Commits · 04ab0eb41f1bb02eb722ce662bd9e6f47cca83a3 · Chen Yisong / angle

03 Dec, 2014 1 commit

Fix memory leak with default Framebuffers. · 04ab0eb4

authored Nov 22, 2014

When the depth-stencil attachment had no depth or stencil bits, we
would like a gl::Renderbuffer object. This was causing unbounded
memory growth in Chrome, and could happen when you watched a lot
of streaming video.

The bug was originally reported and fixed by Austin Kinross in
commit 44f4d74c.

BUG=403471

Change-Id: I022f01a6d3159f766e65f65f023e6ce9baacbc8e
Reviewed-on: https://chromium-review.googlesource.com/231583Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Hendrik Wagenaar <hendrikw@chromium.org>
Tested-by: Hendrik Wagenaar <hendrikw@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

04ab0eb4

21 Oct, 2014 1 commit

Fix typo disabling NPOT on AMD D3D9 · 5af475ed

authored Oct 20, 2014

BUG=425668,angle:799

Change-Id: Ifd4200774665f22d5f6ed196862192585d0799c0
Reviewed-on: https://chromium-review.googlesource.com/224481Reviewed-by: Brandon Jones <bajones@chromium.org>
Tested-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/224790Reviewed-by: Geoff Lang <geofflang@chromium.org>

5af475ed

22 Sep, 2014 1 commit

Use a D24S8 format to back GL_DEPTH_COMPONENT32_OES in D3D9. · df647a2a

authored Sep 19, 2014

Looks like the D3D9 D32 format isn't available on most GPUs (and doesn't
exist in D3D11) so back GL_DEPTH_COMPONENT32_OES with D24S8 instead to
match D3D11.

BUG=angle:750

Change-Id: I0f7a124544c1c14ba21db20fbf6765d07e244966
Reviewed-on: https://chromium-review.googlesource.com/219080Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

df647a2a

19 Sep, 2014 6 commits

Fix a NULL dereference on buffer initialization. · ee009b8e

authored Sep 19, 2014

In some cases, where the user would create a buffer with NULL (empty)
data, we would attempt to dereference NULL when drawing with the
buffer as a vertex attribute.

BUG=angle:749

Change-Id: Ied5ecbab4608c85890cdf7cc32a8dae46989e33b
Reviewed-on: https://chromium-review.googlesource.com/219090Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>

ee009b8e

Remove rx::TextureStorageInterface entirely. · 2f06dbfb

authored Sep 18, 2014

This class provides no added benefit now that we've scrapped the
type-specialized classes. We can squash it's functionality into
TextureStorage instead.

BUG=angle:741

Change-Id: I04984806719a01f299986265fe09b3cccf04dca5
Reviewed-on: https://chromium-review.googlesource.com/218316Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

2f06dbfb

Squash the TextureStorageInterface heirarchy. · 856d9d4b

authored Sep 18, 2014

Since there are no longer any significant differences in the child
classes of TextureStorageInterface, we can scrap the different
implementations. This involves changing the function names of some
Renderer functions which relied on overloading.

BUG=angle:741

Change-Id: I074797e4ac0bf3ecf8fe4bddf84eaed28b1a2c23
Reviewed-on: https://chromium-review.googlesource.com/218315Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

856d9d4b

Move generateMipmap to storage interface base class. · 5e48c034

authored Sep 18, 2014

Using a little bit more of code repetition allows us to move the
implementation of the mipmaps to a simple virtual interface in
TextureStorage.

BUG=angle:741

Change-Id: I7661d0cc3bdb02b646da0802cba285650d7832b2
Reviewed-on: https://chromium-review.googlesource.com/218314Reviewed-by: Brandon Jones <bajones@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

5e48c034

Use ImageIndex in getRenderTargetSerial. · c483326b

authored Sep 18, 2014

Also move getRenderTargetSerial to the TextureStorageInterface
base class, since it shares a common interface.

BUG=angle:741

Change-Id: I1bc1cfac6426e241ac91d373884a7dd8a1c5b188
Reviewed-on: https://chromium-review.googlesource.com/218313Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

c483326b

Updated the vertex buffer classes to use Error objects. · f7100b98

authored Sep 08, 2014

BUG=angle:520

Change-Id: Id003e66b2acbf37dbbe66aaca2fa336c3c884be2
Reviewed-on: https://chromium-review.googlesource.com/217102Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

f7100b98

18 Sep, 2014 4 commits

Fix Renderer11 signed/unsigned warning in 32-bit. · d5a796ca

authored Sep 18, 2014

Change-Id: I1e11f54b0ae245cdd7e090ab0748821738ad6314
Reviewed-on: https://chromium-review.googlesource.com/218761Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

d5a796ca

Use dynamically sized containers for texture and sampler bindings. · 76b10c9a

authored Sep 05, 2014

BUG=angle:685

Change-Id: I7af97a95deee69fbdebca2b57403244f45516e67
Reviewed-on: https://chromium-review.googlesource.com/216564Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

76b10c9a

Use ImageIndex in getRenderTarget. · ac7579c2

authored Sep 17, 2014

Also change ImageIndex to allow invalid indexes for layer. In 3D
indexes, sometimes the layer parameter is invalid, indicating we
index an entire mip level instead of a layer of a mip level.

BUG=angle:741

Change-Id: I7b410f9aaf568f215bb93ae063934669a8fa09bd
Reviewed-on: https://chromium-review.googlesource.com/218312Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

ac7579c2

Don't skip draw calls on zero-sized viewports. · 02f18b88

authored Sep 17, 2014

If transform feedback is active, primitives still need to be rendered even
if no pixels are written to the framebuffer.

Instead of checking for active transform feedback, simply remove the draw
call skipping optimization since it is most likely an application mistake to
draw with a zero sized viewport and we shouldn't optimize for this case.

This change doesn't affect the clear calls because the viewport is set to the
framebuffer size which is non-zero.

BUG=angle:743

Change-Id: I04af9d6de5aad3040e3c6b3c24990e107e21ad36
Reviewed-on: https://chromium-review.googlesource.com/218508Reviewed-by: Brandon Jones <bajones@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

02f18b88

17 Sep, 2014 9 commits

Improve D3D11 varying packing when there are more varyings than registers. · af875527

authored Aug 25, 2014

BUG=angle:738

Change-Id: I0599840fc79d571230acf26105d512322bcffdcd
Reviewed-on: https://chromium-review.googlesource.com/214108Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

af875527

Fix incorrect error check. · bdee2d59

authored Sep 17, 2014

The registers variable became a pointer but this error check was not
updated.

BUG=angle:743

Change-Id: Id2cd7b5ba39180281a3bbc408cd7c31bc663500c
Reviewed-on: https://chromium-review.googlesource.com/218507Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

bdee2d59

Merge the sampler uniform application into ProgramBinary::setUniform. · be4fdb3f

authored Sep 16, 2014

 * Fixes incorrect dirty checks for the dirty sampler mappings flag.
 * Fixes WebGL tests:
   * conformance_canvas_texture_bindings_unaffected_on_resize
   * conformance_reading_read_pixels_test
   * conformance_uniforms_uniform_default_values

BUG=414450

Change-Id: I132dbc301a236aef153fb6da4c6b64c36b0ba3f3
Reviewed-on: https://chromium-review.googlesource.com/218501Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

be4fdb3f

Fix enum in FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE query. · ee85d1bb

authored Sep 17, 2014

We were using DEPTH_STENCIL instead of DEPTH_STENCIL_ATTACHMENT.

BUG=angle:739

Change-Id: I0c746c374577dca2b965dfd7ccfac793f7075efb
Reviewed-on: https://chromium-review.googlesource.com/218592Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

ee85d1bb

Accept ImageIndex in TextureD3D::subImage and getImage. · feda4d29

authored Sep 17, 2014

This paves the way for setting data on the TextureStorage directly
instead of working through the Image objects.

BUG=angle:741

Change-Id: I3be3d5f9b2e45707c1630b74ad3f4789e034c3fd
Reviewed-on: https://chromium-review.googlesource.com/218311Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

feda4d29

Fix the varying sort comparator in ShaderD3D. · e6256f87

authored Sep 17, 2014

The comparator was actually a <= operator, while sort requires a
strict < operator. This was causing a potential assertion failure.

Bug report from Kerim Borchaev.

BUG=angle:742

Change-Id: I37c2925ab0b85e70ee1b2be3c72c6ddc062e8d28
Reviewed-on: https://chromium-review.googlesource.com/218506Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

e6256f87

Move validation of ANGLE_instanced_arrays to the validation layer. · 87a93308

authored Sep 16, 2014

BUG=angle:520

Change-Id: Idb3c50235a7029e72c58bc202aba0cfab735202a
Reviewed-on: https://chromium-review.googlesource.com/218510Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

87a93308

Don't modify the function parameters in ProgramBinary::setUniform. · 61f54180

authored Sep 16, 2014

BUG=414450

Change-Id: Ib5ebf18e57af671b8694ffcfc3cb152557f7f002
Reviewed-on: https://chromium-review.googlesource.com/218500Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

61f54180

Add several missing header guard comments · 08a59f85

authored Sep 16, 2014

These were causing build failures under GCC.

Change-Id: Iba19ed166f909fe559d19731f31849770619f412
Reviewed-on: https://chromium-review.googlesource.com/218580Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Shannon Woods <shannonwoods@chromium.org>

08a59f85

16 Sep, 2014 2 commits

gn: Fix source paths · f8a16d7b

authored Sep 16, 2014

Change-Id: Idf0c746d7e4e9f03b176744e8ff027be2d5f25ad
Reviewed-on: https://chromium-review.googlesource.com/218520Reviewed-by: Brett Wilson <brettw@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Commit-Queue: Cem Kocagil <ckocagil@chromium.org>
Tested-by: Cem Kocagil <ckocagil@chromium.org>

f8a16d7b

Update Queries to return Error objects instead of calling gl::error. · 5aad9673

authored Sep 08, 2014

BUG=angle:520

Change-Id: If8f2bb1c4de7b9cc30861a06aab1d89c97305b26
Reviewed-on: https://chromium-review.googlesource.com/216699Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

5aad9673

12 Sep, 2014 4 commits

Moving more D3D-only logic into ProgramBinaryD3D · 22502d52

authored Aug 29, 2014

BUG=angle:731

Change-Id: Ia62ec9a912c8bd63fcee51ca206d52688dd999da
Reviewed-on: https://chromium-review.googlesource.com/217495Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Brandon Jones <bajones@chromium.org>

22502d52

Remove dead RenderbufferStorage::getDepthBuffer. · e1f0d00c

authored Sep 12, 2014

This code was unused.

BUG=angle:732

Change-Id: I764d7c5624bdb81650e05a22c4ef0431c2002b18
Reviewed-on: https://chromium-review.googlesource.com/217331Reviewed-by: Brandon Jones <bajones@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

e1f0d00c

Add a helper to replace FBO attachment's getSerial. · 612e2e44

authored Sep 12, 2014

Ideally we could store a render target serial in the render target
itself. For now, this helper function allows us to finally stop
exposing rx::RenderTarget from gl::Texture.

BUG=angle:732

Change-Id: I4ffc5e0af6f0542212ad4e74adced21ebbd89462
Reviewed-on: https://chromium-review.googlesource.com/217330Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

612e2e44

Add an attachment helper to replace getRenderTarget. · 9f0b42a6

authored Sep 12, 2014

We would like to stop exposing rx::RenderTarget to any GL-layer
code, like the Context, or Framebuffer. Then we can pull any
necessary information inside the Renderer back-ends.

BUG=angle:732

Change-Id: Ifa35218f097c528e616d732bda4bcd8e5f1c3843
Reviewed-on: https://chromium-review.googlesource.com/217029Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>

9f0b42a6

10 Sep, 2014 10 commits

Fix signed-unsigned mismatch · 231943bf

authored Sep 10, 2014

Change-Id: Ie7fba8a0fd0475a395205139c4f7169cc94e61b7
Reviewed-on: https://chromium-review.googlesource.com/217510Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Shannon Woods <shannonwoods@chromium.org>

231943bf

Updated the index buffer classes to use Error objects. · c9e69b19

authored Sep 08, 2014

BUG=angle:520

Change-Id: Ifc249058a3ed3ffffe163a9e3ec21d6fc8c75bd0
Reviewed-on: https://chromium-review.googlesource.com/217101Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

c9e69b19

Only allocate the streaming index buffers when they are first needed. · 1c134e6c

authored Sep 08, 2014

BUG=angle:520

Change-Id: I111e4fd830ad19f6a4ff50749ba891fc14f9154b
Reviewed-on: https://chromium-review.googlesource.com/217100Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

1c134e6c

Move the counting IB from IndexDataManager to Renderer9. · 23ba30e4

authored Sep 08, 2014

BUG=angle:520

Change-Id: Ice0a04b296af6c0a61a604f629b08603e594bb0b
Reviewed-on: https://chromium-review.googlesource.com/216919Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

23ba30e4

Update the RenderStateCache to use Error objects. · 157f9374

authored Sep 08, 2014

BUG=angle:520

Change-Id: I14e2a84c6d2e6f98a50395b63ac206e32bc10f8b
Reviewed-on: https://chromium-review.googlesource.com/216918Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>

157f9374

Add queries for attachment targets. · 5b5d1244

authored Sep 09, 2014

Queries for the FBO attachment Textures and Renderbuffers allow us
more options than specific methods for querying the texture
storage or texture serial.

BUG=angle:732

Change-Id: Ieb4ddca3955fcf716dbf54331524d0c1e25fe946
Reviewed-on: https://chromium-review.googlesource.com/217028Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Brandon Jones <bajones@chromium.org>

5b5d1244

Add a TexSubImage benchmark. · bbffd556

authored Sep 10, 2014

Currently we are about 2x as slow on this micro-benchmark in D3D11.
Preliminary testing suggests using CopySubresourceRegion gives
improved performance across the board, but testing could disprove
this hypothesis.

Modified from a sample from bajones.

BUG=angle:705

Change-Id: Iedf36a4e7b9b9bbed308302fd2bf3912acacbd2c
Reviewed-on: https://chromium-review.googlesource.com/216272Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

bbffd556

Fix Win32Window::resize breaking angle_tests. · 9e16d40d

authored Sep 08, 2014

We were messing up the client rect on resize, which caused the
ReadPixels checks to mess up around the window edges. Disabling
the window styles on the test windows masked this bug. Fix this
by using a style-less child window inside the parent window.
This gives us access to window styles for the samples project,
along with the ability to use tiny 1x1 windows for testing.

BUG=angle:730

Change-Id: Ic6dd931df7b4e32fbbcacbb004d3bbc49917f658
Reviewed-on: https://chromium-review.googlesource.com/217024Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

9e16d40d

Get stencil buffer clear parameters format directly. · dd8488d3

authored Sep 03, 2014

The state layer should not be querying render targets directly,
since they are not a GL concept.

BUG=angle:732

Change-Id: I5ee96ae54cced8a32c5082ddd9f8796a975e2591
Reviewed-on: https://chromium-review.googlesource.com/213971Reviewed-by: Brandon Jones <bajones@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>

dd8488d3

Only use direct buffers for static data in D3D11. · 82cceb2d

authored Sep 09, 2014

For highly dynamic data, which gets updated every frame, or almost
every frame, we're better off using our existing dynamic buffer
path. We could further optimize the dynamic buffer path by only
uploading changed data every frame.

BUG=angle:705

Change-Id: Icbb357b889be789b30f73067f75b13664c806929
Reviewed-on: https://chromium-review.googlesource.com/217280Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>

82cceb2d

09 Sep, 2014 2 commits

Starting refactor of ProgramBinary · c9610c51

authored Aug 25, 2014

This is gonna take a while...

BUG=angle:731
Change-Id: Ief72c3361b6429f3f6e0bc2d2ea0810d523ff178
Reviewed-on: https://chromium-review.googlesource.com/215661Tested-by: Brandon Jones <bajones@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>

c9610c51

Squash the Texture attachment classes into one. · eeb7b0e9

authored Sep 03, 2014

BUG=angle:732

Change-Id: Ib6b26fe1351bc09e729178f6ec8b8d2ec1f7ff58
Reviewed-on: https://chromium-review.googlesource.com/213970Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>

eeb7b0e9