Loading program binaries across different architectures would cause
runtime errors due to use of size_t.
Also fix the ANGLE major and minor version that were clobbered in an
earlier commit.
BUG=angle:647
BUG=371435
Change-Id: I9473cc7e119592fce336aa47881839543e337b69
Reviewed-on: https://chromium-review.googlesource.com/199633
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Gather all bound textures and samplers first and then pass them to
generateSwizzles and applyTextures to reduce redundant get calls.
BUG=260069
Change-Id: I5b10dddb01a6bfd756c58b76218838cfc1fd59b6
Reviewed-on: https://chromium-review.googlesource.com/199343
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
Part of our effort to relocate all validation logic to a preliminary
pass outside of our state tracking or rendering layers.
BUG=angle:571
Change-Id: I5241daef6249910a781e78fd066debe0ab1d3f2b
Reviewed-on: https://chromium-review.googlesource.com/199348
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
The actual D3D header filenames are all lowercase.
Change-Id: I8f78d618d44360d34b5b01eb5a1972b35f2f15b7
Reviewed-on: https://chromium-review.googlesource.com/199692
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Shannon Woods <shannonwoods@chromium.org>
Compare all uniform data to the values currently stored in the uniform and
only mark the uniform as dirty if the data is new. This saves lots of
buffer map calls when the same data is set on a uniform many times.
BUG=260069
Change-Id: Ic4df8a276a81b074211712ff50e5cc4d0d9bb612
Reviewed-on: https://chromium-review.googlesource.com/199346
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
Reduce the number of IASetInputLayout calls to 1 in all cases and set the
fewest number of buffers at once.
BUG=260069
Change-Id: I059337fc6601ecefd801ef37546935becaa0d355
Reviewed-on: https://chromium-review.googlesource.com/199345
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
Only update the storage on levels that exist. Especially helps for cube
maps which have 90 images that would be iterated over in all cases. Added
an additional early-out check if the image is dirty for cube maps to match
the other texture types.
BUG=260069
Change-Id: I3d298f4f9ebdc23136a31138a3ef8364667fa78c
Reviewed-on: https://chromium-review.googlesource.com/199344
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
The translator gl::Varying struct does not need to know about register
allocation. We can put that entirely in the API library. This makes
exposing Varying variables with the new translator types cleaner.
BUG=angle:466
Change-Id: Ib85ac27de003913a01f76d5f4dc52454530859e6
Reviewed-on: https://chromium-review.googlesource.com/199736
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
We only assigned to elementIndex, and never referenced the value
at any point in our code.
BUG=angle:466
Change-Id: I541a353110bcd1863b07b21bcf21fad38be19ca1
Reviewed-on: https://chromium-review.googlesource.com/199735
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
I came across these three classes in the last CL, so do some
cleanup as I touched these classes.
BUG=
TEST=webgl conformance, no regressions
Change-Id: I2819bab5965fcb4917f85c2eded571a7f95ab9a2
Reviewed-on: https://chromium-review.googlesource.com/199423
Reviewed-by: Zhenyao Mo <zmo@chromium.org>
Tested-by: Zhenyao Mo <zmo@chromium.org>
Basically we should be able to reuse a struct name once the previous
use is out of the scope, or we can reuse it if we are in a inner scope
than where it is defined before.
BUG=368910
BUG=angle:618
TEST=conformance/glsl/bugs/nested-structs-with-same-name.html
Change-Id: Icfac76b844deaca0c8a0501d7426ff3802e657dc
Reviewed-on: https://chromium-review.googlesource.com/199137
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Zhenyao Mo <zmo@chromium.org>
Loading program binaries across different architectures would cause
runtime errors due to use of size_t.
Also fix the ANGLE major and minor version that were clobbered in an
earlier commit.
BUG=angle:647
BUG=371435
Change-Id: If2ad4c8fc246cc9ff0b4d95ba87b7d1ed109a064
Reviewed-on: https://chromium-review.googlesource.com/199102
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Some shaders fail to compile with validation error messages that make no sense.
BUG=angle:648,chromium:372016
Change-Id: Ief0e28efa6bc7eaa483c95f45c00ce4f9f20e541
Reviewed-on: https://chromium-review.googlesource.com/199311
Tested-by: Nicolas Capens <nicolascapens@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
We were not taking the state of GL_POLYGON_OFFSET_FILL into account.
BUG=371604
Change-Id: I9b490ed043faeece86665acaef0b865bf1f7b60d
Reviewed-on: https://chromium-review.googlesource.com/199121
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@chromium.org>
If there are only buffers with data revision 0, getLatestStorage will skip
over them. Fixed the if statement so that it will always find a buffer if
one exists.
BUG=angle:644
Change-Id: I24fc426383c9ae18b7ccf2f8937a9daa7a071ebe
Reviewed-on: https://chromium-review.googlesource.com/198826
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
Full support for using pack buffers as vertex buffers or other
binding points requires us to copy through the staging buffer.
BUG=angle:511
Change-Id: Ife0f5390a1a83dfbe0fa0b19db72f3a8b82b6b83
Reviewed-on: https://chromium-review.googlesource.com/198435
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Pack buffers used in different binding points when combined
with other buffer operations require us to be able to read
from the pack buffer then write to the native buffer.
BUG=angle:511
Change-Id: I2e859695235eb649ddd5dfc1332ecb19b57c1933
Reviewed-on: https://chromium-review.googlesource.com/198434
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
With our fix to huge buffer allocations, an extra check was preventing
pixel pack from running on the first try.
BUG=angle:511
Change-Id: Idabafe4e204ad3b7bf2010595dc455affe93dae0
Reviewed-on: https://chromium-review.googlesource.com/198825
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Making the contructors for shader variables inline saves us from
exporting them across DLL boundaries. This makes it easier to use
them in the translator when building the translator as a DLL.
BUG=angle:466
Change-Id: Iee0556e06dc1f9e98fe9eea6577819305de0dd0b
Reviewed-on: https://chromium-review.googlesource.com/198555
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Since we incur some kind of type confusion that stops us from properly
catching bad_alloc exceptions generated when compiled with exceptions
or when ANGLE is compiled with C++ exceptions, use catch-all clauses.
BUG=angle:634
Change-Id: I6d6efbf1fc7a4ebf8f13150b2814bfcd5edaad91
Reviewed-on: https://chromium-review.googlesource.com/198421
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
When we would allocate huge buffers, we would not always gracefully
fail and return a GL error back to the application. We should be able
to better handle buffer allocation failures with this patch.
BUG=angle:634
Change-Id: Ic3edce6d22b731ec52666a2a0e82f9a4fbbb23ae
Reviewed-on: https://chromium-review.googlesource.com/197994
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
The pack pixels operation is split in two halves: a GPU
copy then a CPU readback. We defer the CPU readback until
we can't wait any longer. That is, we do the readback when
the user maps the buffer, or there's another readback or
internal operation.
This offers the benefits of doing as much GPU work as
asynchronously as possible, and only doing the readback
work on related API calls (map or subsequent pack calls).
BUG=angle:511
Change-Id: I95a01da2b0e842438b180e8cdbb382c9a46ae210
Reviewed-on: https://chromium-review.googlesource.com/197836
Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
We would generate multiple vertex binaries that result in the same
HLSL code, eg for vec2 and vec3 vertex attributes. Eliminate
duplicates by using the converted attribute type for testing
for matching binaries.
BUG=angle:599
Change-Id: I061588164577ff9fa69ebb7d8a3f2bf6bb6fe013
Reviewed-on: https://chromium-review.googlesource.com/197830
Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
It seems we weren't catching bad_alloc despite using the correct syntax.
Using ellipses solves the problem, and since we only expect bad_alloc
it does the equivalent result of catching and returning out of memory.
I'm currently not aware of the proper syntax to catch bad_alloc.
BUg=angle:634
Change-Id: Ib3154546d1b72d180cb565c71f35716696863cdd
Reviewed-on: https://chromium-review.googlesource.com/198245
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Keeping a persistent storage allows us to treat the memory
buffer as the canonical copy of the pack buffer data, and
simplifies corner cases.
BUG=angle:511
Change-Id: Ied4dcb2edce2b85f68645f5a65603cbf18ad5a16
Reviewed-on: https://chromium-review.googlesource.com/197701
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Update the BufferStorage classes to use size_t for all sizes and offsets.
BufferStorage9 now uses a std::vector to manage its memory.
BUG=angle:631
Change-Id: Iea4e7b33ede59a5b9c6a1245690c4b7865096fc3
Reviewed-on: https://chromium-review.googlesource.com/197819
Reviewed-by: Nicolas Capens <nicolascapens@chromium.org>
Tested-by: Geoff Lang <geofflang@chromium.org>
Any time the git index changes, we will re-run the commit id hook.
This will ensure we never are left with stale commit ids from
other git commits.
BUG=angle:626
Change-Id: I03c37225fb902b6c1801e8f993d767122e8bfeb6
Reviewed-on: https://chromium-review.googlesource.com/197723
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
The generic alignment and conversion test to check for the possibility of using
direct storage is relatively expensive. Float types can be tested more quickly.
BUG=angle:613
Change-Id: I5ab008ce82c8b652f0392c20a80a5b198b2437a5
Reviewed-on: https://chromium-review.googlesource.com/197721
Tested-by: Nicolas Capens <nicolascapens@chromium.org>
Reviewed-by: Shannon Woods <shannonwoods@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Dan Glastonbury
Jamie Madill
Geoff Lang
Vladimir Vukicevic
Vladimir Vukicevic
Zhenyao Mo
Nicolas Capens
Shannon Woods