Originally we removed the instruction, changed the source, and then
re-inserted it. This works, but nir_instr_rewrite_src is a bit more
obviously correct.
This works around a bug in older version of UE4, where a shader
defines the same structure twice. Although we aren't sure this is correct
GLSL (it most likely isn't) there are enough UE4 based things out there
we should deal with this.
This drops the error to a warning if the struct names and contents match.
v1.1: do better C++ on record_compare declaration (Rob)
v2: restrict this to desktop GL only (Ian)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95005
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Ken suggested instead of a big and complicated optimization pass, to
just recognize the operations here. It's certainly less code and a lot
prettier, but it seems to actually perform worse for currently unknown
reasons.
total instructions in shared programs: 8923452 -> 8904108 (-0.22%)
instructions in affected programs: 814563 -> 795219 (-2.37%)
helped: 3336
HURT: 10
total cycles in shared programs: 66970734 -> 66651476 (-0.48%)
cycles in affected programs: 10582686 -> 10263428 (-3.02%)
helped: 2438
HURT: 691
total spills in shared programs: 1811 -> 1789 (-1.21%)
spills in affected programs: 85 -> 63 (-25.88%)
helped: 4
total fills in shared programs: 3143 -> 3109 (-1.08%)
fills in affected programs: 167 -> 133 (-20.36%)
helped: 4
LOST: 2
GAINED: 36
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The next patch wants to inspect the LOD argument and do something
different if it's 0.0f. But at that point we've emitted a MOV for it and
we just have a register to look at.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Address registers are always loaded right before use. Don't treat them
as "global", which will cause them to be put into the function's
linkage, and will make the register allocator hold onto that
register until the end of the function.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Move a MultisampleTrait static from header to cpp as clang seemed to get
confused with some specializations in the header vs some in cpp.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
With xwayland, vainfo use VA_DISPLAY_WAYLAND as default and it fails
and fails when specify display with `vainfo --display wayland`.
In fact wayland support for libva uses drm path to connect device,
and should use drm pipe loader to create screen.
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
brw_fs.cpp: In function ‘const unsigned int* brw_compile_fs(const [...]
brw_fs.cpp:6093:64: warning: ‘simd16_grf_start’ may be used uninitialized [...]
prog_data->base.dispatch_grf_start_reg = simd16_grf_start;
brw_fs.cpp:5996:29: note: ‘simd16_grf_start’ was declared here
uint8_t simd8_grf_start, simd16_grf_start;
brw_fs.cpp:6094:52: warning: ‘simd16_grf_used’ may be used uninitialized [...]
prog_data->reg_blocks_0 = brw_register_blocks(simd16_grf_used);
brw_fs.cpp:5997:29: note: ‘simd16_grf_used’ was declared here
unsigned simd8_grf_used, simd16_grf_used;
(and more)
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
We don't really support reading/writing of 3D textures since the hardware
doesn't do 3D, but we do need to make sure that a pipe_transfer for them
has enough space to store the image. This was previously not a problem
because the state tracker only mapped a slice at a time until
fb9fe352ea. Fixes glean glsl1 tests, which
all have setup of a 3D texture at the start.
Align with the spec requirement that the range must be at least
[−2 × maxViewportDimensions, 2 × maxViewportDimensions − 1]. Our
hardware supports this.
Fixes dEQP-VK.api.info.device.properties
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94896
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This is my attempt at fixing at least one of the UE4 bugs with GL4.3.
If we are doing intrastage matching and hit anonymous structs, then
we should do a record comparison instead of using the names.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95005
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Trivial fix to improperly handled cleanup during
VK_ERROR_OUT_OF_HOST_MEMORY.
Identified by Coverity: CID 1358908 and 1358909
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
A few changes to support musl libc as well.
In particular fpu_control.h is glibc specific.
fenv.h doesn't enable to do exactly what we want either,
so instead use assembly directly.
Signed-off-by: Wang He <xw897002528@gmail.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Nine already supports the feature.
There are no failing WINE tests for per stage constants.
Enabling D3DPMISCCAPS_PERSTAGECONSTANT as it fixes
https://github.com/iXit/Mesa-3D/issues/205
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
The last remaining issues with thread_submit have been resolved,
thus turn it when on a different device (the case where is is
beneficial).
Signed-off-by: Axel Davy <axel.davy@ens.fr>
pipe_rasterizer multisample bit should be enabled only when really
wanting to do multisampling, thus we should disable when not having
msaa render target.
This fixes some depth calculation precision issues on radeon.
Also disable it when depth and stencil tests are disabled, since in that
case multisampling is same as not multisampled.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
ATOC extension does something only when alpha test is enabled.
Use a second bit to encode the difference with ATIATOC.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
Nine doesn't support vs output/ps input packing.
We haven't found any application requiring that,
and implementing it properly is complex.
Add asserts for now.
Signed-off-by: Axel Davy <axel.davy@ens.fr>
