| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This should be a win for most loops, which tend to have uniform control
flow.
More importantly, it exposes important information to live variables: that
the break/continue here means that our jump target may have access to
values that were live on our input. Previously, we were just setting the
exec mask and letting control flow fall through, so an intervening def
between the break and the end of the loop would appear to live variables
as if it screened off the variable, when it didn't actually.
Fixes a regression in glsl-vs-loop-redundant-condition.shader_test when a
perturbing of register allocation caused a live variable to get stomped.
Cc: 13.0 <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 8e5ec33f1151dd82402bdfdaa4fff7c284e49a1c)
|
|
|
|
|
|
|
| |
Fixes piglit glsl-fs-shadow2D-clamp-z.
Cc: <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 08d51487e3b8cfb14ca2ece9545b2e2ed344e3cc)
|
|
|
|
|
|
|
|
|
|
| |
It's much better to just skip the draw call entirely. Getting this
information out of register allocation will also be useful for
implementing threaded fragment shaders, which will need to retry
non-threaded if RA fails.
Cc: <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 4d019bd703e7c20d56d5b858577607115b4926a3)
|
|
|
|
|
|
|
|
| |
The 1/W was apparently not accurate enough, and we were getting sparklies
in the distance. The closed driver also did a N-R step here.
Cc: <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 283d4d18e598793bbff7d9ba5a601bced9b36542)
|
|
|
|
|
|
|
|
|
| |
If a conditional assignment is only conditioned on the exec mask, that's
still screening off the value in the executed channels (and, since we're
not storing to the unexcuted channels, we don't care what's in there).
Fixes a bunch of extra register pressure on Processing's Ribbons demo,
which is failing to allocate.
|
|
|
|
|
| |
Fixes 100 piglit tests since the assertions were added to nir.h. What's
amazing is that these tests used to pass, even when casting garbage.
|
|
|
|
|
|
|
|
|
|
| |
One of NIR's invariants is that control flow lists always start and end
with blocks. There's no good reason why we should return a cf_node from
these functions since we know that it's always a block. Making it a block
lets us remove a bunch of code.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
VC4 was running into a major performance regression from enabling control
flow in the glmark2 conditionals test, because of short if statements
containing an ffract.
This pass seems like it was was trying to ensure that we only flattened
IFs that should be entirely a win by guaranteeing that there would be
fewer bcsels than there were MOVs otherwise. However, if the number of
ALU ops is small, we can avoid the overhead of branching (which itself
costs cycles) and still get a win, even if it means moving real
instructions out of the THEN/ELSE blocks.
For now, just turn on aggressive flattening on vc4. i965 will need some
tuning to avoid regressions. It does looks like this may be useful to
replace freedreno code.
Improves glmark2 -b conditionals:fragment-steps=5:vertex-steps=0 from 47
fps to 95 fps on vc4.
vc4 shader-db:
total instructions in shared programs: 101282 -> 99543 (-1.72%)
instructions in affected programs: 17365 -> 15626 (-10.01%)
total uniforms in shared programs: 31295 -> 31172 (-0.39%)
uniforms in affected programs: 3580 -> 3457 (-3.44%)
total estimated cycles in shared programs: 225182 -> 223746 (-0.64%)
estimated cycles in affected programs: 26085 -> 24649 (-5.51%)
v2: Update shader-db output.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> (v1)
|
|
|
|
|
| |
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
|
|
|
|
|
| |
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
|
|
|
|
|
| |
This is a preparation step for having multiple jobs being queued up at the
same time.
|
|
|
|
|
| |
I missed this while adding loop support because the discard test inside a
loop was crashing before, anyway. Fixes piglit glsl-fs-discard-04.
|
|
|
|
| |
Based vaguely on a patch by jonasarrow on github.
|
|
|
|
|
|
|
|
|
|
| |
Before this series, the code generation path was:
GLSL IR -> TGSI -> NIR -> NIR clone -> QIR -> QPU
Now it's (generally)
GLSL IR -> NIR -> NIR clone -> QIR -> QPU
|
| |
|
|
|
|
| |
We end up with this when doing GLSL-to-NIR.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In the GLSL-to-NIR conversion of VC4, I had a bit of trouble with what I
was calling the "state uniforms" that I was putting into the NIR fighting
with its other lowering passes. Instead of using magic uniform base
numbers in the backend, follow the lead of load_user_clip_plane and just
define system values for them.
v2: Fix unintended change to channel_num, drop unspecified const_index
value on blend_const_color_r_float.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
|
| |
|
|
|
|
| |
Avoids another multiplication by 4 of the base in the NIR.
|
|
|
|
|
| |
The scalarizing of FS inputs can be done in a non-driver-dependent manner,
so extract it out of the driver.
|
|
|
|
|
| |
The const_index[] values have always felt magic, and this documents them a
bit better.
|
|
|
|
|
|
|
| |
This lets TTN-using drivers handle FRAG_RESULT_DEPTH the same between all
their source paths.
Reviewed-by: Rob Clark <robdclark@gmail.com>
|
|
|
|
| |
In the case of debugging a crash in TTN, this is nice to have.
|
|
|
|
|
|
|
| |
This works out to be a wash in terms of memory usage: We use more memory
to store the separate ALU instructions, but we optimize out a lot of code
as well. The main result, though, is that we do more of our work at link
time rather than draw time.
|
|
|
|
|
|
|
|
|
|
|
|
| |
We don't want to bake the whole array into the FS key, because of the
hashing overhead. But we can keep a set of the arrays seen, and use a
pointer to the copy in as the array's proxy.
Between this and the previous patch, gl-1.0-blend-func now passes on
hardware, where previously it was filling the 256MB CMA area with shaders
and OOMing.
Drops 712 shaders from shader-db.
|
|
|
|
|
|
|
| |
The compiled_fs_id is a proxy for the vc4->prog.fs->input_slots[], but
only the VS dereferences it.
Drops 754 shaders from shader-db.
|
|
|
|
| |
It's a pretty big block, and I was about to make it bigger.
|
|
|
|
|
|
|
|
|
|
| |
We were baking in the LOD of the source level to each shader. Instead,
pass it in as a uniform -- this requires storing it to a temp register,
but that's better than compiling a ton of separate shaders:
total instructions in shared programs: 115032 -> 115036 (0.00%)
instructions in affected programs: 96 -> 100 (4.17%)
LOST: 572
|
|
|
|
| |
I keep wanting to see this version of the NIR.
|
|
|
|
|
| |
We don't tell the hardware whether we're computing depth, so we need
to manage early Z state manually. Fixes piglit early-z.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To support general GL_TEXTURE_BASE_LEVEL we have to copy to a temporary
miptree. However, if a single level is being selected, we can use the
existing miptree and force all the sampling to be from that particular
level.
This avoids a ton of software fallbacks in glGenerateMipmaps(), which uses
base levels in the blit implementation in gallium. Improves "glmark2 -b
terrain" from 2 fps to 3 (perhaps some more precision would be useful?),
and cuts its CPU usage during the benchmarking from ~30% to ~10% (total
CPU time from 8.8s to 7.6s).
|
|
|
|
|
|
| |
The compiler uses the per-stage flags already, so it didn't need this.
vc4_uniforms was using it, so just replace it with both of the stage flags
for now.
|
|
|
|
|
|
|
|
| |
If a block might be entered from multiple locations, then the uniform
stream will (probably) be at different points, and we need to make sure
that it's pointing where we expect it to be. The kernel also enforces
that any block reading a uniform resets uniforms, to prevent reading
outside of the uniform stream by using looping.
|
| |
|
| |
|
|
|
|
|
|
|
| |
Previously, there were occasionally NIR registers in our programs, but
they were always actually used SSA-only. Now that we're trying to support
control flow, we need to actually conditionally move to registers based on
whether channels are active or not.
|
|
|
|
|
| |
For now it's still always false, but I need it in place for kernel
backwards compat support as I extend the backend for control flow.
|
|
|
|
|
|
|
|
| |
We have the prior list_foreach() all over the code, but I need to move
where instructions live as part of adding support for control flow. Start
by just converting to a helper iterator macro. (The simpler
"qir_for_each_inst()" will be used for the for-each-inst-in-a-block
iterator macro later)
|
|
|
|
|
|
|
|
| |
This avoids a bunch of code gen regressions when enabling loops in vc4.
Prior to that, the GLSL that would have generated these optimizable phi
nodes was being lowered to csels between either (undef, a) or (a, a), and
those were being dealt with by nir_opt_undef and nir_opt_algebraic.
|
|
|
|
|
|
| |
Now that we're about to start generating control flow in our NIR, we want
this in place. It optimizes things frequently in the CS, when the GL VS
has control flow that doesn't affect the vertex position.
|
|
|
|
|
| |
This gets us precompile of vertex shaders at the state tracker level as
well.
|
|
|
|
|
| |
Now we have an immutable nir shader in our shader's CSO that we can clone
and lower/optimize.
|
|
|
|
|
|
| |
This allows the same pipe_shader_state to be referenced from multiple
contexts. Since our pipe_shader_state is treated as immutable (other than
the variant number) within the driver, this is no problem.
|
|
|
|
|
|
|
|
| |
This will be generated by glsl_to_nir, and it turns out that this is a
more code-efficient path than the floating point math, anyway.
No change on shader-db, but drops an instruction in piglit's
glsl-fs-frontfacing.
|
|
|
|
| |
Reviewed-by: Eric Anholt <eric@anholt.net>
|
|
|
|
|
| |
This should get us the same decode code generated, but with a lot less
custom code in the driver.
|
|
|
|
|
| |
This means doing Newton-Raphson on the RCP, but it's probably actually a
good thing to be accurate on.
|
|
|
|
|
| |
This makes fewer programs with loops assertion fail, replacing them with
the rendering failure warning.
|
|
|
|
|
| |
In particular it's been hard to find the point where we switch from
dumping pre-optimization QIR and post-optimization QIR.
|
|
|
|
|
|
|
| |
This is one of two uses of the current QIR CSE pass according to
shader-db. The NIR pass means that we'll end up doing Newton-Raphson on
our RCP, which we weren't doing before, but that's probably actually a
good thing.
|