This patch converts the SSE-optimized build_mask_32() and
build_mask_linear_32() to VMX/VSX.
I measured the results on POWER8 machine with 32 cores at 3.4GHz and
16GB of RAM.
FPS/Score
Name Before After Delta
------------------------------------------------
glmark2 (score) 139.8 142.7 2.07%
openarena and xonotic didn't show a significant (more than 1%)
difference.
v2: Make sure code is build only on POWER8 LE machine
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
This patch converts the SSE optimization done in do_triangle_ccw to
VMX/VSX.
I measured the results on POWER8 machine with 32 cores at 3.4GHz and
16GB of RAM.
FPS/Score
Name Before After Delta
------------------------------------------------
glmark2 (score) 136.6 139.8 2.34%
openarena 16.14 16.35 1.30%
xonotic 4.655 4.707 1.11%
v2:
- Convert loads to use aligned loads
- Make sure code is build only on POWER8 LE machine
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
This file provides a portability layer that will make it easier to convert
SSE-based functions to VMX/VSX-based functions.
All the functions implemented in this file are prefixed using "vec_".
Therefore, when converting from SSE-based function, one needs to simply
replace the "_mm_" prefix of the SSE function being called to "vec_".
Having said that, not all functions could be converted as such, due to the
differences between the architectures. So, when doing such
conversion hurt the performance, I preferred to implement a more ad-hoc
solution. For example, converting the _mm_shuffle_epi32 needed to be done
using ad-hoc masks instead of a generic function.
All the functions in this file support both little-endian and big-endian
but currently the file is build only on POWER8 LE machine.
All of the functions are implemented using the Altivec/VMX intrinsics,
except one where I needed to use inline assembly (due to missing
intrinsic).
v2:
- Use vec_vgbbd instead of __builtin_vec_vgbbd
- Add an aligned load function
- Don't use typeof()
- Make file build only on POWER8 LE machine
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
We only want to set the SVGA_NEW_STIPPLE dirty flag when the polygon
stipple state changes. Before, we only set the flag when we were
enabling stipple, but not disabling.
We don't really have to add SVGA_NEW_STIPPLE to the dirty FS state
set since it's a subset of SVGA_NEW_RAST, but let's be explicit.
This doesn't fix any known bugs.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
and svga_set_sampler_views(). If there's no change, return early
and don't set a SVGA_NEW_x dirty state flag.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
It allows to call nouveau_vp3_bsp_next multiple times
between one begin/end.
It is required to support st/va.
https://bugs.freedesktop.org/show_bug.cgi?id=89969
Signed-off-by: Julien Isorce <j.isorce@samsung.com>
[imirkin: create strparm_bsp function, simplified w0 calculation]
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
The counter was not set but used by the nouveau driver.
It is required otherwise visual output is garbage.
Signed-off-by: Julien Isorce <j.isorce@samsung.com>
Reviewed-by: Christian Koenig <christian.koenig@amd.com>
Experimentally, 4M causes corruption and slowness, try to ramp it up
with size instead.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0 11.1" <mesa-stable@lists.freedesktop.org>
H264 doesn't have a bitplane bo. We just need a device reference, so use
the one from the client.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0 11.1" <mesa-stable@lists.freedesktop.org>
It's the same behavior that we use for later LLVM.
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Found during NIR_TEST_CLONE=1 piglit run. We were using block->index
but forgetting to require it. Causing things to not work with a cloned
shader which didn't preserve block_index.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Immediately convert into NIR and do an initial key-agnostic lowering/
optimization pass. This should let us share most of the per-variant
transformations between each variant, and hopefully minimize the draw-
time variant creation part of the compilation process.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It will still hit a compile_assert() in emit_tex, which has the
advantage of dumping out the offending shader.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Nested functions are supported as an extension in GNU C, but Clang
don't support them.
This fixes compilation errors when (manually) building compute.c,
or by setting --enable-gallium-tests to the configure script.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75165
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Instead of iterating over all the buffer resources looking for coherent
buffers, we keep track of a context-wide count. This will save some
iterations (and CPU cycles) in 99.99% case because usually coherent
buffers are not so used.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Diagnostics sent during code generation and the every error message reported
by LLVMTargetMachineEmitToMemoryBuffer are disjoint reporting mechanisms. We
take care of both and also send an explicit message indicating failure at the
end, so that log parsers can more easily tell the boundary between shader
compiles.
Removed an fprintf that could never be triggered.
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This will allow us to send shader debug info via the context's debug callback.
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com> (v1)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The output via stderr is very helpful for ad-hoc debugging tasks, so that remains
unchanged, but having the information available via debug messages as well
will allow the use of parallel shader-db runs.
Shader stats are always provided (if the context is a debug context, that is),
but you still have to enable the appropriate R600_DEBUG flags to get
disassembly (since it is rather spammy and is only generated by LLVM when we
explicitly ask for it).
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This will allow us to send shader debug info.
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com> (v1)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The fixed alignment of u_upload_mgr will go away.
This is the first step.
The motivation is that one u_upload_mgr can have multiple users,
each allocating from the same buffer, but requiring a different alignment.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
The function only aligned the size, but not the offset.
The offset was aligned only when the previous suballocation was aligned.
That yielded the correct offset alignment if the alignment was constant
for all suballocations.
Instead, directly align the offset, but allow an unaligned size.
There is no change in behavior, because the alignment is constant
at the moment.
This a prerequisite for allowing a variable alignment for suballocations.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
First off, we can't flush in the middle of a command. Secondly
requesting the extra push space might cause a flush to happen. If that
flush happens, we'd have to do the PUSH_REFN again. So instead do
PUSH_REFN after the push space request. This helps avoid rare crashes
with supertuxkart in libdrm due to assertion failures.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0 11.1" <mesa-stable@lists.freedesktop.org>
Quads need to respect winding order, too - not just triangles.
Fixes rendering in GFXBench 4.0's tessellation benchmark.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.0 11.1" <mesa-stable@lists.freedesktop.org>
This will allow the state tracker to inform the driver where in a
broken-up multidraw we currently are. This can then be passed into the
vertex shader.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
