| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
|
| |
The kill_emitted variable was duplicating the functionality of
gl_fragment_program::UsesKill. There's no need for both.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
The previous commit implemented the workaround, cited a bug report
about OilRush, but actually only enabled the workaround for the demos.
Turn it on for OilRush too.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50291
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]> (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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
| |
"count" is a more useful name, since most of the time we're using it for
looping over the variables.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
| |
This was accidentally copy-and-pasted inside.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
integer format
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 <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Signed-off-by: Anuj Phogat <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Signed-off-by: Chad Versace <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Signed-off-by: Chad Versace <[email protected]>
|
|
|
|
|
| |
This generalizes and replaces gbm_bo_create_for_egl_image. gbm_bo_import
will create a gbm_bo from either an EGLImage or a struct wl_buffer.
|
|
|
|
|
|
|
|
|
|
| |
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.
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Tested-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* "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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Tested-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* "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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Tested-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Tested-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Tested-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
| |
Reviewed-by: Chad Versace <[email protected]>
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
| |
We use the new miptree offset to pick out the sub-image when we bind
the EGLImage to a texture.
Signed-off-by: Kristian Høgsberg <[email protected]>
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
| |
|
|
|
|
|
|
|
|
| |
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 <[email protected]>
|
|
|
|
|
|
|
|
| |
This allows revising the dri_interface.h separately from adding driver
support.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|