The kernel streamout support was supposed to get into 3.3 along
the tiling change and thus use the same kernel version bump of
2.13 to report userspace that streamout register were supported.
This is not what happen. So as streamout kernel support did not
bump the kernel driver version, rely on kernel 2.14 version bump
to know if streamout is enabled or not. Which means you need at
least 3.4 kernel.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
The error was being set on the non-error path, rather
than the error path.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
For 'non-legacy' contexts we will want to generate an error
if an uninstalled function is called.
The effect of this change will be that we can avoid installing
legacy functions, and they will then generate an error as
needed for deprecated functions in GL >= 3.1.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Commit 2d4b77c7 (automake: Convert src/mesa/drivers/x11/Makefile to
automake, 2012-06-12) dropped the old Makefile, which used GL_LIB, and
replaced it with a Makefile.am hard-coding the name "GL". This broke
handling of --enable-mangling and --with-gl-lib-name options which
depend on GL_LIB to specify the GL library name.
Use "@GL_LIB@" in src/mesa/drivers/x11/Makefile.am to configure the
library name. Also use this approach to simplify src/glx/Makefile.am
and drop the HAVE_MANGLED_GL conditional. While at it, fix the
compatibility link we create in "lib" for the software-only driver to
use version GL_MAJOR instead of hard-coding "1".
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
This fixes the piglit EXT_framebuffer_multisample/bitmap tests.
Note that we must not rely on ctx->DrawBuffer when flushing the cache, because
that's already updated with a new framebuffer. We want to draw into the old
framebuffer where glBitmap was called.
Reviewed-by: Brian Paul <brianp@vmware.com>
Testing shows that the standard JIT engine retrofited with AVX support is quite
stable and as capable to handle AVX instructions as MC-JIT is.
And the old JIT is much more memory efficient, as we don't need to
allocate one engine instance per shader, as we do for MC-JIT due to its
incompleteness.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
When X is running it is neccesary for pipe_loader to authenticate with
DRM, in order to be able to use the device.
This makes it possible to run OpenCL programs while X is running.
v2:
- Fix C++ style comments
- Drop Xlib-xcb dependency
- Close the X connection when done
- Split auth code into separate function
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Calling glDeleteShader() should mark shaders as pending for deletion,
but shouldn't decrement the refcount every time. Otherwise, repeated
glDeleteShader() is not safe.
This is particularly bad since glDeleteProgram() frees shaders: if you
first call glDeleteShader() on the shaders attached to the program (thus
decrementing the refcount), then called glDeleteProgram(), it would try
to free them again (decrementing the refcount another time), causing
a refcount > 0 assertion to fail.
Similar to commit d950a778.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
If the pack type is not supported, use _mesa_problem
rather than asserting.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
_mesa_is_integer_format is moved to formats.c and renamed
as _mesa_is_enum_format_integer.
_mesa_is_format_unsigned, _mesa_is_type_integer,
_mesa_is_type_unsigned, and _mesa_is_enum_format_or_type_integer
are added.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
llvm-3.2svn r160587 moved createBoundsCheckingPass from
lib/Transforms/Scalar to lib/Transforms/Instrumentation.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Except for a couple of explicit uses, _mesa_inv_sqrtf was disabled since
its addition in 2003 (see f9b1e524).
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Temporarily disabled since 2003 (see 386578c5b).
This saves us from calling sqrt() 128 times to generate the sqrttab in
one_time_init().
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Found by compiler warning:
i830_texstate.c:131:28: warning: argument to 'sizeof' in 'memset' call
is the same expression as the destination; did you mean to
dereference it? [-Wsizeof-pointer-memaccess]
memset(state, 0, sizeof(state));
~~~~~ ^~~~~
On 64-bit systems, memset here would write an extra 4 bytes.
Note: This is a candidate for the stable branches.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
This can potentially cut shader program size by a factor of 4 for 4-wide
execution respectively 2 for 8-wide execution and while this ratios aren't
quite reached for more complex shaders it can be close.
Could not really measure a performance difference so far except for trivial
shaders (glxgears).
There seems to be a fair amount of unnecessary move's generated especially
at the beginning it might be possible to optimize those away somehow.
Things aren't quite as clean, some additional stuff needs to be done for
keeping both paths working (though llvm might be able to optimize this away).
glxgears seems to lose about 5-10% of performance, looking at the generated
shaders this is actually less than I'd think it would be - both 4 and 8-wide
shaders, despite containing a loop actually have about 10% more instructions
in total, and will have roughly 50% more executed instructions (though mostly
cheap ones). Need to figure out how to reduce overhead...
v2: keep complex interpolation for 4-wide mode, adapt to interface changes.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
This thread count is only supposed to be enabled when "WIZ Hashing Disable in
GT_MODE register enabled." I've always been confused whether that means the
bit in the register should be 1 or 0. For my IVB GT2's register 0x7008 value
of 0x0, this appears to work fine.
Improves l4d2 performance at 640x480 by 0.88 +/- 0.11% (n=88). Improves
performance with rasterization at 1280x1024 by 1.45% +/- 0.36% (n=6).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Now that we finally have a list of uniform blocks in the linked shader
program, we can tell what their indices are.
Fixes piglit GL_ARB_uniform_buffer_object/getuniformblockindex.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
At this point in the linking, we've totally lost track of the struct
gl_uniform_buffer that this pointed to in the original unlinked
shader, so we do a nasty n^2 walk to find it the new one based on the
variable name.
Note that these point into the shader's list of gl_uniform_buffers,
not the linked program's.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We'll need to propagate the UBO fields to the uniform storage records
before we can handle the other pnames.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This is a single entrypoint that maps from a series of names to the
indices of those names within the active uniforms list. Each index is
like glGetUniformLocation()'s return value, except that it doesn't
encode an array offset.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
With the upcoming GL_ARB_uniform_buffer_object changes, the only
other caller that will want the cooked value is state_tracker.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We're going to need this structure to cross-validate the uniform
blocks between shader stages, since unused ir_variables might get
dropped. It's also the place we store the RowMajor qualifier, which
is not part of the GLSL type (since that would cause a bunch of type
equality checks to fail).
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Someone tried to be clever and "optimized" add_vertex_data2() to just use
two points for the texture coordinates and then reuse individual
components. Sadly this is not how matrix multiplication works.
Fixes rendercheck -t tmcoords
Signed-off-by: Lucas Stach <dev@lynxeye.de>
Previously, on Gen7, when texturing from a depth or stencil surface,
the blorp engine would configure the 3D pipeline as though the input
surface was non-multisampled, and perform the necessary coordinate
transformations in the fragment shader to account for the IMS layout.
This meant outputting a lot of extra fragment shader code, and it
raised some uncertainty about how to deal with very large surfaces.
This patch modifies blorp to configure the 3D pipeline properly for
IMS layout when reading from depth and stencil surfaces.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Previously, on Gen7, compute_msaa_layout_for_pipeline() would verify
that IMS layout is not used. However, now that we configure
SURFACE_STATE correctly for IMS surfaces, IMS layout is available.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This patch modifies gen7_set_surface_num_multisamples() to set up the
SURFACE_STATE appropriately for texturing from IMS format MSAA
surfaces (which are only used on Gen7 for depth and stencil buffers).
Since the function now sets more than just the number of multisamples,
it's been renamed to gen7_set_surface_msaa().
This will make it possible to remove some kludginess from the blorp
engine.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
When downsampling a compressed multisampled surface, we can take a
shortcut to downsample any pixels that were completely covered by a
single primitive. In this case, the first color value we fetch is the
correct final color for the downsampled pixel, so we can skip the rest
of the blending operation.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
When downsampling an integer-format buffer on Gen7, we need to use the
"avg" instruction rather than the "add" instruction, to ensure that we
don't overflow the range of 32-bit integers. Also, we need to use the
proper register type (BRW_REGISTER_TYPE_D or BRW_REGISTER_TYPE_UD) for
intermediate color data and for writing to the render target.
Note: this patch causes blorp to use the proper register type for all
operations (downsampling, upsampling, and ordinary blits). Strictly
speaking, this is only necessary for downsampling, because the other
operations exclusively use MOV instructions on the color data. But
it's simpler to use the proper register type in all cases.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
When downsampling from an MSAA image to a single-sampled image, it is
inevitable that some loss of numerical precision will occur, since we
have to use 32-bit floating point registers to hold the intermediate
results while blending. However, it seems reasonable to expect that
when all samples corresponding to a given pixel have the exact same
color value, there will be no loss of precision.
Previously, we averaged samples as follows:
blend = (((sample[0] + sample[1]) + sample[2]) + sample[3]) / 4
This had the potential to lose numerical precision when all samples
have the same color value, since ((sample[0] + sample[1]) + sample[2])
may not be precisely representable as a 32-bit float, even if the
individual samples are.
This patch changes the formula to:
blend = ((sample[0] + sample[1]) + (sample[2] + sample[3])) / 4
This avoids any loss of precision in the event that all samples are
the same, by ensuring that each addition operation adds two equal
values.
As a side benefit, this puts the formula in the form we will need in
order to implement correct blending of integer formats.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
From the Ivy Bridge PRM, Vol4 Part3 p152:
"The avg instruction performs component-wise integer average of
src0 and src1 and stores the results in dst. An integer average
uses integer upward rounding. It is equivalent to increment one to
the addition of src0 and src1 and then apply an arithmetic right
shift to this intermediate value."
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
The kill_emitted variable was duplicating the functionality of
gl_fragment_program::UsesKill. There's no need for both.
Reviewed-by: Eric Anholt <eric@anholt.net>
Previously, the code for setting this flag for GLSL programs was
duplicated in three places: brw_link_shader(), glsl_to_tgsi_visitor,
and ir_to_mesa_visitor. In addition to the unnecessary duplication,
there was a performance problem on i965: brw_link_shader() set the
flag before doing its final round of optimizations, which meant that
if the optimizations managed to eliminate all the discard operations,
the flag would still be set, resulting (at least in theory) in slower
performance.
This patch consolidates all of the code that sets UsesKill for GLSL
programs into do_set_program_inouts(), which already is doing a
similar job for UsesDFdy, and which occurs after i965's final round of
optimizations.
Non-GLSL programs (ARB programs and the state tracker's glBitmap
program) are unaffected.
Reviewed-by: Eric Anholt <eric@anholt.net>
Move it to native_wayland_drm_bufmgr_helper.c which only gets compiled when
wayland is enabled and which already includes the right headers.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
The cube sampler generates two-dimensional texture coordinates and
hence passes NULL for the array for the third one. The actual 2D
sampler, lower in the pipe, knew not to used that array since it
didn't need it. But the samplers have become single-texel and the
coordinate array dereference has been moved up one step, to a level
where the code does not know only two coordinates are used. Hence the
segfault.
The simplest fix by far is to add a third dummy coordinate array in
the call to the next pipe step, which will be dereferenced to an
harmless 0 which then will be happily ignored by the sampler.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=52250
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
We also reuse EGL_TEXTURE_RGBA and EGL_TEXTURE_RGB, adding only the new
planar YUV texture formats: EGL_TEXTURE_Y_U_V_WL, EGL_TEXTURE_Y_UV_WL and
EGL_TEXTURE_Y_XUXV_WL.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
The i965 back-end needs to compile dFdy() differently for FBOs and
window system framebuffers, because Y coordinates are flipped between
the two (see commit 82d2596: i965: Compute dFdy() correctly for FBOs).
This patch avoids unnecessarily recompiling shaders that don't use
dFdy(), by only setting render_to_fbo in the wm program key if the
shader actually uses dFdy().
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This patch updates the ir_set_program_inouts_visitor so that it also
sets gl_fragment_program::UsesDFdy.
This is a bit of a hack (since dFdy() isn't an input or an output),
but there's no other obvious visitor to squeeze this functionality
into, and it would be silly to create a brand new visitor just for
this purpose.
v2: use local 'fprog' var to avoid repeated casting.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The i965 back-end needs to compile dFdy() differently for FBOs and
window system framebuffers, because Y coordinates are flipped between
the two (see commit 82d2596: i965: Compute dFdy() correctly for FBOs).
This boolean will allow it to avoid unnecessarily recompiling shaders
that don't use dFdy().
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Unigine Heaven (at least) has a bug where it incorrectly uses the
GL_ARB_blend_func_extended extension.
Dual source blending allows two color outputs per render target;
individual shader outputs can be assigned to be either the first or
second blending input by setting the 'index' via one of two methods:
- An API call: glBindFragDataLocationIndexed()
- The GLSL 'layout' qualifier provided by GL_ARB_explicit_attrib_location
Both of these only work on user defined fragment shader outputs; it's an
error to use either on built-in outputs like gl_FragData.
Unigine uses gl_FragData and gl_FragColor exclusively, and doesn't even
attempt to use either method to set index == 1. However, it does set
the blending function to SRC1 enums, which requires a fragment shader
output with index == 1 or else rendering is undefined.
In other words, enabling ARB_blend_func_extended causes Unigine to
render incorrectly, resulting in an apparent regression, even though our
driver code (as far as I can tell) is perfectly fine.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50291
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Previously, if we were spilling the result of a texture call, we would store
all 4 regs, then for each use of one of those regs as the source of an
instruction, we would unspill all 4 regs even though only one was needed.
In both lightsmark and l4d2 with my current graphics config, the shaders that
produce spilling do so on split GRFs, so this doesn't help them out. However,
in a capture of the l4d2 shaders with a different snapshot and playing the
game instead of using a demo, it reduced one shader from 2817 instructions to
2179, due to choosing a now-cheaper texture result to spill instead of piles
of texcoords.
v2: Fix comment noted by Ken, and fix the if condition associated with it for
the current state of what constitutes a partial write of the destination.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> (v1)
There's one instance of a potential behavior change: propagate_constants may
now propagate into a part of a vgrf after a different part of it was
overwritten by a send that returns multiple registers. I don't think we ever
generate IR that meets that condition, but it's something to note if we bisect
behavior change to this.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
In these places, we care about any sort of send that hits more than one reg,
not just textures. We don't yet have anything else returning more than one
reg, so there's no change.
v2: Use mlen instead of is_tex() for the is-it-a-send check.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
"count" is a more useful name, since most of the time we're using it for
looping over the variables.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
OpenGL specification 3.3 (page 196), section 4.1.3 says:
If drawbuffer zero is not NONE and the buffer it references has an
integer format, the SAMPLE_ALPHA_TO_COVERAGE and SAMPLE_ALPHA_TO_ONE
operations are skipped."
This should work properly even if there are other draw buffers that
are not in integer format.
This patch makes following piglit tests pass on mesa:
int-draw-buffers-alpha-to-coverage
int-draw-buffers-alpha-to-one
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
This patch churns a lot because it needs to change 4-wide filters into
single pixel filters, since each fragment may use a different filter.
The only case not entirely supported is the anisotropic filtering.
Not sure what we want to do there, since a full quad is required by
that filter.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
From the GL 3.0 spec, section 4.3.3, in the documentation for
CopyPixels():
"An INVALID_OPERATION error will be generated if the object bound
to READ_FRAMEBUFFER_BINDING is framebuffer complete and the value
of SAMPLE_BUFFERS is greater than zero."
The same applies to CopyTexImage...() and CopyTexSubImage...()
functions, since they are defined in terms of CopyPixels().
Previously we were generating an INVALID_FRAMEBUFFER_OPERATION error
in these cases.
Fixes piglit tests
"EXT_framebuffer_multisample/negative-{copypixels,copyteximage}".
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Issues fixed:
- set_vs_sampler_views for evergreen is now properly implemented.
- Added the missing inval_texture_cache call for evergreen.
- have_depth_texture was sometimes incorrectly set to false on evergreen even
if there were depth textures in other shader stages. To fix this, set it
to true once and never set it to false again. It's stupid, but it matches
the r600 code. The proper fix is left to another patch.
- Optimizaton: The sampler views which aren't changed aren't updated.
This is a leftover from:
commit fe1fd67556
Author: Marek Olšák <maraeo@gmail.com>
Date: Sun Jul 8 03:10:37 2012 +0200
r600g: don't flush depth textures set as colorbuffers
If only some buffers are changed, the other ones don't have to re-emitted.
This uses bitmasks of enabled and dirty buffers just like
emit_constant_buffers does.
* Also add mcjit in the non-OpenCL case.
* Replace hardcoded llvm-config with $LLVM_CONFIG everywhere.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Tom Stellad <thomas.stellard@amd.com>
Helps spotting and removing the obsolete generated files, which otherwise break
the build.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This is neccessary for linking the llvmpipe tests. It appears this
dependency was introduced by the "wider native register" changes.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
It's been broken (using NULL getBuffersWithFormat() instead of
getBuffers()) due to a copy and paste error for a year now.
GetBuffersWithFormat has been around since 2009, so I don't feel any
guilt in not supporting it.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This means that GLX buffer sharing of these no longer works. On the
other hand, just *look* at this code reduction.
v2:
- [chad] Fix intelCreateBuffer for gen < 6. When the branch for
!screen->hw_has_separate_stencil was taken,
intel_create_private_renderbuffer was incorrectly not used.
- [chad] Remove all code in intel_process_dri2_buffer for processing
depth, stencil, and hiz buffers. That code is now dead.
CC: Eric Anholt <eric@anholt.net>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
commit '7250cd506baa0bd4649b30d87509cdd0cbc06a57'
changes struct gbm_bo, renaming it's 'pitch' to 'stride'.
This applies to Gallium.
Signed-off-by: Elvis Lee <kwangwoong.lee@lge.com>
Previously, if you ran make followed by make check it would work, but
if you just ran make check the test program would fail to compile.
Reviewed-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Squashed commit of the following:
commit 7acb7b4f60dc505af3dd00dcff744f80315d5b0e
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Jul 9 17:46:31 2012 +0100
draw: Don't use dynamically sized arrays.
Not supported by MSVC.
commit 5810c28c83647612cb372d1e763fd9d7780df3cb
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Jul 9 17:44:16 2012 +0100
gallivm,llvmpipe: Don't use expressions with PIPE_ALIGN_VAR().
MSVC doesn't accept exceptions in _declspec(align(...)). Use a
define instead.
commit 8aafd1457ba572a02b289b3f3411e99a3c056072
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Jul 9 17:41:56 2012 +0100
gallium/util: Make u_cpu_detect.h header C++ safe.
commit 5795248350771f899cfbfc1a3a58f1835eb2671d
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Jul 2 12:08:01 2012 +0100
gallium/util: Add ULL suffix to large constants.
As suggested by Andy Furniss: it looks like some old gcc versions
require it.
commit 4c66c22727eff92226544c7d43c4eb94de359e10
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Jun 29 13:39:07 2012 +0100
gallium/util: Truly disable INF/NAN tests on MSVC.
Thanks to Brian for spotting this.
commit 8bce274c7fad578d7eb656d9a1413f5c0844c94e
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Jun 29 13:39:07 2012 +0100
gallium/util: Disable INF/NAN tests on MSVC.
Somehow they are not recognized as constants.
commit 6868649cff8d7fd2e2579c28d0b74ef6dd4f9716
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jul 5 15:05:24 2012 +0200
gallivm: Cleanup the 2 x 8 float -> 16 ub special path in lp_build_conv.
No behaviour change intended, like 7b98455fb40c2df84cfd3cdb1eb7650f67c8a751.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 5147a0949c4407e8bce9e41d9859314b4a9ccf77
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jul 5 14:28:19 2012 +0200
gallivm: (trivial) fix issues with multiple-of-4 texture fetch
Some formats can't handle non-multiple of 4 fetches I believe, but
everything must support length 1 and multiples of 4.
So avoid going to scalar fetch (which is very costly) just because length
isn't 4.
Also extend the hack to not use shift with variable count for yuv formats to
arbitrary length (larger than 1) - doesn't matter how many elements we
have we always want to avoid it unless we have variable shift count
instruction (which we should get with avx2).
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 87ebcb1bd71fa4c739451ec8ca89a7f29b168c08
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jul 4 02:09:55 2012 +0200
gallivm: (trivial) fix typo for wrap repeat mode in linear filtering aos code
This would lead to bogus coordinates at the edges.
(undetected by piglit because this path is only taken for block-based
formats).
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit 3a42717101b1619874c8932a580c0b9e6896b557
Author: José Fonseca <jfonseca@vmware.com>
Date: Tue Jul 3 19:42:49 2012 +0100
gallivm: Fix TGSI integer translation with AVX.
commit d71ff104085c196b16426081098fb0bde128ce4f
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Jun 29 15:17:41 2012 +0100
llvmpipe: Fix LLVM JIT linear path.
It was not working properly because it was looking at the JIT function
before it was actually compiled.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
commit a94df0386213e1f5f9a6ed470c535f9688ec0a1b
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Jun 28 18:07:10 2012 +0100
gallivm: Refactor lp_build_broadcast(_scalar) to share code.
Doesn't really change the generated assembly, but produces more compact IR,
and of course, makes code more consistent.
Reviewed-by: Brian Paul <brianp@vmware.com>
commit 66712ba2731fc029fa246d4fc477d61ab785edb5
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Jun 27 17:30:13 2012 +0100
gallivm: Make LLVMContextRef a singleton.
There are any places inside LLVM that depend on it. Too many to attempt
to fix.
Reviewed-by: Brian Paul <brianp@vmware.com>
commit ff5fb7897495ac263f0b069370fab701b70dccef
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 28 18:15:27 2012 +0200
gallivm: don't use 8-wide texture fetch in aos path
This appears to be a slight loss usually.
There are probably several reasons for that:
- fetching itself is scalar
- filtering is pure int code hence needs splitting anyway, same
for the final texel offset calculations
- texture wrap related code, which can be done 8-wide, is slightly more
complex with floats (with clamp_to_edge) and float operations generally
more costly hence probably not much faster overall
- the code needed to split when encountering different mip levels for the
quads, adding complexity
So, just split always for aos path (but leave it 8-wide for soa, since we
do 8-wide filtering there when possible).
This should certainly be revisited if we'd have avx2 support.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit ce8032b43dcd8e8d816cbab6428f54b0798f945d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 27 18:41:19 2012 +0200
gallivm: (trivial) don't extract fparts variable if not needed
Did not have any consequences but unnecessary.
commit aaa9aaed8f80dc282492f62aa583a7ee23a4c6d5
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 27 18:09:06 2012 +0200
gallivm: fix precision issue in aos linear int wrap code
now not just passes at a quick glance but also with piglit...
If we do the wrapping with floats, we also need to set the
weights accordingly. We can potentially end up with different
(integer) coordinates than what the integer calculations would
have chosen, which means the integer weights calculated previously
in this case are completely wrong. Well at least that's what I think
happens, at least recalculating the weights helps.
(Some day really should refactor all the wrapping, so we do whatever is
fastest independent of 16bit int aos or 32bit float soa filtering.)
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit fd6f18588ced7ac8e081892f3bab2916623ad7a2
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Jun 27 11:15:53 2012 +0100
gallium/util: Fix parsing of options with underscore.
For example
GALLIVM_DEBUG=no_brilinear
which was being parsed as two options, "no" and "brilinear".
commit 09a8f809088178a03e49e409fa18f1ac89561837
Author: James Benton <jbenton@vmware.com>
Date: Tue Jun 26 15:00:14 2012 +0100
gallivm: Added a generic lp_build_print_value which prints a LLVMValueRef.
Updated lp_build_printf to share common code.
Removed specific lp_build_print_vecX.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit e59bdcc2c075931bfba2a84967a5ecd1dedd6eb0
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed May 16 15:00:23 2012 +0100
draw,llvmpipe: Avoid named struct types on LLVM 3.0 and later.
Starting with LLVM 3.0, named structures are meant not for debugging, but
for recursive data types, previously also known as opaque types.
The recursive nature of these types leads to several memory management
difficulties. Given that we don't actually need recursive types, avoid
them altogether.
This is an attempt to address fdo bugs 41791 and 44466. The issue is
somewhat random so there's no easy way to check how effective this is.
Cherry-picked from 9af1ba565d
commit df6070f618a203c7a876d984c847cde4cbc26bdb
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 27 14:42:53 2012 +0200
gallivm: (trivial) fix typo in faster aos linear int wrap code
no longer crashes, now REALLY tested.
commit d8f98dce452c867214e6782e86dc08562643c862
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 26 18:20:58 2012 +0200
llvmpipe: (trivial) remove bogus optimization for float aos repeat wrap
This optimization for nearest filtering on the linear path generated
likely bogus results, and the int path didn't have any optimizations
there since the only shader using force_nearest apparently uses
clamp_to_edge not repeat wrap anyway.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit c4e271a0631087c795e756a5bb6b046043b5099d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 26 23:01:52 2012 +0200
gallivm: faster repeat wrap for linear aos path too
Even if we already have scaled integer coords, it's way faster to use
the original float coord (plus some conversions) rather than use URem.
The choice of what to do for texture wrapping is not really tied to int
aos or float soa filtering though for some modes there can be some gains
(because of easier weight calculations).
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 1174a75b1806e92aee4264ffe0ffe7e70abbbfa3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 26 14:39:22 2012 +0200
gallivm: improve npot tex wrap repeat in linear soa path
URem gets translated into series of scalar divisions so
just about anything else is faster.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit f849ffaa499ed96fa0efd3594fce255c7f22891b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 26 00:40:35 2012 +0100
gallivm: (trivial) fix near-invisible shift-space typo
I blame the keyboard.
commit 5298a0b19fe672aebeb70964c0797d5921b51cf0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 25 16:24:28 2012 +0200
gallivm: add new intrinsic helper to deal with arbitrary vector length
This helper will split vectors which are too large for the hw, or expand
them if they are too small, so a caller of a function using intrinsics which
uses such sizes need not split (or expand) the vectors manually and the
function will still use the intrinsic instead of dropping back to generic
llvm code. It can also accept scalars for use with pseudo-vector intrinsics
(only useful for float arguments, all x86 scalar simd float intrinsics use
4vf32).
Only used for lp_build_min/max() for now (also added the scalar float case
for these while there). (Other basic binary functions could use it easily,
whereas functions with a different interface would need different helpers.)
Expanding vectors isn't widely used, because we always try to use
build contexts with native hw vector sizes. But it might (or not) be nicer
if this wouldn't need to be done, the generated code should in theory stay
the same (it does get hit by lp_build_rho though already since we
didn't have a intrinsic for the scalar lp_build_max case before).
v2: incorporated Brian's feedback, and also made the scalar min/max case work
instead of crash (all scalar simd float intrinsics take 4vf32 as argument,
probably the reason why it wasn't used before).
Moved to lp_bld_intr based on José's request, and passing intrinsic size
instead of length.
Ideally we'd derive the source type info from the passed in llvm value refs
and process some llvmtype return type so we could handle intrinsics where
the source and destination type isn't the same (like float/int conversions,
packing instructions) but that's a bit too complicated for now.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 01aa760b99ec0b2dc8ce57a43650e83f8c1becdf
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 25 16:19:18 2012 +0200
gallivm: (trivial) increase max code size for shader disassembly
64kB was just short of what I needed (which caused a crash) hence
increase to 96kB (should probably be smarter about that).
commit 74aa739138d981311ce13076388382b5e89c6562
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 25 11:53:29 2012 +0100
gallivm: simplify aos float tex wrap repeat nearest
just handle pot and npot the same. The previous pot handling
ended up with exactly the same instructions plus 2 more (leave it
in the soa path though since it is probably still cheaper there).
While here also fix a issue which would cause a crash after an assert.
commit 0e1e755645e9e49cfaa2025191e3245ccd723564
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 25 11:29:24 2012 +0100
gallivm: (trivial) skip floor rounding in ifloor when not signed
This was only done for the non-sse41 case before, but even with
sse41 this is obviously unnecessary (some callers already call
itrunc in this case anyway but some might not).
commit 7f01a62f27dcb1d52597b24825931e88bae76f33
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 25 11:23:12 2012 +0100
gallivm: (trivial) fix bogus comments
commit 5c85be25fd82e28490274c468ce7f3e6e8c1d416
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Jun 20 11:51:57 2012 +0100
translate: Free elt8_func/elt16_func too.
These were leaking.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
commit 0ad498f36fb6f7458c7cffa73b6598adceee0a6c
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 19 15:55:34 2012 +0200
gallivm: fix bug for tex wrap repeat with linear sampling in aos float path
The comparison needs to be against length not length_minus_one, otherwise
the max texel is never chosen (for the second coordinate).
Fixes piglit texwrap-1D-npot-proj (and 2D/3D versions).
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit d1ad65937c5b76407dc2499b7b774ab59341209e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Jun 19 16:13:43 2012 +0200
gallivm: simplify soa tex wrap repeat with npot textures and no mip filtering
Similar to what is already done in aos sampling for the float path (but not
the int path since we don't get normalized float coordinates there).
URem is expensive and the calculation is done trivially with
normalized floats instead (at least with sse41-capable cpus).
(Some day should probably do the same for the mip filter path but it's much
more complicated there hence the gain is smaller.)
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit e1e23f57ba9b910295c306d148f15643acc3fc83
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 18 20:38:56 2012 +0200
llvmpipe: (trivial) remove duplicated function declaration
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 07ca57eb09e04c48a157733255427ef5de620861
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 18 20:37:34 2012 +0200
llvmpipe: destroy setup variants on context destruction
lp_delete_setup_variants() used to be called in garbage collection,
but this no longer exists hence the setup shaders never got freed.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit ed0003c633859a45f9963a479f4c15ae0ef1dca3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 18 16:25:29 2012 +0100
gallivm: handle different ilod parts for multiple quad sampling
This fixes filtering when the integer part of the lod is not the same
for all quads. I'm not fully convinced of that solution yet as it just
splits the vector if the levels to be sampled from are different.
But otherwise we'd need to do things like some minify steps, and getting
mip level base address separately anyway hence it wouldn't really look
like much of a win (and making the code even more complex).
This should now give identical results to single quad sampling.
commit 8580ac4cfc43a64df55e84ac71ce1a774d33c0d2
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 14 18:14:47 2012 +0200
gallivm: de-duplicate sample code common to soa and aos sampling
There doesn't seem to be any reason why this code dealing with cube face
selection, lod and mip level calculation is separate in aos and
soa sampling, and I am sick of having it to change in both places.
commit fb541e5f957408ce305b272100196f1e12e5b1e8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Jun 14 18:15:41 2012 +0200
gallivm: do mip filtering with per quad lod_fpart
This gives better results for mip filtering, though the generated code might
not be optimal. For now it also creates some artifacts if the lod_ipart isn't
the same for all quads, since instead of using the same mip weight for all
quads as previously (which just caused non-smooth gradients) this now will
use the right weights but with the wrong mip level in this case (can easily
be seen with things like texfilt, mipmap_tunnel).
v2: use logic helper suggested by José, and fix issue with negative lod_fpart
values
commit f1cc84eef7d826a20fab6cd8ccef9a275ff78967
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Jun 13 18:35:25 2012 +0200
gallivm: (trivial) fix bogus assert in lp_build_unpack_broadcast_aos_scalars
commit 7c17dbae8ae290df9ce0f50781a09e8ed640c044
Author: James Benton <jbenton@vmware.com>
Date: Tue Jun 12 12:11:14 2012 +0100
util: Reimplement half <-> float conversions.
Removed u_half.py used to generate the table for previous method.
Previous implementation of float to half conversion was faulty for
denormalised and NaNs and would require extra logic to fix,
thus making the speedup of using tables irrelevant.
commit 7762f59274070e1dd4b546f5cb431c2eb71ae5c3
Author: James Benton <jbenton@vmware.com>
Date: Tue Jun 12 12:12:16 2012 +0100
tests: Updated tests to properly handle NaN for half floats.
commit fa94c135aea5911fd93d5dfb6e6f157fb40dce5e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 11 18:33:10 2012 +0200
gallivm: do mip level calculations per quad
This is the final piece which shouldn't change the rendering output yet.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 23cbeaddfe03c09ca18c45d28955515317ffcf4c
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Jun 9 00:54:21 2012 +0200
gallivm: do per-quad cube face selection
Doesn't quite fix the piglit cubemap test (not sure why actually)
but doing per-quad face selection is doing the right thing and
definitely an improvement.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit abfb372b3702ac97ac8b5aa80ad1b94a2cc39d33
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Jun 11 18:22:59 2012 +0200
gallivm: do all lod calculations per quad
Still no functional change but lod is now converted to scalar after
lod calculations.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 519368632747ae03feb5bca9c655eccbc5b751b4
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 16:46:10 2012 +0100
gallivm: Added support for half-float to float conversion in lp_build_conv.
Updated various utility functions to support this change.
commit 135b4d683a4c95f7577ba27b9bffa4a6fbd2c2e7
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 16:02:46 2012 +0100
gallivm: Added function for half-float to float conversion.
Updated lp_build_format_aos_array to support half-float source.
commit 37d648827406a20c5007abeb177698723ed86673
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 14:55:18 2012 +0100
util: Updated u_format_tests to rigidly test half-float boundary values.
commit 2ad18165d96e578aa9046df7c93cb1c3284d8c6b
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 14:54:16 2012 +0100
llvmpipe: Updated lp_test_format to properly handle Inf/NaN results.
commit 78740acf25aeba8a7d146493dd5c966e22c27b73
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 14:53:30 2012 +0100
util: Added functions for checking NaN / Inf for double and half-floats.
commit 35e9f640ae01241f9e0d67fe893bbbf564c05809
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu May 24 21:05:13 2012 +0200
gallivm: Fix calculating rho for 3d textures for the single-quad case
Discovered by accident, this looks like a very old typo bug.
commit fc1220c636326536fd0541913154e62afa7cd1d8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu May 24 21:04:59 2012 +0200
gallivm: do calcs per-quad in lp_build_rho
Still convert to scalar at the end of the function.
commit 50a887ffc550bf310a6988fa2cea5c24d38c1a41
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon May 21 23:21:50 2012 +0200
gallivm: (trivial) return scalar in lp_build_extract_range for length 1 vectors
Our type system on top of llvm's one doesn't generally support vectors of
length 1, instead using scalars. So we should return a scalar from this
function instead of having to bitcast the vector with length 1 later elsewhere.
commit 80c71c621f9391f0f9230460198d861643324876
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 17:49:15 2012 +0100
draw: Fixed bad merge error
commit c47401cfad0c9167de20ff560654f533579f452c
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 15:29:30 2012 +0100
draw: Updated store_clip to store whole vectors instead of individual elements.
commit 2d9c1ad74b0b0b41861fffcecde39f09cc27f1cf
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 15:28:32 2012 +0100
gallivm: Added lp_build_fetch_rgba_aos_array.
A version of lp_build_fetch_rgba_aos which is targeted at simple array formats.
Reads the whole vector from memory in one, instead of reading each element
individually.
Tested with mesa tests and demos.
commit ff7805dc2b6ef6d8b11ec4e54aab1633aef29ac8
Author: James Benton <jbenton@vmware.com>
Date: Tue May 22 15:27:40 2012 +0100
gallivm: Added lp_build_pad_vector.
This function pads a vector with undef to a desired length.
commit 701f50acef24a2791dabf4730e5b5687d6eb875d
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 17:27:19 2012 +0100
util: Added util_format_is_array.
This function checks whether a format description is in a simple array format.
commit 5e0a7fa543dcd009de26f34a7926674190fa6246
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 19:13:47 2012 +0100
draw: Removed draw_llvm_translate_from and draw/draw_llvm_translate.c.
This is "replaced" by adding an optimised path in lp_build_fetch_rgba_aos
in an upcoming patch.
commit 8c886d6a7dd3fb464ecf031de6f747cb33e5361d
Author: James Benton <jbenton@vmware.com>
Date: Wed May 16 15:02:31 2012 +0100
draw: Modified store_aos to write the vector as one, not individual elements.
commit 37337f3d657e21dfd662c7b26d61cb0f8cfa6f17
Author: James Benton <jbenton@vmware.com>
Date: Wed May 16 14:16:23 2012 +0100
draw: Changed aos_to_soa to use lp_build_transpose_aos.
commit bd2b69ce5d5c94b067944d1dcd5df9f8e84548f1
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 19:14:27 2012 +0100
draw: Changed soa_to_aos to use lp_build_transpose_aos.
commit 0b98a950d29a116e82ce31dfe7b82cdadb632f2b
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 18:57:45 2012 +0100
gallivm: Added lp_build_transpose_aos which converts between aos and soa.
commit 69ea84531ad46fd145eb619ed1cedbe97dde7cb5
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 18:57:01 2012 +0100
gallivm: Added lp_build_interleave2_half aimed at AVX unpack instructions.
commit 7a4cb1349dd35c18144ad5934525cfb9436792f9
Author: José Fonseca <jfonseca@vmware.com>
Date: Tue May 22 11:54:14 2012 +0100
gallivm: Fix build on Windows.
MC-JIT not yet supported there.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
commit afd105fc16bb75d874e418046b80d9cc578818a1
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:17:26 2012 +0100
llvmpipe: Added a error counter to lp_test_conv.
Useful for keeping track of progress when fixing errors!
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit b644907d08c10a805657841330fc23db3963d59c
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:16:46 2012 +0100
llvmpipe: Changed known failures in lp_test_conv.
To comply with the recent fixes to lp_bld_conv.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit d7061507bd94f6468581e218e61261b79c760d4f
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:14:38 2012 +0100
llvmpipe: Added fixed point types tests to lp_test_conv.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit 146b3ea39b4726dbe125ac666bd8902ea3d6ca8c
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:26:35 2012 +0100
llvmpipe: Changed lp_test_conv src/dst alignment to be correct.
Now based on the define rather than a fixed number.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit f3b57441f834833a4b142a951eb98df0aa874536
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:06:44 2012 +0100
gallivm: Fixed erroneous optimisation in lp_build_min/max.
Previously assumed normalised was 0 to 1, but it can be -1 to 1
if type is signed.
Tested with lp_test_conv and lp_test_format, reduced errors.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit a0613382e5a215cd146bb277646a6b394d376ae4
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:04:49 2012 +0100
gallivm: Compensate for lp_const_offset in lp_build_conv.
Fixing a /*FIXME*/ to remove errors in integer conversion in lp_build_conv.
Tested using lp_test_conv and lp_test_format, reduced errors.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit a3d2bf15ea345bc8a0664f8f441276fd566566f3
Author: James Benton <jbenton@vmware.com>
Date: Fri May 18 16:01:25 2012 +0100
gallivm: Fixed overflow in lp_build_clamped_float_to_unsigned_norm.
Tested with lp_test_conv and lp_test_format, reduced errors.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit e7b1e76fe237613731fa6003b5e1601a2e506207
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon May 21 20:07:51 2012 +0100
gallivm: Fix build with LLVM 2.6
Trivial, and useful.
commit d3c6bbe5c7f5ba1976710831281ab1b6a631082d
Author: José Fonseca <jfonseca@vmware.com>
Date: Tue May 15 17:15:59 2012 +0100
gallivm: Enable MCJIT/AVX with vanilla LLVM 3.1.
Add the necessary C++ glue, so that we don't need any modifications
to the soon to be released LLVM 3.1.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
commit 724a019a14d40fdbed21759a204a2bec8a315636
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon May 14 22:04:06 2012 +0100
gallivm: Use HAVE_LLVM 0x0301 consistently.
commit af6991e2a3868e40ad599b46278551b794839748
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon May 14 21:49:06 2012 +0100
gallivm: Add MCRegisterInfo.h to silence benign warnings about missing implementation.
Trivial.
commit 6f8a1d75458daae2503a86c6b030ecc4bb494e23
Author: Vinson Lee <vlee@freedesktop.org>
Date: Mon Apr 2 22:14:15 2012 -0700
gallivm: Pass in a MCInstrInfo to createMCInstPrinter on llvm-3.1.
llvm-3.1svn r153860 makes MCInstrInfo available to the MCInstPrinter.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit 62555b6ed8760545794f83064e27cddcb3ce5284
Author: Vinson Lee <vlee@freedesktop.org>
Date: Tue Mar 27 21:51:17 2012 -0700
gallivm: Fix method overriding in raw_debug_ostream.
Use matching type qualifers to avoid method hiding.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit 6a9bd784f4ac68ad0a731dcd39e5a3c39989f2be
Author: Vinson Lee <vlee@freedesktop.org>
Date: Tue Mar 13 22:40:52 2012 -0700
gallivm: Fix createOProfileJITEventListener namespace with llvm-3.1.
llvm-3.1svn r152620 refactored the OProfile profiling code.
createOProfileJITEventListener was moved from the llvm namespace to the
llvm::JITEventListener namespace.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit b674955d39adae272a779be85aa1bd665de24e3e
Author: Vinson Lee <vlee@freedesktop.org>
Date: Mon Mar 5 22:00:40 2012 -0800
gallivm: Pass in a MCRegisterInfo to MCInstPrinter on llvm-3.1.
llvm-3.1svn r152043 changes createMCInstPrinter to take an additional
MCRegisterInfo argument.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit 11ab69971a8a31c62f6de74905dbf8c02884599f
Author: Vinson Lee <vlee@freedesktop.org>
Date: Wed Feb 29 21:20:53 2012 -0800
Revert "gallivm: Change getExtent and readByte to non-const with llvm-3.1."
This reverts commit d5a6c17254.
llvm-3.1svn r151687 makes MemoryObject accessor members const again.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit 339960c82d2a9f5c928ee9035ed31dadb7f45537
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon May 14 16:19:56 2012 +0200
gallivm: (trivial) fix assertion failure for mipmapped 1d textures
In lp_build_rho, we may end up with a 1-element vector (for mipmapped 1d
textures), but in this case we require the type to be a non-vector type,
so need a cast.
commit 9d73edb727bd6d196030dc3026b7bf0c574b3e19
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu May 10 18:12:07 2012 +0200
gallivm: prepare for per-quad lod calculations for large vectors
to be able to handle multiple quads at once in texture sampling and still
do lod calculations per quad, it is necessary to get the per-quad derivatives
into the lp_build_rho function.
Until now these derivative values were just scalars, which isn't going to work.
So we now use vectors, and since the interface needs to change we also do some
different (slightly more efficient) packing of the values.
For 8-wide vectors the packed derivative values for 3 coords would look like
this, this scales to a arbitrary (multiple of 4) vector size:
ds1dx ds1dy dt1dx dt1dy ds2dx ds2dy dt2dx dt2dy
dr1dx dr1dy _____ _____ dr2dx dr2dy _____ _____
The second vector will be unused for 1d and 2d textures.
To facilitate future changes the derivative values are put into a struct, since
quite some functions just pass these values through.
The generated code seems to be very slightly better for 2d textures (with
4-wide vectors) than before with sse2 (if you have a cpu with physical 128bit
simd units - otherwise it's probably not a win).
v2: suggestions from José, rename variables, add comments, use swizzle helper
commit 0aa21de0d31466dac77b05c97005722e902517b8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu May 10 18:10:31 2012 +0200
gallivm: add undefined swizzle handling to lp_build_swizzle_aos
This is useful for vectors with "holes", it lets llvm choose the most
efficient shuffle instructions if some elements aren't needed without having to
worry what elements to manually pick otherwise.
commit 00faf3f370e7ce92f5ef51002b0ea42ef856e181
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri May 4 17:25:16 2012 +0100
gallivm: Get the LLVM IR optimization passes before JIT compilation.
MC-JIT engine compiles the module immediately on creation, so the optimization
passes were being run too late.
So now we create a target data layout from a string, that matches the
ABI parameters reported by the compiler.
The backend optimization passes were always been run, so the performance
improvement is modest (3% on multiarb mesa demo).
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
commit 40a43f4e2ce3074b5ce9027179d657ebba68800a
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed May 2 16:03:54 2012 +0200
gallivm: (trivial) fix wrong define used in lp_build_pack2
should fix stack-smashing crashes.
commit e6371d0f4dffad4eb3b7a9d906c23f1c88a2ab9e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Apr 30 21:25:29 2012 +0200
gallivm: add perf warnings when not using intrinsics with 256bit vectors
Helper functions using integer sse2 intrinsics could split the vectors with AVX
instead of using generic fallback (which should be faster).
We don't actually expect to hit these paths (hence don't fix them up to actually
do the vector splitting) so just emit warnings (for those functions where it's
obvious doing split/intrinsic is faster than using generic path).
Only emit warnings for 256bit vectors since we _really_ don't expect to hit
arbitrary large vectors which would affect a lot more functions.
The warnings do not actually depend on avx since the same logic applies to
plain sse2 too (but of course again there's _really_ no reason we should hit
these functions with 256bit vectors without avx).
commit 8a9ea701ea7295181e846c6383bf66a5f5e47637
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue May 1 20:37:07 2012 +0200
gallivm: split vectors manually for avx in lp_build_pack2 (v2)
There's 2 reasons for this:
First, there's a llvm bug (fixed in 3.1) which generates tons of byte
inserts/extracts otherwise, and second, more importantly, we want to use
pack intrinsics instead of shuffles.
We do this in lp_build_pack2 and not the calling code (aos sample path)
because potentially other callers might find that useful too, even if
for larger sequences of code using non-native vector sizes it might be
better to manually split vectors.
This should boost texture performance in the aos path considerably.
v2: fix issues with intrinsics types with old llvm
commit 27ac5b48fa1f2ea3efeb5248e2ce32264aba466e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue May 1 20:26:22 2012 +0200
llvmpipe: refactor lp_build_pack2 (v2)
prettify, and it's unnecessary to assert when there's no intrinsic due to
unsupported bit width - the shuffle path will work regardless.
In contrast lp_build_packs2, should only rely on lp_build_pack2 doing the
clamping for element sizes for which there is a sse2 intrinsic.
v2: fix bug spotted by Jose regarding the intrinsic type for packusdw
on old llvm versions.
commit ddf279031f0111de4b18eaf783bdc0a1e47813c8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue May 1 20:13:59 2012 +0200
gallivm: add src width check in lp_build_packs2()
not doing so would skip clamping even if no sse2 pack instruction is
available, which is incorrect (in theory only, such widths would also always
hit a (unnecessary) assertion in lp_build_pack2().
commit e7f0ad7fe079975eae7712a6e0c54be4fae0114b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Apr 27 15:57:00 2012 +0200
gallivm: (trivial) fix crash-causing typo for npot textures with avx
commit 28a9d7f6f655b6ec508c8a3aa6ffefc1e79793a0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Apr 25 19:38:45 2012 +0200
gallivm: (trivial) remove code mistakenly added twice.
commit d5926537316f8ff67ad0a52e7242f7c5478d919b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Apr 24 21:16:15 2012 +0200
gallivm: add a new avx aos sample path (v2)
Try to avoid mixing float and int address calculations. This does texture wrap
modes with floats, and then the offset calculations still with ints (because
of lack of precision with floats, though we could do some effort to make it work
with not too large (16MB) textures).
This also handles wrap repeat mode with npot-sized textures differently than
either the old soa or aos int path (likely way faster but untested).
Otherwise the actual address wrap code is largely similar to the soa path (not
quite the same as this one also has some int code), it should get used by avx
soa sampling later as well but doesn't handle more complex address modes yet
(this will also have the benefit that we can use aos sampling path for all
texture address modes).
Generated code for that looks reasonable, but still does not split vectors
explicitly for fetch/filter which means still get hit by llvm (fixed upstream)
which generates hundreds of pinsrb/pextrb instead of two shuffles.
It is not obvious though if it's much of a win over just doing address calcs
4-wide but with ints, even if it is definitely much less instructions on avx.
piglit's texwrap seems to look exactly the same but doesn't test
neither the non-normalized nor the npot cases.
v2: fix comments, prettify based on Brian's and Jose's feedback.
commit bffecd22dea66fb416ecff8cffd10dd4bdb73fce
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Apr 19 01:58:29 2012 +0200
gallivm: refactor aos lp_build_sample_image_nearest/linear
split them up to separate address calculations and fetching/filtering.
Need this for being able to do 8-wide float address calcs and 4-wide
fetch/filter later (for avx). Plus the functions were very big scary monsters
anyway (in particular lp_build_sample_image_linear).
commit a80b325c57529adddcfa367f96f03557725c4773
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Apr 16 17:17:18 2012 +0200
gallivm: fix lp_build_resize when truncating width but expanding vector size
Missed this case which I thought was impossible - the assertion for it was
right after the division by zero...
(AoS) texture sampling may ask us to do this, for things like 8 4x32int
vectors to 1 32x8int vector conversion (eventually, we probably don't want
this to happen).
commit f9c8337caa3eb185830d18bce8b95676a065b1d7
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sat Apr 14 18:00:59 2012 +0200
gallivm: fix cube maps with larger vectors
This makes the branchless cube face selection code work with larger vectors.
Because the complexity is quite high (cannot really be improved it seems,
per-face selection would reduce complexity a lot but this leads to errors
unless the derivatives are calculated all from the same face which almost
doubles the work to be done) it is still slower than the branching version,
hence only enable this with large vectors.
It doesn't actually do per-quad face selection yet (only makes sense with
matching lod selection, in fact it will select the same face for all pixels
based on the average of the first four pixels for now) but only different
shuffles are required to make it work (the branching version actually should
work with larger vectors too now thanks to the improved horizontal add but of
course it cannot be extended to really select the face per-quad unless doing
branching per quad).
commit 7780c58869fc9a00af4f23209902db7e058e8a66
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 30 21:11:12 2012 +0100
llvmpipe: (trivial) fix compiler warning
and also clarify comment regarding availability of popcnt instruction.
commit a266dccf477df6d29a611154e988e8895892277e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 30 14:21:07 2012 +0100
gallivm: remove unneeded members in lp_build_sample_context
Minor cleanup, the texture width, height, depth aren't accessed in their
scalar form anywhere. Makes it more obvious those values should probably be
fetched already vectorized (but this requires more invasive changes)...
commit b678c57fb474e14f05e25658c829fc04d2792fff
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Mar 29 15:53:55 2012 +0100
gallivm: add a helper for concatenating vectors
Similar to the extract_range helper intended to get around slow code generated
by llvm for 128bit insertelements.
Concatenating two 128bit vectors this way will result in a single vinsertf128
operation rather than two 64bit stores plus one 128bit load, though it might be
mildly useful for other purposes as well.
commit 415ff228bcd0cf5e44a4c15350a661f0f5520029
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 28 19:41:15 2012 +0100
gallivm: add a custom 2x8f->1x16ub avx conversion path
Similar to the existing 4x4f->1x16ub sse2 path, shaves off a couple
instructions (min/max mostly) because it relies on pack intrinsics clamping.
commit 78c08fc89f8fbcc6dba09779981b1e873e2a0299
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 28 18:44:07 2012 +0100
gallivm: add avx arithmetic intrinsics
Add all avx intrinsics for arithmetic functions (with the exception
of the horizontal add function which needs another look).
Seems to pass basic tests.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
commit a586caa2800aa5ce54c173f7c0d4fc48153dbc4e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 28 15:31:35 2012 +0100
gallivm: add avx logic intrinsics
Add the blend intrinsics for 8-wide float and 4-wide double vectors.
Since we lack 256bit int instructions these are used for int vectors as well,
though obviously not for byte or word element values.
The comparison intrinsics aren't extended for avx since these are only used
for pre-2.7 llvm versions.
commit 70275e4c13c89315fc2560a4c488c0e6935d5caf
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 28 00:40:53 2012 +0100
gallivm: new helper function for extract shuffles.
Based on José's idea as we can need that in a couple places.
Note that such shuffles should not be used lightly, since data layout
of <4 x i8> is different to <16 x i8> for instance, hence might cause
data rearrangement.
commit 4d586dbae1b0c55915dda1759d2faea631c0a1c2
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 27 18:27:25 2012 +0100
gallivm: (trivial) don't overallocate shuffle variable
using wrong define meant huge array...
commit 06b0ec1f6d665d98c135f9573ddf4ba04b2121ad
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 27 17:54:20 2012 +0100
gallivm: don't do per-element extract/insert for vector element resize
Instead of doing per-element extract/insert if the src vectors
and dst vector differ in total size (which generates atrocious code)
first change the src vectors size by using shuffles to destination
vector size.
We can still do better than that on AVX for packing to color buffer
(by exploiting pack intrinsics characteristics hence eleminating the
need for some clamps) but this already generates much better code.
v2: incorporate feedback from José, Keith and use shuffle instead of
bitcasts/extracts. Due to llvm deficiencies the latter cause all data
to get moved to GPRs and back in pieces (even though the data in the
regs actually stays the same...).
commit c9970d70e05f95d3f52fe7d2cd794176a52693aa
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 23 19:33:19 2012 +0000
gallivm: fix bug in simple position interpolation
Accidental use of position attribute instead of just pixel coordinates.
Caused failures in piglit glsl-fs-ceil and glsl-fs-floor.
commit d0b6fcdb008d04d7f73d3d725615321544da5a7e
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 23 15:31:14 2012 +0000
gallivm: fix emission of ceil opcode
lp_build_ceil seems more appropriate than lp_build_trunc.
This seems to be never hit though someone performs some ceil
to floor magic.
commit d97fafed7e62ffa6bf76560a92ea246a1a26d256
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Mar 22 11:46:52 2012 +0000
gallivm: new vectorized path for cubemap calculations
should be faster when adapted to multiple quads as only selection masks need to be different.
The code is more or less a per-pixel version adapted to only do it per quad.
A per pixel version would be much simpler (could drop 2 selects, 6 broadcasts and the messy
horizontal add of 3 vectors at the expense of only 2 more absolute value instructions -
would also just work for arbitary large vectors).
This version doesn't yet work with larger vectors because the horizontal add isn't adjusted
to be able to work with 2x4 vectors (and also because face selection wouldn't be done per
quad just per block though that would be only a correctness issue just as with lod selection).
The downside is this code is quite a bit slower. On a Core2 it can be sped up by disabling the
hw blend instructions for selection and using logicop fallbacks instead, but it is still slower
than the old code, hence leave that in for now. Probably will chose one or the other version
based on vector length in the end.
commit b375fbb18a3fd46859b7fdd42f3e9908ea4ff9a3
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 21 14:42:29 2012 +0000
gallivm: fix optimized occlusion query intrinsic name
commit a9ba0a3b611e48efbb0e79eb09caa85033dbe9a2
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Mar 21 16:19:43 2012 +0000
draw,gallivm,llvmpipe: Call gallivm_verify_function everywhere.
commit f94c2238d2bc7383e088b8845b7410439a602071
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 20 18:54:10 2012 +0000
gallivm: optimize calculations for cube maps a bit
this does some more vectorized calculations and uses horizontal adds if possible.
A definite win with sse3 otherwise it doesn't seem to make much of a difference.
In any case this is arithmetically identical, cannot handle larger vectors.
Should be useful as a reference point against larger vector version later...
commit 21a2c1cf3c8e1ac648ff49e59fdc0e3be77e2ebb
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 20 15:16:27 2012 +0000
llvmpipe: slight optimization of occlusion queries
using movmskps when available.
While this is slightly better for cpus without popcnt we should
really sum the vectors ourselves (it is also possible to cast to i4 before
doing the popcnt but that doesn't help that much neither since llvm
is using some optimized popcnt version for i32)
commit 5ab5a35f216619bcdf55eed52b0db275c4a06c1b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 20 13:32:11 2012 +0000
llvmpipe: fix occlusion queries with larger vectors
need to adjust casts etc.
commit ff95e6fdf5f16d4ef999ffcf05ea6e8c7160b0d5
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Mar 19 20:15:25 2012 +0000
gallivm: Restore optimization passes.
commit 57b05b4b36451e351659e98946dae27be0959832
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 19:34:22 2012 +0000
llvmpipe: use existing min2 macro
commit bc9a20e19b4f600a439f45679451f2e87cd4b299
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 19:07:27 2012 +0000
llvmpipe: add some safeguards against really large vectors
As per José's suggestion, prevent things from blowing up if some cpu
would have 1024bit or larger vectors.
commit 0e2b525e5ca1c5bbaa63158bde52ad1c1564a3a9
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 18:31:08 2012 +0000
llvmpipe: fix mask generation for uberwide vectors
this was the only piece preventing 16-wide vectors from working
(apart from the LP_MAX_VECTOR_WIDTH define that is), which is the maximum
as we don't get more pixels in the fragment shader at once.
Hence adjust that so things could be tested properly with that size
even though there seems to be no practical value.
commit 3c8334162211c97f3a11c7f64e9e5a2a91ad9656
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 18:19:41 2012 +0000
llvmpipe: fix the simple interpolation method with larger vectors
so both methods actually _really_ work now. Makes textures look
nice with larger vectors...
commit 1cb0464ef8871be1778d43b0c56adf9c06843e2d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 17:26:35 2012 +0000
llvmpipe: fix mask generation and position interpolation with 8-wide vectors
trivial bugs, with these things start to look somewhat reasonable.
Textures though have some swizzling issues it seems.
commit 168277a63ef5b72542cf063c337f2d701053ff4b
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 16:04:03 2012 +0000
llvmpipe: don't overallocate variables
we never have more than 16 (stamp size) / 4 (minimum possible vector size).
(With larger vectors those variables are still overallocated a bit.)
commit 409b54b30f81ed0aa9ed0b01affe15c72de9abd2
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 15:56:48 2012 +0000
llvmpipe: add some 32f8 formats to lp_test_conv
Also add the ability to handle different sized vectors.
commit 55dcd3af8366ebdac0af3cdb22c2588f24aa18ce
Author: Roland Scheidegger <sroland@vmware.com>
Date: Mon Mar 19 15:47:27 2012 +0000
gallivm: handle different sized vectors in conversion / pack
only fully generic path for now (extract/insert per element).
commit 9c040f78c54575fcd94a8808216cf415fe8868f6
Author: Roland Scheidegger <sroland@vmware.com>
Date: Sun Mar 18 00:58:28 2012 +0100
llvmpipe: fix harmless use of unitialized values
commit 551e9d5468b92fc7d5aa2265db9a52bb1e368a36
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 16 23:31:21 2012 +0100
gallivm: drop special path in extract_broadcast with different sized vectors
Not needed, llvm can handle shuffles with different sized result vector just
fine. Should hopefully generate the same code in the end, but simpler IR.
commit 44da531119ffa07a421eaa041f63607cec88f6f8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 16 23:28:49 2012 +0100
llvmpipe: adapt interpolation for handling multiple quads at once
this is still WIP there are actually two methods possible not quite
sure what makes the most sense, so there's code for both for now:
1) the iterative method as used before (compute attrib values at upper left
corner of stamp and upper left corner of each quad initially).
It is improved to handle more than one quad at once, and also do some more vectorized
calculations initially for slightly better code - newer cpus have full throughput with
4 wide float vectors, hence don't try to code up a path which might be faster if there's
just one channel active per attribute.
2) just do straight interpolation for each pixel.
Method 2) is more work per quad, but less initially - if all quads are executed
significantly more overall though. But this might change with larger vector lengths.
This method would also be needed if we'd do some kind of active quad merging when
operating on multiple quads at once.
This path contains some hack to force llvm to generate better code, it is still far
from ideal though, still generates far too many unnecessary register spills/reloads.
Both methods should work with different sized vectors.
Not very well tested yet, still seems to work with four-wide vectors, need changes
elsewhere to be able to test with wider vectors.
commit be5d3e82e2fe14ad0a46529ab79f65bf2276cd28
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Mar 16 20:59:37 2012 +0000
draw: Cleanup.
commit f85bc12c7fbacb3de2a94e88c6cd2d5ee0ec0e8d
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Mar 16 20:43:30 2012 +0000
gallivm: More module compilation refactoring.
commit d76f093198f2a06a93b2204857e6fea5fd0b3ece
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Mar 15 21:29:11 2012 +0000
llvmpipe: Use gallivm_compile/free_function() in linear code.
Should had been done before.
commit 122e1adb613ce083ad739b153ced1cde61dfc8c0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 13 14:47:10 2012 +0100
llvmpipe: generate partial pixel mask for multiple quads
still works with one quad, cannot be tested yet with more
At least for now always fixed order with multiple quads.
commit 4c4f15081d75ed585a01392cd2dcce0ad10e0ea8
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Mar 8 22:09:24 2012 +0100
llvmpipe: refactor state setup a bit
Refactor to make it easier to emit (and potentially later fetch in fs)
coefficients for multiple attributes at once.
Need to think more about how to make this actually happen however, the
problem is different attributes can have different interpolation modes,
requiring different handling in both setup and fs (though linear and
perspective handling is close).
commit 9363e49722ff47094d688a4be6f015a03fba9c79
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Mar 8 19:23:23 2012 +0100
llvmpipe: vectorize tri offset calc
cuts number of instructions in quad-offset-factor from 107 to 75.
This code actually duplicated the (scalar) code calculating the determinant
except it used different vertex order (leading to different sign but it doesn't
matter) hence llvm could not have figured out it's the same (of course with
determinant vectorized in the other place that wouldn't have worked any longer
neither).
Note this particular piece doesn't actually vectorize well, not many arithmetic
instructions left but tons of shuffle instructions...
Probably would need to work on n tris at a time for better vectorization.
commit 63169dcb9dd445c94605625bf86d85306e2b4297
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Mar 8 03:11:37 2012 +0100
llvmpipe: vectorize some scalar code in setup
reduces number of arithmetic instructions, and avoids loading
vector x,y values twice (once as scalars once as vectors).
Results in a reduction of instructions from 76 to 64 in fs setup for glxgears
(16%) on a cpu with sse41.
Since this code uses vec2 disguised as vec4, on old cpus which had physical
64bit sse units (pre-Core2) it probably is less of a win in practice (and if
you have no vectors you can only hope llvm eliminates the arithmetic for
unneeded elements).
commit 732ecb877f951ab89bf503ac5e35ab8d838b58a1
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Mar 7 00:32:24 2012 +0100
draw: fix clipping
bug introduced by 4822fea3f0440b5205e957cd303838c3b128419c broke
clipping pretty badly (verified with lineclip test)
commit ef5d90b86d624c152d200c7c4056f47c3c6d2688
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 6 23:38:59 2012 +0100
draw: don't store vertex header per attribute
storing the vertex header once per attribute is totally unnecessary.
Some quick look at the generated assembly says llvm in fact cannot optimize
away the additional stores (maybe due to potentially aliasing pointers
somewhere).
Plus, this makes the code cleaner and also allows using a vector "or"
instead of scalar ones.
commit 6b3a5a57b0b9850854cfbd7b586e4e50102dda71
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Mar 6 19:11:01 2012 +0100
draw: do the per-vertex "boolean" clipmask "or" with vectors
no point extracting the values and doing it per component.
Doesn't help that much since we still extract the values elsewhere anyway.
commit 36519caf1af40e4480251cc79a2d527350b7c61f
Author: Roland Scheidegger <sroland@vmware.com>
Date: Fri Mar 2 22:27:01 2012 +0100
gallivm: fix lp_build_extract_broadcast with different sized vectors
Fix the obviously wrong argument, so it doesn't blow up.
commit 76d0ac3ad85066d6058486638013afd02b069c58
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Mar 2 12:16:23 2012 +0000
draw: Compile per module and not per function (WIP).
Enough to get gears w/ LLVM draw + softpipe to work on AVX doing:
GALLIUM_DRIVER=softpipe SOFTPIPE_USE_LLVM=yes glxgears
But still hackish -- will need to rethink and refactor this.
commit 78e32b247d2a7a771be9a1a07eb000d1e54ea8bd
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Feb 29 12:01:05 2012 +0000
llvmpipe: Remove lp_state_setup_fallback.
Never used.
commit 6895d5e40d19b4972c361e8b83fdb7eecda3c225
Author: José Fonseca <jfonseca@vmware.com>
Date: Mon Feb 27 19:14:27 2012 +0000
llvmpipe: Don't emit EMMS on x86
We already take precautions to ensure that LLVM never emits MMX code.
commit 4822fea3f0440b5205e957cd303838c3b128419c
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Feb 29 15:58:19 2012 +0100
draw: modifications for larger vector sizes
We want to be able to use larger vectors especially for running the vertex
shader. With this patch we build soa vectors which might have a different
length than 4.
Note that aos structures really remain the same, only when aos structures
are converted to soa potentially different sized vectors are used.
Samplers probably don't work yet, didn't look at them.
Testing done:
glxgears works with both 128bit and 256bit vectors.
commit f4950fc1ea784680ab767d3dd0dce589f4e70603
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Feb 29 15:51:57 2012 +0100
gallivm: override native vector width with LP_NATIVE_VECTOR_WIDTH env var for debug
commit 6ad6dbf0c92f3bf68ae54e5f2aca035d19b76e53
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Feb 29 15:51:24 2012 +0100
draw: allocate storage with alignment according to native vector width
commit 7bf0e3e7c9bd2469ae7279cabf4c5229ae9880c1
Author: José Fonseca <jfonseca@vmware.com>
Date: Fri Feb 24 19:06:08 2012 +0000
gallivm: Fix comment grammar.
Was missing several words. Spotted by Roland.
commit b20f1b28eb890b2fa2de44a0399b9b6a0d453c52
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 19:22:09 2012 +0000
gallivm: Use MC-JIT on LLVM 3.1 + (i.e, SVN)
MC-JIT
Note: MC-JIT is still WIP. For this to work correctly it requires
LLVM changes which are not yet upstream.
commit b1af4dfcadfc241fd4023f4c3f823a1286d452c0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Feb 23 20:03:15 2012 +0100
llvmpipe: use new lp_type_width() helper in lp_test_blend
commit 04e0a37e888237d4db2298f31973af459ef9c95f
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Feb 23 19:50:34 2012 +0100
llvmpipe: clean up lp_test_blend a little
Using variables just sized and aligned right makes it a bit more obvious
what's going on.
The test still only tests vector length 4.
For AoS anything else probably isn't going to work.
For SoA other lengths should work (at least with floats).
commit e61c393d3ec392ddee0a3da170e985fda885a823
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 17:48:30 2012 +0000
gallivm: Ensure vector width consistency.
Instead of assuming that everything is the max native size.
commit 330081ac7bc41c5754a92825e51456d231bf84dd
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 17:44:14 2012 +0000
draw: More simd vector width consistency fixes.
commit d90ca002753596269e37297e2e6c139b19f29f03
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 17:43:00 2012 +0000
gallivm: Remove unused lp_build_int32_vec4_type() helper.
commit cae23417824d75869c202aaf897808d73a2c1db0
Author: Roland Scheidegger <sroland@vmware.com>
Date: Thu Feb 23 17:32:16 2012 +0100
gallivm: use global variable for native vector width instead of define
We do not know the simd extensions (and hence the simd width we should use)
available at compile time.
At least for now keep a define for maximum vector width, since a global
variable obviously can't be used to adjust alignment of automatic stack
variables.
Leave the runtime-determined value at 128 for now in all cases.
commit 51270ace6349acc2c294fc6f34c025c707be538a
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 15:41:02 2012 +0000
gallivm: Add a hunk inadvertedly lost when rebasing.
commit bf256df9cfdd0236637a455cbaece949b1253e98
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 14:24:23 2012 +0000
llvmpipe: Use consistent vector width in depth/stencil test.
commit 5543b0901677146662c44be2cfba655fd55da94b
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 14:19:59 2012 +0000
draw: Use a consistent the vector register width.
Instead of 4x32 sometimes, LP_NATIVE_VECTOR_WIDTH other times.
commit eada8bbd22a3a61f549f32fe2a7e408222e5c824
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 12:08:04 2012 +0000
gallivm: Remove garbagge collection.
MC-JIT will require one compilation per module (as opposed to one
compilation per function), therefore no state will be shared,
eliminating the need to do garbagge collection.
commit 556697ea0ed72e0641851e4fbbbb862c470fd7eb
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 10:33:41 2012 +0000
gallivm: Move all native target initialization to lp_set_target_options().
commit c518e8f3f2649d5dc265403511fab4bcbe2cc5c8
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 09:52:32 2012 +0000
llvmpipe: Create one gallivm instance for each test.
commit 90f10af8920ec6be6f2b1e7365cfc477a0cb111d
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 09:48:08 2012 +0000
gallivm: Avoid LLVMAddGlobalMapping() in lp_bld_assert().
Brittle, complex, and unecesary. Just use function pointer constant.
commit 98fde550b33401e3fe006af59db4db628bcbf476
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 09:21:26 2012 +0000
gallivm: Add a lp_build_const_func_pointer() helper.
To be reused in all places where we want to call C code.
commit 6cfedadb62c2ce5af8d75969bc95a607f3ece118
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 09:44:41 2012 +0000
gallivm: Cleanup/simplify lp_build_const_string_variable.
- Move to lp_bld_const where it belongs
- Rename to lp_build_const_string
- take the length from the argument (and don't count the zero terminator twice)
- bitcast the constant to generic i8 *
commit db1d4018c0f1fa682a9da93c032977659adfb68c
Author: José Fonseca <jfonseca@vmware.com>
Date: Thu Feb 23 11:52:17 2012 +0000
gallivm: Set NoFramePointerElimNonLeaf to true where supported.
commit 088614164aa915baaa5044fede728aa898483183
Author: Roland Scheidegger <sroland@vmware.com>
Date: Wed Feb 22 19:38:47 2012 +0100
llvmpipe: pass in/out pointers rather scalar floats in lp_bld_arit
we don't want llvm to potentially optimize away the vectors (though it doesn't
seem to currently), plus we want to be able to handle in/out vectors of arbitrary
length.
commit 3f5c4e04af8a7592fdffa54938a277c34ae76b51
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Feb 21 23:22:55 2012 +0100
gallivm: fix lp_build_sqrt() for vector length 1
since we optimize away vectors with length 1 need to emit intrinsic
without vector type.
commit 79d94e5f93ed8ba6757b97e2026722ea31d32c06
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Feb 22 17:00:46 2012 +0000
llvmpipe: Remove lp_test_round.
commit 81f41b5aeb3f4126e06453cfc78990086b85b78d
Author: Roland Scheidegger <sroland@vmware.com>
Date: Tue Feb 21 23:56:24 2012 +0100
llvmpipe: subsume lp_test_round into lp_test_arit
Much simpler, and since the arguments aren't passed as 128bit values can run
on any arch.
This also uses the float instead of the double versions of the c functions
(which probably was the intention anyway).
In contrast to lp_test_round the output is much less verbose however.
Tested vector width of 32 to 512 bits - all pass except 32 (length 1) which
crashes in lp_build_sqrt() due to wrong type.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
commit 945b338b421defbd274481d8c4f7e0910fd0e7eb
Author: José Fonseca <jfonseca@vmware.com>
Date: Wed Feb 22 09:55:03 2012 +0000
gallivm: Centralize the function compilation logic.
This simplifies a lot of code.
Also doing this in a central place will make it easier to carry out the
changes necessary to use MC-JIT in the future.
gallivm: Fix typo in explicit derivative shuffle.
Trivial.
draw: make DEBUG_STORE work again
adapt to lp_build_printf() interface changes
Reviewed-by: José Fonseca <jfonseca@vmware.com>
draw: get rid of vecnf_from_scalar()
just use lp_build_broadcast directly (cannot assign a name but don't really
need it, vecnf_from_scalar() was producing much uglier IR due to using
repeated insertelement instead of insertelement+shuffle).
Reviewed-by: José Fonseca <jfonseca@vmware.com>
llvmpipe: fix typo in complex interpolation code
Fixes position interpolation when using complex mode
(piglit fp-fragment-position and similar)
Reviewed-by: José Fonseca <jfonseca@vmware.com>
draw: fix clipvertex/position storing again
This appears to be the result of a bad merge.
Fixes piglit tests relying on clipping, like a lot of the interpolation tests.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
gallivm: Fix explicit derivative manipulation.
Same counter variable was being used in two nested loops. Use more
meanigful variable names for the counter to fix and avoid this.
gallivm: Prevent buffer overflow in repeat wrap mode for NPOT.
Based on Roland's patch, discussion, and review .
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
gallivm: Fix dims for TGSI_TEXTURE_1D in emit_tex.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
gallivm: Fix explicit volume texture derivatives.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
gallivm: fix 1d shadow texture sampling
Always r coordinate is used, hence need 3 coords not two
(the second one is unused).
Reviewed-by: José Fonseca <jfonseca@vmware.com>
gallivm: Enable AVX support without MCJIT, where available.
For now, this just enables AVX on Windows for testing. If the code is
stable then we might consider prefering the old JIT wherever possible.
No change elsewhere.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
The vertex element state isn't in registers any more, so
remove that old code. That fixes a memory corruption with
the blend state and gets eglgears partially working.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Some application calls eglCreateWindowSurface with
EGLNativeWindowType parameter having zero value. It causes SEGV
and disturbs error handling like EGL_NO_SURFACE.
Signed-off-by: Elvis Lee <kwangwoong.lee@lge.com>
Signed-off-by: Brian Paul <brianp@vmware.com>
a112ca5d rather crassly smashed all the compiler flags together into AM_CFLAGS.
Separate them out the way they were before, putting pre-processor flags into
AM_CPPFLAGS, so assembly source gets preprocessed with the correct pre-processor
flags as well.
Also, remove unneeded CFLAGS from AM_CFLAGS, and CXXFLAGS from AM_CXXFLAGS
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Tested-by: Brian Paul <brianp@vmware.com>
I suck at resolving merge conflicts and broke the build in a5a34b1.
This patch adds the missing field intel_mipmap_tree::wraps_etc1.
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
Enable it for all hardware.
No current hardware supports ETC1, so this patch implements it by
translating the ETC1 data to RGBX data during the call to
glCompressedTexImage2D(). For details, see the doxygen for
intel_mipmap_tree::wraps_etc1.
Passes the Piglit test spec/OES_compressed_ETC1_RGB8_texture/miptree and
the ETC1 test in the GLES2 conformance suite.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
Add function _mesa_etc1_unpack_rgba8888. It is intended to be used by
glCompressedTexSubImage2D to decode ETC1 textures into RGBA.
CC: Chia-I <olv@lunarg.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
Move the body of util_etc1_rgb8_unpack_rgba_unorm8 into a new function
that can be shared between gallium and dri drivers,
texcompress_etc_tmp.h:etc1_unpack_rgba8888.
CC: Chia-I <olv@lunarg.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
lp_delete_setup_variants() used to be called in garbage collection,
but this no longer exists hence the setup shaders never got freed.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
When we don't intend to texture from or render to a __DRIimage we
use __DRI_IMAGE_FORMAT_NONE. In that case, we just create the __DRIimage
to reference the underlying buffer, and will create usable __DRIimages
from it using createSubImage later.
If we try to use _mesa_get_format_bytes() on MESA_FORMAT_NONE in
a debug build, we hit an assertion, so let's not do that.
Commit 68e04cc6 was tested using automake-1.11. Unfortunately, automake-1.12
made a "slightly backward-incompatible change" in the use of yacc with C++, and
for a .yy file, the generated header file is now named .hh, not .h
To work with both, write our own rule for running yacc, which generates a
header file named .h, rather than using automake's rule.
Also, remove things from BUILD_SOURCES which don't need to be there
Also, update EXCLUDE rules in doxygen/glsl.doxy, for change of generated files
from .cpp -> .cc, and glsl_lexer.h has never existed.
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Commit defadf2b1 erroneously tries to make gallium drivers link with libdricore
as a static library, not a shared library
Also, change uses of DRI_LIB_DEPS in gallium driver Makefiles to
GALLIUM_DRI_LIB_DEPS, so the libraries added are used in the linking the gallium
driver
Also, fix the path to the libdricore.so symlink, it's made in LIB_DIR, not in
the libdricore directory
Also repair quoting of dricore settings of DRI_LIB_DEPS and GALLIUM_DRI_LIB_DEPS
variables so VERSION is interpolated in configure but TOP and LIB_DIR are
interpolated later (where they are known, but VERSION isn't)
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Signed-off-by: Tom Stellard <thomas.stellard@amd.com>
- Use LLVM limits when LLVM is being used, instead of TGSI limits
- Provide draw_get_shader_param_no_llvm for when llvm is never used (softpipe)
- Eliminate several of the hacks around draw shader caps in several drivers
Unfortunately the hack for PIPE_MAX_VERTEX_SAMPLERS is still necessary.
Reviewed-by: Marek Olšák <maraeo@gmail.com>
The libmesa convenience library is linked with the libglsl convenience
library. libOsmesa is linked with libmesa, and also directly with libglsl.
When using libtool, this gives rise to duplicate symbol errors.
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
* "configure substitutions are not allowed in _SOURCES variables" in automake,
so remove the AC_SUBST'ed GLAPI_ASM_SOURCES and instead use some AM_CONDITIONALS
to choose which asm sources are used
* Change GLAPI_LIB to point to the .la file in other Makefile.am files, and make a link
to the .a file for the convenience of other Makefiles which have not yet been converted
to automake
v2:
- Use AM_CPPFLAGS for cleaner build output
- EXTRA_SOURCES is not needed
- Remove libglapi.a compatibility link on clean
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
Now mesa/drivers/dri is converted to automake, we want to update DRI_LIB_DEPS
so that we link with the libmesa or libdricore libtool library, as appropriate.
However, this is complicated by the fact that gallium/targets is not (yet)
converted, so we can't share the DRI_LIB_DEPS autoconf variable with that anymore.
Add an additional autoconf variable GALLIUM_DRI_LIB_DEPS, which is now used in
gallium/targets/Makefile.dri, to link with the libdircore or libmesa native library.
v2: libdricore$VERSION.a needs to be libdricore$(VERSION).a
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
* "configure substitutions are not allowed in _SOURCES variables" in automake, so instead of
MESA_ASM_FILES, use some AM_CONDITIONALS to choose which architecture's asm sources are used
in libmesa_la_SOURCES. (Can't remove MESA_ASM_FILES autoconf variable as it's still used in
sources.mak)
* Update to link with the .la file in other Makefile.am files, and make a link to the
.a file for the convenience of other Makefiles which have not yet been converted to automake
v2: Remove stray -static from LDFLAGS
v3: Remove .a compatibility link on clean
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
Automake can't handle having both clip.S and clip.c, even though they have different paths
"src/mesa/Makefile.am: object `clip.lo' created by `$(SRCDIR)/sparc/clip.S' and `$(SRCDIR)/main/clip.c'"
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
v2: Use AM_V_GEN to silence generated code rules. Add BUILT_SOURCES to CLEANFILES
v3:
- Fix an accidental // in a path
- Use automake make rules for lex/yacc rather than writing our own
- Update .gitignore appropriately
- Build a libglcpp convenience library rather than awkwardly including
the files in libglsl and delegating the generation
- Remove libglsl.a compatibility link on clean
v4:
- Automake's rules for lex/yacc make .cc if source is .ll or .yy, and apparently we
must use those extensions "because of scons", so update everywhere glsl_parser.cpp
-> glsl_parser.cc and glsl_lexer.cpp -> glsl_lexer.cc. This fixes 'make tarballs'
and building with dricore enabled.
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
This also currently fix the installation of libOSmesa.
v2: Remove old Makefile, libOSmesa is now versioned, fix typos
v3: Keep config substitution alphabetized
v4: Update .gitignore
v5: Libraries will be in the builddir, not the srcdir.
Reviewed-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Matt Turner <mattst88@gmail.com>
This was not implemented, because the spec was changed just recently.
Everything has been in place already.
Gallium has PIPE_FORMAT_B5G6R5_UNORM, while Mesa has MESA_FORMAT_RGB565.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The whole reason I avoided this was because it might operate on a
brw_vertex_program or a brw_fragment_program. However, that isn't a
problem: all we need is the gl_program base type.
This avoids awkwardly passing the loop counter 'i' as a parameter,
simplifies both callers, and also plumbs prog in place for future use.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
If alpha-testing is enabled, we need to send alpha down the pipeline
even if nr_color_buffers == 0. However, tracking whether alpha-testing
is enabled in the WM program key is expensive: it causes us to compile
multiple specializations of the same shader, using program cache space.
This patch removes the check for alpha-testing, and simply emits alpha
whenever nr_color_buffers == 0. We believe this will also be necessary
for alpha-to-coverage, and it should add minimal overhead to an uncommon
case. Saving the recompiles should more than make up the difference.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Previously we only did this pre-Gen6, and used pwrite on Gen6+.
In one workload, this cuts significant amount of overhead.
v2: Simplify the function based on Eric's suggestions.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
We rely on proper IEEE 754 behavior in too many places for this.
See also commit 2fdbbeca43 with equivalent
change for autoconf.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Without that, people with buggy apps that looked at just the server
string for GLX_ARB_create_context would call this function that just
threw an error when you tried to make a context. Google shows plenty
of complaints about this.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This function assumes that lp_build_context::type is a vector type,
which is not true for r600 or radeonsi.
This fixes an assertion failure using glamor 2D accel.
It had many problems:
- The shadow comparison was done post-filtering.
- It required state-dependent recompiles whenever the comparison
function changed.
- It didn't even work: many cases hit assertion failures.
- I never implemented it for the VS.
The new lowering pass which converts textureGrad to textureLod by
computing the LOD value works much better.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Intel hardware doesn't natively support textureGrad with shadow
comparisons. So we need to generate code to handle it somehow.
Based on the equations of page 205 of the OpenGL 3.0 specification,
it's possible to compute the LOD value that would be selected given the
gradient values. Then, we can simply convert the TXD to a TXL.
Currently, this passes 34/46 of oglconform's shadow-grad subtests;
four cubemap tests are regressed. We should investigate this in the
future.
v2: Apply abs() to the scalar case (thanks to Eric).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
This swizzles away unwanted components, while preserving the order of
the ones that remain.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
I needed to compute logs and square roots in a patch I was working on,
and wanted to use the convenient interface. We already have a similar
constructor for binops; adding one for unops seems reasonable.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
I ran into this while trying to create a TXS query, which doesn't have a
coordinate. Since it didn't get initialized to NULL, a bunch of
visitors tried to access it and crashed.
Most of the time, this won't be a problem, but it's just a good idea.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
The only case a depth buffer can be set as a color buffer is when flushing.
That wasn't always the case, but now this code isn't required anymore.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
- maintain a mask of which mipmap levels are dirty (instead of one big flag)
- only flush what was requested at a given point and not the whole resource
(most often only one level and one layer has to be flushed)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
we can just update the state when decompressing, there's no need to add
additional info into the DSA state
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
to remove some overhead from draw_vbo. This is a derived state.
BTW, I've got no idea how compute interacts with 3D here, but it should
use cb_misc_state, so that 3D and compute don't conflict.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Because u_blit couldn't sample a 1D, 3D, CUBE and ARRAY texture, we created
a 2D texture holding a copy of one slice of the source texture (even for 1D).
Let's just do it right.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
This patch updates the blorp engine to properly handle the case where
the surface being textured from uses Gen7's CMS MSAA layout. The
following changes were necessary:
- Before reading color values from the surface, we need to read from
the MCS buffer using the ld_mcs sampler message. This is done by
the mcs_fetch() function, and the result is stored in the mcs_data
register. This only needs to be done once per pixel, since the MCS
value is shared between all samples belonging to a pixel.
- When reading color values from the surface, we need to use the
ld2dms sampler message instead of the ld2dss message, and we need to
provide the value read from the MCS buffer as an argument.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
When a buffer using Gen7's CMS MSAA layout is bound to a texture or a
render target, the SURFACE_STATE structure needs to point to the MCS
buffer and to indicate its pitch. This patch updates the functions
that emit SURFACE_STATE to handle CMS layout properly.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Previously the DWORD used to control the CMS MSAA layout was just a
pad value, because we didn't use it.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
To implement Gen7's CMS MSAA layout, we need an extra buffer, the MCS
(Multisample Control Surface) buffer. This patch introduces code for
allocating and deallocating the buffer, and storing a pointer to it in
the intel_mipmap_tree struct.
No functional change, since the CMS layout is not enabled yet.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
From the Ivy Bridge PRM, Vol 1 Part 1, p112:
There are three types of multisampled surface layouts designated
as follows:
- IMS Interleaved Multisampled Surface
- CMS Compressed Mulitsampled Surface
- UMS Uncompressed Multisampled Surface
Previously, the i965 driver only used IMS and UMS formats, and
distinguished beetween them using the boolean
intel_mipmap_tree::msaa_is_interleaved. To facilitate adding support
for the CMS format, this patch replaces that boolean (and other
booleans derived from it) with an enum
INTEL_MSAA_LAYOUT_{IMS,CMS,UMS}. It also updates the terminology used
in comments throughout the driver to match the IMS/CMS/UMS terminology
used in the PRM. CMS layout is not yet used.
The enum has a fourth possible value, INTEL_MSAA_LAYOUT_NONE, which is
used for non-multisampled surfaces.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
On Gen6, MSAA buffers always use an interleaved layout and non-MSAA
buffers always use a non-interleaved layout, so it is not strictly
necessary to keep track of the layout of the texture and render target
surfaces in the blorp program key. However, it is cleaner to do so,
since (a) it makes the blorp compiler less dependent on implicit
knowledge about how the GPU pipeline is configured, and (b) it paves
the way for implementing compressed multisampled surfaces in Gen7.
This patch won't cause any redundant compiles, because the layout of
the texture and render target surfaces depends on other parameters
that are already in the blorp program key.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
We don't generate public entrypoints for GLES extensions, so move the
GL_NV_draw_buffers definition from ARB_draw_buffers.xml to es_EXT.xml.
When the extension is defined in ARB_draw_buffers.xml, we end up with a
public entry point for it, but no prototype, which gives an error when
compiled with --disable-asm and --disable-shared-glapi.
Instead, just move the GLES extension to es_EXT.xml so this doesn't happen.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
This lets us specify an offset into the bo where the miptree starts,
which will let us set up a texture for a single plane in a planar buffer.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
The additions in version 5 enables creating EGLImages for different planes
of a YUV buffer. createImageFromName is still used to create the containing
__DRIimage, and createSubImage can then be used no that __DRIimage to create
__DRIimages that correspond to the y, u, and v planes (__DRI_IMAGE_FORMAT_R8)
or the uv planes (__DRI_IMAGE_FORMAT_RG88) for formats such as NV12 where
the u and v components are interleaved. Packed formats such as YUYV etc
doesn't require any special treatment, we just sample those as a regular
ARGB texture.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
The code for growing the memory pool (which is used for storing all of
the global buffers) wasn't working. There seem to be two separate issues
with the memory pool code. The first was the way it was growing the pool.
When the memory pool needed more space, it would:
1. Copy the data from the memory pool's backing texture to system memory.
2. Delete the memory pool's texture
3. Create a bigger backing texture for the memory pool.
4. Copy the data from system memory into the bigger texture.
The copy operations didn't seem to be working, and I suspect that since
they were using fragment shaders to do the copy, that there might have
been a problem with the mixing of compute and 3D state.
The other issue is that the size of 1D textures is limited, and I was
having trouble getting 2D textures to work.
I think these problems will be easier to solve once more code is shared
between 3D and compute, which is why I decided to disable it for now
rather than continue searching for a fix.
The original strategy for handling floating point loads, which was to
lower (f32 load) to (f32 bitcast (i32 load)) wasn't really working. The
main problem was that the DAG legalizer couldn't handle replacing a node
with two results (load) with a node with only one result (bitcast).
It didn't change performance on Lightsmark or Nexuiz, which both used
DYNAMIC_DRAW buffers, but it was killing performance (40% CPU wasted pwriting
buffers) on a closed-source app we're looking at.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Add the infrastructure required for this extension. There is no
xserver support and no driver support yet. Drivers can enable this be
advertising DRI2 version 4 and accepting the
__DRI_CTX_FLAG_ROBUST_BUFFER_ACCESS flag and the
__DRI_CTX_ATTRIB_RESET_STRATEGY attribute in create context.
Some additional Mesa infrastructure is needed before drivers can do
this. The GL_ARB_robustness spec, which all Mesa drivers already
advertise, requires:
"If the behavior is LOSE_CONTEXT_ON_RESET_ARB, a graphics reset
will result in the loss of all context state, requiring the
recreation of all associated objects."
It is necessary to land this infrastructure now so that the related
infrastructure can land in the xserver. The xserver has very long
release schedules, and the remaining Mesa parts should land long, long
before the next xserver merge window opens.
v2: Expose robustness as a DRI2 extension rather than bumping
__DRI_DRI2_VERSION.
v3: Add a comment explaining why dri2->base.version >= 3 is also
required for GLX_ARB_create_context_robustness.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
This allows revising the dri_interface.h separately from adding driver
support.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We neglected to list the deprecation model/forward compatible context
support.
inverse() has been done for a while.
None of us know what "highp change" means; GLSL 1.30 already added the
ability to recognize precision keywords, and it doesn't look like 1.40
has any new requirements there (precision keywords still have no meaning).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Use r600_resource_texture::flished_depth_texture for GPU access, and
allocate it in the VRAM. For transfers we'll allocate texture in the GTT
and store it in the r600_transfer::staging.
Improves performance when flushed depth texture is frequently used by the
GPU, e.g. in Lightsmark (~30%)
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
With fixes and updates from Ben Widawsky and comments from Paul Berry.
v2: Use drm_intel_gem_context_destroy to destroy hardware context;
remove useless initialization of hw_ctx, both suggested by Eric.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Acked-by: Paul Berry <stereotype441@gmail.com>
This doesn't do anything with the uniform block declarations yet, so
usage of those uniforms finds them to be undeclared.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
I've been trying to derive from this for UBO support, and the slightly
obfuscated types were putting me over the edge.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The got_one variable was set iff one of the bits in flags.i was set.
v2: Fix incorrect dropping of the ARB_conservative_depth warning.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> (v1)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This function is used when dispatching compute shader in order to avoid
mixing compute and 3D registers in the context's dirty list. This
allows the compute code to resuse 3D functions like evergreen_cb, which
return a struct r600_pipe_state and still have control over when and how
the register writes are emitted.
The start_compute_cs atom initializes some config and context registers
to the values needed for running compute shaders. When a compute shader
is dispatched, this atom is emitted after the start_cs_cmd atom, which
initializes registers that are common to both 3D and compute.
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Some packets require the shader type bit (bit 1) to be set when
used for compute shaders. The pkt_flag will be initialized to
RADEON_CP_PACKET3_COMPUTE_MODE for any struct r600_command_buffer used
for dispatching compute shaders and it will be or'd against the result of
the PKT3 macro when adding a new packet to a struct r600_command buffer.
Reviewed-by: Marek Olšák <maraeo@gmail.com>
For copy propgation, we've dropped the use of a GRF in favor of a
(probably later) use of a different GRF. This definitely requires
invalidating intervals.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Since live intervals are based on ip, removing an instruction trashes
the intervals unless we were to go do some surgery. These happen to
usually remove a use of a grf, so it's time to recalculate, anyway.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
NOTE: This is a candidate for the 8.0 release branch.
This has less impact than for the FS (4k savings), because it was partially
done already, but makes things more consistent.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We factor out all the EGL book-keeping into dri2_create_image() and
simplify the wayland case by using dupImage.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
We have the same switch and allocation code in two places.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This reverts commit cbffaf20e9.
Use the PRIx64 macro in the fprintf() call instead, as suggested
by Dylan Noblesmith.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
ROUND and TRUNC are implemented with one function to reduce code duplication.
Note: ROUND isn't actually used yet, but probably will be soon.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Converting CMP to SLT+LRP didn't work when src2 or src3 was Inf/NaN.
That's the case for GLSL sqrt(0). sqrt(0) actually happens in many
piglit auto-generated tests that use the distance() function.
v2: remove debug/devel code, per Jose
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Was previously implemented with FLOOR.
Fixes quite a few piglit tests of float->int conversion, integer
division, etc.
v2: clean up left over debug/devel code, per Jose
Reviewed-by: José Fonseca <jfonseca@vmware.com>
If the 'dst' register is the same as the 'pass' register we'll generate
invalid code. Use a temporary register in that case.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Redo this commit, and remove the inclusion of gl2ext.h
from src/mapi/glapi/glapi_priv.h. The include was added in
8f3be33985 to fix a missing prototype for
glDrawBuffersNV and others, but it's not possible to include both
glext.h and gl2ext.h from the same file.
I don't see the missing prototype here (with or without shared glapi)
so I'm just removing the offending #include.
Also, since we're redoing this, update to the most recent gl2ext.2.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
That old bug was hidden but the clipper always interpolating in 3d space
no matter what it should have been doing. Now that the interpolation
has been fixed, the bug shows up.
Fixes fdo 51364.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: José Fonseca <jfonseca@vmware.com>
Calling glGenerateMipmap could overwrite vertex buffer state, leading
to incorrect rendering or crashes depending on the Gallium driver.
This was happening on WebGL Conformance test texture-size.
Before 784dd51198 this was covered up
by redundant vertex buffer validation.
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Marek Olšák <maraeo@gmail.com>
This reverts commit 8818b88748.
I get a lot of errors like this one:
In file included from ../../../src/mapi/glapi/glapi_priv.h:49:0,
from glapi_dispatch.c:40:
../../../include/GLES2/gl2ext.h:1074:28: error: redefinition of typedef ‘PFNGLRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC’
../../../include/GL/glext.h:10237:25: note: previous declaration of ‘PFNGLRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC’ was here
This with a clean build (with git clean -fdX).
I don't get the errors on my other machine. I didn't investigate why,
a wild guess is that this depends on the version of gcc.
This is a big win for savage2, hon and yofrankie. 62 new programs for
savage2/hon get 16-wide mode, along with one for humus demos and two
for tropics. Even a few shaders from tropics see reductions of 15% or
more.
total instructions in shared programs: 216536 -> 207353 (-4.24%)
instructions in affected programs: 123941 -> 114758 (-7.41%)
In benchmarking Tropics, only a .040% +/- 034% performance improvement
was observed (n=90). Rather disappointing, but I was primarily
motivated to do this patch by a regression in the number of 16-wide
shaders compiled after a GRF texturing on IVB patch I'm working on.
Hopefully this helps avoid that regression.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This shaves a few instructions off of a ton of programs. For 12
shaders from tropics and sanctuary, it's enough reduction in register
pressure to get 16-wide mode. 7 shaders from heroes of newerth and
savage2 are hurt by about 1.1%, where copy propagation of negates ends
up preventing coalescing, but we could regain that by doing dataflow
analysis in our copy propagation.
No significant performance difference in tropics (n=11)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The meta-ops _mesa_meta_Clear() and _mesa_meta_glsl_Clear() need to
ignore the state of GL_SAMPLE_ALPHA_TO_COVERAGE,
GL_SAMPLE_ALPHA_TO_ONE, GL_SAMPLE_COVERAGE, GL_SAMPLE_COVERAGE_VALUE,
and GL_SAMPLE_COVERAGE_INVERT when clearing multisampled buffers. The
easiest way to accomplish this is to disable GL_MULTISAMPLE during the
clear meta-ops.
Note: this patch also causes GL_MULTISAMPLE to be disabled during
_mesa_meta_GenerateMipmap() and _mesa_meta_GetTexImage() (since those
two meta-ops use MESA_META_ALL). Arguably this isn't strictly
necessary, since those meta-ops use their own non-MSAA fbo's, but it
shouldn't do any harm.
Fixes Piglit tests "EXT_framebuffer_multisample/clear {2,4}
{color,stencil}" on i965.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
From the Ivy Bridge PRM, Vol 2 Part 1 p280-281 (3DSTATE_WM:
Barycentric Interpolation Mode):
"Errata: When Centroid Barycentric mode is required, HW may
produce incorrect interpolation results when a 2X2 pixels have
unlit pixels."
To work around this problem, after doing centroid interpolation, we
replace the centroid-interpolated values for unlit pixels with
non-centroid-interpolated values (which are interpolated at pixel
centers). This produces correct rendering at the expense of a slight
increase in shader execution time.
I've conditioned the workaround with a runtime flag
(brw->needs_unlit_centroid_workaround) in the hopes that we won't need
it in future chip generations.
Fixes piglit tests "EXT_framebuffer_multisample/interpolation {2,4}
{centroid-deriv,centroid-deriv-disabled}". All MSAA interpolation
tests pass now.
Reviewed-by: Eric Anholt <eric@anholt.net>
In order to compute centroid varyings correctly, the fragment shader
needs to be able to load the current pixel/sample mask into a flag
register. This patch adds an opcode to the fragment shader back-end
to do this; the opcode gets translated into the instruction
mov(1) f0<1>UW g1.14<0,1,0>UW { align1 WE_all }
Since this instruction clobbers f0, instruction scheduling has to
treat it the same as instructions that have a conditional modifier.
Reviewed-by: Eric Anholt <eric@anholt.net>
When querying GL_PRIMITIVES_GENERATED, if primitive restart
is also used, then take the software primitive restart
path so GL_PRIMITIVES_GENERATED is returned correctly.
GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN is also updated
since it will also affected by the same issue.
As noted in brw_primitive_restart.c, with further work we
should be able to move this situation back to a hardware
handled path.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Commit d73f6375f5 fixed the cause of the Piglit failure with
ARB_color_buffer_float fragment clamp modes. Now that it's fixed,
there's no reason to leave snorm format rendering disabled.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Commit 0c005bd7 intended to make ir_loop_jump::mode public, but also
accidentally added a new pointer to the enclosing loop. Furthermore, it
tried to initialize the new field by adding "this->loop = loop;" to the
constructor, but since there is no loop parameter, this only initialized
the field to itself---so it will likely be a garbage pointer.
A lot of code, such as lower_jumps, allocates new loop jumps without
setting this field appropriately, so any uses would probably just crash.
Thankfully, there were none, so we can just delete the field.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=51574
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
DrawPixels uses the MESA_META_CLAMP_FRAGMENT_COLOR flag to save/restore
the fragment color clamp mode. This is unnecessary since it never
alters it. It's also harmful: when the clamp mode is GL_FIXED_ONLY,
setting this flag causes _mesa_meta_begin to force it to GL_FALSE,
breaking clamping on SNORM formats.
DrawPixels should use the user-specified clamp mode and not change it.
Fixes Piglit's spec/ARB_color_buffer_float/GL_RGBA8_SNORM-drawpixels
test on i965/Sandybridge (with SNORM render targets re-enabled).
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Add "-f $(srcdir)/gl_API.xml" to the arguments of all
the scripts that by default look for gl_API.xml in the
working directory when run with no arguments, and prepend
$(srcdir) to those scripts that are already using an
explicit -f argument.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
tgsi_ureg was recently enhanced to support local temporaries, and as result
temps are declared individually.
This change avoids many TEMP register declarations on common shaders.
(And fixes performance regression due to mismatches against performance
sensitive shaders.)
Reviewed-by: Brian Paul <brianp@vmware.com>
The templated copy constructor doesn't prevent the compiler from
emitting a default copy constructor, which leads to inconsistent
memory handling and was reported to cause segfaults when doing event
manipulation.
Reported-by: Tom Stellard <thomas.stellard@amd.com>
The function internalizer pass marks non-kernel functions as internal,
which enables optimizations like function inlining and global dead-code
elimination.
v2:
- Pass vector arguments by const reference
Removed u_half.py used to generate the table for previous method.
Previous implementation of float to half conversion was faulty for
denormalised and NaNs and would require extra logic to fix,
thus making the speedup of using tables irrelevant.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Some parameters need to be checked only once.
check_valid_to_render needs to be called only once.
The validate function is based on the one for DrawElements.
Reviewed-by: Brian Paul <brianp@vmware.com>
This is a cleanup for ARB_transform_feedback3, where
GL_MAX_TRANSFORM_FEEDBACK_BUFFERS is introduced for interleaved attribs and
has the same meaning as GL_MAX_.._SEPARATE_ATTRIBS for separate attribs.
Also, the maximum number of TFB buffers is reduced from 32 to 4, which makes
this patch useful even without the extension.
I don't know of any hardware which can do more than 4.
Reviewed-by: Brian Paul <brianp@vmware.com>
Doesn't really change the generated assembly, but produces more compact IR,
and of course, makes code more consistent.
Reviewed-by: Brian Paul <brianp@vmware.com>
For some reason regular gcc on Linux didn't catch these but the mingw
compiler did (generated errors, not warnings).
v2: include the changes in src/mapi/ too
Fixes the es2 build with gcc.
Note: in glext.h the prototypes for glShaderSource() and glShaderSourceARB()
disagree: only the former has the extra const qualifier.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Set the step_rate value when drawing to implement
ARB_instanced_arrays for gen >= 4.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Previously, we were counting gl_FrontFacing, gl_FragCoord and gl_PointCoord
against the limit of varying variables. This prevented some valid shaders
from linking.
The other potential solution to this is to have the driver advertise
more varying vars or set the GLSLSkipStrictMaxVaryingLimitCheck flag.
But the above-mentioned variables aren't conventional varying attributes
so it doesn't seem right to count them.
Reviewed-by: Eric Anholt <eric@anholt.net>
Updated lp_build_printf to share common code.
Removed specific lp_build_print_vecX.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Since we don't have them in hw we emulate them in the shader. Although not
recommended by the spec it is legit.
As a side effect we also get GL 2.1. I think this is as far as we can take
the i915.
The most recent commit adds support for comments and macro expansion
on #line directives. Add testing to verify the new features.
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The GLSL specification requires that #line directives be interpreted
after macro expansion. Our existing implementation of #line macros in
the lexer prevents conformance on this point.
Moving the handling of #line from the lexer to the parser gives us the
macro expansion we need. An additional benefit is that the
preprocessor also now supports comments on the same line as #line
directives.
Finally, the preprocessor now emits the (fully-macro-expanded) #line
directives into the output. This allows the full GLSL compiler to also
see and interpret these directives so it can also generate correct
line numbers in error messages.
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This function is currently used only in the expansion of #if lines,
but we will soon be using it more generally (for the expansion of
(_glcpp_parser_expand_and_lex_from) and some more documentation.
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Commit b823b99ec0 switched from using
functions such as ralloc_asprintf and ralloc_strcat to
ralloc_asprintf_rewrite_tail. This change maintains the string's
length as a aparamter that is updated by the ralloc functions (rather
than recomputing it with strlen over and over).
However, the change failed to updated two locations (glcpp_error and
glcpp_warning), with the result that the string's length wasn't
updated by these calls. Then, subsequent calls to other
ralloc_asprintf_rewrite_tail would overwrite the text appended by
glcpp_error.
This commit fixes the two missing updates, and restores line numbers
to the output of glcpp error messages, (as noticed by a glcpp unit
test case that has been failing since the above-mentioned commit).
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
A strict reading of the GLSL specification would have this be an
error, but we've received reports from users who expect the
preprocessor to interepret undefined macros as 0. This is the standard
behavior of the rpeprocessor for C, and according to these user
reports is also the behavior of other OpenGL implementations.
So here's one of those cases where we can make our users happier by
ignoring the specification. And it's hard to imagine users who really,
really want to see an error for this case.
The two affected tests cases are updated to reflect the new behavior.
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
DUAL_EXPORT can be enabled on r6xx/r7xx when all CBs use 16-bit export
and there is no depth/stencil export.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
It seems DUAL_EXPORT on evergreen may be enabled when all CBs use 16-bit export
mode (EXPORT_4C_16BPC), also there should be at least one CB, and the PS
shouldn't export depth/stencil.
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
In some cases TGSI shader has more color outputs than the number of CBs,
so it seems we need to limit the number of color exports. This requires
different shader variants depending on the nr_cbufs, but on the other hand
we are doing less exports, which are very costly.
v2: fix various piglit regressions
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Shader variants are stored in the list, the key for lookup is based on the
states that require different hw shaders - currently it's rctx->two_side (all
gpus) and rctx->nr_cbufs (evergreen/cayman, when writes_all property is set).
v2:
- use simple list instead of keymap as suggested by Marek on irc
- call r600_adjust_gprs from r600_bind_vs_shader for r6xx/r7xx
(r600_shader_select isn't used for vertex shaders currently)
v3:
- fix call to r600_adjust_gprs - do it after updating current shader
Improves performance for some apps, e.g. FlightGear -
see https://bugs.freedesktop.org/show_bug.cgi?id=50360
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
As with the previous commit for softpipe.
v2: remove 'default' case to get compile-time warning
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
These all return zero. Add a debug_printf() to catch the default case so
we don't accidently mishandle something important in the future.
v2: remove 'default' case to get compile-time warning
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
This is actually required for GL_ARB_framebuffer_object, but the state
tracker doesn't currently check it.
Direct3D 9 allows mixed format color buffers with some restrictions.
Setting this allows Unigine Heaven 2.5 and 3.0 to run. Tested both on
GL and D3D hosts.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
The type is the destination type (i.e. float vector) and not the
source type. Fixes piglit fs-{in,de}crement-uint.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: José Fonseca <jfonseca@vmware.com>
i965 hardware needs to be informed of situations in which it's
possible for pixels (or samples) to be discarded for reasons other
than depth/stencil testing (e.g. due to an explicit "discard" in the
fragment shader). One of these situations is when
GL_ALPHA_TO_COVERAGE is enabled, since that can cause samples to be
discarded by the color calculator when the pixel's alpha value is less
than 1.0.
Without this patch, GL_ALPHA_TO_COVERAGE does not take effect on depth
buffers.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This patch enables the multisampling parameters
GL_SAMPLE_ALPHA_TO_COVERAGE and GL_SAMPLE_ALPHA_TO_ONE, which allow
the fragment shader's alpha output to be converted into a sample
coverage mask and ignored for blending. i965 supports these
parameters through the BLEND_STATE structure.
The GL spec allows, but does not require, the implementation to dither
the conversion from alpha to a sample coverage mask, so that alpha
values that aren't a multiple of 1/num_samples result in the correct
proportion of samples being lit. A bit exists in the BLEND_STATE
structure to enable this functionality, but according to the hardware
docs it must be disabled on Sandy Bridge (see the Sandy Bridge PRM,
Vol2, Part1, p379: AlphaToCoverage Dither Enable). So it is enabled
for Gen7 only.
Fixes piglit tests
"EXT_framebuffer_multisample/sample-alpha-to-{coverage,one} {2,4}".
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This patch enables glSampleCoverage() functionality, which allows the
client program to specify that only a portion of the samples be lit up
when performing multisampled rendering. i965 supports
glSampleCoverage() through the 3DSTATE_SAMPLE_MASK command packet,
which allows the driver to specify a bitfield indicating which samples
to light up.
Fixes piglit tests "EXT_framebuffer_multisample/sample-coverage {2,4}
{inverted,non-inverted}".
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Fixes gles2conform GL.equal.equal_bvec2_frag.
This fixes brw_fs_visitor's translation of ir_unop_f2b. It used CMP to
convert the float to one of 0 or ~0. However, the convention in the
compiler is that true is represented by 1, not ~0. This patch adds an AND
to convert ~0 to 1.
By inspection, a similar problem existed with ir_unop_i2b, with a similar
fix.
[v2 kayden]: eliminate extra temporary register.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=49621
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
This patch causes the fragment shader to be configured correctly (and
the correct code to be generated) for centroid interpolation. This
required two changes: brw_compute_barycentric_interp_modes() needs to
determine when centroid barycentric coordinates need to be included in
the pixel shader thread payload, and
fs_visitor::emit_general_interpolation() needs to interpolate using
the correct set of barycentric coordinates.
Fixes piglit tests "EXT_framebuffer_multisample/interpolation {2,4}
centroid-edges" on i965.
Reviewed-by: Eric Anholt <eric@anholt.net>
To save time, we only instruct the clip stage of the pipeline to
compute noperspective barycentric coordinates if those coordinates are
needed by the fragment shader. Previously, we would determine whether
the coordinates were needed by seeing whether the fragment shader used
the BRW_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC interpolation mode.
However, with MSAA, it's possible that the fragment shader might use
BRW_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC instead. In the future,
when we support ARB_sample_shading, it might use
BRW_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC.
This patch modifies the upload_clip_state() functions to check for all
three possible noperspective interpolation modes.
Reviewed-by: Eric Anholt <eric@anholt.net>
This bitfield tells the back-ends which of a fragment shader's inputs
require centroid interpolation. It is only set for GLSL fragment
shaders, since assembly fragment shaders don't support centroid
interpolation.
Reviewed-by: Eric Anholt <eric@anholt.net>
It was only no-oping the clear() function, not actual triangle
rasterization. Move the no_rast field from lp_context down into
lp_rasterizer so it's accessible where it's needed.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Fixes this build failure on Solaris.
Compiling build/sunos-debug/glsl/glcpp/glcpp-lex.c ...
"src/glsl/glcpp/glcpp-lex.l", line 30: cannot find include file: "glcpp-parse.h"
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
$CLANG_RESOURCE_DIR is the directory that contains all resources
needed by clang to compile programs. When clover uses clang to
compile kernels it needs to specify a resource dir, so that clang
can find its internal headers (e.g. stddef.h).
clang defines $CLANG_RESOURCE_DIR as $CLANG_LIBDIR/clang/$CLANG_VERSION
This patch adds the --with-clang-libdir option in order to accommodate
clang intalls to non-standard locations, and it also adds a check
to the configure script to verify that $CLANG_RESOURCE_DIR/include
contains the necessary header files.
On i965, dFdx() and dFdy() are computed by taking advantage of the
fact that each consecutive set of 4 pixels dispatched to the fragment
shader always constitutes a contiguous 2x2 block of pixels in a fixed
arrangement known as a "sub-span". So we calculate dFdx() by taking
the difference between the values computed for the left and right
halves of the sub-span, and we calculate dFdy() by taking the
difference between the values computed for the top and bottom halves
of the sub-span.
However, there's a subtlety when FBOs are in use: since FBOs use a
coordinate system where the origin is at the upper left, and window
system framebuffers use a coordinate system where the origin is at the
lower left, the computation of dFdy() needs to be negated for FBOs.
This patch modifies the fragment shader back-ends to negate the value
of dFdy() when an FBO is in use. It also modifies the code that
populates the program key (brw_wm_populate_key() and
brw_fs_precompile()) so that they always record in the program key
whether we are rendering to an FBO or to a window system framebuffer;
this ensures that the fragment shader will get recompiled when
switching between FBO and non-FBO use.
This will result in unnecessary recompiles of fragment shaders that
don't use dFdy(). To fix that, we will need to adapt the GLSL and
NV_fragment_program front-ends to record whether or not a given shader
uses dFdy(). I plan to implement this in a future patch series; I've
left FIXME comments in the code as a reminder.
Fixes Piglit test "fbo-deriv".
NOTE: This is a candidate for stable release branches.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
It's not optimal, but it's better than the register pressure scheduler
that was previously being used. The VLIW scheduler currently ignores
all the complicated instruction groups restrictions and just tries to
fill the instruction groups with as many instructions as possible.
Though, it does know enough not to put two trans only instructions in
the same group.
We are able to ignore the instruction group restrictions in the LLVM
backend, because the finalizer in r600_asm.c will fix any illegal
instruction groups the backend generates.
Enabling the VLIW scheduler improved the run time for a sha1 compute
shader by about 50%. I'm not sure what the impact will be for graphics
shaders. I tested Lightsmark with the VLIW scheduler enabled and the
framerate was about the same, but it might help apps that use really
big shaders.
The rest of the TFB implementation remains in transformfeedback.c, and
this will be shared with UBOs.
v2: Move the size/offset checks shared with UBOs to common code as
well. (Kenneth's review)
Reviewed-by: Brian Paul <brianp@vmware.com> (v1)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2: Fix a typo spotted by Eric Anholt.
v3: Fix missing "GL" on types, fix style, fix Studly_Caps extension name,
drop commented code duplicated with GL3x.xml [anholt]
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Our intention is still that it's not abi stable, so make the package
version number get included in the library name. Now you can parallel
install dricore-using drivers from multiple mesa versions. We can put
it into lib now that we're following library versioning rules
(assuming that ABIs don't change within a single Mesa point release).
LD_LIBRARY_PATH still doesn't work with a non-/, non-/usr prefix
because libtool uses rpath instead of runpath for nonstandard
prefixes.
The weird versioning of the libGL where the package version was sort
of expressed as a big integer is dropped. libtool didn't like the 0
prefix, and it didn't really make sense anyway -- if you interpret it
as an integer version number, old Mesa 071200 was bigger than current
Mesa 08100. Instead, just bump the minor version and drop the
patchlevel.
Except for the deleted linux-cell target, these were just the target
cc/cflags. The only usage was for gen_matypes, which wants the
target's structure packing, not the host, anyway.
Every place that uses ASM_FLAGS already uses DEFINES. Not including
it in DEFINES is just a way to screw up potential users, as I've done
several times while working on the build system.
Even pre-automake, we rely on gmake features for pattern
substitutions, and replacing those with reams more make code is not
interesting. This will let us turn the old Makefiles using pattern
substitutions into automake without spewing warnings.
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
1) We need to insert a barrier between consecutive transform feedback calls.
2) VBO cache needs to be flushed when TFB output is used as VBO draw input.
Fixes Piglit test EXT_transform_feedback/immediate-reuse.
Thanks to Christoph Bumiller for pointing out bugs in previous versions
of this patch.
gl_ClipDistance needs special treatment in form of lowering pass
which transforms gl_ClipDistance representation from float[] to
vec4[]. There are 2 implementations - at glsl linker level (enabled
by LowerClipDistance option) and at glsl_to_tgsi level (enabled
unconditionally for gallium drivers). Second implementation is
incomplete - it does not take into account transform feedback (see
commit 642e5b413e "mesa: Fix transform
feedback of unsubscripted gl_ClipDistance array" for details).
There are 2 possible fixes:
- adding transform feedback support into glsl_to_tgsi version
- ripping gl_ClipDistance support from glsl_to_tgsi and enabling
gl_ClipDistance lowering on glsl linker side
This patch implements 2nd option. All it does is:
- reverts most of the commit 59be691638
"st/mesa: add support for gl_ClipDistance"
- changes LowerClipDistance to true
Fixes Piglit tests "EXT_transform_feedback/builtin-varyings
gl_ClipDistance[{2,3,4,5,6,7,8}]-no-subscript" at least on nv50
and evergreen cards.
From the GL 3.0 spec (p.116):
"Multisample rasterization is enabled or disabled by calling
Enable or Disable with the symbolic constant MULTISAMPLE."
Elsewhere in the spec, where multisample rasterization is described
(sections 3.4.3, 3.5.4, and 3.6.6), the following text is consistently
used:
"If MULTISAMPLE is enabled, and the value of SAMPLE_BUFFERS is
one, then..."
So, in other words, disabling GL_MULTISAMPLE should prevent
multisample rasterization from occurring, even if the draw framebuffer
is multisampled. This patch implements that behaviour by setting the
WM and SF stage's "multisample rasterization mode" to
MSRAST_ON_PATTERN only when the draw framebuffer is multisampled *and*
GL_MULTISAMPLE is enabled.
Fixes piglit test spec/EXT_framebuffer_multisample/enable-flag.
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Due to hardware limitations, MSAA is unsupported on Gen6 for formats
containing >64 bits of data per pixel. From the Sandy Bridge PRM,
vol4 part1, p72 ("Surface Format"):
If Number of Multisamples is set to a value other than
MULTISAMPLECOUNT_1, this field cannot be set to the following
formats:
- any format with greater than 64 bits per element
- any compressed texture format (BC*)
- any YCRCB* format
Gen7 has a similar, but less stringent limitation: formats with >64
bits of data per pixel only support 4x MSAA.
This patch causes the unsupported formats to report
GL_FRAMEBUFFER_UNSUPPORTED.
Fixes piglit "multisample-formats" tests on Gen6.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Sandy Bridge and later don't use this field, so there's no point in
setting it. It can only cause harmful state-based recompiles.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
The system array values concept doesn't really because it expects the
system values to be fixed per call, which is wrong for gl_VertexID and
iffy for gl_SampleID. So this patch does two things:
- kill the array, have emit_fetch_system_value directly pick the
values it needs (only gl_InstanceID for now, as the previous code)
- correctly handle the expected type in emit_fetch_system_value
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
This includes:
- picking up correctly which attributes are flatshaded and which are
noperspective
- copying the flatshaded attributes when needed, including the
non-built-in ones
- correctly interpolating the noperspective attributes in screen-space
instead than in a 3d-correct fashion.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
z or stencil texture should not be created with the z/stencil
flags for surface creation as they are intended to be bound
as texture.
v2: remove broken code
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Solaris Studio C compiler does not support anonymous structs and
anonymous unions.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
The idea here is to rewrite comparisons like 2 >= x with x <= 2; we want
to simply exchange arguments, not negate the condition. If equality was
part of the original comparison, it should remain part of the swapped
version.
This is the true cause of bug #50298. It didn't manifest itself on
Sandybridge because we embed the conditional modifier in the IF
instruction rather than emitting a CMP. All other platforms use CMP.
It also didn't manifest itself on the master branch because commit
be5f27a84d ("glsl: Refine the loop instruction counting.") papered over
the problem.
NOTE: This is a candidate for stable release branches.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50298
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Fixes build error on Cygwin and Solaris. _R, _G, and _B are used in
ctype.h on those platforms.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
This fixes a bug where a sampler view was using stale texture/resource
data when the texture was modified through a surface (render to texture).
Bumping the texture and layer ages triggers sampler view revalidation.
Fixes piglit fbo-blit failure.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
This lets us select the front buffer for reading under GLES2.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This extra condition checks the API not the version of the API, so rename
to reflect that.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This is failing sometimes, probably because TargetData keeps a structure layout
cache, which can becomes bogus, ever since the InvalidateStructLayoutInfo API
was removed in LLVM r135245.
This change merely makes the problem easier to diagnose (an assertion
failure instead of a random crash).
instead of failing to allocate a renderbuffer.
This also fixes piglit/get-renderbuffer-internalformat with non-renderable
formats.
Reviewed-by: Brian Paul <brianp@vmware.com>
This allows drivers not to do any allocation in AllocStorage if the storage
cannot be allocated because of an unsupported internalformat + samples combo.
The little ugliness is that AllocStorage is expected to return TRUE in this
case.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This requires the latest streamout kernel patches.
Streamout is disabled by default on r7xx, so this patch is safe for regular
users.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Note: for the moment TGSI_OPCODE_F2U is implemented using
lp_build_itrunc() (the same function used to implement
TGSI_OPCODE_F2I). In the long run, we should create an
lp_build_utrunc() function to do the proper conversion. But this
should allow us to limp along with mostly correct behaviour for now.
Previously, we performed conversions from float->uint by a two step
process: float->int->uint. However, on platforms that use saturating
conversions (e.g. i965), this didn't work, because if the source value
was larger than the maximum representable int (0x7fffffff), then
converting it to an int would clamp it to 0x7fffffff.
This patch just adds the new opcode; further patches will adapt
optimization passes and back-ends to use it, and then finally the
ast_to_hir logic will be modified to emit the new opcode.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch modifies blorp blits (which are used for MSAA) to properly
account for clipping of source coordinates. Previously, if we
detected the possibility of source clipping, we would fall back to the
blit meta-op, which doesn't support MSAA and is very slow for depth
and stencil buffers.
Fixes piglit tests
"EXT_framebuffer_multisample/clip-and-scissor-blit" on i965/Gen6+.
Also substantially speeds up the Humble Bundle V game "Psychonauts" on
Gen6+ (without this patch, the game's depth buffer blits use the slow
blit meta-op).
Reviewed-by: Carl Worth <cworth@cworth.org>
This allows to submit things to the compute only
rings on cayman+
v2: rebased on current master and actually make use
of the new flag in evergreen_compute.c
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
When drawing a depth image the fragment shader also needs to emit the
current raster color.
The new piglit drawpix-z test exercises this.
NOTE: This is a candiate for the 8.0 branch.
This patch updates .gitignore files to account for the new build
artifacts introduced by the following commits:
ae376f0 glx/tests: Rename test as glx-test
8fecdcc mesa/tests: Add tests for _mesa_lookup_enum_by_{name,nr} functions
a29ad2b mesa/tests: Add tests for the generated dispatch table
Haiku targets the Pentium or higher processor.
To ensure compatibility we can do march 586 and
mtune 686. Mesa will still use sse however if
the cpu supports it (and the stack is properly
aligned). These flags only effect the internal
compiler optimizations.
Previously, rbug_*.c would fail to compile with incomplete prototype
errors when make was run from the command line on my machine. My IDE
always built fine, and still does after this patch (Netbeans 7.1.2).
Most of the includes from files in gallium/auxiliary/rbug/* were
assuming an rbug/ subdirectory, while the headers are actually in the
same directory as the .c files.
The build error was also previously a problem for me on Ubuntu 11.10
and Mint 12.
Fixes build for the following configuration: ./autogen.sh
--enable-debug --enable-texture-float --with-gallium-drivers=r600
--with-dri-drivers=radeon --enable-r600-llvm-compiler
Signed-off-by: Brian Paul <brianp@vmware.com>
In single precision, 1.5707963 becomes 1.5707962513 which is too
small. However, 1.5707964 becomes 1.5707963705 which is just right.
The value 1.5707964 is already used in asin.ir.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
There is no GLX protocol for these functions. Open-source Linux
driver have not supported this extension for many years, and it seems
unlikely at this point that this support will return. There's no
reason to have slots for these functions in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
There is no GLX protocol for these functions. No open-source Linux
driver has ever supported this extension, and it seems unlikely at
this point that one ever will. There's no reason to have slots for
these functions in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
There is no GLX protocol for these functions. No open-source Linux
driver has ever supported this extension, and it seems unlikely at
this point that one ever will. There's no reason to have slots for
these functions in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
There is no GLX protocol for these functions, and no Linux driver has
ever supported this extension. There's no reason to have slots for
these functions in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
There is no GLX protocol for this function. Open-source Linux driver
have not supported this extension for many years, and it seems
unlikely at this point that this support will return. There's no
reason to have slots for this function in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
There is no GLX protocol for these functions, and no Linux driver has
ever supported this extension. There's no reason to have slots for
these functions in the dispatch table.
The unit tests (GetProcAddress::TableDidntShrink and others) are also updated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
These are from OpenGL 3.1 and ARB_uniform_buffer_object. I only added
them to 3.1 because that required the least work.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
These are from OpenGL 3.3, ARB_texture_swizzle, and
EXT_texture_swizzle (with different names). I only added them to 3.3
because that required the least work.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Determines whether it's a basis vector, i.e., a vector with one element
equal to 1 and all other elements equal to 0.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When a value was replaced, the new key was strdup'd and leaked.
To fix this, we modify the hash table implementation to return
whether the value was replaced and free() the (now useless)
duplicate string.
When we have multiple shared contexts, and one of them is
long-running, this will lead to never freeing those resources
since they are shared. Instead, free them right away on context
destruction since we know the other context isn't using them.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
NOTE: This is a candidate for the 8.0 branch.
From the GL_NV_primitive_restart spec:
"PrimitiveRestartIndexNV is not compiled into display lists, but is
executed immediately."
Prior to this patch, calls to glPrimitiveRestartIndex would hit the noop
dispatch stub.
+2 oglconforms.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
From the GL_ARB_copy_buffer spec:
"An INVALID_VALUE error is generated if any of readoffset, writeoffset,
or size are negative [...]"
Fixes oglconform's copybuffer/negative.CNNegativeValues test.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The warnings appear to occur with newer automake (probably 1.12).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
These allow one to mangle the library names, without also mangling the
symbol names, to make them distinct from other GL libraries on the
system.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Because these classes are used entirely from their own source files
and not from separate DSOs, the linker gets to produce massively less
code. This cuts about 13k of text in the libdricore case. In the
non-libdricore case, the additional linkage information allows the
compiler to inline some code, so libglsl.a size actually increases by
about 300 bytes.
For a dricore build, improves shader_runner runtime on
glsl-fs-copy-propagation-texcoords-1 by 0.21% +/- 0.03% (n=353574,
outliers removed). No statistically significant difference with n=322
on glslparsertest on a yofrankie shader intended to test compiler
performance.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Now we have just one library of "all of Mesa core" instead of both
libdricore and libglsl that drivers link against.
I did this change in a sort of nonrecursive make fashion: the
generated files are still produced in the non-automake build, like the
rest of dricore, but the GLSL files are stuffed into libdricore
without building a convenience library in src/glsl (even though we
could now). This would make a bit more sense if glsl was just another
dir under src/mesa, because right now I had to contort the prefix
variable name to look another ../ level up.
This is part of a series to fix our build issues in the automake case
by hooking up the automatic Makefile regeneration support. The
extract_git_sha1 is moved into src/mesa/Makefile so that we get
correct dependency generation.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
I tried to update all the old Makefiles that included the default
config to be sure they had a default target if they didn't previously
have one, since this new all target will always point at it. Almost
everything had one.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Some more of the files are now autogenerated, this caused build breakage,
patch adds generation of these missing files. Patch also changes existing
make so that the files are created to be part of the local source
(not intermediate directory, this causes several problems).
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
This patch fixes a copy/paste error and masking of depth/stencil (stencil
is in the top 8 bits), and makes glean/readPixSanity happy.
Both the stencil and the depth buffer piglit test also pass if
glClear(DEPTH | STENCIL) is executed instead of
glClear(DEPTH)/glClear(STENCIL).
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Christopher Egert <cme3000@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
remove archaic .cvsignore
*.pyo is already in toplevel .gitignore
*.pyc is already in toplevel .gitignore
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Previously, blits using the "blorp" mechanism only worked for 8-bit
RGBA color buffers, 24-bit depth buffers, and 8 bit stencil buffers.
This was not enough, because the blorp mechanism must be used for
blitting whenever MSAA is in use. This patch allows all formats to be
used, provided the source and destination formats match.
So far I have confirmed that the following formats work properly with
MSAA:
- GL_RGB
- GL_RGBA
- GL_ALPHA
- GL_ALPHA4
- GL_ALPHA8
- GL_R3_G3_B2
- GL_RGB4
- GL_RGB5
- GL_RGB8
- GL_RGB10
- GL_RGB12
- GL_RGB16
- GL_RGBA2
- GL_RGBA4
- GL_RGB5_A1
- GL_RGBA8
- GL_RGB10_A2
- GL_RGBA12
- GL_RGBA16
Fixes piglit tests "EXT_framebuffer_multisample/formats {2,4}" on
Sandy Bridge and Ivy Bridge.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Previously the blorp engine only supported RGBA8 color buffers and
24-bit depth buffers. This patch adds support for any color buffer
format that is supported as a render target, and for 16-bit and 32-bit
depth buffers.
This required threading the brw_context struct through into
brw_blorp_surface_info::set() so that it can consult the
brw->render_target_format array.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Even though brw_blorp_surface_info is derived from brw_blorp_mip_info,
this function doesn't need to be virtual, because it is never accessed
through a base class pointer. Making the function non-virtual will
allow it to take additional parameters in the brw_blorp_surface_info
case.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch moves the responsibility for deciding on the format of the
source and destination surfaces from the
gen{6,7}_blorp_emit_surface_state() functions to
brw_blorp_surface_info::set(), which is shared between Gen6 and Gen7.
This will make it possible to add support for more surface formats
without code duplication.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
TGSI doesn't need an opcode, since registers are untyped (but beware
once doubles come into the scene). Mesa IR doesn't handle native
integers, so trying to handle them there is worthless, the case
entries are only added for warning reasons.
It was only tested with softpipe, since llvmpipe doesn't support glsl
1.3 yet.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
That adds support for activating the extension. It doesn't actually
*do* anything yet, of course.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
From the issues section of the GL_ARB_texture_compression_rgtc extension:
15) What should glGetTexLevelParameter return for
GL_TEXTURE_GREEN_SIZE and GL_TEXTURE_BLUE_SIZE for the RGTC1
formats? What should glGetTexLevelParameter return for
GL_TEXTURE_BLUE_SIZE for the RGTC2 formats?
RESOLVED: Zero bits.
These formats always return 0.0 for these respective components
and have no bits devoted to these components.
Returning 8 bits for red size of RGTC1 and the red and green
sizes of RGTC2 makes sense because that's the maximum potential
precision for the uncompressed texels.
Thus, we need to return 8 bits for GL_TEXTURE_RED_SIZE on all RGTC formats
and 8 bits for GL_TEXTURE_GREEN_SIZE on RGTC2 formats. BLUE should be 0.
Fixes oglconform/rgtc/advanced.texture_fetch.tex_param.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
While ~loop_state() is already freeing the loop_variable_state objects
via ralloc_free(this->mem_ctx), the ~loop_variable_state() destructor
was never getting called, so the hash table inside loop_variable_state
was never getting destroyed.
Fixes a memory leak in any shader with loops.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
The functions for handling 1D, 2D and 3D texture images were nearly
identical. This folds them all together.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We can't remove this pass yet, because we need it to convert AMDIL
registers in BRANCH* instructions, but we don't need it for
instruction conversion any more.
OpenGL allows you to declare user-defined fragment shader outputs with
less than four components:
out ivec2 color;
This makes sense if you're rendering to an RG format render target.
Previously, we assumed that all color outputs had four components (like
the built-in gl_FragColor/gl_FragData variables). This caused us to
call emit_color_write for invalid indices, incrementing the output
virtual GRF's reg_offset beyond the size of the register.
This caused cascading failures: split_virtual_grfs would allocate new
size-1 registers based on the virtual GRF size, but then proceed to
rewrite the out-of-bounds accesses assuming that it had allocated enough
new (contiguously numbered) registers. This resulted in instructions
that accessed size-1 GRFs which register numbers beyond
virtual_grf_next (i.e. registers that were never allocated).
Finally, this manifested as live variable analysis and instruction
scheduling accessing their temporary array with an out of bounds index
(as they're all sized based on virtual_grf_next), and the program would
segfault.
It looks like the hardware's Render Target Write message requires you to
send four components, even for RT formats such as RG or RGB. This patch
continues to use all four MRFs, but doesn't bother to fill any data for
the last few, which should be unused.
+2 oglconforms.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Commit 4650aea7a5 fixed texelFetchOffset()
on Ivybridge, but didn't update the Ironlake/Sandybridge code.
+18 piglits on Sandybridge.
NOTE: This and 4650aea7a5 are both candidates for stable branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Commit f41ecade7b fixed texelFetchOffset()
on Ivybridge, but didn't update the Ironlake/Sandybridge code.
+15 piglits on Sandybridge.
NOTE: This and f41ecade7b are both candidates for stable branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
This isn't saved/restored by _mesa_meta_begin, so we need to do it
manually (like we do for the read/draw framebuffers). Additionally,
we neglected to re-bind before the glRenderbufferStorage call.
+13 oglconforms.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
DeleteBuffer needs to unbind from these binding points as well, based on
the same rationale as the previous patch.
+51 oglconforms (together with the last patch).
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
_mesa_lookup_bufferobj returns NULL for 0, which caused us to say
"there's no such buffer object" and raise an error, rather than
correctly binding the shared NullBufferObj.
Now you can unbind your buffers.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
According to the GL 3.1 spec, section 2.9 ("Buffer Objects"):
"If a buffer object is deleted while it is bound, all bindings to that
object in the current context (i.e. in the thread that called
DeleteBuffers) are reset to zero."
The code already checked for a number of cases, but neglected these
newer binding points.
+21 oglconforms.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
We were incorrectly assuming that the coordinate's dimensionality is
equal to the gradient's dimensionality. For array types, the coordinate
has one more component.
Fixes 12 subcases of oglconform's glsl-bif-tex-grad test.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Currently, if you pass --with-egl-platforms=x11 but xcb-dri2 isn't available
we just silently fail and disables building the EGL DRI2 driver.
This commit cleans up the EGL platfrom checking and fails if a selected
platform can't find its required dependencies.
Reviewed-by: Eric Anholt <eric@anholt.net>
Commit a07cf3397e added support for TBOs
on Gen7, but missed Gen6.
Passes piglit -t texture_buffer and oglconform's buffermapping
basic.read.texture tests.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
According to Table 6.17 in the GL 2.1 specification, DEPTH_TEXTURE_MODE,
TEXTURE_COMPARE_MODE, and TEXTURE_COMPARE_FUNC need to be restored on
glPopAttrib(GL_TEXTURE_BIT).
Makes a number of oglconform tests happier.
v2: Make restoration conditional on the ARB_shadow and ARB_depth_texture
extensions, as suggested by Brian. I'm not sure that any
implementations still remain that don't support those, but why not?
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
this fixes libdricore directory build with --enable-32-bit on a x86_64 system
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
The VTX_READ instructions were using the ADDRParam ComplexPattern which
allows a load instruction's offset to be a register, but VTX_READ
instructions can only handle an immediate offset.
Also, the load_param pattern fragment had an erroneous return true;
statement that was causing it to match the wrong load instructions.
Tungsten Graphics has not existed for several years, and the majority of
ongoing development and support is done by Intel. I chose to include
"Open Source Technology Center" to distinguish it from, say, the closed
source Windows OpenGL driver.
The one downside to this patch is that applications that pattern match
against "Intel" may start applying workarounds meant for the Windows
driver. However, it does seem like the right thing to do.
This does change oglconform behavior.
Acked-by: Eric Anholt <eric@anholt.net>
Acked-by: Ian Romanick <ian.d.romanick@intel.com>
Acked-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
These look like debug messages from the switch-statement development.
NOTE: This is a candidate for the 8.0 release branch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Tom Stellard:
- Updated for gallium interface changes
- Fixed a few bugs:
+ Set the loop counter
+ Calculate the correct number of pipes
- Added hooks into the LLVM compiler
v2:
-Separate IR type and LLVM triple
-Do the OpenCL C->LLVM IR and linking steps for all PIPE_SHADER_IR
types.
v3:
- Coding style fixes
- Removed compatibility code for LLVM < 3.1
- Split build_module_llvm() into three functions:
compile(), link(), and build_module_llvm()
v4:
- Use struct pipe_compute_program
v5:
- Don't malloc memory for struct pipe_llvm_program
v6:
- Fix serialization of llvm bytecode
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This structure is used as a header that precedes LLVM bytecode programs
that are passed to the drivers.
v2:
- s/pipe_compute_program/pipe_llvm_program/
v3:
- Rename to struct pipe_llvm_program_header
- Drop the char * prog member
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This is for the llvm code that can't use extended initializers.
v2:
- Use const references for vector arguments
- Move constructor defs before data members
- Initialize all values in the default constructors
v3:
- Fix typo
A device now has two function for getting information about the IR
it needs to return.
ir_format() => returns the preferred IR
ir_target() => returns the triple for the target that is understood by
clang/llvm.
v2:
- renamed ir_target() to ir_format()
- renamed llvm_triple() to ir_target()
v3:
- Remove unnecessary include
- Do proper conversion from std::vector<char> to std::string
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
v2: Tom Stellard
- Update CAP description
v3: Tom Stellard
- TGSI targets should pass an empty string for this CAP.
v4: Tom Stellard
- TGSI targets can ignore this CAP.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
TEX instructions can't do saturation. Do the TEX into a temp reg w/out
saturation, then do a MOV_SAT.
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Some distributions (like Arch Linux) make /usr/bin/python Python 3,
rather than Python 2. Since compare_ir uses /usr/bin/env python,
such systems will fail to run optimization-test, causing 'make check' to
always fail.
Automake's TESTS_ENVIRONMENT variable provides a mechanism to run
programs or set environment variables in the test environment.
Ideally, I think we would want to use AM_TESTS_ENVIRONMENT, since
TESTS_ENVIRONMENT is supposed to be user-overridable. However, it isn't
supported using the default/serial test runner.
Fixes 'make check' on Arch Linux and Gentoo.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Matt Turner <mattst88@gmail.com>
I started writing unit tests for a new piece of code, and discovered
they all failed due to a bug in ralloc. Clearly it needs a test suite.
v2: Rename to 'ralloc-test' and fix copyright date. (idr review)
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
If an object is allocated out of the NULL context, info->parent will be
NULL. Using the PTR_FROM_HEADER macro would be incorrect: it would say
that ralloc_parent(ralloc_context(NULL)) == sizeof(ralloc_header).
Fixes the new "null_parent" unit test.
NOTE: This is a candidate for the 7.9, 7.10, 7.11, and 8.0 branches.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Discovered while running the Khronos conformance test suite and
receiving "implementation error: meta program compile failed."
This bug was recently introduced by the i965 clear patch set and would
only be detected while using the ES2 API and only on gen6+ hardware.
Signed-off-by: Oliver McFadden <oliver.mcfadden@linux.intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This is performed in a subdirectory to avoid needing to convert all of
src/mesa/Makefile in one go.
I can now cherry-pick a commit containing glapi XML changes, do "(cd
src/mapi/glapi/gen && make) && make", and get a working driver.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
In order to do the minimal change for libdricore conversion to
automake, I need to put its Makefile.am in a subdirectory. Automake
gets whiny/broken if you use GNU make features like "addprefix" or
"$(FILES:%=../%)" to munge your *_SOURCES. So, use a plain old
variable to be able to substitute in that "../"
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
*_SOURCES is reserved for files lists for particular automake targets.
Also, "-" in the variable names is not allowed.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This variable won't be set when called from non-automake makefiles,
but it cleans up shared-glapi's output.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Mesa already always depends on python to build. The checked in
changes are not reviewed (because any trivial change rewrites the
world). We also have been pushing commits between xml change and
regen where at-build-time xml-generated code disagrees with committed
xml-generated code. And worst of all, sometimes we ("I") check in
*stale* xml-generated code.
Acked-by: Ian Romanick <ian.d.romanick@intel.com>
commit 87f12bb2d9 tried to fix rb->mt
being NULL, but change this case wrong.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We now model loading uses sgpr values with LLVM IR load instructions that
use the USER_SGPR address space.
The definition of the sgpr parameter to the use_sgpr() helper function
in radeonsi_shader.c has changed so that you can pass raw sgpr values
rather than having to divide the sgpr value you want to use by the dword
width of the type you want to load.
This function was causing compile errors in the tablegen'd code for
some intrinsic definitions. I don't think we really need this function,
so I'm removing the function body just as a temporary solution. I'll
look into removing the entire AMDILIntrinsicInfo class later.
v2: use a define for the maximum sample count
v3: also test odd sample counts (r300 supports MS3)
While multisample renderbuffers are supported by mesa, MS visuals
are not, so we need a way to tell dri/st not to advertise them even
if the gallium driver does support multisampled surfaces.
Otherwise applications selecting these non-functional visuals would
run into trouble ...
Reviewed-by: Brian Paul <brianp@vmware.com>
The code which scans the index buffer for restart indexes wasn't adding
the index buffer offset so we were always starting at offset=0. The
offset is usually zero so it wasn't noticed before.
Fixes a failure in the piglit primitive-restart test when testing
vertex data + index data in a single VBO.
NOTE: This is a candidate for the 8.0 branch.
Basic 4x MSAA support now works on Gen7. This patch enables it.
As with Gen6, MSAA support is still fairly preliminary. In
particular, the following are not yet supported:
- 8x oversampling (Gen7 has hardware support for this, but we do not
yet expose it).
- Fully general blits between MSAA and non-MSAA buffers.
- Formats other than RGBA8, DEPTH24, and STENCIL8.
- Centrold interpolation.
- Coverage parameters (glSampleCoverage, GL_SAMPLE_ALPHA_TO_COVERAGE,
GL_SAMPLE_ALPHA_TO_ONE, GL_SAMPLE_COVERAGE, GL_SAMPLE_COVERAGE_VALUE,
GL_SAMPLE_COVERAGE_INVERT).
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
On Gen6, the blending necessary to blit an MSAA surface to a non-MSAA
surface could be accomplished with a single texturing operation. On
Gen7, the WM program must fetch each sample and blend them together
manually. From the Bspec (Shared Functions/Messages/Initiating
Message/Message Types/sample):
[DevIVB+]:Number of Multisamples on the associated surface must be
MULTISAMPLECOUNT_1.
This patch implements the manual blend operation.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Since blorp uses color textures and render targets to do all its work
(even when blitting stencil and depth data), it always has to
configure the Gen7 GPU to use the new "sliced" MSAA layout. However,
when blitting stencil or depth data, the actual MSAA layout is
interleaved (as in Gen6). Therefore, blorp has to do extra coordinate
transformation work to account for the interleaving manually.
This patch causes blorp to perform the necessary extra coordinate
transformations.
It also modifies the blorp SURFACE_STATE setup code for Gen7, so that
it does not try to correct the surface width and height to account for
MSAA, since "sliced" MSAA layout doesn't affect the surface width or
height.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
When a Gen7 SURFACE_STATE is configured for MSAA, a number of
additional constaints come in to play. This patch adds a function
gen7_check_surface_setup() which verifies that all of those
constraints are met.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Starting in Gen7, there are two possible layouts for MSAA surfaces:
- Interleaved, in which additional samples are accommodated by scaling
up the width and height of the surface. This is the only layout
available in Gen6. On Gen7 it is used for depth and stencil
surfaces only.
- Sliced, in which the surface is stored as a 2D array, with array
slice n containing all pixel data for sample n. On Gen7 this layout
is used for color surfaces.
The "Sliced" layout has an additional requirement: it must be used in
ARYSPC_LOD0 mode, which means that the surface doesn't leave any extra
room between array slices for miplevels other than 0.
This patch modifies the surface allocation functions to use the
correct layout when allocating MSAA surfaces in Gen7, and to set the
array offsets properly when using ARYSPC_LOD0 mode. It also modifies
the code that populates SURFACE_STATE structures to ensure that
ARYSPC_LOD0 mode is selected in the appropriate circumstances.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Gen7 support for blorp (blits using the render bath) now works for
non-MSAA purposes. This patch enables it.
Since blorp operations re-use the logic for HiZ ops, this required
adding a case to the switch statement in gen7_blorp_emit_wm_config(),
to allow for the case where no HiZ op is being performed.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
On Gen6, texel fetch is always accomplished using the SAMPLE_LD
message, which accepts arguments (u, v, r, lod, si). On Gen7, there
are two* texel fetch messages: SAMPLE_LD for non-MSAA surfaces, taking
arguments (u, lod, v), and SAMPLE_LD2DSS for MSAA surfaces, taking
arguments (si, u, v).
*Technically, there are other texel fetch messages, but they are used
for "compressed" MSAA surfaces, which we don't yet support.
This patch adds the proper message types and argument orderings for
Gen7.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Gen7 hardware requires us to enable at least one WM dispatch mode,
even if there is no program being dispatched to. When this code was
only used for HiZ operations (which don't use a WM program), we used
32-pixel dispatch, because it didn't matter. But blit programs are
compiled for 16-pixel dispatch. So just enable 16-wide dispatch
unconditionally.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2: Enable 16-wide dispatch unconditionally rather than add the
unnecessary complication of using 32-wide dispatch when there is no WM
program.
On Gen7, push constants for shader programs are stored in the URB, so
blorp code needs to set aside space for them. This was previously
unnecessary because blorp code was based on HiZ operations, which
don't require any shaders.
This patch adds a call from gen7_blorp_exec() to
gen7_allocate_push_constants(), to ensure that push constants are
assigned the correct location in the URB. It also extracts a new
function gen7_emit_urb_state() from gen7_upload_urb(), which is
re-used by gen7_blorp_emit_urb_config() to ensure that the URB regions
used by all the pipeline stages leave room for the push constants.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We know from previous bug fixes (commits
c25e5300cb and
b2ace06cbb) that texture border color
doesn't work if the dynamic state upper bound is set to 0. Although
the blorp engine doesn't make use of texture borders, it seems like we
ought to err on the safe side and set this value properly.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch separates out the portions of gen6_blorp_emit_batch_head()
that emit 3DSTATE_MULTISAMPLE, 3DSTATE_SAMPLE_MASK, and
STATE_BASE_ADDRESS. This paves the way for making the blorp code work
on Gen7, where additional command packets
(3DSTATE_PUSH_CONSTANT_ALLOC_VS and 3DSTATE_PUSH_CONSTANT_ALLOC_PS)
need to be emitted before 3DSTATE_MULTISAMPLE.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch modifies the "blorp" WM program so that it can be run in
MSDISPMODE_PERSAMPLE (which means that every single sample of a
multisampled render target is dispatched to the WM program, not just
every pixel).
Previously we were using the ugly hack of configuring multisampled
destination surfaces as single-sampled, and generating sample indices
other than zero by swizzling the pixel coordinates in the WM program.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This patch modifies the function brw_blorp_blit_program::texel_fetch()
to emit the SI (sample index) argument to the SAMPLE_LD message when
reading from a sample index other than zero.
Previously we were using the ugly hack of configuring multisampled
source surfaces as single-sampled, and accessing sample indices other
than zero by swizzling the texture coordinates in the WM program.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch generalizes the function
brw_blorp_blit_program::texture_lookup() so that it prepares the
arguments to the sampler message based on a caller-provided array
rather than assuming the argument order is always (u, v).
This paves the way for the messages we will need to use in Gen7, which
use argument orders (u, lod, v) and (si, u, v) (si=sample index).
It will also will allow us to read from arbitrary sample indices on
Gen6, by supplying the arguments (u, v, r, lod, si) to the SAMPLE_LD
message instead of just (u, v).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Gen6 MSAA buffers (and Gen7 MSAA depth/stencil buffers) interleave
MSAA samples in a complex pattern that repeats every 2x2 pixel block.
Therefore, when allocating an MSAA buffer, we need to make sure to
allocate an integer number of 2x2 blocks; if we don't, then some of
the samples in the last row and column will be cut off.
Fixes piglit tests "EXT_framebuffer_multisample/unaligned-blit {2,4}
color msaa" on i965/Gen6.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Without passing the -ldflags parameter before $(LDFLAGS) in some cases
flags will be passed to MKLIB which it does not understand.
This might be -m64, -m32 or similar.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Thomas Gstädtner <thomas@gstaedtner.net>
Signed-off-by: Brian Paul <brianp@vmware.com>
This patch gets the FreeBSD SCons build working again. The build still
fails though.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
We need to return immediately after inserting instructions that require
S_WAITCNT so that the parent class' custom inserter won't try to insert
them again.
Fix uninitialized scalar variable defects report by Coverity.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
We should just set the bits of functionality that we support; the
GL/ES1/ES2 flags in extensions.c will take care of advertising the
appropriate extensions for the current API.
This enables the GL_EXT_texture_compression_dxt1 extension on ES1/ES2
when libtxc_dxtn is installed or the force_s3tc driconf option is set.
The main extension code set this up properly, but the ES-specific code
failed to do so.
Otherwise, the extension strings reported by es1_info, es2_info, and
glxinfo all remain the same.
This patch manually disables the ARB_framebuffer_object bit on ES
to preserve the behavior of 1c0f5d8324.
v2: Rebase, fix the i915 Makefile, and unconditionally set the
OES_draw_texture bit as core Mesa will only apply it to ES1 now.
Tested-by: Daniel Charles <daniel.charles@intel.com> [v1]
Reviewed-by: Chad Versace <chad.versace@linux.intel.com> [v1]
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
If the primitive restart index and the primitive type can
be handled by the cut index feature, then use the hardware
to handle the primitive restart feature.
The VBO module's software handling of primitive restart is
used as a fall back.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
For newer hardware we disable the VBO module's software handling
of primitive restart. We now handle primitive restarts in
brw_handle_primitive_restart.
The initial version of brw_handle_primitive_restart simply calls
vbo_sw_primitive_restart, and therefore still uses the VBO
module software primitive restart support.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
When considering which components of a variable were killed by an
assignment, constant propagation would previously just use the write
mask of the assignment. This worked if the LHS of the assignment was
simple, e.g.:
v.xy = ...; // (assign (xy) (var_ref v) ...)
But it did the wrong thing if the LHS of the assignment involved an
array indexing operator, since in this case the write mask is always
(x):
v[i] = ...; // (assign (x) (deref_array (var_ref v) (var_ref i)) ...)
In general, we can't predict which vector component will be selected
by array indexing, so the only safe thing to do in this case is to
kill the entire variable.
Fixes piglit tests {fs,vs}-vector-indexing-kills-all-channels.shader_test.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Now that the linker handles initializers of samplers just like any
other uniform, a bunch of this annoying code is unnecessary.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The linker may have set initial values for uniforms. Propagate these
values to the driver's backing storage when it is first associated.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2: Fix handling of arrays-of-structure. Thanks to Eric Anholt for
pointing this out.
v3: Minor comment change based on feedback from Ken.
Fixes piglit glsl-1.20/execution/uniform-initializer/fs-structure-array
and glsl-1.20/execution/uniform-initializer/vs-structure-array.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2: Add support for gen6, and don't turn it on if blending is
disabled. (fixes GPU hang), and note it in docs/GL3.txt
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The i965 driver needed this as well for hardware setup, so instead of
duplicating the logic, just save it off.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Dave Airlie <airlied@redhat.com>
While it doesn't have the same warning in the simulator as in gen7,
let's emit it out of paranoia. We wouldn't want our resolves of some
previous clear to get clamped to some current clamping value.
Suggested-by: pretty much everyone
When doing fast clears, a fulsim warning said that the batch was being
emitted without the viewport set up. While the fast clear pass I was
looking at doesn't use the clear value, the later resolves which also
didn't set up the vieport would trigger the same. It's not obvious
from the error message whether it meant "fast clear value gets clamped
to something you haven't defined" or "fast clear value doesn't get
clamped, and I saw it was out of the current (uninitialized) range,
and you probably wanted it clamped to that (uninitialized) range". Be
paranoid and assume the first case.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Having this enum separate caused us to need a bunch of helper
functions to translate to the op to be executed.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
The GLSL clear path doesn't need any buffer presence checks, since
those are already handled in the normal drawing path code.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Our understanding is that the 3D engine is supposed to be faster
anyway. We used to have more overhead in our tri clear path than we
do today, which would have led to this choice. But given that we
almost always see a depth clear along with a color clear, the path was
hardly exercised anyway.
Also, the color mask logic was broken in the presence of
GL_EXT_draw_buffers2's per-buffer colormask.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Previously, when the environment variable INTEL_DEBUG=aub was set,
mesa would simply instruct DRM to start dumping data to an .aub file,
but we would not provide DRM with any information about the format of
the data in various buffers. As a result, a lot of the data in the
generate .aub file would be unannotated, making further data analysis
difficult.
This patch causes the entire contents of each batch buffer to be
annotated using the data in brw->state_batch_list (which was
previously used only to annotate the output of INTEL_DEBUG=bat). This
includes data that was allocated by brw_state_batch, such as binding
tables, surface and sampler states, depth/stencil state, and so on.
The new annotation mechanism requires DRM version 2.4.34.
Reviewed-by: Eric Anholt <eric@anholt.net>
When we are generating an AUB dump, we make a final call to
aub_dump_bmp() as the context is being destroyed, to ensure that any
rendering performed before the application exits can be seen during a
simulation run. However, we were doing this before flushing the batch
buffer; as a result simulation runs would not always see the effect of
all rendering commands.
This patch flushes the batch buffer just before making the final call
to aub_dump_bmp(), to ensure that all rendering is properly captured
in the final bitmap.
This is a long standing problem, that recently surfaced with the change
to enable perspective correct color interpolation.
A fix for all possible formats is left to the future.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Previously assumed normalised was 0 to 1, but it can be -1 to 1
if type is signed.
Tested with lp_test_conv and lp_test_format, reduced errors.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
Fixing a /*FIXME*/ to remove errors in integer conversion in lp_build_conv.
Tested using lp_test_conv and lp_test_format, reduced errors.
Signed-off-by: José Fonseca <jfonseca@vmware.com>
This patch removes two Clang warnings in GLU:
The first one seems to be an actual bug in mapdesc.cc: Clang complains
that sizeof(dest) will return the size of REAL*[MAXCOORDS], instead of
the intended REAL[MAXCOORDS][MAXCOORDS]. The second one is just
cosmetic because Clang doesn't like extra parentheses.
NOTE: This is a candidate for the 8.0 branch
Reviewed-by: Brian Paul <brianp@vmware.com>
Fixes another case of sampler views being created by one context,
shared by another, then deleted by the first, leaving a dangling
pipe context pointer.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Use it where performance matters more and the exact method of float->int
conversion/rounding isn't terribly important. There should no net change
here since F_TO_I() is the new name of the old IROUND() function.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
The different implementations of IROUND() behaved differently and in
the case of fistp, depended on the current x86 FPU rounding mode.
This caused some tests like piglit roundmode-pixelstore and
roundmode-getintegerv to fail on 32-bit x86 but pass on 64-bit x86.
Now IROUND() always rounds to the nearest integer (away from zero).
The new F_TO_I function converts a float to an int by whatever means
is fastest. We'll use this where we're more concerned with performance
and not too worried to how the conversion is done.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
The IROUND converted all arguments to 0 or 1. That's not what we wanted.
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
For zero-stride vertex arrays, the svga driver copies the value into
the constant value and uses that value in the shader. The recent
gallium-userbuf changes caused a regression in this. An example
symptom was per-primitive glColor3f() calls getting ignored.
Where we copied the vertex value from the vertex buffer to the
constant buffer we neglected to take into account the
pipe_vertex_buffer::buffer_offset field. Adding that value to the
source offset fixes the problem. Actually, it looks like we should
have been doing this all along, but it never was an issue before for
some reason.
If the MESA_GLSL env var contains "errors", GLSL compilation and
link errors will be reported to stderr.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Fix uninitialized scalar variable defect reported by Coverity.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Piglits test for fragment shaders pass, vertex shaders fail. The
actual failure seems to be in the interpolators, and not the
textureSize query.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: José Fonseca <jose.r.fonseca@gmail.com>
Fixes a bunch of piglit tests related to flat interpolation of floats.
Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: José Fonseca <jose.r.fonseca@gmail.com>
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.
