Marek Olšák [Thu, 23 Apr 2020 02:45:45 +0000 (22:45 -0400)]
ac: generate FMA for inexact instructions for radeonsi
NIR mostly does this already.
Totals:
SGPRS:
2588520 ->
2591784 (0.13 %)
VGPRS:
1666984 ->
1666888 (-0.01 %)
Spilled SGPRs: 4074 -> 4131 (1.40 %)
Spilled VGPRs: 38 -> 38 (0.00 %)
Private memory VGPRs: 2176 -> 2176 (0.00 %)
Scratch size: 2228 -> 2228 (0.00 %) dwords per thread
Code Size:
52726872 ->
52715468 (-0.02 %) bytes
LDS: 92 -> 92 (0.00 %) blocks
Max Waves: 479872 -> 479897 (0.01 %)
Wait states: 0 -> 0 (0.00 %)
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4696>
Marek Olšák [Thu, 23 Apr 2020 02:38:14 +0000 (22:38 -0400)]
ac: update and document fast math flags used by radeonsi
This should have no effect, because we never use FP division, but
it's safer for the future.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4696>
Marek Olšák [Thu, 23 Apr 2020 01:05:36 +0000 (21:05 -0400)]
ac: force enable -structurizecfg-skip-uniform-regions for LLVM 11
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4696>
Danylo Piliaiev [Thu, 23 Apr 2020 09:41:34 +0000 (12:41 +0300)]
st/mesa: Treat vertex inputs absent in inputMapping as zero in mesa_to_tgsi
After updating vertex inputs being read based on optimized NIR, they may go out
of sync with inputs in mesa IR. Which is translated to TGSI and used together
with NIR if draw doesn't have llvm.
It's much easier to treat such inputs as zero because there is no pass to
entirely get rid of them and they don't contribute to shader's output.
Fixes: d684fb37bfbc47d098158cb03c0672119a4469fe
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2815
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4705>
Samuel Pitoiset [Mon, 20 Apr 2020 07:02:18 +0000 (09:02 +0200)]
gitlab-ci: add lists of expected failures for RADV CI
Currently only supports PITCAIRN, POLARIS10, VEGA10 and NAVI10
with ACO only, but it's a start.
Unfortunately, we have to duplicate and we will have to try to
keep these lists up-to-date, but it's better than nothing.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4689>
Samuel Pitoiset [Thu, 23 Apr 2020 12:05:07 +0000 (14:05 +0200)]
radv: fix robust_buffer_access if enabled via VkPhysicalDeviceFeatures2
It can be enabled via pEnabledFeatures or via vkPhysicalDeviceFeatures2.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4706>
Dave Airlie [Fri, 24 Apr 2020 01:23:19 +0000 (11:23 +1000)]
gallivm: fix stencil border
Fixes:
dEQP-GLES31.functional.texture.border_clamp.unused_channels.depth32f_stencil8_sample_stencil
dEQP-GLES31.functional.texture.border_clamp.sampler.uint_stencil
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Tue, 21 Apr 2020 05:28:38 +0000 (15:28 +1000)]
llvmpipe: clamp color storage for integer types.
If storing to an integer for lower bit size (i.e. 16-bit uint to
10-bit uint), we need to clamp to the maximum value not truncate.
Fixes:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.r16_uint.a2b10g10r10_uint_pack32.optimal_optimal_nearest
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Tue, 3 Dec 2019 06:01:37 +0000 (16:01 +1000)]
llvmpipe: enable stencil only formats. (v2)
This fixes two bugs, one in clearing and one in sign extensions
for S8 only types, and enables it for use.
These are useful for vulkan support later.
v2: move casting to same place as Z casting.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Thu, 16 Apr 2020 07:15:28 +0000 (17:15 +1000)]
llvmpipe/setup: add point size clamping
Fixes
dEQP-GLES2.functional.rasterization.limits.points
dEQP-VK.rasterization.primitive_size.points.point_size*
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Wed, 22 Apr 2020 00:05:59 +0000 (10:05 +1000)]
llvmpipe: fix d32 unorm depth conversions.
When the depth value was 1.0 and was being converted to Z32_UNORM
the conversion would scale it up to INT32_MAX + 1 which would
cause FPToSI to give incorrect results, changing it to use
FPToUI for the unsigned 32-bit case only fixes it.
Fixes:
GTF-GL45.gtf30.GL3Tests.depth_texture.depth_texture_fbo_clear
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Mon, 6 Apr 2020 06:40:04 +0000 (16:40 +1000)]
draw/tess: fix TES patch vertices in.
Fixes CTS KHR-GL45.tessellation_shader.single.max_patch_vertices
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Mon, 6 Apr 2020 07:00:19 +0000 (17:00 +1000)]
llvmpipe: fix ssbo alignment
KHR-GL45.geometry_shader.api.max_shader_storage_blocks
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Tue, 7 Apr 2020 02:08:20 +0000 (12:08 +1000)]
llvmpipe: bump max images to 16
This is needed to make some tests run, and helps for vulkan later.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Mon, 20 Apr 2020 05:39:50 +0000 (15:39 +1000)]
util/indirect: handle stride less than number of parameters.
It's legal to have a stride less than the num of parameters,
in this case no need to try and over map the buffer which asserts
Fixes:
GTF-GL45.gtf43.GL3Tests.multi_draw_indirect.multi_draw_indirect_stride
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Fri, 27 Mar 2020 05:27:26 +0000 (15:27 +1000)]
gallivm/nir: add helper invocation support
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Dave Airlie [Tue, 7 Apr 2020 05:06:26 +0000 (15:06 +1000)]
gallivm/nir: fix image store conversions
This fixes a few of the image store paths, to do the
correct clamping of unsigned/signed values
Fixes: KHR-GLES31.core.layout_binding.block_layout_binding_block_ComputeShader
KHR-GL45.shader_image_load_store.basic-allFormats-store
KHR-GL46.shader_image_load_store.multiple-uniforms
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4574>
Connor Abbott [Fri, 24 Apr 2020 13:53:19 +0000 (15:53 +0200)]
tu: Don't invert point coords
We shouldn't need to invert them, and the Vulkan blob doesn't either.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4733>
Connor Abbott [Thu, 23 Apr 2020 14:23:18 +0000 (16:23 +0200)]
ir3: Remove VARYING_SLOT_PNTC remapping hack
The st now does this for us.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4732>
Connor Abbott [Thu, 23 Apr 2020 14:08:21 +0000 (16:08 +0200)]
st/nir: Fix assigning PointCoord location with !PIPE_CAP_TEXCOORD
This was trying to emulate the effect of mapping GL -> TGSI -> NIR,
but failed to handle VARYING_SLOT_PNTC which led to a kludgy workaround
in freedreno.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4732>
Connor Abbott [Thu, 23 Apr 2020 09:56:07 +0000 (11:56 +0200)]
freedreno/a6xx: Implement PrimID passthrough
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4704>
Connor Abbott [Wed, 22 Apr 2020 13:27:24 +0000 (15:27 +0200)]
tu: Implement PrimID passthrough
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4704>
Connor Abbott [Wed, 22 Apr 2020 15:54:41 +0000 (17:54 +0200)]
ir3: Skip missing VS outputs in VS out map when linking
The hardware is capable of automatically filling in certain values in
the VPC without writing them from the last geometry stage, like
gl_PointCoord or gl_PrimitiveID when there is no GS. However, we *do*
have to enable these outputs (i.e. set the VPC_VAR_DISABLE bit to 0) as
VPC_VAR_DISABLE is really about FS inputs rather than VS outputs. To do
this, we move the computation of the enable bits to ir3_link_add(),
which is also a nice refactor anyway. In addition we detect the PrimID
case specifically so that the driver can program the location.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4704>
Connor Abbott [Wed, 22 Apr 2020 13:04:25 +0000 (15:04 +0200)]
freedreno/a6xx: Document PrimID passthrough registers
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4704>
Joshua Ashton [Fri, 24 Apr 2020 10:13:51 +0000 (11:13 +0100)]
radv: Pass logical device to si_emit_graphics
We'll need this in order to retrieve the va of a bo for a future ext.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4728>
Kristian H. Kristensen [Thu, 23 Apr 2020 18:34:31 +0000 (11:34 -0700)]
freedreno/ir3: Print @tex write mask using 0x%x
That way we can parse it again with the assembler.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4741>
Kristian H. Kristensen [Thu, 23 Apr 2020 18:33:57 +0000 (11:33 -0700)]
freedreno/ir3: Reset lex line number when we start parsing
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4741>
Kristian H. Kristensen [Thu, 23 Apr 2020 18:33:04 +0000 (11:33 -0700)]
freedreno/ir3: Parse, but ignore @in, @out and @tex headers
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4741>
Kristian H. Kristensen [Wed, 22 Apr 2020 23:57:52 +0000 (16:57 -0700)]
freedreno/ir3: Move ir3 assembler to backend compiler
For easier reuse.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4741>
Kristian H. Kristensen [Wed, 22 Apr 2020 23:52:46 +0000 (16:52 -0700)]
freedreno/computerator: Decouple ir3 assembler
Specifically, don't include ir3_asm.h in the parser as that's
computerator specific.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4741>
Andres Gomez [Wed, 22 Apr 2020 13:36:33 +0000 (16:36 +0300)]
Revert "meson,ci: Disable sparse_array tests on windows"
The Wine version in the build image has been upgraded.
This reverts commit
6be65b077743fc80efe061b1e05cb13b2ff1a6b1.
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4678>
Andres Gomez [Wed, 22 Apr 2020 13:51:48 +0000 (16:51 +0300)]
gitlab-ci: install winehq-stable to get 5.0 instead of 4.0
Additionally, purge the winehq-stable package and its dependencies to
avoid crashing when building for s390x.
v2:
- Remove winehq-stable and dependencies for s390x.
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2657
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Daniel Stone <daniels@collabora.com> [v1]
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4678>
Marek Vasut [Mon, 9 Mar 2020 03:14:26 +0000 (04:14 +0100)]
etnaviv: Fix depth stencil ops on GC880/GC2000
This patch fixes depth stencil ops on MX6S GC880 and MX6Q GC2000.
The following dEQPs now pass:
dEQP-GLES2.functional.fragment_ops.depth_stencil.stencil_depth_funcs.*
dEQP-GLES2.functional.fragment_ops.depth_stencil.stencil_ops.*
which is roughly 600 fixed dEQP tests.
The problem is that if the front-facing stencil has a value mask 0x00 and
the back-facing stencil has some non-zero value mask, then the stencil part
of the depth stencil buffer is written with 0x00 unconditionally. The blob
replicates the value mask of the back-facing stencil to the value mask of
the front-facing stencil to achieve correct rendering, replicate the same
behavior here.
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4275>
Rhys Perry [Mon, 6 Apr 2020 10:22:55 +0000 (11:22 +0100)]
radv/aco: enable 8/16-bit storage and int8/int16 on GFX8+
With this, Doom Eternal should now run with ACO on GFX8+.
The generated 8/16-bit storage code is okay but the generated int8/int16
code is currently pretty bad but it works and apparently Doom Eternal
doesn't actually use it (even though it requires it).
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4707>
Rhys Perry [Mon, 6 Apr 2020 10:15:00 +0000 (11:15 +0100)]
aco: lower 8/16-bit integer arithmetic
dEQP-VK.spirv_assembly.type.* passes with the features and extensions
enabled.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4707>
Rhys Perry [Thu, 23 Apr 2020 16:23:21 +0000 (17:23 +0100)]
aco: improve sub-dword emit_split_vector() with sgprs
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 23 Apr 2020 16:17:49 +0000 (17:17 +0100)]
aco: clobber scc in s_bfe_u32 in get_alu_src()
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 19:18:23 +0000 (20:18 +0100)]
aco: handle undef p_create_vector operands in the optimizer
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 19:15:00 +0000 (20:15 +0100)]
aco: vectorize global loads/stores
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 15:01:31 +0000 (16:01 +0100)]
aco: allow 8/16-bit shared loads
These should work now
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:46:09 +0000 (19:46 +0100)]
aco: add and use get_buffer_store_op() helper
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:27:13 +0000 (19:27 +0100)]
aco: refactor visit_store_scratch() to use new helpers
Should support 8/16-bit stores now
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:25:06 +0000 (19:25 +0100)]
aco: refactor visit_store_global() to use new helpers
Should support 8/16-bit stores now
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:22:23 +0000 (19:22 +0100)]
aco: refactor visit_store_ssbo() to use new helpers
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:20:26 +0000 (19:20 +0100)]
aco: refactor store_vmem_mubuf() to use new helpers
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:12:19 +0000 (19:12 +0100)]
aco: refactor store_lds() to use new helpers
It should also work correctly for 8/16-bit stores
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:07:06 +0000 (19:07 +0100)]
aco: add helpers for splitting stores
split_store_data() splits a vector and p_as_uniforms it if needed.
scan_write_mask()/advance_write_mask() are similar to
u_bit_scan_consecutive_range(), but makes it easier to only clear part of
the range and will also give ranges for zero'd bits.
split_buffer_store() is a helper for splitting VMEM/SMEM stores.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Wed, 15 Apr 2020 14:39:44 +0000 (15:39 +0100)]
aco: use emit_load helper for VMEM/SMEM loads
Also implements 8/16-bit loads for scratch/global.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 19:02:31 +0000 (20:02 +0100)]
aco: refactor load_lds to use new helpers
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Thu, 16 Apr 2020 18:51:02 +0000 (19:51 +0100)]
aco: add emit_load helper
This helper is used for recombining split loads, passing the result to
p_as_uniform, aligning the offset down and shifting it right if needed and
handling large constant offsets.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Tue, 14 Apr 2020 19:32:39 +0000 (20:32 +0100)]
aco: add and use RegClass::get() helper
Eventually, we'll probably want to replace the current
RegClass(type, size) constructor with this.
This has a functional change in that get_reg_class() now creates v1/v2
instead of v4b/v8b.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Mon, 6 Apr 2020 19:15:36 +0000 (20:15 +0100)]
aco: be more careful about using SMEM for load_global
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Tue, 14 Apr 2020 19:15:46 +0000 (20:15 +0100)]
radv: allocate larger shader memory slabs if needed
Fixes dEQP-VK.ssbo.phys.layout.random.16bit.scalar.13 hang with ACO
(features needed for the test are implemented in a later commit)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Rhys Perry [Fri, 10 Apr 2020 13:25:46 +0000 (14:25 +0100)]
radv: align buffer descriptor sizes to dword
This is needed to prevent bounds checking issues when load 8/16-bit values
with dword loads.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4639>
Timur Kristóf [Thu, 23 Apr 2020 13:53:09 +0000 (15:53 +0200)]
aco: Move s_setprio to correct place after the gs_alloc_req.
Previously the setprio was inside the branch, so it would only reset
the priority on the first wave, but not the others.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Thu, 23 Apr 2020 13:50:56 +0000 (15:50 +0200)]
aco: Use 24-bit multiplication for NGG wave id and thread id.
Both of them should always fit 24 bits anyway.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Thu, 23 Apr 2020 13:39:56 +0000 (15:39 +0200)]
aco: Use 24-bit multiplication in TCS I/O
The TCS inputs and outputs must always fit into the LDS,
which implies that their addresses also always fit 24 bits.
On AMD GPUs, 24-bit multiplication is much faster than 32-bit
multiplication, so we can take the opportunity to use that
for TCS I/O instead.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Thu, 23 Apr 2020 13:24:47 +0000 (15:24 +0200)]
aco: Const correctness for aco_print_ir.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Wed, 18 Mar 2020 09:40:06 +0000 (10:40 +0100)]
aco: Const correctness for get_barrier_interaction.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Mon, 10 Feb 2020 15:34:56 +0000 (16:34 +0100)]
aco: Abort when RA can't find a register.
Previously, it was just unreachable, which means it will generate
invalid shaders when it encounters a situation when it can't allocate
registers for eg. a large load.
This commit makes it slightly easier to notice such problems without
triggering a GPU hang.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Thu, 23 Apr 2020 13:17:11 +0000 (15:17 +0200)]
aco: Increase barrier_count to 7 to include barrier_barrier.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Mon, 13 Apr 2020 17:03:55 +0000 (19:03 +0200)]
aco: Only store TCS outputs to VMEM when they are read by TES.
Totals from affected shaders (GFX10):
Code Size: 10832 -> 10736 (-0.89 %) bytes
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Timur Kristóf [Mon, 13 Apr 2020 16:35:57 +0000 (18:35 +0200)]
radv: Add inputs read by TES to radv_shader_info.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4536>
Jonathan Marek [Wed, 22 Jan 2020 02:12:57 +0000 (21:12 -0500)]
turnip: add adreno 650
Tile alignment is 96, with gmem alignment of 0x6000
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4608>
Jonathan Marek [Fri, 17 Apr 2020 16:41:14 +0000 (12:41 -0400)]
turnip: use RESOLVE_TS event
This is required on a650 to flush the GMEM store.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4608>
Jonathan Marek [Fri, 17 Apr 2020 16:35:53 +0000 (12:35 -0400)]
turnip: remove unused RB_UNKNOWN_8E04_blit
New blit code doesn't change this value, and different values seem to be
related to the driver version and not the GPU version.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4608>
Mike Blumenkrantz [Thu, 23 Apr 2020 18:21:05 +0000 (14:21 -0400)]
zink: set UBO alignments in nir_intrinsic_load_uniform lowering
resolves this error
error: nir_intrinsic_align_offset(instr) < nir_intrinsic_align_mul(instr) (../src/compiler/nir/nir_validate.c:582)
in ext_packed_depth_stencil-readdrawpixels piglit test
port of
f5b14d983e5afa1b8f75e6f3692830a1ee46d1df
Fixes: fb64954d9dd ("nir: Validate that memory load/store ops work on whole bytes")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4711>
Fritz Koenig [Fri, 24 Apr 2020 03:32:17 +0000 (20:32 -0700)]
freedreno: allow FMT6_8_UNORM as a UBWC format
FMT6_8_UNORM is necessary for NV12 textures.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4722>
Jason Ekstrand [Thu, 13 Feb 2020 05:30:58 +0000 (23:30 -0600)]
spirv: Rewrite CFG construction
This commit completely rewrites the way we extract a structured CFG from
SPIR-V. The new approach is different in a few ways:
1. It does a breadth-first search instead of depth-first. This means
that we've visited the merge node for a construct before we visit
any of the nodes inside the construct. This makes it easier to
validate things like loop and switch nesting.
2. We record more information in the CFG. Earlier commits added a
parent pointer to vtn_cf_node but we now record all of the merge and
other special blocks for each CFG node. This lets us validate
things more precisely.
3. It makes heavy use of merge blocks for walking the CFG. Previously,
we sort of used them as hints for trying to guess the CFG structure
but things got dicey whenever a merge was missing. We had some
heuristics for how to handle short-circuiting if statements but it
was a bunch of special cases.
Now, we make them a fundamental part of walking the CFG. When we
encounter a control-flow construct, we add the body components of
the construct to the BFS work list and then jump to the merge block
if one exists to continue scanning the current CFG nesting level.
If no merge block exists, we assume that means that control-flow
never re-converges in a normal way and that the only way to get back
to normality is with a direct jump such as a loop break or continue.
This should make things far more robust when trying to deal with the
more creative placement (or lack thereof) of merge instructions.
Reviewed-by: Alan Baker <alanbaker@google.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3820>
Closes: #2760
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4446>
Jason Ekstrand [Fri, 6 Mar 2020 22:11:00 +0000 (16:11 -0600)]
anv: Add support for HiZ+CCS
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 22:49:39 +0000 (16:49 -0600)]
intel/isl: Refactor isl_surf_get_ccs_surf
This refactor breaks out a new isl_surf_supports_ccs function which does
most of the validity checking. The isl_surf_get_ccs_surf function calls
this function and then dives right into constructing the CCS aux_surf.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 22:43:10 +0000 (16:43 -0600)]
intel/isl: Delete a misleading comment
Untyped messages are only use on Gen9+ for UBOs and SSBOs. They will
never be used on anything using an isl_surf.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 22:10:11 +0000 (16:10 -0600)]
anv/cmd_buffer: Move anv_image_init_aux_tt higher
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Sat, 7 Mar 2020 01:22:57 +0000 (19:22 -0600)]
anv: Simplify a case in layout_to_aux_usage
If it's depth, the only possible value of planes[plane].aux_usage is
ISL_AUX_USAGE_HIZ at least right now.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Sat, 7 Mar 2020 02:15:08 +0000 (20:15 -0600)]
intel/blorp: Allow more HiZ usages in hiz_clear_depth_stencil
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 22:21:55 +0000 (16:21 -0600)]
anv: Generalize some aux usage checks
For the checks dealing with fast-clear values, we change them to check
for the depth aspect because the distinction there really is between
color and depth more than between HiZ and CCS.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 22:21:30 +0000 (16:21 -0600)]
anv/blorp: Do less hard-coding of aux usages
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 23:14:04 +0000 (17:14 -0600)]
anv: Rework anv_layout_to_aux_state
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 23:31:37 +0000 (17:31 -0600)]
anv: Be more conservative about image view usage
We were ORing together the image and stencil usage rather than actually
following the formula in the spec. This can lead to assertions in other
parts of the driver if we're not careful.
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 17:56:59 +0000 (11:56 -0600)]
anv: Move vk_image_layout_is_read_only higher
While we're at it, we drop some _KHR suffixes
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Jason Ekstrand [Fri, 6 Mar 2020 17:56:27 +0000 (11:56 -0600)]
anv: Add a vk_image_layout_to_usage_flags helper
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4100>
Rafael Antognolli [Thu, 16 Apr 2020 18:47:37 +0000 (11:47 -0700)]
anv: Enable HiZ on multi-layer depth buffers.
Improves The Witcher 3 fps by 2-10% on ICL (depending on the configs and
system).
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4661>
Christian Gmeiner [Sun, 11 Aug 2019 12:03:13 +0000 (14:03 +0200)]
etnaviv: support for using generic blit path
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1641>
Christian Gmeiner [Sun, 11 Aug 2019 11:48:43 +0000 (13:48 +0200)]
etnaviv: call util_blitter_save_fragment_constant_buffer_slot(..)
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1641>
Christian Gmeiner [Fri, 24 Apr 2020 10:47:04 +0000 (12:47 +0200)]
etnaviv: drop default state for FE_HALTI5_ID_CONFIG
It gets emitted when needed - see emit_halti5_only_state(..).
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4729>
Christian Gmeiner [Fri, 24 Apr 2020 07:02:48 +0000 (09:02 +0200)]
docs/features: mark GL_ARB_texture_filter_anisotropic as done for etnaviv
Needs GPUs with HALT0.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4725>
Jonathan Marek [Thu, 23 Apr 2020 16:31:15 +0000 (12:31 -0400)]
freedreno/ir3: fix incorrect conversion folding
Fixes dEQP-VK.glsl.builtin.function.pack_unpack.unpackhalf2x16_compute
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4708>
Jonathan Marek [Thu, 23 Apr 2020 03:21:20 +0000 (23:21 -0400)]
freedreno/ir3: set even bit for f2f16_rtne
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4708>
Jonathan Marek [Thu, 23 Apr 2020 03:20:26 +0000 (23:20 -0400)]
freedreno/ir3: fix 16-bit ssbo access
Update cat6 instruction type, and shift 1 in lower_offset_for_ssbo.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4708>
Rhys Perry [Fri, 24 Apr 2020 10:58:17 +0000 (11:58 +0100)]
aco: fix outdated label_vec from p_create_vector labelling
Fixes random dEQP-VK.transform_feedback.fuzz.* crashes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 2dc550202e82c5da198ad0a416a5d24dd89addd8
('aco: copy-propagate p_create_vector copies of vectors')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4730>
Jason Ekstrand [Thu, 23 Apr 2020 02:35:48 +0000 (21:35 -0500)]
nir/lower_subgroups: Mask off unused bits in ballot ops
Thanks to VK_EXT_subgroup_size_control, we can end up with
gl_SubgroupSize being as low as 8 on Intel.
Fixes: d10de253097 "anv: Implement VK_EXT_subgroup_size_control"
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4694>
Jason Ekstrand [Thu, 23 Apr 2020 02:18:03 +0000 (21:18 -0500)]
anv: Drop an assert
Ever since Vulkan 1.2, this feature has been in core so enabling the
extension is no longer required.
Fixes: 4ef3f7e3d37e "anv: Enable Vulkan 1.2 support"
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4694>
Marek Olšák [Thu, 23 Apr 2020 05:37:51 +0000 (01:37 -0400)]
radeonsi: use pipe_blend_state::max_rt to update fewer blend registers
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4698>
Marek Olšák [Fri, 27 Mar 2020 01:48:48 +0000 (21:48 -0400)]
ac,radeonsi: simplify checking for Navi1x chips
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4698>
Marek Olšák [Thu, 16 Apr 2020 17:32:27 +0000 (13:32 -0400)]
ac: out-of-order rasterization is not supported on gfx10
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4698>
Jonathan Marek [Thu, 23 Apr 2020 16:10:07 +0000 (12:10 -0400)]
turnip: divide cube map depth by 6
This matches the GL driver and fixes these tests:
dEQP-VK.glsl.texture_functions.query.texturesize.samplercubearray*
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4709>
Jason Ekstrand [Wed, 22 Apr 2020 20:49:25 +0000 (15:49 -0500)]
spirv: Fix passing combined image/samplers through function calls
Fixes dEQP-VK.spirv_assembly.instruction.function_params.sampler_param
cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4684>
Jason Ekstrand [Wed, 22 Apr 2020 20:47:28 +0000 (15:47 -0500)]
nir/opt_deref: Remove certain sampler type casts
The SPIR-V parser sometimes generates casts from specific sampler types
like sampler2D to the bare sampler type. This results in a cast which
causes heartburn for drivers but is harmless to remove.
cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4684>
Jason Ekstrand [Wed, 22 Apr 2020 19:05:13 +0000 (14:05 -0500)]
spirv: Use nir_const_value for spec constants
When we originally wrote spirv_to_nir we didn't have a good scalar value
union to handily use so we rolled our own thing for spec constants. Now
that we have nir_const_value, we can use that and simplify a bunch of
the spec constant logic.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4675>
Jason Ekstrand [Wed, 22 Apr 2020 18:43:51 +0000 (13:43 -0500)]
turnip: Properly handle all sizes of specialization constants
cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4675>
projects / mesa.git / log