Since we can have per-pixel lod we should also honor the filter per-pixel
(in fact we didn't honor it per quad neither in the multiple quad case).
Do this by running the linear path and simply beating the weights into shape
(the sample with the higher weight is the one which should have been chosen
with nearest filtering hence adjust filter weight to 1.0/0.0 based on that).
If all pixels use nearest filter (either min and mag) then still run just a
nearest filter as this is way cheaper (probably around 4 times faster for 2d,
more for 3d case) and it should be relatively rare that pixels really need
different filtering. OTOH if all pixels would require linear don't do anything
special since the linear path with filter adjustments shouldn't really be all
that much more expensive than ordinary linear, and we think it's rare that
min/mag filters are configured differently so there doesn't seem much value
in trying to optimize this further.
This does not yet fix the AoS path (though currently AoS is only used for
single quads hence it could be considered less broken, just never honoring
per-pixel filter decision but doing it per quad).
v2: simplify code a bit (unify min linear and min nearest cases)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
While a sqrt here and there shouldn't hurt much (depending on the cpu) it is
possible to completely omit it since rho is only used for calculating lod and
there log2(x) == 0.5*log2(x^2). Depending on the exact path taken for
calculating lod this means we get a simple mul instead of sqrt (in case of
nearest mip filter in fact we don't need to replace the sqrt with something
else at all), only in some not very useful path this doesn't work (combined
brilinear calculation of int level and fractional lod, accurate rho calc but
brilinear filtering seems odd).
Apart from being faster as an added bonus this should increase our crappy
fractional accuracy of lod, since fast_log2 is only good for ~3bits and this
should increase accuracy by one bit (though not used if dimension is just one
as we'd need an extra mul there as we never had the squared rho in the first
place).
v2: use separate ilog2_sqrt function if we have squared rho.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
This is just preparation for per-pixel (or per-quad in case of multiple quads)
min/mag filter since some assumptions about number of miplevels being equal
to number of lods no longer holds true.
This change does not change behavior yet (though theoretically when forcing
per-element path it might be slower with different min/mag filter since the
code will respect this setting even when there's no mip maps now in this case,
so some lod calcs will be done per-element just ultimately still the same
filter used for all pixels).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Fixes "Missing break in switch" defect reported by Coverity.
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The downstream android kernel driver is "kgsl", the upstream drm/kms
driver is called "msm". Since libdrm_freedreno handles the differences
between the two, we need to load the same thing for either device.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We need to set the flag on all the .xyzw components that are written by
the instruction, not just on .x. Otherwise a later use of rN.y (for
example) will not trigger the appropriate sync bit to be set.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Seems like most/all instructions have some restrictions about const src
registers. In seems like the 2 src (cat2) instructions can take at most
one const, and the 3 src (cat3) instructions can take at most one const
in the first 2 arguments. And so on. Handle this properly now.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
MSAA was tested by one user on RS690 and it works for him with color
compression (CMASK) disabled. Our theory is that his chipset lacks CMASK RAM.
Since we don't have hardware documentation about which chipsets actually have
CMASK RAM, I had to take a guess based on the presence of HiZ.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
In particular noone is interested in the vertex count, so drop that,
and also drop the duplicated num_primitives_generated /
so.primitives_storage_needed variables in drivers. I am unable for now to figure
out if primitives_storage_needed in SO stats (used for d3d10) should
increase if SO is disabled, though the equivalent num_primitives_generated
used for OpenGL definitely should increase. In any case we were only counting
when SO is active both in softpipe and llvmpipe anyway so don't pretend there's
an independent num_primitives_generated counter which would count always.
(This means the PIPE_QUERY_PRIMITIVES_GENERATED count will still be wrong just
as before, should eventually fix this by doing either separate counting for this
query or adjust the code so it always counts this even if SO is inactive depending
on what's correct for d3d10.)
Reviewed-by: Brian Paul <brianp@vmware.com>
The command is submitted once the event has been triggered, but it might not
have completed yet. Therefore, we have to add it to deps in order to wait on it.
Signed-off-by: Niels Ole Salscheider <niels_ole@salscheider-online.de>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
I previously fixed this partly in 9e8400f4c9,
however I didn't go far enough in testing it, now when I parse a TGSI shader
with arrays in it my iterator can see the ArrayID set to the proper value.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
I assume this should have been part of commit
7727fbb7c5. This (obviously) fixes a lot tests.
Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Commit 53e20b8b introduced the use of a template to initialize some
common fields. Move this copying of fields to before the common vp3
fields are initialized.
Reported-by: Martin Peres <martin.peres@labri.fr>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Christian König <christian.koenig@amd.com>
The cmps.f.* instruction doesn't actually seem to give a float 1.0 or
0.0 output. It either needs a cov.u16f16 or add.s + sel.f16. This
makes SGT/SLT/etc more similar to CMP, so handle them in trans_cmp().
This fixes a bunch of piglit tests.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It seems there are a number of cases where instructions have limitations
about taking reading src's from const register file, so make
get_unconst() a bit easier to use.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We probably should get rid of assert() entirely, but at this stage it is
more useful for things to crash where we can catch it in a debugger.
With compile_error() we have a single place to set an error flag (to
bail out and return an error on the next instruction) so that will be a
small change later when enough of the compiler bugs are sorted.
But re-arrange/cleanup the error/assert stuff so we at least get a dump
of the TGSI that triggered it. So we see some useful output in piglit
logs.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Don't crash when no color buffer bound. Something caught when starting
to run piglit, fixes a hanful of piglit tests.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Category 4 instructions (rsq, rcp, sqrt, etc) seem to be unable to take
a const register as src. In these cases we need to move the src to a
temporary gpr first.
This is the second case of such a restriction, where the instruction
encoding appears to support a const src, but in fact the hw appears to
ignore that bit. So split things out into a helper that can be re-used
for any instructions which have this limitation.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Our current (rather naive) register assignment is based on mapping
different register files (INPUT, OUTPUT, TEMP, CONST, etc) based on the
max register index of the preceding file. But in some cases, the lowest
used register in a file might not be zero. In which case
file_count[file] != file_max[file] + 1.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Sometimes things other than color dst need saturating, like if there is
a 'clamp(foo, 0.0, 1.0)'. So for saturated dst add the extra
instructions to fix up dst.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The 1st src to add.s needs (r) flag (repeat), otherwise it will end up:
add.s dst.xyzw, tmp.xxxx -1
instead of:
add.s dst.xyzw, tmp.xyzw, -1
Also, if we are using a temporary dst to avoid clobbering one of the src
registers, we actually need to use that as the dst for the sel
instruction.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This patch adds support for:
PIPE_COMPUTE_CAP_MAX_INPUT_SIZE
PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE
Return the values reported by the closed source driver for now.
Signed-off-by: Niels Ole Salscheider <niels_ole@salscheider-online.de>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Previously, the min/mag switchover point when using nearest/none mip
filter was effectively -0.5 which can't be right. Looks like new OpenGL
thinks it's ok if it's always 0.0 (older versions required 0.5 in some
cases), let's hope everybody else thinks that's fine too.
Refactor this slightly and get the per-quad/per-pixel min/mag decision
values further down to sampling, though still only the first component
is used yet.
While here also fix code trying to skip lod bias application etc. when
mipfilter is none, as this is still needed for determining min/mag filter.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
As of "2f142d59 build: Add --enable-gallium-osmesa flag." the pkgconfig
file from classic osmesa is no longer installed when building gallium
osmesa, so copy it to gallium osmesa and install the copy instead.
CC: "9.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Except for explicit derivs with cube maps which are very bogus anyway.
Just like explicit lod this is only used if no_quad_lod is set in
GALLIVM_DEBUG env var.
Minification is terrible on cpus which don't support true vector shifts
(but should work correctly). Cannot do the min/mag filter decision (if
they are different) per pixel though, only selecting different mip levels
works.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
