|
25390156
|
2025-08-21T00:13:19
|
|
Suppress unsafe buffers on a file-by-file basis in src/ [1 of N]
In this CL, we suppress many files but stop short of actually
enabling the warning by not removing the line from the
unsafe_buffers_paths.txt file. That will happen in a follow-on
CL, along with resolving any stragglers missed here.
This is mostly a manual change so as to familiarize myself with
the kinds of issues faced by the Angle codebase when applying buffer
safety warnings.
-- Re-generate affected hashes.
-- Clang-format applied to all changed files.
-- Add a few missing .reserve() calls to vectors as noticed.
-- Fix some mismatches between file names and header comments.
-- Be more consistent with header comment format (blank lines and
trailing //-only lines when a filename comment adjoins license
boilerplate).
Bug: b/436880895
Change-Id: I3bde5cc2059acbe8345057289214f1a26f1c34aa
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6869022
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
fa29f604
|
2025-07-02T13:23:31
|
|
Remove sharedContextLock from {Enable|Disable}VertexAttribArray
VertexArray objects are per context objects. In theory they do not need
to protected by shared context lock. The reason we are taking locks
because all these functions end up accessing Buffer object which are
shared. In prior CLs we have removed subject observer usage from
VertexArray which means VertexArray no longer accessed from other
thread. In prior CLs we also split VertexArray into two classes:
VertexArrayPrivate which has no buffer, and VertexArray which is
subclass from VertexArrayPrivate and owns buffer. In this CL,
glEnableVertexAttribArray and glDisableVertexAttribArray calls no longer
take shared context lock. ContextPrivateEnableVertexAttribArray and
ContextPrivateDisableVertexAttribArray are called from these two APIs
and they only have access to StatePrivate. State Private holds a
VertexArrayPrivate pointer, which means they do not have anyway to
access buffer objects. The main challenge I run into here is
mCachedActiveClientAttribsMask, mCachedActiveBufferedAttribsMask,
mCachedActiveDefaultAttribsMask, mCachedHasAnyEnabledClientAttrib,
mCachedNonInstancedVertexElementLimit,
mCachedInstancedVertexElementLimit. These StateCache variable needs to
be updated when these two APIs are called, and calculating these
variable needs access to buffer object. The solution here is adding a
bool mIsCachedActiveAttribMasksValid in the PrivateStateCache so that
instead of immediately update these mCached* variable, we just set
mIsCachedActiveAttribMasksValid to false. Then whenever any of these
mCached* variable is needed, we will check
mIsCachedActiveAttribMasksValid and calculate these cached variables. It
adds one if check when accessing these caches, but the other benefit is
that we may have avoided duplicated calculation when multiple states
changed.
Bug: b/433331119
Change-Id: I3227c72bc40501712db93fb3d540b835f07150b5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4514436
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
|
|
|
b4d84458
|
2025-05-23T18:08:19
|
|
Move Buffer from VertexBinding to VertexArray
In later CL we will not taking shared context lock for certain
VertexArray API calls. VertexArray itself is per context, so this sounds
reasonable to do. The main challenge here is a lot of VertexArray
function end up accessing gl::Buffer object, which could be modified by
other shared contexts. In order to safely not taking the shared context
lock, we need to separate out Buffer object out of VertexArray itself so
that these lockless APIs will take VertexArray that does not have access
to buffer.
In this CL, VertexArray is split into two classes: VertexArrayPrivate is
everything in VertexArray except buffers. VertexArray is a subclass of
VertexArrayPrivate and owns all the buffers. Buffer is removed from
gl::VertexBinding class. In order to let back end access to buffers,
VertexArrayImpl holds a weak reference to
VertexArray::mVertexArrayBuffers (which is a vector of buffers).
Further, VertexArrayBufferBindingMask mBufferBindingMask is moved from
VertexArrayState into VertexArray class well, since it tracks which
index has a non-null buffer. The bulk of change are due to the
VertexARrayImpl constructor change, since it now takes
vertexArrayBuffers argument. Other bulk of changes are due to
VertexBinding no long has the buffer, but you need to get it directly
from VertexArray or VertexArrayImpl.
This CL also reverts some of the change in crrev.com/c/6758215 that
mVertexBindings no longer contains kElementArrayBufferIndex.
BYPASS_LARGE_CHANGE_WARNING
Bug: b/433331119
Change-Id: I15f4576f7c5c8d8f4d9c9c07d38a60ce539bfeea
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6774702
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
8dca0efe
|
2025-07-21T15:29:10
|
|
Replace VertexArray::DIRTY_BIT_LOST_OBSERVATION with API call
This dirty bit was added so that back end can inspect buffers and set
proper VertexArray::DirtyBitType. The same thing can achieved by add a
virtual function on VertexArrayImpl class. The advantage of virtual
function on VertexArrayImpl is that all back end essentially have the
same implementation and we can just implemented in VertexArrayImpl
instead of duplicate in each back end. The other advantage is after this
CL DIRTY_BIT_BINDING_n and DIRTY_BIT_BUFFER_DATA_n will be well aligned
instead of offset by 1 caused by DIRTY_BIT_LOST_OBSERVATION. The other
motivation of this change is in later CL I want to move
mBufferBindingMask out of VertexArrayState, which means back end will
not have access to it. By using VertexArrayImpl API, I can pass
mBufferBindingMask directly to the back end via function parameter. So,
this CL removed DIRTY_BIT_LOST_OBSERVATION, added
VertexArrayImpl::checkBufferForDirtyBits().
Bug: b/433331119
Change-Id: I5c8cbc9bace63db416e86c2ae3631f74a12b20b8
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6775986
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
2ac49bb6
|
2025-07-01T12:11:12
|
|
Reland "Vulkan: Move VertexArray::ElementBuffer away from observer"
This reverts commit 79ac1a8cd767a32cce6401203e20c4bd4ca4d539.
Reason for revert: the regression bug is fixed in PS6
The regression bug with the original CL is caused by when we bind a
vertex array without element buffer rebind, we missed to reset
mIndexRangeInlineCache. The other bug is that VertexArray::mDiryBits is
64 bit but VertexArrayBufferBindingMask is 16 bit, in
VertexArray::setDependentDirtyBits(), bufferBindingMask.to_ulong() <<
DIRTY_BIT_BUFFER_DATA_0 is only producing the 32 bit value on windows
platform due to unsigned long is 32 bit value. bits() is used and bit
shift is operated on to uint64_t here to avoid dropping high bits on
windows.
Two tests are added that reproduce the regression bug caused by
the original CL.
Bug: angleproject:400711938
Original change's description:
> Revert "Vulkan: Move VertexArray::ElementBuffer away from observer"
>
> This reverts commit 3f012a43ee2c101543785720eedfeaa80708479d.
>
> Reason for revert: https://issues.chromium.org/427064102
>
> Bug: angleproject:400711938
> Original change's description:
> > Vulkan: Move VertexArray::ElementBuffer away from observer
> >
> > Right now, VertexArray's element buffer is always observing buffer's
> > change. In previous CLs, we have moved vertex array away from
> > subject/observer usage. This CL moves element buffer away from
> > subject/observer as well. Since the gl::Buffer tracks buffer's binding
> > to each context's current vertex array's binding point,
> > kElementArrayBufferIndex is added to VertexArrayBufferBindingMask bits
> > so that the element buffer is tracked exactly the same as other vertex
> > array buffer bindings. The VerextArray code has been modified to handle
> > this special bit, since element buffer has its own binding point
> > VertexArrayState::mElementArrayBuffer as opposed to
> > VertexArrayState::mVertexBindings. After this CL, VertexArray object
> > should be completely off subject/observer usages.
> >
> > Bug: angleproject:400711938
> > Change-Id: I662ddfabc95034bdc7734939c944ab033f41801c
> > Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6552160
> > Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
> > Reviewed-by: Geoff Lang <geofflang@chromium.org>
> > Commit-Queue: Charlie Lao <cclao@google.com>
>
> Bug: angleproject:400711938
> Change-Id: I9487ba8b108baaeda1c8a27189dba64f77616774
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6663539
> Commit-Queue: Charlie Lao <cclao@google.com>
> Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Bug: angleproject:400711938
Change-Id: I3f47ad1238c41f12b5cbd7a59b84be3fce1e9562
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6664004
Reviewed-by: Amirali Abdolrashidi <abdolrashidi@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
937cf23c
|
2025-05-13T16:11:47
|
|
Vulkan: Remove VertexArrayBufferContentsObservers
When vertex array needs to convert buffer's data, right now it uses
Subject/Observer to subscribe notifications from buffers about the data
change. Since we always dirty all binding point when we bind
VertexArray, or app has to rebind buffer to vertex array if its on the
other context, this notification really is only needed for the current
context's vertex array. In prior CLs we already moved notification from
buffer to the current context's vertex array away from Subject/Observer
usage pattern. This CL did similar things to
VertexArrayBufferContentsObservers::mContentsObservers.
VertexArrayBufferContentsObservers has been deleted in this CL. Each
VertexArrayImpl now tracks the need of content observer with a bit mask
of each bindingIndex (which is tracked by
mContentsObserverBindingsMask). When a buffer's content changes,
gl::Buffer will retrieve this bit mask from backend and pass it to the
current gl::Context, which sends to current VertexArray object, which
then set proper DATA dirty bits on VertexArray based on the binding bit
mask. If back end think it does not need any data conversion, then the
bit mask is zero and nothing will be done. This further removes
dependence on subject observer, which enables us to avoid taking shared
context lock for glEnableVertexAttribArray and
glDisableVertexAttribArray.
Bug: angleproject:400711938
Change-Id: Ieb0c09c042a560dd121242b63ec24478482399b3
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6549157
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
4d478713
|
2025-04-03T08:10:40
|
|
Revert "Metal: Make StateCache descriptors hash consistent"
This reverts commit 4334125be506e0c2c5dde4cfc29b31a7ff2bdec6.
Reason for revert: Pipeline creation failures.
Bug: chromium:407310859
Original change's description:
> Metal: Make StateCache descriptors hash consistent
>
> Various descriptors would compare equality and do hashing based on the
> underlying storage. The storage contains alignment padding and
> bitfields. The contents of these are not defined, even when the
> constructors tried to memset and memcpy.
>
> The various hash and compare functions also seemed to be missing some
> elements, like rasterSampleCount.
>
> Missing == on fields and hashing unused memory may lead to inconsistency
> where a==b is true but hash(a) != hash(b).
>
> Fix by:
> Remove the memset/memcpy and write out the operator== and
> hash() consistently.
>
> Initialize the members to their default values.
>
> Use uint32_t : 1 instead of bool if the struct tries to optimize for
> size, since mixing types stops the packing.
>
> Use uint32_t to get the alignment to uint32_t, which is mostly what
> is expected.
>
> Use uint32 someVar : N where N is payload bits + padding bits for
> the last member. This way the assignment clears the padding bits.
> The goal os to get the operator== to compare the padding bits, which
> lets the compiler elide the memberwise compares and use word-wise
> compares where it makes sense.
>
> This is an attempt to fix rare crashes related to map inserts
> and lookups. std::unordered_map will crash if operator== is not
> consistent with hash.
>
> Bug: angleproject:403372465
> Change-Id: I0a3a2ff327ac9f65e7d41bc9585cda54ce93ab86
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6355828
> Commit-Queue: Kimmo Kinnunen <kkinnunen@apple.com>
> Auto-Submit: Kimmo Kinnunen <kkinnunen@apple.com>
> Reviewed-by: Alexey Knyazev <lexa.knyazev@gmail.com>
> Reviewed-by: Geoff Lang <geofflang@chromium.org>
Bug: angleproject:403372465
Change-Id: Id111c0728b603776fd8085fd395819bb5aa14722
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6428466
Commit-Queue: Quyen Le <lehoangquyen@chromium.org>
Reviewed-by: Quyen Le <lehoangquyen@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
4334125b
|
2025-03-14T16:12:32
|
|
Metal: Make StateCache descriptors hash consistent
Various descriptors would compare equality and do hashing based on the
underlying storage. The storage contains alignment padding and
bitfields. The contents of these are not defined, even when the
constructors tried to memset and memcpy.
The various hash and compare functions also seemed to be missing some
elements, like rasterSampleCount.
Missing == on fields and hashing unused memory may lead to inconsistency
where a==b is true but hash(a) != hash(b).
Fix by:
Remove the memset/memcpy and write out the operator== and
hash() consistently.
Initialize the members to their default values.
Use uint32_t : 1 instead of bool if the struct tries to optimize for
size, since mixing types stops the packing.
Use uint32_t to get the alignment to uint32_t, which is mostly what
is expected.
Use uint32 someVar : N where N is payload bits + padding bits for
the last member. This way the assignment clears the padding bits.
The goal os to get the operator== to compare the padding bits, which
lets the compiler elide the memberwise compares and use word-wise
compares where it makes sense.
This is an attempt to fix rare crashes related to map inserts
and lookups. std::unordered_map will crash if operator== is not
consistent with hash.
Bug: angleproject:403372465
Change-Id: I0a3a2ff327ac9f65e7d41bc9585cda54ce93ab86
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6355828
Commit-Queue: Kimmo Kinnunen <kkinnunen@apple.com>
Auto-Submit: Kimmo Kinnunen <kkinnunen@apple.com>
Reviewed-by: Alexey Knyazev <lexa.knyazev@gmail.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
7d37d970
|
2025-01-30T13:47:59
|
|
Make error messages end in a dot
Some messages had dot and some not.
Fix by adding dot to all error message string literals.
Best effort, for free form string literals, check ANGLE_CHECK invocation
sites only.
Bug: angleproject:392938089
Change-Id: Idedc3fe1d1e186fb91b5e54407df87f6830995a0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/6218460
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
59adc191
|
2024-05-23T00:00:00
|
|
Metal: Ensure valid vertex buffer layout stride
Metal Validation causes runtime failures if the sum
of the vertex buffer binding offset and the related
vertex layout stride is greater than the buffer's
length.
Fixed: angleproject:342350849
Change-Id: I531f092c6130b37cb8b0b89b20804a222160ceb7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5565528
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
Reviewed-by: Quyen Le <lehoangquyen@chromium.org>
|
|
|
6cf3c96d
|
2024-05-23T00:00:00
|
|
Metal: Handle empty vertex buffers
Use the current attribute value when
the bound array buffer has no data
to avoid Metal Validation errors.
Drive-by: do not emit empty
client data upload commands.
Fixed: angleproject:40050025
Change-Id: I813725322c52bc16ca432f617a40fe7cee24dd6d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5569091
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
Reviewed-by: Quyen Le <lehoangquyen@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
|
|
|
8f966559
|
2023-12-07T00:00:00
|
|
Metal: Fix matrix vertex attributes with mixed sources
Updated VertexArrayMtl::setupDraw to support matrix
vertex attributes that use current and array values
simultaneously.
Simplified the helper function.
Fixed: angleproject:8456
Change-Id: I09a26a978cda4b9ac3747325ad571d5ad2fff72d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5116500
Commit-Queue: Alexey Knyazev <lexa.knyazev@gmail.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Quyen Le <lehoangquyen@chromium.org>
|
|
383df961
|
2023-11-06T22:00:52
|
|
Metal: only convert vertex if offset is not multiple of 4.
Previously we always convert vertex attribute if its binding offset
is not multiple of the attribute's size. This requirement seems to be
unnecessary.
This CL removes that requirement so the only requirement left for offset
is that it must be multiple of 4.
Bug: chromium:1496807
Change-Id: I35c421951c7817b77bd0c006ed4b72cd04b5a8d7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5006359
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Quyen Le <lehoangquyen@chromium.org>
|
|
|
bfc764c5
|
2023-11-03T03:16:40
|
|
Metal: Optimize vertex conversion using compute shader.
When converting vertex data with compute shader,
getComputeCommandEncoderWithoutEndingRenderEncoder() should be used
instead of getComputeCommandEncoder(). Because vertex data doesn't
depend on previous render commands, hence ending the current render
command encoder is unnecessary.
The only render command that could write vertex data is the one with
transform feedback (XFB). However, the XFB written data should already
be synchronized by glEndTransformFeedback() priorly if any.
Using getComputeCommandEncoderWithoutEndingRenderEncoder() ensures that
the conversion will happen in a compute pass before any draw calls in
the render pass.
Bug: chromium:1496807
Change-Id: I4613902b2ae83814e7cda712de89c7d8f285e5d1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5004821
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Quyen Le <lehoangquyen@chromium.org>
|
|
ae67a24b
|
2023-08-25T14:51:50
|
|
Metal: Move program state to ProgramExecutableMtl
Bug: angleproject:8297
Change-Id: I1d13f7aee1ff5b0ce799b486d8a57c83c4481983
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4812047
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
1e1c9d9d
|
2023-08-11T15:16:26
|
|
Pack and reduce Program::mProgramInput size
Right now mProgramInput is std::vector<sh::ShaderVariable>. It really
only need a subset of ShaderVariable struct. This CL adds a ProgramInput
struct so that we can add data members that actually required. This CL
also makes bools into bitfield and some variables to uint16_t to further
compact the size. This CL also groups the data memebers other than
string to basicDataTypeStruct which only contains basic data types and
the entire struct is memcpied during program binary load and save. This
not just reduces number of memcpy calls, but also improves reliability
so that when someone adds a new member into the struct, it will
automatically load/save correctly.
Bug: b/275102061
Change-Id: Ic055c986453ed46e56057a0122c9926245fef4d1
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4776267
Reviewed-by: Roman Lavrov <romanl@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
aea88562
|
2023-05-19T16:52:43
|
|
Reland "Metal: Optimized BufferSubData per device"
This reverts commit ee64836f702332adaca58d9f452063a04b2da955 ,
relanding the patch stack described there.
Between patchsets 1 and 5:
- The shadow buffer allocation has been replaced with a multimap of
precisely-sized buffers, rather than rounding up buffer sizes.
- Garbage collection of shadow buffers is triggered in three situations:
- A certain number of context switches have occurred; this number
was hand-tuned to avoid GC every frame.
- A certain number of command buffer submissions has occurred; this
number was hand-tuned to GC no more often than every few seconds
on representative workloads.
- The total size of the allocated shadow buffers is more than 1 MB,
and either more than twice the size at the last garbage
collection, or 64 MB more than at the last garbage collection. In
this case, aggressive GC is performed in order to reclaim shadow
buffers more quickly.
Performance before and after these changes appears identical on
microbenchmarks. On one Figma test case, comparing GPU memory
allocated inside the BufferManager, peak consumption is decreased by
over 75%, and steady-state consumption decreases by over 88%.
Patchset 6 adds a needed workaround for a bug in the
AMDMTLBronzeDriver affecting uploads of client-side data, and
therefore some dEQP tests. It also streamlines the aggressive GC.
Bug: angleproject:7544
Change-Id: I81b061f0b33c27fa403527fa12d626f4e9c88ebe
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4497413
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
0561884e
|
2023-04-26T17:26:42
|
|
Vulkan: Dirty VertexArray binding bit if buffer storage change
In crrev.com/c/3669603, we did optimization for black_desert_mobile that
when vertex array is unbound, we remove vertex array from buffer's
observer list to reduce overhead of observer notifications when buffer
is been modified. To compensate for the lost notification, when vertex
array is bound, we always assume every buffer that is bound to vertex
array has been dirtied, for the simplicity at that time. This CL further
the optimization of that CL. In this CL, I moved the dirty bit set into
backend and improves vulkan backend by checking buffer's serial number
and only dirty the binding if the serial has changed. Given this, now we
can also remove all the non-current vertex array from buffer's observer
list (previously it is heuristic based with a hard coded observer count
limit). This and the previous CL improves asphalt_9 by ~1%.
Bug: b/277644512
Change-Id: Ibc3f8e3df9fe70c6879e0b2bca86d8487a9dba73
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4481241
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yuxin Hu <yuxinhu@google.com>
Commit-Queue: Charlie Lao <cclao@google.com>
|
|
|
ee64836f
|
2023-02-11T17:56:06
|
|
Revert "Metal: Optimized BufferSubData per device"
This reverts commit 968041b54770af8917001d8fe9b52a881cfed0b2.
Includes the following patches:
git revert -n 995db1f66bcf87fc9e47d908fb2a885e810d2567 \
9a6c90c8f802b4d107a081bfccaf4be007e7af54 \
dbd47e378582ef86db52c7379cd220cf0b2c8193 \
369b320f92f54774879e8b8faff834fc8db0793e \
4abae6f97586448712e2dc1cced4a678b0901d7b \
968041b54770af8917001d8fe9b52a881cfed0b2
Several conflicts with top-of-tree were resolved during this revert.
The aim is to reland this with additional code which will reduce the
amount of excess buffer memory allocated, and release the resources
associated with temporary buffer allocations.
Bug: angleproject:7544
Change-Id: Ib7a6bc2ab1c2f23cb43112cd980106e2898c3826
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4240556
Reviewed-by: Kimmo Kinnunen <kkinnunen@apple.com>
Reviewed-by: Gregg Tavares <gman@chromium.org>
Commit-Queue: Kenneth Russell <kbr@chromium.org>
|
|
968041b5
|
2022-08-19T12:11:23
|
|
Metal: Optimized BufferSubData per device
Adds a staging buffer path which means there are 4 paths
for bufferSubData.
1. direct copy
* get a pointer to the buffer
* copy the new data to the buffer
* if the buffer is managed, tell metal which part was updated
2. use a shadow copy
* copy the data to a shadow copy
* copy the entire shadow to a new buffer
* start using the new buffer
3. use a new buffer
* get a new buffer (or unused)
* put the new data in the new buffer
* blit any unchanged data from the old buffer to the new buffer
* start using the new buffer
4. use a staging buffer
* get a staging buffer
* put the new data in the staging buffer
* blit from the staging buffer to the existing buffer.
Further, there are 3 types of memory storage modes.
Managed, Staged, Private.
Based on the GPU type different storage modes and different
paths in different sitatutions are more performant.
So, add feature flags to select paths by GPU.
Bug: angleproject:7544
Change-Id: I741dd1874201043416374194bd2001ded8dbd9b4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3842641
Reviewed-by: Kyle Piddington <kpiddington@apple.com>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Quyen Le <lehoangquyen@chromium.org>
Commit-Queue: Gregg Tavares <gman@chromium.org>
|
|
1670c6dd
|
2022-02-17T14:15:08
|
|
Metal: Buffer data is not correctly updated
When using a dynamic draw buffer, Metal creates
more buffers to avoid in-flight synchronization issues.
However, without correctly rebinding the buffers from the VAO,
we get missing data.
This patch adds a fix and a test to reproduce
the issue.
Upstream of https://bugs.webkit.org/show_bug.cgi?id=236427
and https://bugs.webkit.org/show_bug.cgi?id=236733
Bug: angleproject:7122
Change-Id: I879ff688af04a8215df6134400d0aab582b92842
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3472691
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Kyle Piddington <kpiddington@apple.com>
|
|
|
1642b4a3
|
2021-12-10T16:01:10
|
|
Metal: Fix Instanced Draw feature support on Simulator
Bug: angleproject:6786
Change-Id: I4c567bf2ec61df581ae0e8eeb1c36c7d37e15d04
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3331637
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Kyle Piddington <kpiddington@apple.com>
Commit-Queue: Gregg Tavares <gman@chromium.org>
|
|
|
a0924015
|
2021-11-22T19:44:05
|
|
Metal:Fix GLSLTest_ES3.GLVertexIDIntegerTextureDrawElements Metal
The test was failing if the "first" argument to glDrawElements
was odd. When odd the offset was not a multiple of 2 bytes
and the code was going down a different path than when the
offset was event.
The fix for that just removes a check for alignment to 4 bytes.
Checking with UNSIGNED_BYTE indices though and it still
failed. That was because it was encoding the index conversion
path on one command buffer and the provoking index prep
on another and so they were happening out of order.
Putting both on the same command buffer fixes that.
Bug: angleproject:6688
Change-Id: I89b68ba965d073e0fa70b6d5b444dc743b5e34a7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3296631
Commit-Queue: Gregg Tavares <gman@chromium.org>
Reviewed-by: Kyle Piddington <kpiddington@apple.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
77eb8a49
|
2021-10-29T10:36:06
|
|
Metal: Reduce memory usage of attribute re-writing
Rework vertex buffer caching to allow us to reuse parts
of converted buffers.
Reusing conversions dramatically drops the memory usage
when drawing index ranges by offset with
unaligned attributes.
Bug: angleproject:6638
Change-Id: I79797da202629b1632e1397ce1227ee3d7a1c9d7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3251467
Commit-Queue: Kyle Piddington <kpiddington@apple.com>
Reviewed-by: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
|
|
|
af5ea7f9
|
2021-10-07T12:47:41
|
|
PrimitiveRestart tests fail due to incorrect draw commands.
Primitive restart draw commands were being encoded
incorrectly. Single element ranges were being disregarded,
and a maximum draw size was not being enforced when the
index buffer was aliased / reused.
Bug: angleproject:6535
Change-Id: I5affad4df46c4ab2b3702caa89b483ae685618a7
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3212629
Commit-Queue: Kyle Piddington <kpiddington@apple.com>
Reviewed-by: Gregg Tavares <gman@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
|
|
|
da3db87e
|
2021-07-06T14:00:58
|
|
Upstream latest changes to Metal backend from Apple to 7/1/2021
This CL merges in the ANGLE changes between these two WebKit commits:
https://git.webkit.org/?p=WebKit.git;a=commit;h=8648b353ab1d7730438c2e08319e1a4d64982c31
https://git.webkit.org/?p=WebKit.git;a=commit;h=166e4924a52971d6a32ad48247a439b16c00e062
Include provoking vertex buffer out of bounds fix
from https://bugs.webkit.org/show_bug.cgi?id=230107
Fix bad merge of resetting of dirty bits, breaking
DepthStencilFormatsTest.DepthTextureRender test and perhaps others.
Disable GL_APPLE_clip_distance when the direct-to-Metal compiler is
active. It can not yet handle the gl_ClipDistance array.
Disable use of rectangular textures for IOSurfaces. Metal can bind
IOSurfaces to 2D textures, and this was passing all tests in the
SPIR-V Metal backend. Introducing rectangular textures breaks the
SPIR-V Metal backend, and the tests currently fail on the
direct-to-Metal backend.
Fix several bugs with ProvokingVertex, which was causing
both the SpirV and Direct backends to incorrectly draw
indices.
(https://bugs.webkit.org/show_bug.cgi?id=230107)
Skip the following tests on the Metal backend which is still failing
RobustResourceInitTestES3.BlitDepthStencilAfterClearBuffer
GLSLTest_ES3.GLVertexIDIntegerTextureDrawArrays/ES3_Metal
With these changes, angle_end2end_tests again runs to completion.
Bug: angleproject:6395
Change-Id: I3cc58f531426a95fc8f177a4ad87f56c1855a546
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3167010
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Kyle Piddington <kpiddington@apple.com>
|
|
cf8c5678
|
2021-09-17T13:16:36
|
|
Vulkan: Don't sync VAOs after BufferSubData calls.
We still need to syncState after buffers that contain converted
attributes are updated. Includes a perf regression test.
Bug: angleproject:6371
Change-Id: I54227fc43e7b3fe79072da7783dab0177ccb0486
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/3182706
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
d33a2222
|
2021-04-26T16:56:15
|
|
Upstream Apple's direct-to-Metal backend: compile libANGLE.
This change is meant to merge the metal backend changes from Apple's
direct-to-Metal backend. Taken from Kyle Piddington's CL:
https://chromium-review.googlesource.com/c/angle/angle/+/2857366/
The goal of this CL is to merge the metal backend code in a state
that compiles, but not to switch the Metal backend over to using
the direct-to-metal backend yet.
Bug: angleproject:5505
Bug: angleproject:6127
Change-Id: If6783e06e0086b3a1dd25c6f53caca5cfc96cb86
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2950067
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
|
|
cb6176f3
|
2020-09-29T01:08:00
|
|
Metal: Support tri-fan & line-loop with primitive restart
Triangle fan:
- If primitive restart is NOT enabled and there is no active render
pass, use Compute Shader to generate indices.
- If primitive restart is enabled, use CPU to generate indices.
Line loop:
- If draw non-instanced without primitive restart, generate and
draw only one additional last segment (fastest).
- If draw instanced, primitive restart is NOT enabled, and there is no
active render pass, use Compute Shader to generate indices (OK).
- Otherwise, use CPU to generate indices (slowest).
Also Disable OcclusionQueriesTest.ClearNotCounted failure on NVIDIA.
Bug: angleproject:2634
Bug: angleproject:5307
Change-Id: Ia5529825807a964f5fcb2a4af8844778896cd42a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2435859
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
|
|
|
8a50b42b
|
2020-10-24T19:29:12
|
|
Metal: Convert index & vertex format on GPU when possible.
- When converting vertex buffer:
- if there is no render pass active, use compute shader to convert.
- if there is a render pass active and device supports explicit memory
barrier then convert the buffer in vertex shader with direct buffer
write and insert a memory barrier.
- if there is a render pass active and device doesn't support explicit
memory barrier then convert the buffer on CPU.
Bug: angleproject:2634
Change-Id: I5346e3a2adb855f40e46a3912d9db404a4482e0f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2434025
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
|
|
|
ed23dc84
|
2020-10-24T19:12:32
|
|
Metal: default integer attribs & offset mod for idx conv.
- Support default integer attributes.
- When converting index buffer, use offset modulo instead of offset to
reduce number of conversions if application uses many different
offsets to the index buffer.
Bug: angleproject:2634
Change-Id: I97aa9ea1826ffc9dbe5784fe5b5af2f99df63e2c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2494524
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
69da0b92
|
2020-09-13T20:00:19
|
|
Metal: Use shared memory for small dynamic buffers.
- If BufferMtl is static or large size, don't use shadow copy. Use one
MTLBuffer and map directly on it.
- If BufferMtl is dynamic and small size, use shadow copy and buffer
pool of 10 MTLBuffer (s). The MTLBuffer is allocated in shared memory
in this case (PCI-E memory for example). MTLBuffer in shared memory
region doesn't need to sync content between CPU and GPU.
- When copyBuffer, if BufferMtl is being used by GPU use blit command to
do the copy on GPU side.
- Also implemented GL_MAP_UNSYNCHRONIZED_BIT.
Bug: angleproject:2634
Change-Id: I7a5aab309d24c76106a7087358ee5883ee05d250
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2408592
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
d7149f2a
|
2020-08-26T19:44:54
|
|
Metal: Disable unused attribute slots.
If an attribute slot is unused and user enables it but doesn't bind any
buffer, Metal validation layer will complain that a buffer slot is
missing (even though the slot is not used inside shader).
Fix: if shader program doesn't use an attribute slot or there is no
buffer bound to it, then disable that attribute.
Bug: angleproject:2634
Change-Id: Ib56e0bf64bab13efce8b2972bfe386e26f45e50e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2392171
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
bdecaf33
|
2020-08-04T20:16:27
|
|
Metal: Implement PBO.
Bug: angleproject:2634
Change-Id: I77f085227298bf46361825d1886e04830dc9987a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2336558
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
1c654d54
|
2020-08-26T00:16:52
|
|
Metal: Fix handling of vertex attrib offset not multiple of 4
Metal requires the vertex attribute offset to be multiples of 4, and its
stride must not be less than attribute's size.
This should fix the WebGL's test:
conformance/attribs/gl-vertexattribpointer-offsets.html
Bug: angleproject:4846
Bug: angleproject:2634
Change-Id: I0784a8ccaedd5e6c58a266243bfa94ba36e53e11
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2374829
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
ab372311
|
2020-05-14T23:27:06
|
|
Metal: refactor RenderUtils to split into multiple util classes.
This is useful for later modifications where blit/clear could be further
categorized based on texture format type (float/integer).
Bug: angleproject:2634
Change-Id: I877abd21761af9e91657686a60e189a43a33e3f4
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2193195
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
ab42afa6
|
2019-11-21T10:13:44
|
|
Metal: fix vertex attribute's conversion lost after changing buffer binding.
After vertex buffer's attribute is converted and stored in conversion
buffer. Binding the same attribute to another buffer, then binding it
back to previous buffer will result in previous conversion
information lost. The conversion method would skip the conversion due to
buffer's content hadn't been changed, however it didn't reuse the old
conversion result.
This CL also changed the way binding offset is used in Metal backend.
- Previous, the offset would be assigned to the offset field of
MTLVertexAttributeDescriptor, then the buffer would simply be bound to
the command encoder with offset=0
i.e. setVertexBuffer(buffer, index, 0)
- However this approach has several disadvantages. Since Metal doesn't
allow MTLVertexAttributeDescriptor's offset to be larger than the
vertex attribute's stride, the old approach would force the back-end
to convert the attribute and store in conversion buffer.
New approach:
- MTLVertexAttributeDescriptor's offset will be zero. The offset will be
used to bind the buffer itself to the render command encoder.
i.e. setVertexBuffer(buffer, index, offset)
This way the "offset <= stride" restriction no longer exists. The only
restriction is the offset must be multiple of attribute's size.
Added 3 new tests:
- SimpleStateChangeTest.RebindTranslatedAttribute
- VertexAttributeTest.DrawWithLargeBufferOffset
- VertexAttributeTest.DrawWithLargeBufferOffsetAndLessComponents
Bug: angleproject:2634
Change-Id: I6c2fa8091436e4a24405d791f86d17d97df02d64
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1940009
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
|
|
|
e5b474c6
|
2019-11-19T23:04:00
|
|
Metal: support ANGLE_instance_arrays/EXT_instance_arrays
Also added TRIANGLE_FAN & LINE_LOOP instanced draws test cases.
Bug: angleproject:2634
Change-Id: I84740a7221ab49710cf23767c81fa2d303aad364
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1919280
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
6fcc0bb8
|
2019-11-21T01:19:40
|
|
Metal: Re-add end2end test configs (running test is still disabled)
angle_test_instantiate.cpp & angle_test_instantiate_apple.mm:
- Disabled metal platform selection on pre-10.13 mac devices for
Bug: angleproject:4153
Explicitly disabled tests on metal:
- DifferentStencilMasksTest.DrawWithDifferentMask
- PointSpritesTest.PointSizeAboveMaxIsClamped
- WebGL2ReadOutsideFramebufferTest.CopyTexSubImage3D
This requires the crash fix in http://crrev.com/c/1924101
Bug: angleproject:4153
Bug: angleproject:2634
Change-Id: I95046d731a8ba7414cf1a1f4b6f2940282725872
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1926389
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
|
|
8392b118
|
2019-11-20T16:57:08
|
|
Revert "Metal: Enable end2end tests."
This reverts commit 0bb42e091b77f174632434a05789b2ce632bd902.
Reason for revert: https://bugs.chromium.org/p/chromium/issues/detail?id=1026633
Causing failures on Mac bots.
Original change's description:
> Metal: Enable end2end tests.
>
> Explicitly disabled tests:
> - DifferentStencilMasksTest.DrawWithDifferentMask
> - PointSpritesTest.PointSizeAboveMaxIsClamped
> - WebGL2ReadOutsideFramebufferTest.CopyTexSubImage3D
>
> Bug: angleproject:2634
> Change-Id: I76ab450c06d0e1be1e7469a8b9c6497684c3ca54
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1906607
> Commit-Queue: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Jamie Madill <jmadill@chromium.org>
> Reviewed-by: Geoff Lang <geofflang@chromium.org>
TBR=geofflang@chromium.org,jonahr@google.com,jmadill@chromium.org,le.hoang.q@gmail.com
Change-Id: Iaa4264834170a49c274f186d3d74f57714c84b32
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: angleproject:2634
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1926378
Reviewed-by: Courtney Goeltzenleuchter <courtneygo@google.com>
Commit-Queue: Courtney Goeltzenleuchter <courtneygo@google.com>
|
|
|
0bb42e09
|
2019-11-19T23:04:00
|
|
Metal: Enable end2end tests.
Explicitly disabled tests:
- DifferentStencilMasksTest.DrawWithDifferentMask
- PointSpritesTest.PointSizeAboveMaxIsClamped
- WebGL2ReadOutsideFramebufferTest.CopyTexSubImage3D
Bug: angleproject:2634
Change-Id: I76ab450c06d0e1be1e7469a8b9c6497684c3ca54
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1906607
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
|
|
|
fe26bae4
|
2019-10-29T18:38:53
|
|
Metal backend implementation pt 2
This is without Metal specific shader translator implemented yet.
Bug: angleproject:2634
Change-Id: I95d589442251c9ba111bd05a2dc379a36739046c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1855069
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
d200a77a
|
2019-10-10T00:44:01
|
|
Metal backend skeleton implementation.
Bug: angleproject:2634
Change-Id: I34be82f4a80a6851fecb53a51e069b134d82613a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1849079
Commit-Queue: Le Hoang Quyen <le.hoang.q@gmail.com>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|