On gen >= 8 one doesn't provide ending address but number of bytes
available. This is relative to the given offset.
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Also change the interface to use start and end offsets.
Signed-off-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Devices with smaller GMEM size need more tiles. On db410c at 2048x1152,
glmark2 shadow needed ~330 tiles for fullscreen. Lets bump it up to
512. (Maybe with MRT you could end up needing more, but at that point
things are probably going to be painfully slow.)
Signed-off-by: Rob Clark <robdclark@gmail.com>
Some games are sloppy.. perhaps because it is defined behavior for DX or
perhaps because nv blob driver defaults things to zero.
So add driconf param to force uninitialized variables to default to zero.
This issue was observed with rust, from steam store. But has surfaced
elsewhere in the past.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
For .vert/.frag, now multiple can be specified on the cmdline for
purposes of linking, and the last one specified is the one that is
fed into the ir3 backend (and dumped along the way if --verbose is
specified)
Without this, varyings in frag shaders would appear as undefined.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Rather than doing a separate submit at context create, move these cmds
to before first tile, as is done on a3xx/a4xx. Otherwise state can
be overwritten by other contexts.
Signed-off-by: Rob Clark <robdclark@gmail.com>
gcc6 does not like the trick where we point to one entry before the
array start and then start a while with a pre-increment.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
These are all new false positives with gcc6.
In nouveau_compiler.c: gcc6 no longer assumes that passing a pointer
to a variable into a function initialises that variable.
In nv50_ir_from_tgsi.cpp op and mode are not set if there are 0
enabled dst channels, this never happens, but gcc cannot know this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Add support for SV_WORK_DIM for nvc0 and nve4.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This brings it inline with the other macros like NVC0_CB_AUX_UBO_INFO
and NVC0_CB_AUX_TEX_INFO.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
In order to implement get_work_dim() the driver may need to know the
clEnqueueNDRangeKernel() work_dim parameter, so pass it to the driver.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Add a new WORK_DIM SV type, this is will return the grid dimensions
(1-4) for compute (opencl) kernels.
This is necessary to implement the opencl get_work_dim() function.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
For the case (both src or dst) where we had a texobject, but the
texobject target was not the same that the method target, this spec
paragraph was appplied:
/* Section 18.3.2 (Copying Between Images) of the OpenGL 4.5 Core
* Profile spec says:
*
* "An INVALID_VALUE error is generated if either name does not
* correspond to a valid renderbuffer or texture object according
* to the corresponding target parameter."
*/
But for that case, the correct spec paragraph should be:
/* Section 18.3.2 (Copying Between Images) of the OpenGL 4.5 Core
* Profile spec says:
*
* "An INVALID_ENUM error is generated if either target is
* not RENDERBUFFER or a valid non-proxy texture target;
* is TEXTURE_BUFFER or one of the cubemap face selectors
* described in table 8.18; or if the target does not
* match the type of the object."
*/
specifically the last sentence: "or if the target does not match the
type of the object".
This patch fixes the error returned (s/INVALID/ENUM) for that case,
and moves up the INVALID_VALUE spec paragraph, as that case (invalid
texture object) was handled before.
Fixes:
GL44-CTS.copy_image.target_miss_match
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Immediates are stored into a separate table, and are
consolidated, so if we get an immediate we don't need
to offset it as the index it has is correct.
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Basically we just have to scale up the coordinates and then add the
relevant sample offset. The code to handle this was already largely
present from Christoph's earlier attempts to pipe images through back in
the dark ages, this just hooks it all up.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Fixes a regression introduced by commit 42624ea83 which triggered
an assertion in
dEQP-GLES2.functional.texture.completeness.cube.not_positive_level_0
While stImage must have a non-zero size as verified by the caller, we also
look at the size of the base image in an attempt to make a better guess at
the level0 size (this is important when the base image size is odd). However,
the base image may have a zero size even when it exists.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96629
Cc: 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
use bicubic filtering as high quality scaling L1.
v2: fix a typo and add a newline to code
v3: -render the unscaled image on a temporary surface (Christian)
-apply noise reduction and sharpness filter on
unscaled surface
-render the final scaled surface using bicubic
interpolation
v4: support high quality scaling
v5: set dst_area and dst_clip in bicubic filter
v6: set buffer layer before setting dst_area
v6.1: add PIPE_BIND_LINEAR when creating resource
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This is a shader based bicubic interpolater which uses cubic
Hermite spline algorithm.
v2: set dst_area and dst_clip during scaling (Christian)
v3: clear the render target before rendering
v4: intialize offsets while initializing shaders
use a constant buffer to send dst_size to frag shader
small changes to reduce calculation in shader
v5: send half pixel offset instead of sending dst_size
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Fix this build error with GCC 4.4.
CC state_tracker/st_nir_lower_builtin.lo
In file included from state_tracker/st_nir_lower_builtin.c:61:
state_tracker/st_nir.h:34: error: redefinition of typedef ‘nir_shader’
../../src/compiler/nir/nir.h:1830: note: previous declaration of ‘nir_shader’ was here
Suggested-by: Rob Clark <robdclark@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96235
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>
silent, flush, incomplete_tex and incomplete_fbo flags were not
documented (see src/mesa/main.debug.c for more info).
FP is not checked anymore.
v2 (Brian Paul):
* MESA_DEBUG accepts a comma-separated list of parameters.
* Clarify how MESA_DEBUG behaves with mesa debug and release builds.
* Updated wording.
v3: Better wording for one paragraph (Brian Paul)
Reviewed-by: Brian Paul <brianp@vmware.com>
Fixes:
GL44-CTS.texture_barrier_ARB.same-texel-rw-multipass
On Haswell, Broadwell and Skylake (note that in order to execute that
test, it is needed to override GL and GLSL versions).
On gen6 this test was already working without this change. It keeps
working after it.
This commit replaces the call to brw_emit_mi_flush for gen6+ with two
calls to brw_emit_pipe_control_flush:
* The first one with RENDER_TARGET_FLUSH and CS_STALL set to initiate
a render cache flush after any concurrent rendering completes and
cause the CS to stop parsing commands until the render cache
becomes coherent with memory.
* The second one have TEXTURE_CACHE_INVALIDATE set (and no CS stall)
to clean up any stale data from the sampler caches before rendering
continues.
Didn't touch gen4-5, basically because I don't have a way to test
them.
More info on commits:
0aa4f99f5672473658c5
Thanks to Curro to help to tracking this down, as the root case was a
hw race condition.
v2: use two calls to pipe_control_flush instead of a combination of
gen7_emit_cs_stall_flush and brw_emit_mi_flush calls (Curro)
v3: no need to const cache invalidation (Curro)
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
The output of draw requires a null viewport transform, which the regular
code is ill-equiped to do. Reinstate the original settings in the render
path, and add setting of the viewport clip polygon based on fb
width/height (as that is all taken care of by draw).
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
This aligns the 4-element color float array to 16 byte boundaries. This
should allow compiler vectorizers to generate better optimizations.
Also fixes broken vectorization generated by Intel compiler.
v2: Fixed indentation and added a lengthy comment explaining the
reason for the alignment.
Cc: <mesa-stable@lists.freedesktop.org>
Reported-by: Tim Rowley <timothy.o.rowley@intel.com>
Tested-by: Tim Rowley <timothy.o.rowley@intel.com>
Signed-off-by: Chuck Atkins <chuck.atkins@kitware.com>
Acked-by: Roland Scheidegger <sroland@vmware.com>
Encapsulate the test for which flags are needed to get a compiler to
support certain features. Along with this, give various options to try
for AVX and AVX2 support. Ideally we want to use specific instruction
set feature flags, like -mavx2 for instance instead of -march=haswell,
but the flags required for certain compilers are different. This
allows, for AVX2 for instance, GCC to use -mavx2 -mfma -mbmi2 -mf16c
while the Intel compiler which doesn't support those flags can fall
back to using -march=core-avx2.
This addresses a bug where the Intel compiler will silently ignore the
AVX2 instruction feature flags and then potentially fail to build.
v2: Pass preprocessor-check argument as true-state instead of
false-state for clarity.
v3: Reduce AVX2 define test to just __AVX2__. Additional defines suchas
__FMA__, __BMI2__, and __F16C__ appear to be inconsistently defined
w.r.t thier availability.
v4: Fix C++11 flags being added globally and add more logic to
swr_require_cxx_feature_flags
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
Tested-by: Tim Rowley <timothy.o.rowley@Intel.com>
Signed-off-by: Chuck Atkins <chuck.atkins@kitware.com>
The linker deals with atomic counters in terms of uniforms but the
data structure are called after the atomic counters.
Renamed the data structures used in the linker for disambiguation.
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
Currently the linker uses the uniform count for the total number of
atomic counters. However uniforms don't include the innermost array
dimension in their count, but atomic counters are expected to include
them.
Although the spec doesn't directly state this, it's clear how offsets
will be assigned for arrays.
From OpenGL 4.2 (Core Profile), page 98:
" * Arrays of type atomic_uint are stored in memory by element
order, with array element member zero at the lowest offset. The
difference in offsets between each pair of elements in the
array in basic machine units is referred to as the array
stride, and is constant across the entire array. The stride can
be queried by calling GetIntegerv with a pname of
ATOMIC_COUNTER_- ARRAY_STRIDE after a program is linked."
From that it is clear how arrays of atomic counters will interact with
GL_MAX_ATOMIC_COUNTER_BUFFER_SIZE.
For other kinds of uniforms it's also clear that each entry in an
array counts against the relevant limits.
Hence, although inferred, this is the expected behavior.
Fixes GL44-CTS.arrays_of_arrays_gl.AtomicDeclaration
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
So that we do copies host-side rather than in the guest with map/memcpy.
Tested with piglit arb_copy_buffer-subdata-sync test and new
arb_copy_buffer-intra-buffer-copy test.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Acked-by: Roland Scheidegger <sroland@vmware.com>
With host-side buffer copies (via SVGA3D_vgpu10_BufferCopy()) we have
to make sure any pending map-write operations are completed before reading
if the buffer is dirty. Otherwise the ReadbackSubResource operation could
get stale data from the host buffer.
This allows the piglit arb_copy_buffer-subdata-sync test to pass when
we start using the SVGA3D_vgpu10_BufferCopy command.
v2: check the sbuf->dirty flag in the outer conditional, per Charmaine.
Acked-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
We previously could do blits with util_resource_copy_region() when doing
'loose' format checking. Also do blits with util_resource_copy_region()
when the blit src/dst formats (not the underlying resources) exactly
match. Needed for GL_ARB_copy_image.
Acked-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
v2: remove extra svga_define_texture_level() call, per Charmaine.
Acked-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Do texture->texture copies host-side with this command when possible.
Use the previous software fallback otherwise.
Reviewed-by: Brian Paul <brianp@vmware.com>
Acked-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
