Not (currently) state that is overwridden by u_blitter itself, but
drivers with custom blit/clear which are reusing part of the u_blitter
infrastructure will use it.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Apply the median and matrix filter before the compostioning
we apply the deinterlacing first to avoid the extra overhead
in processing the past and the future surfaces in deinterlacing.
v2: apply the filters on all the surfaces (Christian)
v3: use get_sampler_view_planes() instead of
get_sampler_view_components() and iterate over
VL_MAX_SURFACES (Christian)
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
These are the same for a3xx and later. (a2xx could probably use them
too, but due to limited hw support and ancient downstream kernels, it
isn't so easy to test.)
Signed-off-by: Rob Clark <robdclark@gmail.com>
For various tex fetch instructions, coord's get fixed up in different
ways. But modifying the array returned from get_src() has side-effects
if the same SSA src is used again.. the later instruction will see the
previous fixups.
Fix this, and const'ify things to prevent this sort of mistake in the
future.
Noticed by Varad when adding support for txf_ms.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Separated FE stats out into its own structure. There are 17 FE vs 3 BE
stat fields. Since there is only one FE thread per DC then we don't have
to loop over all threads and sum up FE stats over all the worker threads.
This also reduces size of DC since we only need to store one copy of the
FE stats and not one per worker. Finally, we can use the new FE callback
mechanism to update these.
Signed-off-by: Tim Rowley <timothy.o.rowley@intel.com>
- Remove HYPERTHREADED_FE support
- Add threading info as optional data passed to SwrCreateContext.
If supplied this data will override any KNOB thread settings.
Signed-off-by: Tim Rowley <timothy.o.rowley@intel.com>
This is a blocking call that waits until all FE work is complete.
This is useful for waiting for FE work to complete such as for streamout.
Signed-off-by: Tim Rowley <timothy.o.rowley@intel.com>
enable/disable raster tile trivial accept test based on scissor enable trait.
Can be optimized further.
Signed-off-by: Tim Rowley <timothy.o.rowley@intel.com>
1. SoWriteOffset is no longer treated as a stat
2. Added callback from core to update streamout write offset
Signed-off-by: Tim Rowley <timothy.o.rowley@intel.com>
This is the case when the "00 00 03" is very close to the beginning of
nal unit header
v2: move the check to rbsp init
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This can help enable some blend optimizations (see the register spec).
Vulkan always sets this.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
This is the proper fix for Overlord and Witcher 2 hangs.
The hang condition is that 1 app must write to MRT0 and MRT1 from a pixel
shader while MRT1 is disabled in CB_TARGET_MASK (does this generate
unflushable pixel quads? I don't know), and another app (e.g. Glamor)
must enable dual source blending in both MRT0 and MRT1. The hw gets
confused, which leads to corruption and hangs.
Cc: 12.0 11.2 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Since commit 6d7177f01b, radeonsi
would take a different path if info->indirect_params was not
initialized properly. Nine was not initializating this field.
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
+23% Bioshock Infinite performance.
v2: - use the new fence_finish interface
- allow deferred fences with multiple contexts
- clear the ctx pointer after a deferred flush
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
required by glClientWaitSync (GL 4.5 Core spec) that can optionally flush
the context
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Some ideas copied from Jakob Sinclair's implementation, but the color
clearing is completely different.
v2: remove leftover code, disable conditional rendering
disable render condition cleanly
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
We need this for 'inline'.
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Tested-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
As requested with the initial creation of util/bitscan.h
now move other bitscan related functions into util.
v2: Split into two patches.
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Tested-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
This enables GL_shader_draw_parameters and GL_ARB_indirect_parameters as well
as a properly accelerated implementation of GL_ARB_multi_draw_indirect.
Enabling the feature requires a sufficiently uptodate firmware -- those have
already been released a long time ago, although this does mean that the
feature only works with the amdgpu kernel module, since the radeon module
doesn't have a way to query the firmware version.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Note that for indirect draws, the new MULTI firmware packets are required.
There's also no need to reset last_{start_instance,sh_base_reg}, since
resetting last_base_vertex is sufficient.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
