RB3D_DEBUG_CTL doesn't help, so I resolve to a tiled temporary texture and
then blitting it to the destination one, which we also do in other situations.
The handling of the CAP is broken in st/mesa anyway. Let's just kill it.
This commit pretty much enables fast Z clear for FBOs with Z24S8.
The driver falls back to clearing with a quad if the fast clear cannot be
used. It can still do fast color clear, for example.
Effectively this path would always assert. Move the break statement to
the (probable) intended place.
Note: This is a candidate for the stable branches.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Include LLVM_LDFLAGS when building with LLVM. Fixes the following build
errors:
CXXLD swrast_dri.la
/usr/bin/ld: cannot find -lLLVMR600CodeGen
/usr/bin/ld: cannot find -lLLVMR600Desc
/usr/bin/ld: cannot find -lLLVMR600Info
/usr/bin/ld: cannot find -lLLVMR600AsmPrinter
Reviewed-by: Andreas Boll <andreas.boll.dev@gmail.com>
v2: Andreas Boll <andreas.boll.dev@gmail.com>
- don't remove compatibility with scripts for the old build system
v3: Andreas Boll <andreas.boll.dev@gmail.com>
- remove more obsolete hacks
v4: Andreas Boll <andreas.boll.dev@gmail.com>
- add a previously removed TOP variable to fix vgapi build
According to bug #54524, I regressed oglconform's multicontext test
when I reenabled the fragment shader precompile.
However, these test cases only passed by miraculous coincedence. We
assign each fragment program a unique ID (brw_fragment_program::id which
becomes brw_wm_prog_key::program_string_id) which we obtain by storing a
per-context counter.
The test case uses GLX context sharing to access the same fragment
program from two different contexts. This means that we share a program
cache. Before the precompile, if both contexts happened to use the same
shaders in the same order, we'd obtain the same program_string_ids (by
virtue of doing the same computation twice). However, the more likely
scenario is that they completely disagree on program_string_id.
This meant that we'd have two completely different fragment shaders in
the cache with the same ID, tricking us to think they were the same
(aside from NOS), so we'd render using the wrong program.
This patch implements a simple fix suggested by Eric: it moves the
global counter out of brw_context and into intel_screen, which is shared
across all contexts. A mutex protects it from concurrent access.
This is also the first direct usage of pthreads in the i965 driver.
Fixes 10 subcases of oglconform's multicontext test.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54524
Reviewed-by: Eric Anholt <eric@anholt.net>
Technically, variable sized arrays are a required feature of C99,
redacted to be optional in C11, and not actually part of C++ whatsoever.
Gcc allows using them in C++ unless you specify -pedantic, and Clang
appears to allow them for simple/POD types.
exec_list is arguably POD, since it doesn't have virtual methods, but I
can see why Clang would be like "meh, it's a C++ struct, say no", seeing as
it's meant to support C99.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=58970
Reviewed-by: Matt Turner <mattst88@gmail.com>
The simulator gets very angry about our i2b code:
cmp.ne(16) g3<1>D g2<0,1,0>D 0F
We can't mix integer DWord and float types. The only reason to use 0F
here was to share code with f2b. Split it and use 0D instead.
While we don't believe anything bad will actually happen because of
this, it's nice to fix the warnings and easy enough to do.
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Often when debugging, I don't want to see SIMD16 shaders. It makes
INTEL_DEBUG=vs/fs output much easier to read, especially when a program
dumps many shaders. Plus, I also want to verify that SIMD8 works before
even considering SIMD16.
v2: Fix the likeliness check (caught by Chris and Eric).
Reviewed-by: Eric Anholt <eric@anholt.net>
Choose MESA_FORMAT_ARGB2101010 when storing
GL_RGBA + GL_UNSIGNED_INT_2_10_10_10_REV or
GL_RGB + GL_UNSIGNED_INT_2_10_10_10_REV.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Note: This is a candidate for the stable branches.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The ARB_get_program_binary spec says "OpenGL 3.0 is required." The
nearly identical OES_get_program_binary extension is available for
OpenGL ES 2.0, so I don't see how / why OpenGL 3.0 is a requirement for
the ARB version. Let's just enable whenever GL_ARB_shader_objects is
available.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
After recent changes in the XML, the dispatch generators will expect
this function to be named _mesa_ProgramParameteri.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
There were two bugs here. First, this and several other queries were
not available in a desktop GL context with GL_ARB_ES2_compatibility.
Second, GL_NUM_SHADER_BINARY_FORMATS returns zero, but
GL_SHADER_BINARY_FORMATS writes one element of data to the buffer. If
NUM is zero, no data should be written.
Fixes piglit test 'arb_get_program_binary-overrun shader'.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This adds TBO support to r600g, and with GLSL 1.40 enabled,
we now get 3.1 core profiles advertised for r600g.
The r600/700 implementation is a bit different from the evergreen one,
as r6/7 hw lacks vertex fetch swizzles. So we implement it by passing 5
constants per sampler to the shader, the shader uses the first 4 as masks
for each component and the 5th as the alpha value to OR in.
Now TXQ is also broken so we have to pass a constant for the buffer size,
on evergreen we just pass this, on r6/7 we pass it as the 6th element
in the const info buffer.
v1.1: drop return as DDX doesn't use a texture type
v2: add r600/700 support.
Signed-off-by: Dave Airlie <airlied@redhat.com>
This adds 12 more constant buffers for use as UBOs,
along with adding relative constant fetching for 2D indices.
This with GLSL 1.40 enabled passes all the same tests as softpipe
on my evergreen system.
Signed-off-by: Dave Airlie <airlied@redhat.com>
First we test that line continuations are honored within a comment, (as
recently changed in glcpp), then we test that line continuations can be
disabled via an option within the context. This is tested via the new support
for a test-specific command-line option passed to glcpp.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Previously, we were only supporting line-continuation backslash characters
within lines of pre-processor directives, (as per the specification). With
OpenGL 4.2 and GLES3, line continuations are now supported anywhere within a
shader.
While changing this, also fix a bug where the preprocessor was ignoring
line continuation characters when a line ended in multiple backslash
characters.
The new code is also more efficient than the old. Previously, we would
perform a ralloc copy at each newline. We now perform copies only at each
occurrence of a line-continuation.
This commit fixes the line-continuation.vert test in piglit.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This will allow testing of disabled line-continuation on a case-by-case basis,
(with the option communicated to the preprocessor via the GL context).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This will allow the test exercising disabled line continuations to arrange
for the --disable-line-continuations argument to be passed to the standalone
glcpp.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>