Just setting builder->exact isn't sufficient because that only applies to
instructions that are built with the builder but instructions created
manually and only inserted using the builder are left alone.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
This pass is similar to propagate_invariance in the GLSL compiler. The
real "output" of this pass is that any algebraic operations which are
eventually consumed by an invariant variable get marked as "exact".
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
While mathematically correct, these two optimizations result in an
expression with substantially lower precision than the original. For any
positive finite floating-point value, log2(x) is well-defined and finite.
More precisely, it is in the range [-150, 150] so any sum of logarithms
log2(a) + log2(b) is also well-defined and finite as long as a and b are
both positive and finite. However, if a and b are either very small or
very large, their product may get flushed to infinity or zero causing
log2(a * b) to be nowhere close to log2(a) + log2(b).
This imprecision was causing incorrect rendering in Talos Principal because
part of its HDR rendering process involves doing 8 texture operations,
clamping the result to [0, 65000], taking a dot-product with a constant,
and then taking the log2. This is done 6 or 8 times and summed to produce
the final result which is written to a red texture. In cases where you
have a region of the screen that is very dark, it can end up getting a
result value of -inf which is not what is intended.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96425
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
A nearly identical block already exists in the gen >= 6 block above.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Be consistent with the rest of the "set_xyz" state interfaces.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This yields a small performance improvement in Unigine Heaven.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The old calculation treated too many RBs as disabled.
Cc: 11.0 11.1 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The old limit, introduced in commit afa752d3f0,
was exceeded by 4 SE configurations which hit si_write_harvested_raster_configs.
Cc: 11.1 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Apparently, these are deprecated. There's some AutoUpgrade feature which
is supposed to promote these to cmp/select, which apparently doesn't work
with jit code. It is possible it's not actually even meant to work (see
the bug filed against llvm which couldn't provide an answer neither)
but in any case this is meant to be only temporary unless the intrinsics
are really illegal. So, just use the fallback code (which should be cmp/select,
we're actually doing cmp/sext/trunc/select, but in any case llvm 3.9 manages
to optimize this back to pmin/pmax in the end).
This addresses https://llvm.org/bugs/show_bug.cgi?id=28176
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Tested-by: Vinson Lee <vlee@freedesktop.org>
Tested-by: Aaron Watry <awatry@gmail.com>
This makes the check match up what we do on nv50 as well - there's no
point in switching over the push path if everything's in managed
buffers. This can happen when a shader uses a vertex without an enabled
array - we end up passing it a constant attribute.
This also has the effect of "fixing" some flickering in Talos. I have no
idea why. I've stared at the push logic forwards, backwards, and
sideways. By always forcing the push path (which is slow), the
flickering also goes away, but other rendering is still wrong
(specifically draw 383068 as identified in the bug). However by not
switching over to the push path, draw 383068 is correct.
Note that other flickering remains in Talos, like the red/green
walls/floors. This takes care of the shadow flickering though.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90513
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
If we have a loop, instructions before the tex might be added as tex
uses, and those may in fact dominate all other uses of the tex results.
This however doesn't mean that we don't need a texbar after the tex.
Only check if uses dominate each other they are dominated by the tex.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96565
Fixes: 7752bbc44 (gk104/ir: simplify and fool-proof texbar algorithm)
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Make sure to pass the requisite information in draws, blits, and clears
that work on the context's draw buffer.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This says how many window rectangles are supported by the
implementation, although it may not exceed PIPE_MAX_WINDOW_RECTANGLES.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>
Window rectangles apply to all framebuffer operations, either in
inclusive or exclusive mode. They may also be specified as part of a
blit operation.
In exclusive mode, any fragment inside any of the specified rectangles
will be discarded.
In inclusive mode, any fragment outside every rectangle will be
discarded.
The no-op state is to have 0 rectangles in exclusive mode.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This is pretty useful for debugging purposes and those should
not be omitted.
Fixes: 517a93b3 ("nvc0: add ARB_shader_draw_parameters support")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
A pipe pointer in the screen allows for access to current device context
in flush_frontbuffer and resource_destroy. This wasn't tracking current
context in multi-context situations.
v2: More caffeine. Corrected compare, removed unnecessary set of
screen-pipe in create_context, and added a few comments.
To match what's done in the automake build.
v2: Use git rev-parse to get a 10-character hash ID
Fix Python imports
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
From the Cherryview's PRM, Volume 7, 3D Media GPGPU Engine, Register Region
Restrictions, page 844:
"When source or destination datatype is 64b or operation is integer DWord
multiply, indirect addressing must not be used."
v2:
- Fix it for Broxton too.
v3:
- Simplify code by using subscript() and not creating a new num_components
variable (Kenneth).
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95462
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
From the Cherryview PRM, Volume 7, 3D Media GPGPU Engine,
Register Region Restrictions:
"When source or destination is 64b (...), regioning in Align1
must follow these rules:
1. Source and destination horizontal stride must be aligned to
the same qword.
(...)"
v2:
- Fix it for Broxton too.
v3:
- Remove inst->regs_written change as it is not necessary (Ken)
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95462
Tested-by: Mark Janes <mark.a.janes@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
I recently experimented with performing rasterizer discard in the SOL
unit instead of the clipper, and as far as I can tell, it's basically
the same performance. The clipper comes directly after SOL anyway,
and setting the clipper to REJECT_ALL should be pretty darn cheap.
Keep the perf_debug on Sandybridge, where the GS actually does work.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
There are almost no tests in any test suite, but what little I've found
seems to work. Ilia believes everything is in place.
v2: Predicate the enable on ES 3.1 being available (Gen8+) and also
ARB_compute_shader being available (requested by Ilia).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
There are quite a few pipelines that desktop applications (including a
bunch of piglit test) can expect to have run but don't meet the GLES
requirements. Instead of failing validation, just emit a debug message.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96358
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Cc: Gregory Hainaut <gregory.hainaut@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Previously some callers of precision_qualifier_allowed would strip the
arrayness from the type and some would not. As a result, some places
would not notice that float[6], for example, needed a precision
qualifier.
Fixes the new piglit test no-default-float-array-precision.frag.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96358
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Cc: Gregory Hainaut <gregory.hainaut@gmail.com>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
The use of a bitmask makes functions iterating only active
attributes less visible in profiles.
v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
The use of a bitmask makes functions iterating only active
attributes less visible in profiles.
v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Replaces an iterate and test bit in a bitmask loop by a
loop only iterating over the bits set in the bitmask.
v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Replaces an iterate and test bit in a bitmask loop by a
loop only iterating over the bits set in the bitmask.
v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
The change helps to use the bitmask/ffs in the next change.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Replaces an iterate and test bit in a bitmask loop by a
loop only iterating over the bits set in the bitmask.
v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
