When working on tessellation shaders, I created some vec4 virtual
opcodes for creating message headers through a sequence like:
mov(8) g7<1>UD 0x00000000UD { align1 WE_all 1Q compacted };
mov(1) g7.5<1>UD 0x00000100UD { align1 WE_all };
mov(1) g7<1>UD g0<0,1,0>UD { align1 WE_all compacted };
mov(1) g7.3<1>UD g8<0,1,0>UD { align1 WE_all };
This is done in the generator since the vec4 backend can't handle align1
regioning. From the visitor's point of view, this is a single opcode:
hs_set_output_urb_offsets vgrf7.0:UD, 1U, vgrf8.xxxx:UD
Normally, there's no hazard between sources and destinations - an
instruction (naturally) reads its sources, then writes the result to the
destination. However, when the virtual instruction generates multiple
hardware instructions, we can get into trouble.
In the above example, if the register allocator assigned vgrf7 and vgrf8
to the same hardware register, then we'd clobber the source with 0 in
the first instruction, and read back the wrong value in the last one.
It occured to me that this is exactly the same problem we have with
SIMD16 instructions that use W/UW or B/UB types with 0 stride. The
hardware implicitly decodes them as two SIMD8 instructions, and with
the overlapping regions, the first would clobber the second.
Previously, we handled that by incrementing the live range end IP by 1,
which works, but is excessive: the next instruction doesn't actually
care about that. It might also be the end of control flow. This might
keep values alive too long. What we really want is to say "my source
and destinations interfere".
This patch creates new infrastructure for doing just that, and teaches
the register allocator to add interference when there's a hazard. For
my vec4 case, we can determine this by switching on opcodes. For the
SIMD16 case, we just move the existing code there.
I audited our existing virtual opcodes that generate multiple
instructions; I believe FS_OPCODE_PACK_HALF_2x16_SPLIT needs this
treatment as well, but no others.
v2: Rebased by mattst88.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
We've apparently always been botching JIP for sequences such as:
do
cmp.f0.0 ...
(+f0.0) break
...
if
...
else
...
endif
...
while
Normally, UIP is supposed to point to the final destination of the jump,
while in nested control flow, JIP is supposed to point to the end of the
current nesting level. It essentially bounces out of the current nested
control flow, to an instruction that has a JIP which bounces out another
level, and so on.
In the above example, when setting JIP for the BREAK, we call
brw_find_next_block_end(), which begins a search after the BREAK for the
next ENDIF, ELSE, WHILE, or HALT. It ignores the IF and finds the ELSE,
setting JIP there.
This makes no sense at all. The break is supposed to skip over the
whole if/else/endif block entirely. They have a sibling relationship,
not a nesting relationship.
This patch fixes brw_find_next_block_end() to track depth as it does
its search, and ignore anything not at depth 0. So when it sees the
IF, it ignores everything until after the ENDIF. That way, it finds
the end of the right block.
I noticed this while reading some assembly code. We believe jumping
earlier is harmless, but makes the EU walk through a bunch of disabled
instructions for no reason. I noticed that GLBenchmark Manhattan had
a shader that contained a BREAK with a bogus JIP, but didn't measure
any performance improvement (it's likely miniscule, if there is any).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This just splits out a common pattern into an inline function
to make things cleaner to read.
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We really should initialise HS/LS_2 and SQ_LDS_ALLOC exists
on all evergreen not just cayman, so we should initialise
it as well.
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This adds the defines for a bunch of registers and shader
values that are required to implement tessellation.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Move some common code into one place, tess will also need
to use this function.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
There was only a single user which was using strlen(buf).
As this function is not user facing (i.e. we don't need to feed back
original length via a callback), we can simplify things.
Suggested-by: Timothy Arceri <timothy.arceri@collabora.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Add some checks if the original/dup'd fd is valid and ensure that we
don't leak it on error. The former is implicitly handled within the
pipe_loader, although let's make things explicit and check beforehand.
Spotted by Coverity (CID 1339865)
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
In theory this wouldn't be an issue, as we'll find the correct name and
break out of the loop before we hit the sentinel.
Let's fix this and avoid issues in the future.
Spotted by Coverity (CID 1339869, 1339870, 1339871)
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Earlier commit factored out the mpeg4 IQ matrix handling into separate
function, although it forgot to add a break in its case statement.
Thus the data ended up partially overwritten as the mpeg4 and h265
structs are members of the desc union.
Spotted by Coverity (CID 1341052)
Fixes: 64761a841d "st/va: move MPEG4 functions into separate file"
Cc: Julien Isorce <j.isorce@samsung.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Non-timer queries are suspended during blits. When the blits end, the queries
are resumed, but this resume operation itself might run out of CS space and
trigger a flush. When this happens, we must prevent a duplicate suspend during
preflush suspend, and we must also prevent a duplicate resume when the CS flush
returns back to the original resume operation.
This fixes a regression that was introduced by:
commit 8a125afa6e
Author: Nicolai Hähnle <nhaehnle@gmail.com>
Date: Wed Nov 18 18:40:22 2015 +0100
radeon: ensure that timing/profiling queries are suspended on flush
The queries_suspended_for_flush flag is redundant because suspended queries
are not removed from their respective linked list.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reported-by: Axel Davy <axel.davy@ens.fr>
Cc: "11.1" <mesa-stable@lists.freedesktop.org>
Tested-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Zero length arrays are non standard:
warning C4200: nonstandard extension used : zero-sized array in struct/union
Cannot generate copy-ctor or copy-assignment operator when UDT contains a zero-sized array
And all code does `N * sizeof query_result->batch[0]`, so it should work
exactly the same.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
From Section 11.1.3.11 (Validation) of the GLES 3.1 spec:
"An INVALID_OPERATION error is generated by any command that trans-
fers vertices to the GL or launches compute work if the current set
of active program objects cannot be executed, for reasons including:"
It then goes on to list the rules we validate in the
_mesa_validate_program_pipeline() function.
For ValidateProgramPipeline the only mention of generating an error is:
"An INVALID_OPERATION error is generated if pipeline is not a name re-
turned from a previous call to GenProgramPipelines or if such a name has
since been deleted by DeleteProgramPipelines,"
Which we handle separately.
This fixes:
ES31-CTS.sepshaderobjs.PipelineApi
No regressions on the eEQP 3.1 tests.
Cc: Gregory Hainaut <gregory.hainaut@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
This packet is a different size on gen8 and we hit an assertion when we
try to merge a gen9 size dword array from the pipeline with the gen8
sized array we create from dynamic state.
Use a static assert in the merge macro and fix this issue by using different
wm_depth_stencil arrays on gen8 and gen9.
Rather than assigning inloc up front, when we don't yet know if it will
be unused, assign it last thing before the legalize pass.
Also, realize when inputs are unused (since for frag shader's we can't
rely on them being removed from ir->inputs[]). This doesn't make sense
if we don't also dynamically assign the inloc's, since we could end up
telling the hw the wrong # of varyings (since we currently assume that
the # of varyings and max-inloc are related..)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Make the interpolation / point-sprite replacement mode setup deal with
varying packing.
In a later commit, we switch to packing just the varying components that
are actually used by the frag shader, so we won't be able to assume
everything is vec4's aligned to vec4. Which would highly confuse the
previous vinterp/vpsrepl logic.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The thread scratch space is thread-local so using the full IA-coherent
stateless surface index (255 since Gen8) is unnecessary and
potentially expensive. On Gen8 and early steppings of Gen9 this is
not a functional change because the kernel already sets bit 4 of
HDC_CHICKEN0 which overrides all HDC memory access to be non-coherent
in order to workaround a hardware bug.
This happens to fix a full system hang when running any spilling code
on a pre-production SKL GT4e machine I have on my desk (forcing all
HDC access to non-coherent from the kernel up to stepping F0 might be
a good idea though regardless of this patch), and improves performance
of the OglPSBump2 SynMark benchmark run with INTEL_DEBUG=spill_fs by
33% (11 runs, 5% significance) on a production SKL GT2 (on which HDC
IA-coherency is apparently functional so it wouldn't make sense to
disable globally).
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Unfortunately Gen7 scratch block reads and writes seem to be hardwired
to BTI 255 even on Gen9+ where that index causes the dataport to do an
IA-coherent read or write. This change is required for the next patch
to be correct, since otherwise we would be writing to the scratch
space using non-coherent access and then reading it back using
IA-coherent reads, which wouldn't be guaranteed to return the value
previously written to the same location without introducing an
additional HDC flush in between.
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Since the query names are not very enlightening, and there are thousands
of them, GALLIUM_HUD=help should only show the first and last query name
for each hardware block.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
No drivers currently implement ARB_geometry_shader4, nor are there
any plans to implement it. We only support the version of geometry
shaders that was incorporated into OpenGL 3.2 / GLSL 1.50.
Signed-off-by: Marta Lofstedt <marta.lofstedt@linux.intel.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Patch adds additional mask for tracking which vertex arrays have
associated vertex buffer binding set. This array can be directly
compared to which vertex arrays are enabled and should match when
drawing.
Fixes following CTS tests:
ES31-CTS.draw_indirect.negative-noVBO-arrays
ES31-CTS.draw_indirect.negative-noVBO-elements
v2: update mask in vertex_array_attrib_binding
v3: rename mask and make it track _BoundArrays which matches what
was actually originally wanted (Fredrik Höglund)
v4: code cleanup, check for GLES 3.1 (Fredrik Höglund)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Fredrik Höglund <fredrik@kde.org>
This was missed in commit 59cfb21d ("targets: use the non-inline sw
helpers").
Fixes build failure:
CXXLD libXvMCgallium.la
../../../../src/gallium/auxiliary/pipe-loader/.libs/libpipe_loader_static.a(libpipe_loader_static_la-pipe_loader_sw.o):(.data.rel.ro+0x0): undefined reference to `sw_screen_create'
collect2: error: ld returned 1 exit status
Makefile:756: recipe for target 'libXvMCgallium.la' failed
make[3]: *** [libXvMCgallium.la] Error 1
Trivial.
Analogous to previous commit. As we no longer have anyone who uses NIR
we can drop the link.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Samuel Iglesias Gonsálvez