Currently we decide not to support multisampled 2D array textures
because on Metal they are only available on macOS version greater than
10.14.
This patch also removes the compatibility check between cube map texture
views and sample counts because currently the sample count of 2D array
textures is always equal to 1.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: I9736c977192409c79572f061fa1b7681b5b3e6c8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6000
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Same idea as for buffers, Destroy can be used to free GPU memory
associated with resources without waiting for javascript garbage
collection to occur.
Bug: dawn:46
Change-Id: Ia796b06b5228cbec4cfe8d78a500f967181d8c1f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5540
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
The range of the Z-coordinate in clipping volume is [-w, w] on OpenGL,
while it is [0, w] on D3D12, Metal and Vulkan.
In this patch, the "fixup_clipspace" flag of SPIRV-Cross is enabled on
OpenGL backend and disabled on D3D12 backend to unify the behaviour of
clip space on all Dawn backends. An end2end test is also added for this
fix.
This patch also fix a bug when clearing depth stencil attachments on
OpenGL backend. Before clearing depth stencil attachments, we should
enable depth stencil writing by properly setting depth and stencil
masks. We do not need to set the depth and stencil masks back because
they will be set again when applying the render pipeline. The newly
added test will fail without this fix when running the test together
with all the end2ends.
BUG=dawn:122
TEST=dawn_end2end_tests
Change-Id: I4f50ce3eb1f16d731ee4cffc12a56e17844b4675
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5860
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Existing validation requiring copy depth = 1 is not accurate. Copies
with a depth of 0 should be allowed.
Bug: dawn:18
Change-Id: Ib7607ee0965935127b3a8b66bc49b38ddcc56953
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5940
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
This patch adds a workaround on the GL error when we use glTextureView()
on a GL texture which is created outside DAWN and not configured by
glTexStorage*d(). glTextureView() is only allowed to be used on the
textures configured by glTexStorage*D(). When the external GL texture
is configured by glTexImage2D() (for example, textures from GLFW),
calling glTextureView() will cause an INVALID_OPERATION error.
To workaround this issue, we refer the solution on the Metal backend
that we avoid calling glTextureView() on the following senarios:
1. We may call glTextureView() only when the usage of the texture
includes Sampled or Storage.
2. We won't call glTextureView() if the view uses the same format as the
original texture, the whole mipmap levels and array slices.
BUG=dawn:16
Change-Id: Ibdfaa122ac061a2e2bb47f76e0030f1d0fc548a2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5780
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This is to allow proper synchronization with other devices and APIs on
macOS. There is a global graphics queue so we usually don't need
synchronization but on Metal, commands are inserted on this queue only
when the command buffer is scheduled.
Metal's schedule and completed handlers can be fired on a different
thread so this CL also makes the code there data-race free.
BUG=chromium:938895
BUG=dawn:112
Change-Id: Id45a66fb4d13216b9d01f75e0766732f6e09ddf0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5700
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch adds the support of multisampling in the render pipeline and
the validations that the sample count of the render pipeline must be
equal to the ones in render pass color and depth stencil attachments.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: I823c565bf9466ac6029f2797b31368bbdd6b8280
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5622
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Update stride value to max(attribute.offset + sizeof(attribute) for
each attribute) when input stride is 0 in Metal backend
BUG=dawn:75
TEST=dawn_end2end_tests
Change-Id: Ic0b2bfc685b8aeb92d1d4401174a819279386ea3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5221
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch simply added dynamic buffer offset in dawn idl and
modify the shape of SetBindGroup.
BUG=dawn:55
Change-Id: I516e08f3ee558ba375a87d98eaea6d60e93d4514
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5600
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch adds a missing validation on the sample count of the depth
stencil attachment when calling CommandEncoder.BeginRenderPass().
According to D3D SPEC, when MSAA is used, all bound render targets and
depth buffers must have the same sample count and quality.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: I719995b80f89e5139711c97a4bf76ba479919ad1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5621
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
If a GL texture is externally-managed, don't make any GL calls
during the Texture constructor.
Bug: dawn:114
Change-Id: I8ece5cde433fb036f37bda7d18ddb81d9b50a89d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5640
Commit-Queue: Stephen White <senorblanco@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch adds validations to the multisampled color attachments and
resolve targets.
The resolve target must be a valid texture view, and it must follow
the following validations:
According to Vulkan SPEC:
1. If we set resolve target, the texture of the color attachment must
be a multisampled texture, and the texture of the resolve target
must be a non-multisampled texture.
2. The format of the resolve target must be the same as that of the
color attachment.
3. The sample count of all the color attachments must be same.
4. The resolve target must have dawn::TextureUsage::OutputAttachment
usage bit set when it was created.
On D3D12:
1. ID3D12GraphicsCommandList::ResolveSubresource() method only allow
resolving in a subresource of a texture, thus the level count and
mipmap level count of the resolve target can only be 1.
2. ID3D12GraphicsCommandList::ResolveSubresource() method requires
the dimension of the source and destination must match.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: I080415bef0d600600083a95f641815188798dca3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5340
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This makes it so that default adapter discovery finds
Metal/D3D12/Vulkan first.
Change-Id: I3795bcc8a3d47eb3c53881b7fc43e8d56b4b89fd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5481
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This reverts commit 9bf529ec94.
Reason for revert:
Fixed test failure by submitting basic render pass to clear out texture
before running the tests.
The test was failing previously because the texture pixel color was not
cleared before running the tests, causing unexpected
pixel colors to be compared. Creating a basic render pass clears
the texture, but since the first test fails on submit expectedly,
the pixel is never cleared.
Bug: dawn:46
Change-Id: Ic190c2d8d6af3f9d8def3370b92c6974a82a0096
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5500
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
This is to match Chromium style.
Change-Id: Ic97cc03e2291c653ade9662ba3d5e629872b10ad
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5482
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This reverts commit b6a80b321e.
Reason for revert: dawn_end2end_tests are failing on the Chromium GPU FYI bots. Example here: https://ci.chromium.org/p/chromium/builders/ci/Win10%20FYI%20Release%20%28NVIDIA%29/4226
Original change's description:
> Destroy backend implementation for Buffers
>
> Destroy can be used to free the GPU memory associated with resources
> without waiting for javascript garbage collection to occur.
> The buffer is validated at submission to the queue.
> So any buffer that has been destroyed before submission, will then
> invalidate the submit and result in an error.
>
> Bug: dawn:46
> Change-Id: I40df56ce97baef01deea7552d7a6d40b558fc985
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5320
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
> Commit-Queue: Natasha Lee <natlee@microsoft.com>
TBR=cwallez@chromium.org,kainino@chromium.org,enga@chromium.org,rafael.cintron@microsoft.com,natlee@microsoft.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: dawn:46
Change-Id: Iadf37a8a6675c744207ec7daaa3fd2fde7da3714
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5480
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Destroy can be used to free the GPU memory associated with resources
without waiting for javascript garbage collection to occur.
The buffer is validated at submission to the queue.
So any buffer that has been destroyed before submission, will then
invalidate the submit and result in an error.
Bug: dawn:46
Change-Id: I40df56ce97baef01deea7552d7a6d40b558fc985
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5320
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
This patch is a follow-up of the descriptorization of render pass
descriptor. In this patch we changes the return value of
BeginComputePass from nullptr to an error compute pass encoder when
there is any error in BeginComputePass() to keep it consistent with what
we do in BeginRenderPass().
This patch also provides functions to create error render/compute pass
encoders. With this patch we can create a pass encoder in error by
specifying ErrorTag in the constructor, which is more staightforward
and human readable than the current implementation.
BUG=dawn:6
Change-Id: I1899ae65804f8cecd3079dc313e7e18acb88e37c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5140
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch fixes a Dawn crash issue when using textures in error state
in the copy commands of CommandEncoder.
In Dawn's copy commands (copyBufferToTexture and CopyTextureToBuffer),
we should check if the texture is in error state or not, or the assert
ASSERT(!IsError()) in texture->GetFormat() will fail and a crash will
occur.
In current Dawn code the validations on the buffer and texture objects
in the copy commands are executed in CommandEncoder::Finish(), which
is too late for textures according to the previous investigation. This
patch moves all these validations to the call of copy commands. The
checks on buffers are also moved away to keep the consistency of the
ones on textures.
BUG=chromium:937628
TEST=dawn_unittests
Change-Id: I0bc44e76262fba5927df20c6a7551b107bad5ca1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/5240
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch adds validations on the sample count of the textures used in
buffer-to-texture and texture-to-buffer copies. Vulkan SPEC requires the
textures used in vkCmdCopyBufferToImage and VkCmdCopyImageToBuffer must
have a sample count equal to VK_SAMPLE_COUNT_1_BIT.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: I189923eee2d8734d5ae3b57aea1a55533e8d98b7
Reviewed-on: https://dawn-review.googlesource.com/c/5220
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch adds validations on the creation of the multisampled textures
with sampleCount == 4. The validations include:
1. Only accept 1 or 4 as valid value of sampleCount.
2. According to Vulkan SPEC, when sampleCount > 1:
- The value of mipLevelCount can only be 1
- We cannot create cube map or cube map array texture views on this
texture.
BUG=dawn:56
TEST=dawn_unittests
Change-Id: Iac7cbe7cd7af16216b9185afd37a80eef0712f6b
Reviewed-on: https://dawn-review.googlesource.com/c/5160
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
It was essentially checking "currentPtr + commandSize <= endPtr" and
commandSize could make currentPtr overflow, making the comparison
succeed when it shouldn't have. This was caught through flakiness of the
LargeCommands allocator test.
Added a test provoking an overflow in Allocate and checking nullptr is
returned.
BUG=
Change-Id: I8ae4dad5b33c9d2005027c4d45b110ee0c65dd9a
Reviewed-on: https://dawn-review.googlesource.com/c/2841
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This allows wrapping IOSurfaces in a dawn::Texture so a Dawn application
can sample from, or render to an IOSurface. It uses Metal's
functionality for wrapping textures in MTLTexture.
Support for single-plane BGRA8, RG8 and R8 IOSurfaces is added as well
as tests for sampling and using BeginRenderPass to clear them.
BUG=dawn:112
Change-Id: I367dbd1a75a0c7b81901fb0aae05f1cd46af3f3a
Reviewed-on: https://dawn-review.googlesource.com/c/5101
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
When uploading textures from an unpack buffer, some drivers count an
extra row padding when checking if the pixel unpack buffer is big
enough. This will generate error about totalUsedBytes is larger than the
buffer.
Using split copy to take care of the last row of the unpack buffer will
workaround this issue.
BUG=dawn:74
Change-Id: I9d521d65e1f9e8039ae0388092d5e650cc4e65ce
Reviewed-on: https://dawn-review.googlesource.com/c/4940
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch removes RenderPassDescriptorBuilder completely from Dawn.
With this patch, RenderPassDescriptor is a structure instead of a Dawn
object, and all the checks in RenderPassDescriptorBuilder are moved into
CommandEncoder.cpp.
This patch also updates the helper functions and structures related to
RenderPassDescriptor because RenderPassDescriptor is no longer an
object but a structure with members in pointers.
BUG=dawn:6
Change-Id: Ic6d015582031891f35ffef912f0e460a9c010f81
Reviewed-on: https://dawn-review.googlesource.com/c/4902
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Metal needs buffer to buffer copy size must be a multiple of 4 bytes.
Adding validation to check this.
BUG=dawn:73
Change-Id: I9a4685d75439502017efa5455f7c2920a77f7a6d
Reviewed-on: https://dawn-review.googlesource.com/c/4900
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This change sets limit for stride in VertexInputDescriptor and
offset in VertexAttributeDescriptor, and adds validation code
for them.
It also uses existing descriptors to replace redundant definitions.
BUG=dawn:107
Change-Id: Ifbb07f48ec9a5baae8ae8d21865dc384670b759a
Reviewed-on: https://dawn-review.googlesource.com/c/4901
Commit-Queue: Yunchao He <yunchao.he@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Introduces pushDebugGroup, popDebugGroup, and insertDebugMarker implementations
for Vulkan and Metal using VK_EXT_debug_marker and XCode, respectively.
Bug: dawn:44
Change-Id: I0ae56c4d67aa832123f27a1fcdddf65746261e57
Reviewed-on: https://dawn-review.googlesource.com/c/4241
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Color state descriptors are a sequence of descriptors embedded
in RenderPipeline descriptor. We should use const*const* to
annotate them for supporting potential extended features added
in color state descriptors in future.
BUG=dawn:77
Change-Id: Id755569fa5076668c3db7b219b699845bc2a1d69
Reviewed-on: https://dawn-review.googlesource.com/c/4880
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This removes the CommandBufferBuilder and copies all the logic into
CommandEncoderBase instead. No changes were done to the logic except for
the implementation of CommandEncoderBase::HandleError and Finish.
BUG=dawn:8
Change-Id: I7b6f44c3cf501477422f067bd277cef470073860
Reviewed-on: https://dawn-review.googlesource.com/c/4820
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
std::codecvt has a protected destructor, so it needs some extra boilerplate to
be used with std::wstring_convert. This CL follows the example from [1].
This was working before because MSVC is nonstandard and std::codecvt::~codecvt
is public.
[1] https://en.cppreference.com/w/cpp/locale/codecvt
BUG=chromium:801780
Change-Id: Ic4b5baabcd9bfc9f60231204f050e369e79b4579
Reviewed-on: https://dawn-review.googlesource.com/c/4840
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch removes RenderPassDescriptor from BeginRenderPassCmd so that
all the backends can get the data of render pass from the non-pointer
field of BeginRenderPassCmd instead. In this patch, RenderPassDescriptor
has been removed from all Dawn backends other than the virtual function
DeviceBase::CreateRenderPassDescriptor().
This patch is one of the preparations on completely removing
RenderPassDescriptorBuilder from Dawn.
BUG=dawn:6
Change-Id: I3a78f0b2d5318c2bf85858d6fbe939b7861a2cf8
Reviewed-on: https://dawn-review.googlesource.com/c/4800
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Manages a single persistently mapped GPU heap which is sub-allocated
inside of ring-buffer for uploads. To handle larger buffers without additional
unused heaps, ring buffers are created on-demand.
BUG=dawn:28
TEST=dawn_unittests
Change-Id: If4d3e717186895b1409502c1dea5ab751a4776b2
Reviewed-on: https://dawn-review.googlesource.com/c/4440
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
CommandEncoder is the replacement for CommandBufferBuilder. This commit
adds the dawn.json definition for it and an initial implementation that
wraps CommandBufferBuilder. This is done so that the code can be ported
to CommandEncoder gradually, otherwise the commit would be too big and
would risk many merge conflicts.
This converts all samples and end2end tests to use CommandEncoder.
BUG=dawn:8
Change-Id: If4ce86e6fb39ba4e0c2af6328d40e63be17d18c1
Reviewed-on: https://dawn-review.googlesource.com/c/4741
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This also makes SwapChain support WebGPU-style error handling.
BUG=dawn:8
Change-Id: I5a142ae58600445f0f44f6dbe419cb7c3cdc9464
Reviewed-on: https://dawn-review.googlesource.com/c/4660
Reviewed-by: Yunchao He <yunchao.he@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch removes RenderPassDescriptorD3D12 completely and moves the
creation of RTVs and DSVs to CommandBufferD3D12::RecordCommands(), where
we allocate all RTVs and DSVs used in the current command buffer in one
RTV heap and one DSV heap. Note that the method to allocate RTVs and
DSVs are too simple in this patch, and we will optimize it later.
This patch also adds a test to make sure Dawn works correctly when we
use two different render passes in one command buffer.
This patch is one of the preparations on completely removing
RenderPassDescriptorBuilder from Dawn.
BUG=dawn:6
Change-Id: I02e30c007fb8668a7474a3caf7a858782d0c92df
Reviewed-on: https://dawn-review.googlesource.com/c/4520
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Yunchao He
Jiawei Shao
Natasha Lee
Brandon Jones
Corentin Wallez
Yan, Shaobo
Li Hao
Stephen White
Austin Eng
Tom Anderson
Bryan Bernhart