- 28 Sep, 2016 13 commits
-
-
Nicolas Capens authored
It's only used for creating intrinsics calls, which is an implementation detail. Bug swiftshader:10 Change-Id: Ieac1303e08571e265e239e74c6ea84ece99b53e1 Reviewed-on: https://swiftshader-review.googlesource.com/7278Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: Ide80a6d97f67ae04d9118307984235f8f4c2e3ad Reviewed-on: https://swiftshader-review.googlesource.com/7277Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: Iaed1c9a078b82516db67fe0de287e7ea0362dc03 Reviewed-on: https://swiftshader-review.googlesource.com/7276Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: Ib45d459aa4f68e572b1deaa03e0aaf83002b7881 Reviewed-on: https://swiftshader-review.googlesource.com/7275Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: I86b80cb03ff48ff7273aaf23a8e4995b3436f825 Reviewed-on: https://swiftshader-review.googlesource.com/7274Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Use only llvm::Constant in the interface instead. Bug swiftshader:10 Change-Id: Ideb2ed1ee083e4e6896c4bc8e8830ea02972b487 Reviewed-on: https://swiftshader-review.googlesource.com/7273Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: I69105d8287e52f0f0573c45c59958994a56fc0af Reviewed-on: https://swiftshader-review.googlesource.com/7214Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:10 Change-Id: I24fc66248bd2f66d2e8bc2b2a15e6c8fadc47efb Reviewed-on: https://swiftshader-review.googlesource.com/7213Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
"false" could not be converted to a pointer type. Changed it to nullptr. Also changed a few other "return 0" to "return nullptr". Change-Id: Idd4573d4627bb89f1b79e3b21de5af5bd6063916 Reviewed-on: https://swiftshader-review.googlesource.com/7357Reviewed-by:
Alexis Hétu <sugoi@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
Alexis Hetu authored
Fixed first class array constructors by allowing basic type arrays and structure arrays to be handled properly for the EOpConstruct* operations. This fixes all dEQP.functional.shaders.arrays.* tests. Change-Id: I4fe99ec5256abf6483d3595890ba9c426abc97f8 Reviewed-on: https://swiftshader-review.googlesource.com/7351Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
There was an ambiguous type warning between Float4 and RValue<Float4>. Change-Id: I526fc2430a66b8545e81fe5ae97fb423e7f01aec Reviewed-on: https://swiftshader-review.googlesource.com/7356Reviewed-by:
Alexis Hétu <sugoi@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
Alexis Hetu authored
It appears the Nucleus::createFPToUI function is broken and does not yield the correct values for one pixel out of 4 when the f2u operation is used. By manually rewriting the conversion using signed int conversion and some arithmetic, the issue is solved. This fixes at least 130 failures and also fixes many other tests which were passing only due to the threshold being too lenient, but which were showing obvious visual artefacts. This affects, to the very least: dEQP.functional.fbo.blit.conversion dEQP.functional.shaders.functions.datatypes dEQP.functional.shaders.operator.binary_operator ... and any other test using the f2u operation. Change-Id: If38dad6b6ae8198f40e863d0847fa5080a2997e3 Reviewed-on: https://swiftshader-review.googlesource.com/7354Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
A few types were missing: 8b/16b, signed/unsigned integers Change-Id: I1c168b5df941f6c01801406ba7879221b682fb96 Reviewed-on: https://swiftshader-review.googlesource.com/7353Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 26 Sep, 2016 1 commit
-
-
Alexis Hetu authored
Fixes all failures in dEQP.functional.vertex_arrays.* In OpenGL ES 3.0, vertex attributes streams can be of a different type from the vertex attribute in the shader. For this reason, some conversion may be required. This cl solves this issue by: 1) Sending the information about the vertex attribute's type in the shader to the vertex routine. 2) Handling this information by adding conversion where appropriate. Change-Id: I04a5a34aea12684209e584aa5f15a3edfd57f956 Reviewed-on: https://swiftshader-review.googlesource.com/7254Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 23 Sep, 2016 1 commit
-
-
Alexis Hetu authored
It appears that some bundling operations on Mac actually require that all libraries have a unique name, since output_dir is effectively ignored and overwritten in that scenario, so I have to remove that renaming code on Mac so that SwiftShader doesn't clash with Angle. Change-Id: I3d36df525ad9229da0c35f27d367e7deef7e430d Reviewed-on: https://swiftshader-review.googlesource.com/7271Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 22 Sep, 2016 1 commit
-
-
Meng-Lin Wu authored
To avoid duplicating code, Float4(RValue<Byte4>) and Float4(RValue<SByte4>) now call Int4(RValue<Byte4>) and Int4(RValue<SByte4>) respectively. Change-Id: Ic49fdd151d15cc1c2753f535d3654028e7a1910a Reviewed-on: https://swiftshader-review.googlesource.com/5560Tested-by:
Meng-Lin Wu <marleymoo@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 21 Sep, 2016 3 commits
-
-
Alexis Hetu authored
It appears this issue might be fixable directly in chromium: (https://codereview.chromium.org/2359913002/) For now, removing the mac specific change. BUG=31645700 Change-Id: I3d43c01f2d6500d67a298a047e8d98153f06dfb4 Reviewed-on: https://swiftshader-review.googlesource.com/7230Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com>
-
Alexis Hetu authored
The following cl had broken all fragment uniform buffers: https://swiftshader-review.googlesource.com/6950 Since the null pointer check was only made on the vertex uniform buffer, it was preventing the fragment uniform buffer from being set, which was causing a crash when this happened. There was a second issue with this cl, which was allowing 'uniformBuffers[-1].get()' to be called, which should not have been allowed. Also, it is safer to call VertexProcessor::setUniformBuffer and PixelProcessor::setUniformBuffer unconditionally in order to set the null pointers correctly when no buffers are set. Change-Id: I275b0bca6173ee3df4e0e369926ed0a94500e1ba Reviewed-on: https://swiftshader-review.googlesource.com/7216Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
Since the swiftshader libraries are loaded at runtime by chromium, the correct way to add swiftshader as a dependency is to use 'loadable_module' for mac and 'data_deps' for other platforms. BUG=31645700 Change-Id: Ic393a4132fbfc44e71247fc0294cca2cb57d35c8 Reviewed-on: https://swiftshader-review.googlesource.com/7217Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 16 Sep, 2016 3 commits
-
-
Alexis Hetu authored
Tying up loose ends from the new output formats project. One of the missing outputs was the RG8 format, which I added here. Change-Id: I865ad4c462e892dbc3a7820003b2390b54453ee9 Reviewed-on: https://swiftshader-review.googlesource.com/7171Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
There was a silly mistake in the code that I just found. Instead of being ordered (1,2,3,4), the 4 pixels being outputted were ordered (3,4,1,2), which was within the tolerance threshold for most of the image, but exhibited weird sparse diagonal lines of errors (values outside of threshold). Reordered the pixels properly to fix the issue. Change-Id: I082463d22e585e5ee98836a87ef60efca05e58f0 Reviewed-on: https://swiftshader-review.googlesource.com/7192Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug swiftshader:5 Change-Id: I6aec7237f2db6222702e828a2b9dae1a1a6ac9af Reviewed-on: https://swiftshader-review.googlesource.com/7191Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 15 Sep, 2016 3 commits
-
-
Alexis Hetu authored
The default alpha value for integer types is 1 and not the largest representable value for that type, as it was previously implemented. A value of 1 still means that the texture is opaque. This fixes ~60 fragment output tests. Change-Id: I96650fc8ba4812dcb8a2787979c6829f14f0960e Reviewed-on: https://swiftshader-review.googlesource.com/7190Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
Change-Id: I5e8424e7150d1aaede8f5a23c34c7fb023959f0c Reviewed-on: https://swiftshader-review.googlesource.com/5790Reviewed-by:
Nicolas Capens <capn@google.com> Tested-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
For transform feedback, any vertex shader output can be read from and which outputs will be read by transform feedback buffers is not known at compile time. For that reason, any output being written to in the vertex shader code shouldn't be optimized out. Change-Id: Ia4322c43b7e3308ec5d930650e70aacf032dc6ec Reviewed-on: https://swiftshader-review.googlesource.com/5051Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 13 Sep, 2016 1 commit
-
-
Alexis Hetu authored
Fixed a few things in the BUILD.gn files: 1) Style guide mentions single entries should be on single lines 2) Changed "host_os == ..." for the is_... equivalent values 3) Added missing Mac libs and include dirs Change-Id: I28d35fd86a8cf9c9157a397370a385fe3e5e958b Reviewed-on: https://swiftshader-review.googlesource.com/7152Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 09 Sep, 2016 5 commits
-
-
Alexis Hetu authored
I've found no occurrence of warning 4530 being triggered and only 3 conversion warning 4244, so I fixed them and re-enabled them. Change-Id: I06647466a12faf57c866f88f9124041324fc2490
-
Nicolas Capens authored
BUG=swiftshader:3 Change-Id: I070059f3b4e9e431c7667b98017912b72d9d66a4 Reviewed-on: https://swiftshader-review.googlesource.com/6952Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
BUG=swiftshader:3 Change-Id: I114384fc0e25ee1f7874f726b85ef5bcaf01ed36 Reviewed-on: https://swiftshader-review.googlesource.com/6951Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
Some alignment warnings were popping up on some windows bots. Rather than silencing them, I just fixed them. Change-Id: I21bc558e04498357c5d76a9caf9bd86f0a5cb540 Reviewed-on: https://swiftshader-review.googlesource.com/7131Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
There were only a handful of integer conversion warnings in libGLESv2, so I fixed them rather than silencing them. Change-Id: I94a7f2d4e4c34a48a8f128e6303778ef23ac45c9 Reviewed-on: https://swiftshader-review.googlesource.com/7132Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 08 Sep, 2016 1 commit
-
-
Nicolas Capens authored
Change-Id: I3e3207280ae4b7c66811c9ec11140ece2335112a Reviewed-on: https://swiftshader-review.googlesource.com/7092Reviewed-by:
Nicolas Capens <capn@google.com> Tested-by:
Nicolas Capens <capn@google.com>
-
- 07 Sep, 2016 1 commit
-
-
Nicolas Capens authored
Older Chrome branches still need the Register() function. Bug chromium:644329 Change-Id: I6b07aea6e069604f3e5d9e08c97c287df31c0dd6 Reviewed-on: https://swiftshader-review.googlesource.com/7090Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> (cherry picked from commit 95db80e4269db8ee31432590defa699d6dd36f6b) Reviewed-on: https://swiftshader-review.googlesource.com/7091Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 01 Sep, 2016 2 commits
-
-
Alexis Hetu authored
Windows clang requires that intrinsics be properly declared through intrin.h, otherwise, intrinsics were getting manually declared through a pragma instruction from winnt.h and clang returned an error. Change-Id: Iddfa4e3d6c76388e986dc445e40fd05359c2763c Reviewed-on: https://swiftshader-review.googlesource.com/7050Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
Alexis Hetu authored
Windows clang has no option to do anything similar to "-fno-operator-names", so it generates errors without any way to silence them. Renaming these functions is easy enough, so it was done here. Also removed the now useless flag from the code blocks project files. Change-Id: I9e25e25a72bf24567e3be928e07b187df87398bc Reviewed-on: https://swiftshader-review.googlesource.com/7051Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 31 Aug, 2016 1 commit
-
-
Alexis Hetu authored
Made Windows files work for Windows clang. Verified on Linux and Windows Visual Studio compiler to make sure nothing was broken. Change-Id: I82815491579cdfca602660279f6831c8820249f6 Reviewed-on: https://swiftshader-review.googlesource.com/7052Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Corentin Wallez <cwallez@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com>
-
- 22 Aug, 2016 2 commits
-
-
Nicolas Capens authored
Bug swiftshader:2 Change-Id: I43c55f948a0816cf062b9b2e9d158c9a52fafdbc Reviewed-on: https://swiftshader-review.googlesource.com/6930Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com> Tested-by:
Nicolas Capens <capn@google.com>
-
Nicolas Capens authored
Bug b/30838145 Change-Id: I06aeafa6e7a8aefbc1d4447a9d2607895d07a096 Reviewed-on: https://swiftshader-review.googlesource.com/6950Tested-by:
Nicolas Capens <capn@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <capn@google.com>
-
- 12 Aug, 2016 1 commit
-
-
Anthony Vallee-Dubois authored
This change also fixes a missing include for memcmp(). Bug 14642677 Change-Id: I4207d259b9d7b20e0cb35ded65bacd5c2d882087 Reviewed-on: https://swiftshader-review.googlesource.com/3884Reviewed-by:
Nicolas Capens <capn@google.com> Tested-by:
Nicolas Capens <capn@google.com>
-
- 11 Aug, 2016 1 commit
-
-
Chinmay Garde authored
Change-Id: I6b63d75ee2b18be92544227ff0162936f5a2328f Reviewed-on: https://swiftshader-review.googlesource.com/6910Reviewed-by:
Nicolas Capens <capn@google.com> Tested-by:
Chinmay Garde <chinmaygarde@google.com>
-