The following commits broke things by starting to feed us unhandled
extract_u16/extract_u8 opcodes:
commit 905ff86198
Author: Matt Turner <mattst88@gmail.com>
AuthorDate: Wed Feb 3 14:28:31 2016 -0800
Commit: Matt Turner <mattst88@gmail.com>
CommitDate: Fri Mar 4 11:52:34 2016 -0800
nir: Recognize open-coded extract_u16.
commit 76289fbfa8
Author: Matt Turner <mattst88@gmail.com>
AuthorDate: Thu Jan 21 09:09:48 2016 -0800
Commit: Matt Turner <mattst88@gmail.com>
CommitDate: Fri Mar 4 11:52:34 2016 -0800
nir: Recognize open-coded extract_u8.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
First off, st/mesa lowers DSQRT incorrectly (it uses CMP to attempt to
find out whether the input is less than 0). Secondly the current
approach (x * rsq(x)) behaves poorly for x = inf - a NaN is produced
instead of inf.
Instead we switch to the less accurate rcp(rsq(x)) method - this behaves
nicely for all valid inputs. We still don't do this for DSQRT since the
RSQ/RCP ops are *really* inaccurate, and don't even have Newton-Raphson
steps right now. Eventually we should have a separate library function
for DSQRT that does it more precisely (and perhaps move this lowering to
the post-opt phase).
This fixes a number of dEQP precision tests that were expecting better
behavior for infinite inputs.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
The idea is that a single triangle will cover the whole area being
drawn, allowing the blit shader to do its work. However the max fb size
is 16384x16384, which means that the triangle we draw needs to be twice
that in order to cover the whole area fully. Increase the size of the
triangle to 32768x32768.
This fixes a number of dEQP tests that were failing because a blit was
involved which would miss some of the resulting texture.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Bitfields where shuffled around for the better on a4xx, so we don't need
any patching on this one. It appears to be something we set entirely in
the gmem code so no conflict between tiling and render state like we had
in a3xx.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Move where we pick dummy FS for binning pass, so the whole driver sees
the same dummy/no-op FS stage.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Sampler states don't really make sense with buffer textures, but they
can be set anyway, so we need to be defensive here. This bug was lurking
for a while and was finally noticed due to PBO uploads setting sampler
states.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94284
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Laurent Carlier <lordheavym@gmail.com>
Tested-by: Shawn Starr <shawn.starr@rogers.com>
Since we aren't going to put the function parameters or the return variable
in the list of locals, it won't get a proper declaration. This changes
nir_print to print the type along with each parameter or return variable.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Otherwise, we have a problem when we go to print functions with arguments
because their names get added to the hash table during declaration which
happens after we print the prototype.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
NIR has never been used on IR where we haven't already done function
inlining so this code has been dead from the beginning. Let's just get rid
of it for now. We can always put it back in if we decide to use NIR for
function inlining at some point in the future.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
"Braswell" is a Cherryview based *thing*. It unfortunately requires extra
information to determine its marketing name. Unlike all previous products, and
hopefully all future ones, there is no unique 1:1 mapping of PCI device ID to
brand string.
I put up a fight about adding any complexity to our GL renderer string code for
a very long time. However, a wise man made a comment to me that I couldn't argue
with: if a user installs Windows on their hardware, the brand string should be
the same as what we display in Linux. The Windows driver apparently does this
check, so we should too.
Note that I did manage to find a good use for this info anyway in the compute
shader thread counts.
v2: memcpy instead of strncpy, and some minor changes (Matt)
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com
We have better information now, and 28 was not a valid thing to support. 6 EUs
per sublice with 7 threads per EU is the minimum supported config.
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com
The way we are organizing this code, the statically configured max_cs_threads
should always be the minimum value we actually support (ie. are aware of). As a
result, we can fall back to that if we get invalid numbers from the kernel (ie.
when the query succeeds, but the result is lower than expected).
I was originally planning to use an assert, but there is no reason to be so
mean.
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com
With the previous patches, the code can find out the actual number of available
compute threads. It is enabled only for Cherryview since that is the only
platform I know for a fact has shipped devices which can benefit from this. It
seems like other platforms /might/ benefit from this because of fused
configurations which /might/ have shipped. Fallback code is still there.
v2: Some minor adjustments from Matt
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com
Certain products are not uniquely identifiable based on device id alone. The
kernel exports an interface to help deal with this. This patch merely introduces
the consumer of the interface and makes sure nothing breaks.
It is also possible to use these values for programming GPGPU mode, and I plan
to do that as well.
The interface was introduced in libdrm 2.4.60, which is already required, so it
should all be fine.
v2: Some minor changes recommended by Matt
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Clear DCC flags if necessary when binding a new sampler view.
v2: Do not reset DCC flags of bound sampler views.
v3: Check that we have a real texture (Nicolai)
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Found by inspection of the source based on a bisected bug report.
This bug has been in the code for a long time, but the more recent PBO upload
feature exposed it because it leads to more uses of buffer textures.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94388
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Cc: "11.0 11.1 11.2" <mesa-stable@lists.freedesktop.org>
This will allow the nouveau backend to not try and split up ops that are
fused in GLSL.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Since it is all about calling into blitter functions, it makes more
sense here. This change also reduces the size of the interfaces between
.c files.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
There is an annoying corner case that I stumbled across while looking into
piglit's arb_shader_image_load_store/execution/load-from-cleared-image.shader_test
(which can be easily adapted to demonstrate the bug without the
ARB_shader_image_load_store extension)
When we bind a texture and then clear it using glClear (by attaching it
to the current framebuffer) for the first time, we allocate a separate
cmask for the texture to do fast clear, but the corresponding bit in
compressed_colortex_mask is not set. Subsequent rendering will use
incorrect data.
Conversely, when a currently bound texture with an existing cmask is
exported leading to that cmask being disabled, the compressed_colortex_mask
bit will remain set, leading to an assertion later on in debug builds.
Since iterating through all contexts and/or remembering where every
texture is bound would be costly, and cmask enable/disable should be
rare, we will maintain a global counter to signal contexts that they
must update their compressed_colortex_masks.
This patch introduces the global counter, and subsequent patches will
do the mask update.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
intel_alloc_private_renderbuffer_storage did:
rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat);
Unfortunately, internalFormat was usually an unsized format (such as
GL_DEPTH_COMPONENT). In OpenGL ES, _mesa_base_fbo_format() refuses to
accept unsized formats, and returns 0 rather than a real base format.
This meant that we ended up with a completely bogus rb->_BaseFormat for
window system buffers on OpenGL ES. All other renderbuffer allocation
functions in intel_fbo.c instead use the mesa_format, and do:
rb->_BaseFormat = _mesa_get_format_base_format(...);
We can do likewise, using rb->Format. This appears to work just fine.
dEQP-GLES3.functional.state_query.fbo.framebuffer_attachment_x_size_initial
failed, as it tried to perform a GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE query
on the window system depth buffer. That query relies on a proper
rb->_BaseFormat being set, so it broke because rb->_BaseFormat was 0 due
to the above bug.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94458
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
We were failing to reset our location tracking when encountering a
NEWLINE in the <HASH> state. Rip the code from the <*>{NEWLINE} rule,
which handles this properly.
Also, update 146-version-first-hash.c to have proper expectations.
When I introduced the test, I didn't verify that the line/column
numbers were correct, and it turns out they varied based on the type
of newline ending.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94447
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Because compute support is not enabled by default for these chipsets,
NVF0_COMPUTE=1 needs to be used, along with GALLIUM_HUD to enable
performance counters.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
This is really verbose but most of the configuration will be reused
for SM35 (GK110).
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Rob Clark
Ilia Mirkin
Nicolai Hähnle
Matt Turner
Jason Ekstrand
Boyuan Zhang
Ben Widawsky
Bas Nieuwenhuizen
Marek Olšák
Kenneth Graunke
Tim Rowley
Samuel Pitoiset