Adds a toggle to force the use of small shader-visible heaps and
whitebox tests to verify bindgroup encoding correctness.
BUG=dawn:155
Change-Id: I4118b850d9f2cb445ae805aa68ebf4fab671261b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16960
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch adds all the validations on the creation of bind groups with
read-only and write-only storage textures.
1. Only the textures with STORAGE usage can be used as read-only or
write-only storage textures.
2. The format of the texture view used as read-only or write-only
storage texture must match the corresponding declarations in the
bind group layout.
3. The texture view dimension of the texture view used as read-only or
write-only storage texture must match the corresponding declaration
in the bind group layout.
Note that we don't test the match of the sample count because currently
we don't support sample count > 1 when creating a texture with STORAGE
usage and creating a bind group layout with read-only or write-only
storage textrue binding type.
This patch also adds a unit test to verify that it is invalid to create
a bind group layout with either read-only or write-only storage texture
binding type and dynamic offsets.
This patch also implements the bind group with storage textures on
Vulkan to make the Vulkan fuzzer happy with this patch.
BUG=dawn:267
TEST=dawn_unittests
Change-Id: Iee1b3c49671aae8a5424882b035624248d5fc281
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17583
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
ErrorData should not be discarded. The error must be handled.
Bug: chromium:1063810
Change-Id: I55e782e499f307b91790db10a42878afdc7540dd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17501
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This is a regression that was caught by the WebGPU CTS after
https://dawn-review.googlesource.com/c/dawn/+/17240 where the
BindGroupLayout validation was changed to use a set, and the check for
KMaxBindingsPerGroup removed.
This CL also adds a regression test.
Bug: dawn:354
Change-Id: I0cda545b9df7220ca53216878cf24ac8ce880648
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17620
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This reverts commit 7c24b6b2ff.
Reason for revert: We can reland this CL now as the fix in shaderc has
been merged.
Note that the declaration whether a storage texture is multisampled or not
cannot be extracted correctly in SPVC. The fix in Dawn will be added after
it is fixed in shaderc.
Original change's description:
> Revert "Check bind group layout with storage texture in pipeline descriptors"
>
> This reverts commit 63f2666ee7.
>
> Reason for revert: causes failures in dawn_unittests after Dawn uses SPVC by default. We need a fix in both SPVC and this CL before re-landing.
>
> Original change's description:
> > Check bind group layout with storage texture in pipeline descriptors
> >
> > This patch adds all the validations on the use of bind group layout with
> > read-only storage texture and write-only storage texture in the creation
> > of pipeline objects.
> >
> > 1. GPUBindGroupLayout.bindingType must match the type of the storage
> > texture variable (read-only or write-only) in shader.
> > 2. GPUBindGroupLayout.storageTextureFormat must be a valid texture
> > format that supports STORAGE usage.
> > 3. GPUBindGroupLayout.storageTextureFormat must match the storage
> > texture format declaration in shader.
> > 4. GPUBindGroupLayout.textureDimension must match the storage texture
> > dimension declared in shader.
> >
> > BUG=dawn:267
> > TEST=dawn_unittests
> >
> > Change-Id: Ifa3c2194dc76de14f790a0a73868e69bbb31c814
> > Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17167
> > Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
> > Reviewed-by: Kai Ninomiya <kainino@chromium.org>
>
> TBR=cwallez@chromium.org,kainino@chromium.org,yunchao.he@intel.com,jiawei.shao@intel.com,shaobo.yan@intel.com,hao.x.li@intel.com,enga@chromium.org,jiajie.hu@intel.com
>
> Change-Id: Idb4083b11f22fa7e4c5c8477bc4b65b58900746e
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: dawn:267
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17380
> Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
> Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
TBR=cwallez@chromium.org,kainino@chromium.org,yunchao.he@intel.com,jiawei.shao@intel.com,shaobo.yan@intel.com,hao.x.li@intel.com,enga@chromium.org,jiajie.hu@intel.com
Bug: dawn:267, chromium:1063570
Change-Id: If762cbb206e738f4e54e75c88d506fdf3a44f280
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17461
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
(Manual)?SwapChainTests had a narrowing conversions from double to float
Bug: dawn:269
Change-Id: I5b07f53556fa5461877631904fa2c1ab6c8e6596
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17540
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Also removes a MaybeError from a function that can't return errors.
Bug: chromium:1063810
Change-Id: I9125b2b164a2892006442e26d6070d226fec8665
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17481
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
SPVC only allows SPIR-V 1.0 while WebGPU samples and other existing
content outputs SPIR-V 1.3. Turning SPVC on by default broke all of
them, so we disable SPVC until we decide what to do.
Bug: chromium:1063810
Change-Id: Iedcf31960eb25b0974e8c67030232902cd58a4a1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17480
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Also fixes a bug where we weren't validating duplicating
bindings in the shader, and where dynamic offset validation
could be incorrectly fetching the wrong bindings.
Bug: dawn:354
Change-Id: I93178c34eb4d43119e8b9de5738ae4596e9277cd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17240
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
The webgpu.h surface-based swapchains are implement on Metal which
required adding the present mode to NewSwapChainBase.
Additional automated tests are added which require getting the Instance
so a new getter is added to DawnTest. Additional some the state tracking
of swapchains is performed in the backend, so the
SwapChainValidationTests are turned into regular DawnTests so they can
check backends do the correct state tracking. To not lose coverage of
the Null backend, a NullBackend() DawnTestParam factory is added.
Finally swapchains cannot be entirely tested in an automated fashion, so
a new example is added called "ManualSwapChainTests" that allows
manually checking a number of properties. Documentation of the controls
and a manual test plan is in a comment at the top of the example's
source.
Bug: dawn:269
Change-Id: If62fffc29a6cefdbec62747d01c523e2a5475715
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17181
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
They were failing because services of the Metal backend like the
MapRequestTracker thought that work was still pending since the device
only waited for previous commands to complete, and not also pending
commands.
Bug: dawn:68
Change-Id: I917ecefe90707b0c89f7e9b4b7379a98ed3956d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17320
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This reverts commit 63f2666ee7.
Reason for revert: causes failures in dawn_unittests after Dawn uses SPVC by default. We need a fix in both SPVC and this CL before re-landing.
Original change's description:
> Check bind group layout with storage texture in pipeline descriptors
>
> This patch adds all the validations on the use of bind group layout with
> read-only storage texture and write-only storage texture in the creation
> of pipeline objects.
>
> 1. GPUBindGroupLayout.bindingType must match the type of the storage
> texture variable (read-only or write-only) in shader.
> 2. GPUBindGroupLayout.storageTextureFormat must be a valid texture
> format that supports STORAGE usage.
> 3. GPUBindGroupLayout.storageTextureFormat must match the storage
> texture format declaration in shader.
> 4. GPUBindGroupLayout.textureDimension must match the storage texture
> dimension declared in shader.
>
> BUG=dawn:267
> TEST=dawn_unittests
>
> Change-Id: Ifa3c2194dc76de14f790a0a73868e69bbb31c814
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17167
> Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
TBR=cwallez@chromium.org,kainino@chromium.org,yunchao.he@intel.com,jiawei.shao@intel.com,shaobo.yan@intel.com,hao.x.li@intel.com,enga@chromium.org,jiajie.hu@intel.com
Change-Id: Idb4083b11f22fa7e4c5c8477bc4b65b58900746e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: dawn:267
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17380
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch adds all the validations on the use of bind group layout with
read-only storage texture and write-only storage texture in the creation
of pipeline objects.
1. GPUBindGroupLayout.bindingType must match the type of the storage
texture variable (read-only or write-only) in shader.
2. GPUBindGroupLayout.storageTextureFormat must be a valid texture
format that supports STORAGE usage.
3. GPUBindGroupLayout.storageTextureFormat must match the storage
texture format declaration in shader.
4. GPUBindGroupLayout.textureDimension must match the storage texture
dimension declared in shader.
BUG=dawn:267
TEST=dawn_unittests
Change-Id: Ifa3c2194dc76de14f790a0a73868e69bbb31c814
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17167
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
The CTS has been changed to use the version of SPIR-V that spvc is
expecting, so the issues that previously caused a rollback should no
longer occur.
BUG=dawn:337
Change-Id: Ied5c39059d417359a697930287728faf57289c62
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16900
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This adds parity for reflection features in the spvc-only code path up
to 421684f. Additionally changes will be needed to bring support to
parity up to HEAD.
BUG=dawn:337
Change-Id: I2143587c54018da766934fa60d80aedd1ccd151b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17165
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
CopyResource may only be used for resources that have exactly the same
format, dimension, mips, layers. And it can only be used if the entire
texture region is copied.
Bug: dawn:353
Change-Id: Ia8f96cc10c88fe026e23bce2d0532624725b12e0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16984
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
The roll in Chromium is failing, because Dawn doesn't handle all of the new enum
values, but I cannot land the 'proper' fix in Dawn, because Chromium doesn't yet
know about the new enum values. Yay, cross dependencies.
So my solution is land the default case in Dawn, land the DEPS roll in Chromium,
then patch in the 'proper' cases in Dawn.
Roll third_party/shaderc/ a7657e4fa..95185d920 (2 commits)
a7657e4fa2..95185d920a
$ git log a7657e4fa..95185d920 --date=short --no-merges --format='%ad %ae %s'
2020-03-17 rharrison Rolling 4 dependencies (#1002)
2020-03-16 rharrison Add reflection support for storage textures (#1001)
Created with:
roll-dep third_party/shaderc
Change-Id: Id73b38643d9a6d418be9c75e218fdad7a18839bc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/17164
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Use D3D12's QueryDeviceVideoMemoryInfo to get the OS-determined process
budget. Also introduces an export for reserving some amount of process
memory - which keeps Dawn from using the entire process's budget.
Bug: dawn:193
Change-Id: I6c17bd703d7cb24759bcee89c03add46944fec8c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16383
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch adds the validation on the storage texture format declared in
shaders when we create a rendering or compute pipeline with read-only or
write-only storage textures.
This patch also fixes a typo in the TextureValidationTest.
BUG=dawn:267
TEST=dawn_unittests
Change-Id: Id302b4b7803d7e03b57c61de1290cc71ba940e2c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16940
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch adds the validation on the creation of the bind group
layout with read-only storage texture, write-only storage texture
and read-write storage texture. Currently read-write storage textures
are not supported in any shader stages.
This patch also fixes chromium:1061156.
BUG=chromium:1061156, dawn:267
TEST=dawn_unittests, dawn_end2end_tests
Change-Id: Ib42678719df48565a46e39f21c34ec640960dcdc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16920
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Now that all backends use slab-allocated bind groups, this patch also
moves the BindGroup implementation with owned-data into the Null backend.
Bug: dawn:340
Change-Id: I08a952075b382008fb82f1fbab3f779cc05bc2a3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16747
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Track what heaps are required to be resident upon command list
submission.
Bug: dawn:193
Change-Id: Icefe5aed047140f401d51018937c5364b3ed7c4f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16382
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch adds the basic validation of read-only storage texture,
write-only storage texture and read-write storage texture as new
binding types with no bind group layout provided in the creation of
pipeline state objects.
- Read-only storage textures can be used in vertex, fragment and
compute shaders.
- Write-only storage textures can only be used in compute shaders
due to the limitation on Metal.
- Read-write storage textures are not allowed now and they are
reserved to be supported as an extension in the future.
BUG=dawn:267
TEST=dawn_unittests
Change-Id: Iffc432f29a855b85d59451cb3c50269e03b84627
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16661
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Rather than destory GPU descriptor heaps upon being switched out,
heaps are stored in a list where they can be re-used once the GPU
is no longer using them.
BUG=dawn:155
Change-Id: I2074573e354f114c45afe9895e8515980d325852
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16282
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
When creating a directly allocated resource in D3D12, also create a
dawn_native::d3d12::Heap to represent that allocation alongside the
ResourceHeapAllocation. This matches D3D12's allocation model when using
CreateCommittedResource and makes residency management much easier.
Bug: dawn:193
Change-Id: I2280863dcfca57bad72962a2b097f8f2d4cc7dad
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16381
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This patch intends to fix a crash issue when creating a rendering
pipeline with storage buffer declared in vertex shader and pipeline
layout is not set.
Without this patch, in PipelineLayoutBase::CreateDefault() the
bindingSlot.visibility is always set to Fragment and Compute when it is
a storage buffer, therefore a crash happens at the failure of the
assertion modules->IsCompatibleWithPipelineLayout() when the storage
buffer is actually declared in the vertex shader.
BUG=dawn:276
TEST=dawn_unittests
Change-Id: I56876a97d53ead5ed226dc1b9bbed1a77156b2b2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16564
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Use ExternalImageDescriptor->isCleared to set the clear status of
subresources so it can be correctly lazy cleared when used.
Also remove old Wrap path that uses regular texture descriptors
since we have moved to use ExternalImageDescriptor.
Bug: chromium:1036080
Change-Id: Icb605dbf3cf3f0dc8a30287e8b9b8d9134805112
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16320
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This reverts commit 4ae315b0d1.
Reason for revert: crbug.com/1059205
Bug: chromium:1059205
Original change's description:
> Vulkan: Report and enable subgroup size control device extension.
>
> Certain Vulkan ICDs (Intel ones notably) will compile SPIR-V
> shaders with an liberal, compiler-selected, subgroup size (i.e.
> either 8, 16 or 32). For more context, see [1].
>
> This can be a problem for compute, when one shader stores data
> in device memory using a subgroup-size dependent layout, to be
> consumed by a another shader. Problems arise when the compiler
> decides to compile both shaders with different subgroup sizes.
>
> To work-around this, the VK_EXT_subgroup_size_control device
> extension was introduced recently: it allows the device to
> report the min/max subgroup sizes it provides, and allows
> the Vulkan program to control the subgroup size precisely
> if it wants to.
>
> This patch adds support to the Vulkan backend to report and
> enable the extension if it is available. Note that:
>
> - The corresponding VkStructureType enum values and
> struct types are not rolled to the third-party Vulkan
> headers used by Dawn yet, so vulkan_platform.h has been
> modified to define them if necessary. This can be
> removed in the future when the Vulkan-Headers are
> updated in a different patch.
>
> - This modifies VulkanDeviceInfo::GatherDeviceInfo() to
> use VkGetPhysicalDevice{Properties2,Features2} if the
> VK_KHR_get_device_properties2 instance extension is
> available. Otherwise, the Vulkan 1.0 APIs
> VkGetPhysicalDevice{Properties,Features} are used instead
> (and it is assumed that no subgroup size control is
> possible).
>
> - This changes the definition of VulkanDeviceKnobs to
> make room for the required pNext-linked chains of
> extensions.
>
> - A helper class, PNextChainBuilder is also provided in
> UtilsVulkan.h to make it easy to build pNext-linked
> extension struct chains at runtime, as required when
> probing device propertires/features, or when
> creating a new VkDevice handle.
>
> Apart from that, there is no change in behaviour in this CL.
> I.e. a later CL might force a specific subgroup size for
> consistency, or introduce a new API to let Dawn clients
> select a fixed subgroup size.
>
> [1] https://bugs.freedesktop.org/show_bug.cgi?id=108875
>
> Change-Id: I524af6ff3479f25b0a8bb139a062fe632c826893
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16020
> Reviewed-by: Austin Eng <enga@chromium.org>
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
> Commit-Queue: Corentin Wallez <cwallez@chromium.org>
TBR=cwallez@google.com,cwallez@chromium.org,enga@chromium.org,enga@google.com,david.turner.dev@gmail.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I893d771d7effdf83685dda3edac8a08f98d2f6e5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16522
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This reverts commit 56f1678437.
Reason for revert: crbug.com/1059205
Bug: chromium:1059205
Original change's description:
> Vulkan: Enforce fixed subgroup size for compute shaders.
>
> This CL ensures that, on architectures with a varying subgroup size,
> compute shaders are always compiled with a fixed subgroup size to
> avoid consistency issues when one shader writes data in a subgroup-size
> dependent layout to GPU memory, to be read by another shader in a
> future dispatch.
>
> At the moment, only Intel ICDs are known to implement this [1],
> and the code uses a heuristics to chose the size of 16, which seems to
> be the sweet spot according to Intel engineers.
>
> + Update the PNextChainBuilder class to deal with the fact that
> VkComputePipelineCreateInfo::pNext is defined as a const void*,
> which created compiler errors in the previous implementation.
>
> [1] https://bugs.freedesktop.org/show_bug.cgi?id=108875
>
> Change-Id: I332faa53b9f854a8abe43a7271f30d8c5deb2142
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16021
> Commit-Queue: Corentin Wallez <cwallez@chromium.org>
> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
TBR=cwallez@google.com,cwallez@chromium.org,enga@chromium.org,enga@google.com,david.turner.dev@gmail.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I922eccc310505da4b4a9fc853335733ca4900fc8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16521
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This matches Chromium's logic and make it less likely to break developer
workflow in the subsequent roll of Chromium's buildtools.
Bug: dawn:339
Change-Id: Ic42553827be125985a3d16b4f5d003b6662ead39
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16520
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This CL ensures that, on architectures with a varying subgroup size,
compute shaders are always compiled with a fixed subgroup size to
avoid consistency issues when one shader writes data in a subgroup-size
dependent layout to GPU memory, to be read by another shader in a
future dispatch.
At the moment, only Intel ICDs are known to implement this [1],
and the code uses a heuristics to chose the size of 16, which seems to
be the sweet spot according to Intel engineers.
+ Update the PNextChainBuilder class to deal with the fact that
VkComputePipelineCreateInfo::pNext is defined as a const void*,
which created compiler errors in the previous implementation.
[1] https://bugs.freedesktop.org/show_bug.cgi?id=108875
Change-Id: I332faa53b9f854a8abe43a7271f30d8c5deb2142
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16021
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Certain Vulkan ICDs (Intel ones notably) will compile SPIR-V
shaders with an liberal, compiler-selected, subgroup size (i.e.
either 8, 16 or 32). For more context, see [1].
This can be a problem for compute, when one shader stores data
in device memory using a subgroup-size dependent layout, to be
consumed by a another shader. Problems arise when the compiler
decides to compile both shaders with different subgroup sizes.
To work-around this, the VK_EXT_subgroup_size_control device
extension was introduced recently: it allows the device to
report the min/max subgroup sizes it provides, and allows
the Vulkan program to control the subgroup size precisely
if it wants to.
This patch adds support to the Vulkan backend to report and
enable the extension if it is available. Note that:
- The corresponding VkStructureType enum values and
struct types are not rolled to the third-party Vulkan
headers used by Dawn yet, so vulkan_platform.h has been
modified to define them if necessary. This can be
removed in the future when the Vulkan-Headers are
updated in a different patch.
- This modifies VulkanDeviceInfo::GatherDeviceInfo() to
use VkGetPhysicalDevice{Properties2,Features2} if the
VK_KHR_get_device_properties2 instance extension is
available. Otherwise, the Vulkan 1.0 APIs
VkGetPhysicalDevice{Properties,Features} are used instead
(and it is assumed that no subgroup size control is
possible).
- This changes the definition of VulkanDeviceKnobs to
make room for the required pNext-linked chains of
extensions.
- A helper class, PNextChainBuilder is also provided in
UtilsVulkan.h to make it easy to build pNext-linked
extension struct chains at runtime, as required when
probing device propertires/features, or when
creating a new VkDevice handle.
Apart from that, there is no change in behaviour in this CL.
I.e. a later CL might force a specific subgroup size for
consistency, or introduce a new API to let Dawn clients
select a fixed subgroup size.
[1] https://bugs.freedesktop.org/show_bug.cgi?id=108875
Change-Id: I524af6ff3479f25b0a8bb139a062fe632c826893
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16020
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
A Chromium's LinkedList class to Dawn. Implementation and header are
a direct copy/paste. This is to be used to implement an LRU Cache
for the ResidencyManager class.
Bug: dawn:193
Change-Id: I7cb02649590be4db0fe54c9d80557ac49efc34de
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/16380
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Bryan Bernhart
Jiawei Shao
Austin Eng
Corentin Wallez
Ryan Harrison
Natasha Lee
Brandon Jones
Nico Weber
David Turner