Certain built-in arrays, such as gl_ClipDistance[], gl_CullDistance[],
gl_TessLevelInner[], and gl_TessLevelOuter[] are specified as scalar
arrays. Normal scalar arrays are sparse - each array element usually
occupies a whole vec4 slot. However, most hardware assumes these
built-in arrays are tightly packed.
The new var->data.compact flag indicates that a scalar array should
be tightly packed, so a float[4] array would take up a single vec4
slot, and a float[8] array would take up two slots.
They are still arrays, not vec4s, however. nir_lower_io will generate
intrinsics using ARB_enhanced_layouts style component qualifiers.
v2: Add nir_validate code to enforce type restrictions.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
I've started working on a shader-db alike for Vulkan,
it's based on vktrace and it records pipelines, this
adds support to dump the shader stats exactly like
radeonsi does, so I can reuse the shader-db scripts it
uses.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The sample id is packed into bits 8-12, so adjust
things properly.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This fixes a bunch of crashes in CTS tests looking for this.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This cleans up the ddxy emission along the same lines as
radeonsi. It also means we don't use LDS on VI chips we
use the dspermute interface, it also removes some duplicated
code.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
For the hw resolve there is no need to emit any sort
of texture coordinates, so drop them all in the meta path.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We want to also invalidate the cache when LLVM gets changed. As the
specific LLVM revision is not fixed at build time, we will need to
check at runtime. Computing a checksum for LLVM is going to be very
expensive, so just use the mtime.
Tested on my computer that the returned DSO for the LLVM symbol is
actually the LLVM DSO.
Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
No sense in repeatedly determining it. Also, it might be dependent
on the device as shaders get compiled differently for SI/CIK/VI etc.
Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
The former calculations were for min/max y. The width/height don't take
translate into account.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
This may ultimately be possible to support, but for now it's not hooked
up and the swr core only supports this output from GS.
This normally wouldn't matter, but we lie about supporting GL 3.2, and
also the blitter and st/mesa will make use of this functionality if
claimed.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
Multiple buffers may reference client arrays. When this happens, we
might reach for scratch space multiple times, which could cause later
arrays to invalidate the pointers allocated for the earlier ones.
This fixes copyteximage 2D_ARRAY.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Currently a sequence like draw/map/draw/map will cause the second map to
not wait for the second draw. This is because the first map will clear
the resource business bit, and the second draw won't reset it since no
state has changed.
swr_update_derived does a tiny bit of extra work, including updating the
SWR_BACKEND_STATE as well as waiting for prending fences. If that's a
problem, we could call swr_update_resource_status directly from
draw/clear handlers.
Fixes clearbuffer-stencil, clearbuffer-depth, clearbuffer-depth-stencil,
and clearbuffer-display-lists.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
The minification should happen before alignment, not after. See similar
logic on ComputeLODOffsetY. The current logic requires unnecessarily
large textures when there's an initial NPOT size.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
There's no guarantee that mip width/height will be a multiple of the
compressed block size. Doing a divide by the block size first yields
different results than GL expects, so we do the divide at the end.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tim Rowley <timothy.o.rowley@intel.com>
A past commit added the ability to compile "optimized" shader variants
asynchronously (not stalling the app).
This commit builds upon that and adds what is basically a runtime shader
linker. If a VS output isn't used by the currently-bound PS, a new VS
compilation is started without that output. The new shader variant
is used when it's ready.
All apps using separate shader objects I've seen had unused VS outputs.
Eliminating unused/useless VS outputs also eliminates the corresponding
vertex attribute loads.
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This is the first user of optimized monolithic shader variants.
Cull distances can't be disabled by states.
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
key->part.*: prolog and epilog flags only
key->as_{ls,es}: special flags
key->mono.*: flags for monolithic compilation only
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Company Of Heroes 2 needs only 24.
This saves 512 bytes of CE RAM per shader stage.
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
FORCE_TILING should disable it. It has no effect now, but that may change
soon.
Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
When the uploading of shader fails on si_shader_binary_upload(),
it returns -ENOMEM. We should handle si_shader_binary_upload() failure path
on si_create_compute_state().
CID 1394027
v2: Fixes from Edward O'Callaghan's review
a) Update explicitly return value check with "si_shader_binary_upload() < 0"
b) Update commit message.
Signed-off-by: Mun Gwan-gyeong <elongbug@gmail.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
It turns out that noone actually cares if the address computations overflow,
be it the stride mul or the offset adds.
Wrap around seems to be explicitly permitted even by some other API (which
is a _very_ surprising result, as these overflow computations were added just
for that and made some tests pass at that time - I suspect some later fixes
fixed the actual root cause...). So the requirements in that other api were
actually sane there all along after all...
Still need to make sure the computed buffer size needed is valid, of course.
This ditches the shiny new widening mul from these codepaths, ah well...
And now that I really understand this, change the fishy min limiting
indices to what it really should have done. Which is simply to prevent
fetching more values than valid for the last loop iteration. (This makes
the code path in the loop minimally more complex for the non-indexed case
as we have to skip the optimization combining two adds. I think it should
be safe to skip this actually there, but I don't care much about this
especially since skipping that optimization actually makes the code easier
to read elsewhere.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Don't keep the ofbit. This is just a minor simplification, just adjust
the buffer size so that there will always be an overflow if buffers aren't
valid to fetch from.
Also, get rid of control flow from the instanced path too. Not worried about
performance, but it's simpler and keeps the code more similar to ordinary
fetch.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
The code for elts and linear paths was nearly 100% identical by now - with
the elts path simply having some additional gather for the elements in the
main loop (with some additional small differences before the main loop).
Hence nuke the separate functions and decide this at jit shader execution
time (simply based on the presence of the elts pointer).
Some analysis shows that the generated vs jit functions seem to be just very
minimally more complex than the former elts functions, and almost none of the
additional complexity is in the main loop (basically just the branch logic
for the branch fetching the actual indices).
Compared to linear, the codesize of the function is of course a bit larger,
however the actual executed code in the main loop appears to be near 100%
identical (the additional code looking up indices is skipped as expected).
So, I would not expect a (meaningful) performance difference with the
generated code, neither with elts nor linear, this does however roughly
half the compilation time (the compiled shaders should also use only half
the memory of course).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
This was kind of strange, since it replaced indices which were only
overflowing due to bias with MAX_UINT. This would cause an overflow later
in the shader, except if stride was 0, however the vertex id would be
essentially random then (-1 + eltBias). No test cared about it, though.
So, drop this and just use ordinary int arithmetic wraparound as usual.
This is much simpler to understand and the results are "more correct" or
at least more consistent (vertex id as well as actual fetch results just
correspond to wrapped around arithmetic).
There's only one catch, it is now possible to hit the cache initialization
value also with ushort and ubyte elts path (this wouldn't be an issue if
we'd simply handle the eltBias itself later in the shader). Hence, we need
to make sure the cache logic doesn't think this element has already been
emitted when it has not (I believe some seriously bad things could happen
otherwise). So, borrow the logic which handled this from the uint case, but
not before fixing it up...
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
vsplit_get_base_idx explicitly returned idx 0 and set the ofbit
in case of overflow. We'd then check the ofbit and use idx 0 instead of
looking it up. This was necessary because DRAW_GET_IDX used to return
DRAW_MAX_FETCH_IDX and not 0 in case of overflows.
However, this is all unnecessary, we can just let DRAW_GET_IDX return 0
in case of overflow. In fact before bbd1e60198
the code already did that, not sure why this particular bit was changed
(might have been one half of an attempt to get these indices to actual draw
shader execution - in fact I think this would make things less awkward, it
would require moving the eltBias handling to the shader as well).
Note there's other callers of DRAW_GET_IDX - those code paths however
explicitly do not handle index buffer overflows, therefore the overflow
value doesn't matter for them.
Also do some trivial simplification - for (unsigned) a + b, checking res < a
is sufficient for overflow detection, we don't need to check for res < b too
(similar for signed).
And an index buffer overflow check looked bogus - eltMax is the number of
elements in the index buffer, not the maximum element which can be fetched.
(Drop the start check against the idx buffer though, this is already covered
by end check and end < start).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
v4: Add windows-specific gen_knobs.{cpp|h} changes
v5: remove aggresive squashing of gen_knobs.py to this commit; added
SConscript to EXTRA_DIST in Makefile.am
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Modify gen_knobs.py so that each invocation creates a single generated
file. This is more similar to how the other generators behave.
v5: remove Scoscript edits from this commit; moved to commit that first
adds SConscript
Acked-by: Emil Velikov <emil.velikov@collabora.com>
To buils The SWR driver (currently optional, not compiled by default)
v3: add option as opposed to target
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
- Handle dynamic library loading for windows
- Implement swap for gdi
- fix prototypes
- update include paths on configure-based build for swr_loader.cpp
v2: split to multiple patches
v3: split and reshuffle some more; renamed title
v4: move Makefile.am changes to other commit. Modify header files
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
There are 2 swr_create_screen() functions. One in swr_loader.cpp, which
is used during driver init, and the other is hiding in swr_screen.cpp,
which ends up in the arch-specific .dll/.so.
Rename the second one to swr_create_screen_internal(), to avoid confusion
in header files.
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reorder header files so that we have a chance to defined NOMINMAX before
mesa include files include windows.h
v3: split from bigger patch
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>