When I originally implemented the ARB_copy_image extension, the fast-path
was written in meta using texture views. This path only worked if both
images were uncompressed color images. All of the other cases fell back to
the blitter or, in the worst case, mapping and memcpy on the CPU. Now that
we have the blorp path, it handles all copies ever and the old meta,
blitter, and CPU paths are only used on gen5 and below. The primary reason
why we needed the meta path (apart from having a slow blitter on later
hardware) was to handle multisampling which gen5 and earlier don't support
anyway. Since the blitter is reasonably fast on gen5, we can just delete
the meta path and get rid of all that terrible code.
If we decide that we're ok with just disabling ARB_copy_image on gen5 and
earlier (I personally am), then we could get rid of another 300 lines or so
of semi-hairy code.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
By using emit_miptree_blit which does chunking, this fixes the blitter path
for the case where the image is too tall to blit normally. We also pull it
into intel_blit as intel_miptree_copy. This matches the naming of the
blorp blit and copy functions brw_blorp_blit and brw_blorp_copy.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: "13.0" <mesa-dev@lists.freedesktop.org>
This moves the nir_lower_indirect_derefs() call into
brw_preprocess_nir() so thats is called by both OpenGL and Vulkan
and removes that call to the old GLSL IR pass
lower_variable_index_to_cond_assign()
We want to do this pass in nir to be able to move loop unrolling
to nir.
There is a increase of 1-3 instructions in a small number of shaders,
and 2 Kerbal Space program shaders that increase by 32 instructions.
Shader-db results BDW:
total instructions in shared programs: 8705873 -> 8706194 (0.00%)
instructions in affected programs: 32515 -> 32836 (0.99%)
helped: 3
HURT: 79
total cycles in shared programs: 74618120 -> 74583476 (-0.05%)
cycles in affected programs: 528104 -> 493460 (-6.56%)
helped: 47
HURT: 37
LOST: 2
GAINED: 0
Use llvm provided API based on cpuid rather than our own
manually mantained list of mattr enabling/disabling.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
And declare coded_buf in vlVaContext as "vlVaBuffer *" instead of
"struct vlVaBuffer *".
This fixes several warnings later about assignment from incompatible
pointer type.
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Remove some transitional left overs from the gallium pipe-loader rework
and kill off unneeded AM_CPPFLAGS.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This moves the function to get the LLVM environment variables higher
in the file. It still needs to be below the "--enable-opencl" because
it uses $enable_opencl.
It can be called without condition now as it only throws errors if
openCL is enabled.
v5:
HAVE_MESA_LLVM is only used for gallium. Rename it to HAVE_GALLIUM_LLVM.
In order to only link LLVM when it is needed, HAVE_GALLIUM_LLVM is only
set if "$enable-gallium-llvm" is yes.
Signed-off-by: Tobias Droste <tdroste@gmx.de>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The gop_size in rate control is the budget window for internal rate
control calculation, and shouldn't always equal to idr period. Define
a coefficient to let budget window contains a number of idr period for
proper rate control calculation. Adjust the number of i/p frame remaining
accordingly.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=98005
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
The gop_size in rate control is the budget window for internal rate
control calculation, and shouldn't always equal to idr period. Define
a coefficient to let budget window contains a number of idr period for
proper rate control calculation. Adjust the number of i/p frame remaining
accordingly.
v2: fixed regression issues introduced by previous version
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=98005
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
This fixes:
dEQP-VK.glsl.texture_gather.basic.2d.depth32f.*
Cc: "13.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The active_query count was incorrect for query types that don't require
a begin_query. Removed the unnecessary assert.
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
This has bothered me for about as long as NIR has been around. Why do we
have two different unions for constants? No good reason other than one of
them is a direct port from GLSL IR.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
This is mostly just used during linking however the st uses it
when updating textures.
In order to store gl_program in the CurrentProgram array
rather than gl_shader_program we need to move this field to
the shared gl_shader_program_data struct.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
If app tries to create a fence but there is no rendering to submit, we
need a dummy/no-op submit. Use a string-marker for the purpose.. mostly
since it avoids needing to realize that the packet format changes in
later gen's (so one less place to fixup for a5xx).
Signed-off-by: Rob Clark <robdclark@gmail.com>
Prep-work for next patch, mostly move to tracking last_fence as a
pipe_fence_handle (created now only in fd_gmem_render_tiles()), and a
bit of superficial renaming.
Signed-off-by: Rob Clark <robdclark@gmail.com>
This enables gallium support for EGL_ANDROID_native_fence_sync, for
drivers which support PIPE_CAP_NATIVE_FENCE_FD.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This will be needed for explicit synchronization with devices outside
the gpu, ie. EGL_ANDROID_native_fence_sync.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
With fixes from Chad squashed in, plus fixes for issues that Rafael
found while writing piglit tests.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Chad Versace <chadversary@chromium.org>
Tested-by: Chad Versace <chadversary@chromium.org>
Doesn't work so well when you start having more than one possible
attrib. Prep-work for next patch.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Tested-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Chad Versace <chadversary@chromium.org>
Tested-by: Chad Versace <chadversary@chromium.org>
Reduce the noise in the next patch. For EGL_SYNC_NATIVE_FENCE_ANDROID
the sync condition is conditional on EGL_SYNC_NATIVE_FENCE_FD_ANDROID
attribute.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Chad Versace <chadversary@chromium.org>
Tested-by: Chad Versace <chadversary@chromium.org>
We were previously also verifying that no backing buffers were available
when an array wasn't enabled. This is has no basis in the spec, and it
causes GLupeN64 to fail as a result.
Fixes: c2e146f487 ("mesa: error out in indirect draw when vertex bindings mismatch")
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Noticed in shaders with branching, where we ended up scheduling delay
slots near the start of a block for the uniforms reset setup.
total instructions in shared programs: 93970 -> 93951 (-0.02%)
instructions in affected programs: 3117 -> 3098 (-0.61%)
3DMMES performance +0.423087% +/- 0.133521% (n=9,10)
This helps us get the delay slots between SFU writes and reads filled.
total instructions in shared programs: 94494 -> 93970 (-0.55%)
instructions in affected programs: 59206 -> 58682 (-0.89%)
3DMMES performance +1.89967% +/- 0.157611% (n=10,9)
The latency_between was trying to handle the delay between the coordinate
write ("before") and the corresponding sample read ("after"), but we were
handing in the two instructions swapped.
This meant that we tried to fit things between a tex_s and its *preceding*
tex_result. This made us only interleave normal texture coordinates by
accident, and pessimized UBO reads by pushing the tex_result collection
earlier until there was nothing but it (and then its preceding coordinate
setup) left.
In addition to latency reduction, things end up packing better (probably
due to reduced live ranges of the texture results):
total instructions in shared programs: 98121 -> 94775 (-3.41%)
instructions in affected programs: 91196 -> 87850 (-3.67%)
3DMMES performance +1.15569% +/- 0.124714% (n=8,10)
I'm not sure how I managed to write the SF merge code
(7d8b79f398) without allowing merges with
NOPs. *Everything* we try to merge with will have a NOP on one or the
other side of the instruction, and that's why that commit showed no
benefit.
total instructions in shared programs: 99347 -> 95128 (-4.25%)
instructions in affected programs: 91906 -> 87687 (-4.59%)
3DMMES performance +2.57105% +/- 0.135276% (n=6,8)
This should be a win for most loops, which tend to have uniform control
flow.
More importantly, it exposes important information to live variables: that
the break/continue here means that our jump target may have access to
values that were live on our input. Previously, we were just setting the
exec mask and letting control flow fall through, so an intervening def
between the break and the end of the loop would appear to live variables
as if it screened off the variable, when it didn't actually.
Fixes a regression in glsl-vs-loop-redundant-condition.shader_test when a
perturbing of register allocation caused a live variable to get stomped.
Cc: 13.0 <mesa-stable@lists.freedesktop.org>
This is already supported in genX_state.c, expose the extension string.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
In an attempt to fix 3DSTATE_DEPTH_BUFFER for stencil-only cases, I
accidentally kept setting the SurfaceType to 2D in the stencil-only case
thanks to a copy+paste error.
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
The buffer_size does not take the offset into account. Just add the
offset into the pointer which lines up the structures much better.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
The number has to be less than or equal to the max, not just less than.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
The components count the number of individual values, not the number of
slots.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
