it doesn't interact with compute shaders in any way
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Fix incorrect swizzling in SIMD16 Transpose_16_16 breaking the
two-channel 16-bpc formats like R16G16_FLOAT.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
The T images are composed of effectively swizzled-around blocks of LT (4x4
utile) images, so we can reduce the t_utile_address() calls by 16x by
calling into the simpler LT loop.
This also adds support for calling down with non-utile-aligned
coordinates, which will be part of lifting the utile alignment requirement
on our callers and avoiding the RMW on non-utile-aligned stores.
Improves 1024x1024 TexSubImage by 2.55014% +/- 1.18584% (n=46)
Improves 1024x1024 GetTexImage by 2.242% +/- 0.880954% (n=32)
They now have less of a dependency on the cpp, and don't have to do a
divide.
Hacking up mesa-demos teximage to do only one subtest and not draw
points, I saw 1024x1024 glTexSubImage2D() improve by 4.86939% +/-
1.40408% (n=30) and glGetTexImage() by 2.18978% +/- 0.140268% (n=5).
If we get up toward 256MB (or whatever the CMA area size is),
VC4_GEM_CREATE will start throwing errors. Even if we don't trigger
that, when we flush the kernel's BO allocation for the CLs or bin
memory may end up throwing an error, at which point our job won't get
rendered at all.
Just flush early (half of maximum CMA size) so that hopefully we never
get to that point.
Fix typo using wrong (uninitialized) build context introduced by
4634cb5921. (This only affects very rare
small packed formats which have a PIPE_SWIZZLE_0 channel, such as
r4a4, which is never used by mesa/st. Nevertheless it broke lp_test_format.)
Using bit replication. This path now resembles something which might make
sense. (The logic was mostly copied from llvmpipe fs backend.)
I am not convinced though it is actually faster than SoA sampling (actually
I'm quite certain it's always a loss with AVX).
With SoA it's just shift/mask/cvt/mul for getting the colors, whereas
there's still roughly 3 shifts, 3 or/and per channel for AoS
(i.e. for SoA it's exactly the same as it would be for a rgba8 format,
whereas the extra effort for AoS is significant). The filtering
might still be faster (albeit with FMA the instruction count gets down
quite a bit there on the SoA float filtering path on new cpus). And those
small unorm formats often don't have an alpha channel (which makes things
worse relatively for AoS path).
(This also fixes a trivial bug in the llvmpipe fs code this was derived
from, albeit it was only relevant for 4-bit channels.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
This code uses a vector shift which has to be emulated on x86 unless
there's AVX2. Luckily in some cases we can actually avoid the shift
altogether, so do that.
Also make sure we hit the fast lp_build_conv() path when applicable,
albeit that's quite the hack...
That said, this path is taken for AoS sampling for small unorm (smaller
than rgba8) formats, and it is completely hopeless even with those
changes, with or without AVX.
(Probably should have some code similar to the one in the llvmpipe fs
backend code, using bit replication to extend to rgba8888 - rounding
is not quite 100% accurate but if it's good enough there it should be
here as well.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
If we only feed one source vector at a time, we cannot use pack intrinsics
(as we only have a 64bit destination dst vector). lp_bld_conv_auto is
specifically designed to alter the length and number of destination vectors,
so this works just fine (if we use single source vectors at a time, afterwards
we immediately reassemble the vectors).
For AVX though this isn't really possible, since we expect 128bit output
already for a single 256bit input. (One day we should handle AVX2 which again
would need multiple inputs, however there's the problem that we get different
ordered output there and we don't want to reorder, so would need to be able
to tell build_conv to handle upper and lower halfs independently.)
A similar strategy would probably work for 32->8bit too (if it doesn't hit
the special case) but I'm going to try something different for that...
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
simd instruction sets usually have comparisons for equal, not unequal.
So use a different comparison against the mask itself - which also means
we don't need a all-zero as well as a all-one (for the pxor) reg.
Also add code to avoid scalar expansion of i1 values which we definitely
shouldn't do. There's problems with this though with llvm select
interaction, so it's disabled (basically using llvm select instead of
intrinsics may still produce atrocious code, even in cases where we
figured it should not, albeit I think this could probably be fixed
with some better selection of optimization passes, but I have zero
idea there really).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Draw calls no longer flush SDMA IBs. r600_need_dma_space is
responsible for synchronizing execution between both IBs.
Initial buffer clears and fast clears will stay unflushed in the SDMA IB
(up to 64 MB) as long as the GFX IB isn't flushed either.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Call r600_dma_emit_wait_idle only when there is a possibility of
a read-after-write hazard. Buffers not yet used by the SDMA IB don't
have to wait.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
r600_dma_emit_wait_idle is going away in its current form.
The only difference is that the moved code is executed before DMA calls
instead of after them.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This fixes the mistake introduced in commit
b6737a8bcd
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Old test caused breakage with llvm-svn (4.0.0svn), and not needed as
the minimum required llvm version for swr is 3.6.
Reviewed-by: George Kyriazis <george.kyriazis@intel.com>
wrap lp_bld_type.h around extern "C".
Windows decorates global variables, so when used from .cpp files, need
to use an undecorated version.
Also, removed related and unneeded code from swr_screen.cpp
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
This seems to fix the GPU hangs caused by:
commit ed3190b3f3
Author: Marek Olšák <marek.olsak@amd.com>
Date: Sun Nov 13 18:41:43 2016 +0100
radeonsi: don't export ClipVertex and ClipDistance[] if clipping is disabled
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99219
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>