This will prevent mixing it up with other serial types in the future.
Bug: dawn:442
Change-Id: I74e964708acc62eb0f33127cc48f1b9a7b171d11
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28923
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
A lot of our switches over enum values use the following pattern:
default:
UNREACHABLE();
return foo;
This is problematic because when adding a new value to one of the WebGPU
enums, there is no compilation error for switches that are missing it.
Currently we're supposed to write code and tests and fix UNREACHABLEs when
we see them.
Instead we should strive to have most switches on enums to be complete
and explicitily tag unreachable values as UNREACHABLE. Some switches
might still want to use default: UNREACHABLE() if only a couple values
need to be handled out of very many.
In this CL we go through all the UNRAECHABLEs and change them if need
be. Also an ErrorQueue class is added to avoid having
QueueBase::SubmitImpl just be UNREACHABLE (and force overriding
instead).
Bug: dawn:527
Change-Id: I33dfb4703104912cc5f001f9faf907a61324de68
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28501
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Previously, lazy clearing always added DEPTH_STENCIL or RENDER_TARGET to
textures because we cleared using ClearDepthStencilView or
ClearRenderTargetView. Now, we're able to clear using copies.
This also allows textures to actually use the small resource heap
placement optimization. Doing so generates debug layer warnings when the
small alignment is first tried but rejected. This CL silences those
warnings.
Bug: dawn:145
Change-Id: Id385846536b337cddcfdadc5739561c7adc30c8c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26840
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Adding zero as the optimized clear value to silence a warning in
the D3D12 debug layer. Warning will still occur under non-zero
clears.
Bug: dawn:418
Change-Id: Idada948950f28130618834d966b237dcc39460c7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24300
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This CL checks buffer sizes before creating map read/write handles.
It is an error to map a buffer that can't be addressed on the CPU.
It also changes client-side synchronous errors on mapAsync to be
normal map failures, and not device lost errors. These should be
recoverable.
The CL adds additional testing for really large, but not UINT64_MAX
buffers, and fixes a VVL warning when buffer allocations exceed the
size of their memory heap.
Bug: dawn:450, dawn:398, chromium:1014740
Change-Id: Ieb34c04c3d01c429b7e3b7810729d5e91ecb6270
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/22626
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Adds logic to lock residency for bound descriptor heaps, then unlock and
insert into the LRU cache when no longer bound. Adds a basic functional
test.
Bug: dawn:193
Change-Id: Idfaaee6b873374c07a0b94b1982ad65353218799
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21400
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Returning COM for getters needlessly refcounts which wastes CPU cycles
in critial sections and floods PIX traces with [Add/Release]Ref.
BUG=dawn:212
Change-Id: Ifa853f2d5f78a450fdb7ffb9492f0d08dfbcdd37
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/21364
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
GetResourceAllocationInfo() returns UINT_MAX64 when
the requested size is too large. This corrects the
validation to OOM when D3D considers the size
invalid.
Only validating for zero-size would cause a device
loss as certain D3D drivers may not always consider
zero-sized invalid and NextPowerOfTwo(UINT_MAX64)
would overflow the allocator.
BUG=dawn:393
Change-Id: Idaad10c139f6428d4f48bca24027a6691257aca9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/20400
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Implements logic for managing the NON_LOCAL memory segment for UPLOAD
and READBACK heaps on Non-UMA devices.
Bug: dawn:193
Change-Id: I2426bf6b5f7a7ccd4420f830f344379af9faf73c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/19901
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
When attempting to allocate more than Dawn's budget within a single
serial, all heaps in the LRU will be un-evictable because the last
submission serial is the same as the current serial. We can work
around this by instead using the LockHeap and UnlockHeap functions
instead of EnsureCanMakeResident when calling CreatePlacedResource.
Also added in some additional comments regarding the last submission serial.
Bug: dawn:193
Change-Id: Ie4ec7ed5350b0858ea817431fbf77df6ca8acd96
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18622
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Dawn was using a very uncommon way to do breaks from case statements
when a block was introduced for that case statement. Fix it by running
the following commands:
git grep -l "} break;" | xargs sed -i "" -e "s/} break;/break;}/"
git cl format
Some -Wunreachable-code-break become very apparent in this CL but and are
fixed in a follow-up to keep mechanical and manual changes separate.
Bug: None
Change-Id: I558eda92bb1c9d938cc7cf07b091b733b57d3aca
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/18660
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Nico Weber <thakis@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
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>
Enables use of 64KB texture alignments where permitted.
This saves heap memory (64KB vs 4MB per allocation) and
improves re-use.
BUG=dawn:27
Change-Id: Ief4c531446788284e69ec1646cfe2ea7a25c7bb2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/14683
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Use `resourceDesc.alignment` + GetResourceAllocationInfo
to determine the buffer size and OOM should it return an
empty sized buffer instead of overflowing + INVALID_ARGS.
BUG=dawn:238
Change-Id: I0a2cc7dac629d55624dafa4a3c4a45f16e90049c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/14420
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
ResourceAllocatorManager::DeallocateMemory was correctly invalidating the
passed in allocation object. However, since the subclass
ResourceHeapAllocation class was not overriding the Invalidate method and
clearing out the D3D12Resource pointer, the resource ended up being tied
to the lifetime of the Texture object instead of being released on Destroy.
In Chromium, this bug was particularly egregious as it meant swap chain
texture cleanup was at the whims of the Javascript garbage collector.
Bug: dawn:242
Change-Id: Ia5856c61c8d3b92a2247a9aaa5f91c5de0a99dcb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/13200
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Enables use of 4KB resource alignments where permitted.
This saves heap memory (4KB vs 64KB per allocation) and improves re-use.
BUG=dawn:27
Change-Id: I7a0a30252f480db2d0fa7f5d949207a56e3aa2e9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12900
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Enables mixing of texture and buffers in the same heap.
This allows better heap re-use and reduces internal fragmentation.
A toggle has been added and enabled by default.
BUG=dawn:27
Change-Id: I466dc96240fe1e8de6e3dc56ed5547d7b61ee045
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12821
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This mostly makes the MemoryAllocator not owned by the
BuddyResourceAllocator so that we don't need an extra class for the
dependency injection in the Vulkan backend. (the container for the
BuddyMemoryAllocator can be it's MemoryAllocator at the same time).
Also renames methods of MemoryAllocator to be more explicit.
Also renames the constructor parameter of BuddyMemoryAllocator to be
(subjectively) closer to what the represent.
BUG=dawn:27
Change-Id: I37355ad5b3cded143956f0adc4742fa1b717e9bc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12661
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Allows buffer/texture direct access to underlying resource rather than indirectly with a opaque memory type.
BUG=dawn:27
Change-Id: I2eb69f4e30c96c431dbc96094d671be1e0a29869
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11800
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
