The spec citation says it's just for when the RT write message BTI might
point to a different RT, and if we don't have any color attachments then
we won't have one of those at all.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21508>
Power-of-two SW stack sizes are prone to causing collisions in the
hashing function used by the L3 to map memory addresses to banks,
which can cause stack accesses from most DSSes to bottleneck on a
single L3 bank. Fix it by padding the SW stack stride by a single
cacheline if it was a power of two. This has been reported by Felix
DeGrood to improve Quake2 RTX performance by ~30% on DG2-512 in
combination with other RT patches Lionel Landwerlin has been working
on.
Many thanks to Felix DeGrood for doing much of the legwork and
providing several iterations of Q2RTX performance counter dumps which
eventually prompted me to consider the hash collision theory and
motivated this patch, and for providing additional performance counter
dumps confirming that there is no longer an appreciable imbalance in
traffic across L3 banks after this change.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21461>
This allows us to also inherit `-mfpmath=sse` added in previous commit.
Acked-by: Yonggang Luo <luoyonggang@gmail.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21371>
Now that all color blend bits are dynamic, emit_cb_state() is doing
almost nothing and half of that is wrong.
In the case that color write enable is dynamic, at the time the pipeline
state is emitted, it sees all the color attachments as having write
disabled and stores the WriteDisabled bit for each channel.
When all dynamic state is flushed, we have the right values already but
the values recorded into the command buffer get ORed with the ones
stored in the pipeline, and so WriteDisabled tag along when they
shouldn't.
Since all disabled color attachments are handled already when dynamic
state is flushed, there's no point in doing so at pipeline creation
time too. And since the only other thing done by emit_cb_state() is
writing three hardcoded values, they might as well be taken care of in
the same place as everything else.
Fixes CTS from the future:
dEQP-VK.pipeline.*.extended_dynamic_state.*.color_blend_equation_*dynamic*
dEQP-VK.pipeline.*.extended_dynamic_state.*.color_blend_all_*
Fixes: fc3fd7c69e (anv: dynamic color write mask)
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21509>
Not necessary but, all things being equal, be consistent with Iris.
Now that intel_debug_write_identifiers() already add the padding,
there's no need to include extra "+ 8" to the offset.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21479>
We need to emit 3DSTATE_HS for each primitive with tessellation.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21308>
Patch packs 3DSTATE_HS state during pipeline creation but it
gets emitted only before 3DPRIMITIVE. We will later need this
to implement a workaround.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21308>
From the Haswell PRM Vol. 2b, 3DSTATE_WM::Pixel Shader Kill Pixel:
"This bit is required to be ENABLED in the following situations:
- The API pixel shader program contains "killpix" or "discard"
instructions, or other code in the pixel shader kernel that can
cause the final pixel mask to differ from the pixel mask received
on dispatch.
- A sampler with chroma key enabled with kill pixel mode is used by
the pixel shader.
- Any render target has Alpha Test Enable or AlphaToCoverage Enable
enabled.
- The pixel shader kernel generates and outputs oMask."
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19876>
Tigerlake PRM: Volume 2c: Command Reference: Registers Part 2 - Registers M through Z
RCU_MODE :: Compute Engine Enable
This bit indicates if Compute Engine (a.k.a Dual Context or Multi
Context) is enabled or not. This bit must be treated as global
control for enabling and disabling of compute engine. Hardware
allocates required resources for the compute engine based on this
bit.
....
HW reserves 4KB of URB space...
Right now no gen12 platform has Dual Context enabled in kernel side,
exposing a compute engine but that can change, so here adding
has_compute_engine to intel_device_info and only reserving URB space
if compute engine is available.
While at it also fixing the error path when pb_slabs_init() fails.
Bspec: 46034
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21031>
Hoping that I didn't miss any, this *should* add assertions
to all functions and passes which explicitly handle 'nir_loop'.
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13962>
Rob added these new helpers a while back, which freedreno and radeonsi
both share. We should use them too. The new helpers use variables and
system value intrinsics, so we can drop the explicit binding table
creation and just use the normal paths.
Because we have to rewrite the system value uploading anyway, we drop
the scrambling of the default tessellation levels on upload, and instead
let the compiler go ahead and remap components like any normal shader.
In theory, this results in more shuffling in the shader. In practice,
we already do MOVs for message setup. In the passthrough shaders I
looked at, this resulted in no extra instructions on Icelake (SIMD8
SINGLE_PATCH) and Tigerlake (8_PATCH). On Haswell, one shader grew by
a single instruction for a pittance of cycles in a stage that isn't a
performance bottleneck anyway. Avoiding remapping wasn't so much of an
optimization as just the way that I originally wrote it. Not worth it.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20809>
Since the dirty range started out as 0..0, you would have 0..VBend as the
new dirty range on the first draw, and if your VB was >32b then you'd
flush every time you used it. Instead, if there's no existing dirty range
then just set it to our new VB's range.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21370>
Since the dirty range started out as 0..0, you would have 0..VBend as the
new dirty range on the first draw, and if your VB was >32b then you'd
flush every time you used it. Instead, if there's no existing dirty range
then just set it to our new VB's range.
Perf results with zink+anv on my CFL:
sauerbraten: +24.8182% +/- 0.602077% (n=5)
portal-2-v2.trace: +4.64289% +/- 0.285285% (n=5)
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21370>
VK_LAYER_EXPORT is going away in the next Vulkan header update. We
already have a PUBLIC macro in util/macros.h which does the same thing.
Unlike VK_LAYER_EXPORT, it should work in Windows too.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21225>
