gecko-dev/gfx/angle/cherry_picks.txt

183 строки
6.8 KiB
Plaintext

commit 6c72b64f910d0b0f81456bec26ca295c865c04da
Author: Geoff Lang <geofflang@chromium.org>
Date: Wed May 13 18:50:51 2020 -0400
Unset the ActiveTextureCache entry if the program does not reference it
When changing uniforms of a program, State::onActiveTextureChange is
called to update the ActiveTextureCache. If the sampler uniform type
changes to TextureType::InvalidEnum, the entry in ActiveTextureCache was
not cleared. This causes stale entries in ActiveTextureCache because the
cache no longer matches what textures are bound and the cache does not
add references to the textures in it.
BUG=chromium:1078375
BUG=chromium:1072406
BUG=chromium:1078866
Change-Id: If9719dcd4fc865b2301db450eb8115e7cfe46c4a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2199654
Reviewed-by: Tim Van Patten <timvp@google.com>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
commit eabf2a79aac36223b5ab5c5b0810d5522095f0bc
Author: Jeff Muizelaar <jrmuizel@gmail.com>
Date: Mon May 11 13:01:46 2020 -0400
Use deviceContext instead of deviceContext1 (#24)
Sometimes we don't have a deviceContext1 (on Win7) and would crash. We
on't have any need to use it so just use deviceContext.
commit 19f1bef071c32df60d9d62a7907c42270d0a3dd7
Author: Jeff Muizelaar <jrmuizel@gmail.com>
Date: Fri May 8 15:08:24 2020 -0400
Handle a null blend state. (#23)
Null is used as the default blend state so we have to check for it.
commit 8df54289d717eb0624a0ee16fd681c73a9472af4
Author: Jeff Muizelaar <jrmuizel@gmail.com>
Date: Wed May 6 14:04:42 2020 -0400
Don't use ClearView if we previously used dual source blending on Intel gen6. (#22)
Doing a ClearView after a dual source blend seems to cause a TDR on
Intel SandyBridge. Presumeably this is because the ClearView is
implemented as a regular draw and the driver doesn't properly set up the
state.
If we detect that this is going to happen we fall back to the manual
draw call path. This lets us use ClearView most of the time still.
commit 3885ee272685941f6dfb6cd941d550b107f033d2
Author: Jeff Gilbert <jgilbert@mozilla.com>
Date: Wed Jan 22 18:42:56 2020 -0800
Validate `context` before use in ValidateStreamConsumerGLTextureExternalAttribsNV.
Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1608235
Bug: angleproject:4333
Change-Id: I957b6412f7c744e4bbaa7d8084b3c81ee58d45ca
commit 5eaf76b0bfd1f149d7b03393500c6e49a393c167
Author: Jeff Gilbert <jgilbert@mozilla.com>
Date: Thu Dec 26 20:58:48 2019 -0800
Lost no-error context should gracefully handle GetProgramiv.
Includes test that a lost no-error context doesn't crash on getProgramiv.
Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1601703
Bug: angleproject:4244
Change-Id: Iaf86cabd0e70d644d5b0e9ef2f51e9edeee5d020
commit 85f04c65943e0a7070a1419fffa246e5ec39d1a1
Author: Jeff Gilbert <jgilbert@mozilla.com>
Date: Thu Dec 26 17:43:41 2019 -0800
clang-format
Change-Id: Iadf327c12bc43d7daedcd80fc08683c55c182ccc
commit f75ca830a4276c1b757f866aba3ab2429a27f69c
Author: Dzmitry Malyshau <dmalyshau@mozilla.com>
Date: Thu Dec 5 13:02:14 2019 -0500
Restrict the slow texture upload workaround to 128bit formats. (#21)
The workaround is slow for having a CPU-visible mirror of the texture, but that mirror is not properly recycled with regards to GPU usage, thus causing a CPU stall on update.
Original reports were related to WebRender vertex texture updates that are RGBA32F and RGBA32U formats. Limiting the workaround to these formats would allow the affected platforms to upload regular texture data faster without stalls.
commit 6bbe8f55d3e4b3b44394af323c5842296938ceab
Author: Lee Salzman <lsalzman@mozilla.com>
Date: Fri Aug 2 07:11:16 2019 +1000
Use image upload workaround for Intel Ivy Bridge with D3D11 renderer.
Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1578910
Bug: angleproject:3980
Change-Id: Ia14c5afd989365975bc57b303e59b4ef7669d522
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1845664
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>
commit fe53e2b60af2314d7fe7040915c1499ef3bb7274
Author: Miko Mynttinen <mikokm@gmail.com>
Date: Fri Nov 1 17:23:03 2019 +0100
Remove static_assert for global mutex
commit 7d5a10ac34cac9f629d0080302f1a754e902c66e
Author: Miko Mynttinen <mikokm@gmail.com>
Date: Thu Sep 5 18:04:32 2019 +0200
Do not override symbol visibility attribute on non-windows builds
commit 0f5a52939b91442075a787523fcf394bdccd4d54
Author: Miko Mynttinen <mikokm@gmail.com>
Date: Tue Sep 3 18:54:54 2019 +0200
Disable debug layers
commit fea78b079b9e778e240b06a0b8deadf57fe6d75a
Author: Miko Mynttinen <mikokm@gmail.com>
Date: Tue Sep 3 20:04:28 2019 +0200
Fix MinGW include
commit 811d7beadb57fdd4c999d0702fb48085b770c078
Author: Jeff Gilbert <jgilbert@mozilla.com>
Date: Wed Mar 27 13:52:04 2019 -0700
Don't clamp in glBlendColor if context supports renderable [half-]float.
Add test for glBlendColor >1.0, skipped on D3D9 since it's unsupportable
there.
Bug: angleproject:3323
Change-Id: I3fcf867dae96bdadf8a0183269d8f27b659d0baf
commit ea4db0f4e0c1b70dbe6e204e6afe77b23b8ea256
Author: Glenn Watson <39062770+gw3583@users.noreply.github.com>
Date: Fri Aug 2 07:11:16 2019 +1000
Fix occasional corruption of vertex textures in HD4600 GPUs. (#19)
This appears to be caused by the UpdateSubResource call,
which is worked around by enabling the other image upload
path that ANGLE supports.
The workaround doesn't include a driver version, since the
bug is occurring in the most recent driver that is currently
available (15.40.42.5063, released 19th Mar 2019).
Cherries picked
================================================================================
Merge base from: origin/chromium/3865
commit 08b97da894b278c10e70b6888ac65e3972a13b31
Author: Shahbaz Youssefi <syoussefi@chromium.org>
Date: Tue Jul 23 12:07:13 2019 -0400
Vulkan: Fix staging buffer alignment for pixel sizes > 4
Previously, all non-compressed formats were satisfied with a 4-byte
alignment. The introduction of float formats changes this.
This change calculates the necessary alignment using the same code path
as compressed formats.
Also fixes RGB9E5's pixelByte being calculated as 3 instead of 4.
Bug: angleproject:3731
Change-Id: I6abfd1c294e7f863eb99d2468eaac09e5d714039
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1715205
Reviewed-by: Mohan Maiya <m.maiya@samsung.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>