- 30 Oct, 2017 1 commit
-
-
Alexis Hetu authored
Currently Windows and Linux already use the Subzero backend, while MacOS still uses the LLVM backend. This cl switches MacOS to using the Subzero backend. Change-Id: Ibc20add93c5f413271832fd1dd7fe7ae00cda9f4 Reviewed-on: https://swiftshader-review.googlesource.com/11213Reviewed-by:
Alexis Hétu <sugoi@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
- 27 Oct, 2017 6 commits
-
-
Corentin Wallez authored
BUG=swiftshader:86 Change-Id: Ia57c894170e790cdd1ea2fe3b92bbba7f643ba64 Reviewed-on: https://swiftshader-review.googlesource.com/13368Tested-by:
Corentin Wallez <cwallez@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
Bug swiftshader:86 Change-Id: I8339f3882163daa078ac71034aaf37147e51d725 Reviewed-on: https://swiftshader-review.googlesource.com/13349Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
Change-Id: I119c9bd65624035cbb0afb850b061e140b6a296f Reviewed-on: https://swiftshader-review.googlesource.com/13348Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com>
-
Corentin Wallez authored
We should really have these suppressions in a shared config over all swiftshader. BUG=swiftshader:86 Change-Id: I1b8fa6342090e78e2338a7d748bcb29911f9e393 Reviewed-on: https://swiftshader-review.googlesource.com/13328Reviewed-by:
Alexis Hétu <sugoi@google.com> Tested-by:
Corentin Wallez <cwallez@google.com>
-
Corentin Wallez authored
The fuzzer still needs to parse state and source from `data` before it can actually be useful. Bug swiftshader:86 Change-Id: Iafcbce4748b83a4fb9f829e3e8a473d161d3a696 Reviewed-on: https://swiftshader-review.googlesource.com/13249Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Corentin Wallez authored
Without this the GN shared library doesn't export any of the gl entry-points making it useless. Bug swiftshader:86 Change-Id: I97f6acf4204d6e908ad12c89c79d40769dc824a4 Reviewed-on: https://swiftshader-review.googlesource.com/13308Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
- 26 Oct, 2017 4 commits
-
-
Nicolas Capens authored
This helps clarify dependencies and simplifies include paths. Bug swiftshader:86 Change-Id: I564ee420bb9029fa6428e49b63a86d633301bec2 Reviewed-on: https://swiftshader-review.googlesource.com/13288Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
There's a data race involving threadCount, so it is now an AtomicInt object. Also removed unused extern variables. Bug chromium:778254 chromium:776621 Change-Id: I93174e8f4e5e48f26414ba9c81bb9a2c317dbf73 Reviewed-on: https://swiftshader-review.googlesource.com/13290Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
Wrote a new version of primitive restart entirely handled by the index data manager. Passes all primitive_restart dEQP tests Change-Id: I0ce28764a33300a6625f1c369f95d8d178e1e3c0 Reviewed-on: https://swiftshader-review.googlesource.com/10951Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
Render targets with only a red or alpha component were getting swapped values within pixel quads at the edge of primitives. Also provide format mappings for GL_EXT_texture_rg. Change-Id: Ifb395039fb2958400ab36d5d184d7952d60fa9c6 Reviewed-on: https://swiftshader-review.googlesource.com/13268Reviewed-by:
Alexis Hétu <sugoi@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
- 24 Oct, 2017 1 commit
-
-
Corentin Wallez authored
Bug: swiftshader:86 Change-Id: I351e433123abfaf010e3b1641383054e77dc6383 Reviewed-on: https://swiftshader-review.googlesource.com/13228Tested-by:
Corentin Wallez <cwallez@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
- 19 Oct, 2017 1 commit
-
-
Alexis Hetu authored
Rather than have to download a previous version (10.0.15063.0), we can update the solution to the current version (10.0.16299.0). Built solution in 32b/64b release/debug with no failures. Change-Id: If19d8f6edcce4bcfac21c005f647ab1619837d16 Reviewed-on: https://swiftshader-review.googlesource.com/13208Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
- 11 Oct, 2017 1 commit
-
-
Nicolas Capens authored
Visual Studio 2017 doesn't like configuration elements with empty content. Remove it to use the default value. Change-Id: Idc4f281aff145c7e9a143c0a5dbdd54eb4658ee9 Reviewed-on: https://swiftshader-review.googlesource.com/13128Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com>
-
- 10 Oct, 2017 1 commit
-
-
Nicolas Capens authored
Chrome now uses Visual Studio 2017 as the default compiler on Windows. To ensure we maintain compatibility during standalone development, we should use VS2017 for SwiftShader as well. The 'Community' edition of Visual Studio 2017 is freely available. Bug chromium:683729 Change-Id: I3ed1edaeb9fa786b575202ba5b9c86faf1daa3c0 Reviewed-on: https://swiftshader-review.googlesource.com/13048Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
- 06 Oct, 2017 3 commits
-
-
Alexis Hetu authored
Uniform block members do not get a uniform location, since they cannot be set (or get) through regular glUniform* functions. These are instead always set using uniform buffers. Bug chromium:763675,chromium:763698,chromium:771639 Change-Id: Icf65eba3faab420c63f70af66f828a786e3fc446 Reviewed-on: https://swiftshader-review.googlesource.com/13068Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
Nicolas Capens authored
Blaze (Bazel) requires linker dependencies to have a .ld, .lds, or .ldscript extension. Note that version scripts are just the contents of the VERSION command in (true) linker scripts. However, the gold linker used by Android does not support implicit linker scripts (which amend the default linker script), nor the INSERT command. Therefore these linker scripts are still actually just version scripts. https://sourceware.org/binutils/docs/ld/VERSION.html Bug b/67482470 Change-Id: I3bbe25dab6d916376bdc41bb723279ed2ef62fb0 Reviewed-on: https://swiftshader-review.googlesource.com/13108Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
We perform depth testing before computing scanline invariants, as an optimization. But it was not taking into account that the fixed-function depth value may require clamping due to the depth bias. Bug swiftshader:82 Change-Id: I6f8953f9d060c73b5fe209b8cbad70e7cb7588b4 Reviewed-on: https://swiftshader-review.googlesource.com/13088Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
- 03 Oct, 2017 8 commits
-
-
Casey Dahlin authored
Test: Improved test passes Bug: b/67106219 Change-Id: Iab15a2331a91dffaaeb1627a42d0fb88dfa841d0 Reviewed-on: https://swiftshader-review.googlesource.com/13028Tested-by:
Casey Dahlin <sadmac@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
3D texture sampling did not support width*height > MAX_USHORT(65535). Changed for 32 bit integer computation for that multiplication to solve this issue. Verified with 3D texture related dEQP tests. Bug b/26105892 chromium:763382 Change-Id: Ifd07278f6263d000620a0b525c07ed4fcbdcf2e7 Reviewed-on: https://swiftshader-review.googlesource.com/12988Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
Nicolas Capens authored
Bug b/37496338 Bug b/37495545 Bug b/37496078 Bug b/37496856 Bug b/37496321 Bug b/37496082 Change-Id: I173a296703fcc1f49810742d95ae6abe7c9bcf01 Reviewed-on: https://swiftshader-review.googlesource.com/12868Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
Subzero assumes that stores and loads to/from stack memory can take an immediate offset. This isn't true for sub-vector loads and stores. We can just use regular full vector loads and stores instead. Bug b/37496321 Change-Id: Iaffd60f8077a4c9d602665421f90a7288ab04050 Reviewed-on: https://swiftshader-review.googlesource.com/13008Reviewed-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Casey Dahlin authored
Test: New test passes Bug: b/37496078 Change-Id: Ieff1b080bdde3f9c86a0531566c8b08d7cdf24fe Reviewed-on: https://swiftshader-review.googlesource.com/13009Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Casey Dahlin <sadmac@google.com>
-
Nicolas Capens authored
Use VDUP for replicating a single element. Use VZIP for interleaving vectors. Use VMOV Dd, Dm for rearranging quadword vectors. Bug b/67106219 Change-Id: I0de1457454c1db6d467bf870288b7af7cb59ac09 Reviewed-on: https://chromium-review.googlesource.com/695004Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12968Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
When the first source register is different from the destination register, but equal to the second source, we need to first narrow the second source to prevent overwriting data we still need. Similar when the destination is equal to the first source. Bug b/37496082 Change-Id: I908a8e125a77ec4bf1eb5eab9e48c6112ee4ca13 Reviewed-on: https://chromium-review.googlesource.com/696032Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12949Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
OpenGLES 3.0 allows for more combinations of format/internalformat for floating point types. These types were already supported, just not using the combinations added here. Bug chromium:763384 Change-Id: I146548c2920799c7ea0d5d537d556ba562708147 Reviewed-on: https://swiftshader-review.googlesource.com/12928Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
- 02 Oct, 2017 6 commits
-
-
Casey Dahlin authored
Test: New tests pass Bug: b/37496856 Change-Id: Ifae1f802d61a7d51f6e8f5c135e3c8c67be30125 Reviewed-on: https://swiftshader-review.googlesource.com/12950Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Casey Dahlin <sadmac@google.com>
-
Nicolas Capens authored
We weren't clamping the value when less than zero. Bug b/37496082 Change-Id: I3170d14a622f8571f0142dd4ce80f96617c7e0aa Reviewed-on: https://swiftshader-review.googlesource.com/12948Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Casey Dahlin <sadmac@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
When SSE4.1 isn't available, we were infinitely calling PackUnsigned() recursively. Bug b/37496082 Change-Id: Ie1fa8150d64740776c5b80baf07d997dde9d51be Reviewed-on: https://swiftshader-review.googlesource.com/12929Tested-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Casey Dahlin <sadmac@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Casey Dahlin authored
Test: New tests pass Bug: b/37496082 Change-Id: I4a0aae14318f03b5ce59f12f7542db13270f52ab Reviewed-on: https://swiftshader-review.googlesource.com/12848Tested-by:
Casey Dahlin <sadmac@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
Integer division already had this modification, to avoid division by 0, but integer modulo hadn't been modified to match the other fix, which is now done here. Bug chromium:763435 Change-Id: I40f4ed9f1c4b5889ddf47abc1bb7c70c21cbd7b8 Reviewed-on: https://swiftshader-review.googlesource.com/12908Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Alexis Hétu <sugoi@google.com>
-
Nicolas Capens authored
MultiplyAddPairs is implemented using VMULL+VPADD. MultiplyHighSigned/Unsigned is implemented using VMULL+VSHRN. SubVectorLoad/Store is implemented using VLDR/VLD1/VSTR/VST1. VectorPackSigned/Unsigned is implemented using two VQMOVN. Bug b/37496078 Bug b/37496856 Bug b/37496321 Bug b/37496082 Change-Id: I141fd901d53da24ce780f503dc7ad17b94fc6ba8 Reviewed-on: https://chromium-review.googlesource.com/693049Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12709Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
- 29 Sep, 2017 5 commits
-
-
Casey Dahlin authored
Test: New test passes Bug: b/67106219 Change-Id: Ib21f73d312d39cf2b3aa15a11d2a207885a58978 Reviewed-on: https://swiftshader-review.googlesource.com/12828Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Casey Dahlin <sadmac@google.com>
-
Alexis Hetu authored
If imageSize is 0, inputPitch may be 0, which would cause a division by 0. In any case, if imageSize is 0, there's no work to perform, so we just skip loadCompressedData() entirely in that case. Bug:765094 Change-Id: Iedc6516ef6d025d24a8827597045cb3b83599e4a Reviewed-on: https://swiftshader-review.googlesource.com/12648Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Alexis Hetu authored
WindowSurface::checkForResize() contained an assert which got triggered in some tests. In order to avoid this issue, this cl gracefully handles an invalid window handle. Change-Id: I6bace88370980e94676d4af9de0d88e561175835 Reviewed-on: https://swiftshader-review.googlesource.com/12808Tested-by:
Alexis Hétu <sugoi@google.com> Reviewed-by:
Nicolas Capens <nicolascapens@google.com>
-
Casey Dahlin authored
Bug b/37495545 Test: New tests pass Change-Id: Icc81729d8c65aadccdd685490218d298e671b638 Reviewed-on: https://swiftshader-review.googlesource.com/12768Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Casey Dahlin <sadmac@google.com>
-
Casey Dahlin authored
Bug b/37495545 Change-Id: I1d025f894bb7cf08dcaafd950605781633596ab3 Reviewed-on: https://chromium-review.googlesource.com/689098Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12748Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
- 28 Sep, 2017 3 commits
-
-
Nicolas Capens authored
This also fixes the encoding of right shifts with unsigned elements. Bug b/37496338 Change-Id: I9a1dc91359daea5f4391a137b7f9e03bd941146b Reviewed-on: https://chromium-review.googlesource.com/688057Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12668Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Nicolas Capens authored
Change-Id: I24a2ac536c24732538cc5c7f2aad4e7763989f9e Reviewed-on: https://chromium-review.googlesource.com/568673Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Reviewed-by:
Jim Stichnoth <stichnot@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12690Tested-by:
Nicolas Capens <nicolascapens@google.com>
-
Aaron Gable authored
R=bradnelson@chromium.org Bug: 665593 Change-Id: I1b9a5fd31c5e426bd193a32a2feb291eb17af5a6 Reviewed-on: https://chromium-review.googlesource.com/587993Reviewed-by:
Brad Nelson <bradnelson@chromium.org> Reviewed-on: https://swiftshader-review.googlesource.com/12689Reviewed-by:
Nicolas Capens <nicolascapens@google.com> Tested-by:
Nicolas Capens <nicolascapens@google.com>
-