This allows mod(int, int) to become selected as float mod when doubles
are supported.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: "10.6" <mesa-stable@lists.freedesktop.org>
which it actually implements instead of the newest version defined in
dri_interface.h
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Boyan Ding <boyan.j.ding@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This avoids a security issue where userspace could have written the tile
state/tile alloc behind the GPU's back, and will apparently be necessary
for fixing stability bugs (tile state buffers are missing some top bits
for the tile alloc's address).
We can't use sampler messages with gradient information (like
sample_g or sample_d) to deal with this scenario because according
to the PRM:
"The r coordinate and its gradients are required only for surface
types that use the third coordinate. Usage of this message type on
cube surfaces assumes that the u, v, and gradients have already been
transformed onto the appropriate face, but still in [-1,+1] range.
The r coordinate contains the faceid, and the r gradients are ignored
by hardware."
Instead, we should lower this to compute the LOD manually based on the
gradients and use a different sample message that takes the computed
LOD instead of the gradients. This is already being done in
brw_lower_texture_gradients.cpp, but it is restricted to shadow
samplers only, although there is a comment stating that we should
probably do this also for samplerCube and samplerCubeArray.
Because of this, both dEQP and Piglit test cases for textureGrad with
cube maps currently fail.
This patch does two things:
1) Activates the texturegrad lowering pass for all cube samplers.
2) Corrects the computation of the LOD value for cube samplers.
I had to do 2) because for cube maps the calculations implemented
in the lowering pass always compute a value of rho that is twice
the value we want (so we get a LOD value one unit larger than we
want). This only happens for cube map samplers (all kinds). I am
not sure about why we need to do this, but I suspect that it is
related to the fact that cube map coordinates, when transported
to a specific face in the cube, are in the range [-1, 1] instead of
[0, 1] so we probably need to divide the derivatives by 2 when
we compute the LOD. Doing that would produce the same result as
dividing the final rho computation by 2 (or removing a unit
from the computed LOD, which is what we are doing here).
Fixes the following piglit tests:
bin/tex-miplevel-selection textureGrad Cube -auto -fbo
bin/tex-miplevel-selection textureGrad CubeArray -auto -fbo
bin/tex-miplevel-selection textureGrad CubeShadow -auto -fbo
Fixes 10 dEQP tests in the following category:
dEQP-GLES3.functional.shaders.texture_functions.texturegrad.*cube*
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Triggers an INVALID_OPCODE warning on GK208. Seems rare enough to not
warrant verification on other chips. Fixes the new piglits:
ubo_array_indexing/fs-nonuniform-control-flow.shader_test
ubo_array_indexing/vs-nonuniform-control-flow.shader_test
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Mark GL_ARB_framebuffer_no_attachments as done for i965.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Enable GL_ARB_framebuffer_no_attachments in i965 for Gen7 and higher.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Ensure that the GPU spawns the fragment shader thread for those
fragment shaders with atomic buffer access.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Add helper function that checks if current fragment shader active
of gl_context has atomic buffer access.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Change references to gl_framebuffer::Width, Height, MaxNumLayers
and Visual::samples to use the _mesa_geometry_ convenience functions
for those places where the geometry of the gl_framebuffer is needed
(in contrast to the geometry of the intersection of the attachments
of the gl_framebuffer).
This patch is to pave the way to enable GL_ARB_framebuffer_no_attachments
on Gen7 and higher in i965.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Add helper convenience function that intersects the scissor values
against a passed bounding box. In addition, to avoid replicated code,
make the function _mesa_scissor_bounding_box() use this new function.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Add convenience helper functions for fetching geometry of gl_framebuffer
that return the geometry of the gl_framebuffer instead of the geometry of
the buffers of the gl_framebuffer when then the gl_framebuffer has no
attachments.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Implement GL_ARB_framebuffer_no_attachments in Mesa core
- changes to conditions for framebuffer completenss
- implement set/get functions for framebuffers for
new functions in GL_ARB_framebuffer_no_attachments
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Define the enumeration constants, function entry points and
glGet for the GL_ARB_framebuffer_no_attachments.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
Define the infrastructure for the extension GL_ARB_framebuffer_no_attachments:
- extension table
- additions to gl_framebuffer
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kevin Rogovin <kevin.rogovin@intel.com>
I was thinking of the MIN opcode in terms of unsigned math, but it's
signed, so if you used a negative array index, you could read before the
UBO. Fixes segfaults under simulation in piglit array indexing tests with
mprotect-based guard pages.
I wanted to assert that src1 came from a non-unspilled register in shader
validation, and this easily gets us that. And, as a bonus:
total instructions in shared programs: 93347 -> 92723 (-0.67%)
instructions in affected programs: 60524 -> 59900 (-1.03%)
Disabling miptails fixed the buffer corruption happening in FBO
which use YF/YS tiled renderbuffer or texture as color attachment.
Spec recommends disabling mip tails only for non-mip-mapped surfaces.
But, without disabling miptails I couldn't get correct data out of
mipmapped YF/YS tiled surface.
We need better understanding of miptails before start using them.
For now this patch helps move things forward.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Patch sets the alignments for texture and renderbuffer surfaces.
V3: Make changes inside horizontal_alignment() and
vertical_alignment() (Topi)
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Makes no functional changes in the code.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
This function will be utilised in later patches.
V2: Make both pointers constants (Topi)
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
This patch sets the tiled resource mode for texture and renderbuffer
surfaces.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
The surfaceless platform is for off-screen rendering only. Render node support
is required.
Only consider the render nodes. Do not use normal nodes as they require
auth hooks.
v3: change platform_null to platform_surfaceless
v4: make libdrm required for surfaceless
v5: remove modified include guards with defined(HAVE_SURFACELESS_PLATFORM)
v6: use O_CLOEXEC for drm fd
Signed-off-by: Haixia Shi <hshi@chromium.org>
Signed-off-by: Zach Reizner <zachr@google.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>
Previously when setting up the sample instruction for an indirect
sampler the vec4 backend was directly passing the pseudo opcode's
src0. However vec4_visitor::visit(ir_texture *) doesn't set the
texture operation's src0 -- it's left as BAD_FILE, which when
translated into a brw_reg gives the null register. In brw_SAMPLE,
gen6_resolve_implied_move() inserts a MOV from the inst->base_mrf and
sets the src0 appropriately. The indirect sampler case did not have a
call to gen6_resolve_implied_move().
The fs backend avoids this because the platforms that support dynamic
indexing of samplers (IVB+) have been converted to not use the
fake-MRF hack, and instead send from proper GRFs.
This patch makes it call gen6_resolve_implied_move before setting up
the indirect message. This is similar to what is done for constant
sampler numbers in brw_SAMPLE.
The Piglit tests for sampler array indexing didn't pick this up
because they were using a texture with a solid colour so it didn't
matter what texture coordinates were actually used. The tests have now
been changed to be more thorough in this commit:
http://cgit.freedesktop.org/piglit/commit/?id=4f9caf084eda7
With that patch the tests for gs and vs are currently failing on
Ivybridge, but this patch fixes them. There are no other changes to a
Piglit run on Ivybridge.
On Skylake the gs tests were failing even without the Piglit patch
because Skylake needs the source registers to work correctly in order
to send a message header to select SIMD4x2 mode.
(The explanation in the commit message is partially written by Matt
Turner)
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This reverts commit adee54f826.
Further down in the GLSL ES 3.10 spec it say:
"If an array is declared as the last member of a shader storage block
and the size is not specified at compile-time, it is sized at run-time.
In all other cases, arrays are sized only at compile-time."
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Before 9b5e92f get_gl_override was called only once, but now it is
called for multiple APIs (GLES2, GL), version needs to be set always.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90797
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
Tested-by: Martin Peres <martin.peres@linux.intel.com>
This function was trying to align the width and height to a multiple
of the block size for compressed textures. It was using align_w/h as a
shortcut to get the block size as up until Gen9 this always happens to
match. However in Gen9+ the alignment values are expressed as
multiples of the block size so in effect the alignment values are
always 4 for compressed textures as that is the minimum value we can
pick. This happened to work for most compressed formats because the
block size is also 4, but for FXT1 this was breaking because it has a
block width of 8.
This fixes some Piglit tests testing FXT1 such as
spec@3dfx_texture_compression_fxt1@fbo-generatemipmap-formats
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
The state tracker will pass through requests from buggy applications
which will have the buffer size larger than the max allowed (64k). Clamp
the size to 64k so that we don't get errors when uploading the constbuf
data.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
One of the places we have to insert texbars is in situations where the
result of the tex gets overwritten by a different instruction (e.g. in a
conditional statement). However in some situations it can actually
appear as though the original tex itself is an overwriting instruction.
This can naturally never really happen, so just ignore the tex
instruction when it comes up.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90347
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.5 10.6" <mesa-stable@lists.freedesktop.org>
Back in 2013, a patch was added (with 2 reviewers!) at the end of the
block to early exit the loop in this case, without noticing that the loop
already did. I added another early exit case, again without noticing, but
Rob caught me. Just drop the loop condition that apparently surprises
most of us, instead of leaving the end of the loop conspicuously not
exiting on success.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
