Skip to content

S
swiftshader
Switch branch/tag
  1. 21 Jan, 2021 6 commits
    • Mend git merge history for SPIRV-Headers and SPIRV-Tools · d79f17cf
      Ben Clayton authored 
      https://swiftshader-review.googlesource.com/c/SwiftShader/+/50749
squashed together the changes from two `git subtree pull` commands (for
SPIRV-Headers and SPIRV-Tools), which resulted in a change with a single
parent.

`git subtree` expects a merge commit with a single set of `git-subtree-*`
metadata tags. The squash resulted in a change description with two sets
of these, confusing `git subtree` even further.

The process to fix this is probably not the easiest, but the first one I found that works:

  # Rollback to the parent change of 50749
  git checkout 43bb60e1
  # Perform the subtree pull again for SPIRV-Headers (to 104ecc356)
  git subtree pull --prefix third_party/SPIRV-Headers https://github.com/KhronosGroup/SPIRV-Headers 104ecc356c1bea4476320faca64440cd1df655a3 --squash -m "Update SPIR-V Headers"
  # Note the second parent of the merge for the SPIRV-Headers (Merge: 43bb60e1 ded361b4)
  git show -1
  # Perform the subtree pull again for SPIRV-Tools (to b0e22d28f)
  git subtree pull --prefix third_party/SPIRV-Tools https://github.com/KhronosGroup/SPIRV-Tools b0e22d28f5ec891d6b256703575d08fce4228bd9 --squash -m "Update SPIR-V Tools"
  # Note the second parent of the merge for the SPIRV-Tools (Merge: a00ce432a9d bdb9eea9)
  git show -1
  # Jump back to master
  git checkout origin/master
  # Revert 50749
  git revert 9dff6a3b
  # Re-merge the squashed update change for SPIRV-Headers  ** this change **
  git merge -Xsubtree="third_party/SPIRV-Headers" ded361b4
  # Re-merge the squashed update change for SPIRV-Tools
  git merge -Xsubtree="third_party/SPIRV-Tools" bdb9eea9

Bug: b/174239232
Change-Id: Ibab851eb610b1448193f57a628eb95bdb58057ee
    • Revert 9dff6a3b to mend git merge history for SPIRV-Headers and SPIRV-Tools · 24d6bf7f
      Ben Clayton authored 
      https://swiftshader-review.googlesource.com/c/SwiftShader/+/50749
squashed together the changes from two `git subtree pull` commands (for
SPIRV-Headers and SPIRV-Tools), which resulted in a change with a single
parent.

`git subtree` expects a merge commit with a single set of `git-subtree-*`
metadata tags. The squash resulted in a change description with two sets
of these, confusing `git subtree` even further.

The process to fix this is probably not the easiest, but the first one I found that works:

  # Rollback to the parent change of 50749
  git checkout 43bb60e1
  # Perform the subtree pull again for SPIRV-Headers (to 104ecc356)
  git subtree pull --prefix third_party/SPIRV-Headers https://github.com/KhronosGroup/SPIRV-Headers 104ecc356c1bea4476320faca64440cd1df655a3 --squash -m "Update SPIR-V Headers"
  # Note the second parent of the merge for the SPIRV-Headers (Merge: 43bb60e1 ded361b4)
  git show -1
  # Perform the subtree pull again for SPIRV-Tools (to b0e22d28f)
  git subtree pull --prefix third_party/SPIRV-Tools https://github.com/KhronosGroup/SPIRV-Tools b0e22d28f5ec891d6b256703575d08fce4228bd9 --squash -m "Update SPIR-V Tools"
  # Note the second parent of the merge for the SPIRV-Tools (Merge: a00ce432a9d bdb9eea9)
  git show -1
  # Jump back to master
  git checkout origin/master
  # Revert 50749
  git revert 9dff6a3b
  # Re-merge the squashed update change for SPIRV-Headers
  git merge -Xsubtree="third_party/SPIRV-Headers" ded361b4
  # Re-merge the squashed update change for SPIRV-Tools
  git merge -Xsubtree="third_party/SPIRV-Tools" bdb9eea9

Bug: b/174239232
Change-Id: I0b498184493f0fcb00b1684182b6a89076a5cca0
    • Squashed 'third_party/SPIRV-Headers/' changes from 7845730cab6..104ecc356c1 · ded361b4
      Ben Clayton authored 
      104ecc356c1 Merge pull request #180 from dneto0/issue-179
93effe16473 MeshShadingNV enables builtins PrimitiveId, Layer, and ViewportIndex
4de110ce1c7 Merge pull request #182 from dgkoch/khr_rt_final
6323756078a de-alias/reassign OpIgnoreIntersectionKHR/OpTerminateRayKHR
928a192ee62 Raytracing and Rayquery updates for final
eb5e409129c Updated headers for new trace/executeCallable and acceleration structure cast.
5ab5c96198f Reserve additional loop control bit for Intel extension (NoFusionINTEL) (#175)
793f9199bed Add EmbarkStudios/rust-gpu to vendor list. (#174)

git-subtree-dir: third_party/SPIRV-Headers
git-subtree-split: 104ecc356c1bea4476320faca64440cd1df655a3
    • Target macOS 10.12 when building LLVM · 5b36fbc5
      Nicolas Capens authored 
      When enabling C++17, LLVM uses std::shared_mutex instead of its own
implementation. On macOS it requires version 10.12 (Sierra), while
Chrome must still support 10.11. However, we don't need LLVM as Reactor
backend, except for ARM64, but Mac devices with such CPU have a more
recent OS version.

llvm-10.0/llvm/include/llvm/Support/RWMutex.h:100:8: error: 'shared_mutex' is unavailable: introduced in macOS 10.12

Bug: b/174843857
Change-Id: Ie24095aa17d22072cfb9d655b87252cbea5209d0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/52030Reviewed-by: 's avatarAntonio Maiorano <amaiorano@google.com>
Tested-by: 's avatarNicolas Capens <nicolascapens@google.com>
    • Use aligned allocation for classes with aligned members · 532770bb
      Nicolas Capens authored 
      The legacy PixelProcessor and VertexProcessor classes for OpenGL ES
contain member fields with types like float4, which require 16-byte
alignment. This isn't guaranteed by the compiler until C++17. On macOS,
this also requires OS support, which was added in version 10.14
(Mojave), while Chrome still has to support macOS 10.11.

src/Renderer/PixelProcessor.cpp:75:18: error: aligned deallocation function of type 'void (void *, std::align_val_t) noexcept' is only available on macOS 10.14 or newer

Overriding new and delete for these classes allows us to use our custom
allocator which guarantees alignment. Note that the Renderer class
already used the same approach. Also note that this hasn't caused issues
before because these fields aren't actually accessed by instructions
which demand alignment. However, it's still good for performance and to
align with the intent (pun intended).

Bug: b/174843857
Change-Id: Ia5de5f6fe67a4f54805cdde3fed565c30b9318a8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/52029Reviewed-by: 's avatarAntonio Maiorano <amaiorano@google.com>
Tested-by: 's avatarNicolas Capens <nicolascapens@google.com>
    • Enable and require C++17 compilation · b3e5c440
      Nicolas Capens authored 
      Updates build files to specify using the C++17 standard.

Add a usage of the C++17 exclusive [[maybe_unused]] attribute as a test.
Also unconditionally use <filesystem> in Reactor unit tests.

C++17 deprecates things like std::unary_function<>, which were present
in Subzero's LLVM headers, but the functionality was unused so it has
been deleted.

Bug: b/174843857
Bug: b/155971541
Change-Id: Ib167726535399b0b12751a3d93feaa9db762cd5d
Signed-off-by: 's avatarNicolas Capens <capn@google.com>
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/51948Tested-by: 's avatarNicolas Capens <nicolascapens@google.com>
Reviewed-by: 's avatarAntonio Maiorano <amaiorano@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
  3. 20 Jan, 2021 4 commits
  5. 19 Jan, 2021 1 commit
  7. 18 Jan, 2021 2 commits
  9. 15 Jan, 2021 2 commits
  11. 14 Jan, 2021 2 commits
  13. 13 Jan, 2021 5 commits
  15. 11 Jan, 2021 3 commits
  17. 10 Jan, 2021 1 commit
  19. 08 Jan, 2021 5 commits
  21. 07 Jan, 2021 3 commits
  23. 05 Jan, 2021 1 commit
  25. 30 Dec, 2020 1 commit
    • Add support for Vulkan12Properties · ff29e249
      Sean Risser authored 
      VkPhysicalDeviceVulkan12Properties was added to the core spec by Vulkan
1.2. It allows users to query the following properties simultaneously:
    * VkPhysicalDeviceDriverProperties
    * VkPhysicalDeviceFloatControlsProperties
    * VkPhysicalDeviceDescriptorIndexingProperties
    * VkPhysicalDeviceDepthStencilResolveProperties
    * VkPhysicalDeviceSamplerFilterMinmaxProperties
    * VkPhysicalDeviceTimelineSemaphoreProperties
    * and framebufferIntegerColorSampleCounts

All these structs are a part of the core spec and must be supported.

Bug: b/176498010
Change-Id: I82b06f219d2ee091139964ee8b19f05e1c64e3a8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/51408
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: 's avatarNicolas Capens <nicolascapens@google.com>
Tested-by: 's avatarSean Risser <srisser@google.com>
Commit-Queue: Sean Risser <srisser@google.com>
  27. 17 Dec, 2020 2 commits
    • Optimize reciprocal sqrt operation · 1cc5b335
      Antonio Maiorano authored 
      This change deprecates rr::RcpSqrt_pp with rr::RcpSqrt. As with Rcp,
RcpSqrt computes the result using Newton-Rhapson if it's faster and the
initial approximation intrinsic is available on the current target.
Currently, only LLVM on Intel will use NR for RelaxedPrecision. Note
that passing in Precision::Relaxed will produce a faster, but less
precise reciprocal sqrt.

Also made it so that SprivShader instruction GLSLstd450InverseSqrt now
invokes RcpSqrt(x, Precision::Full) instead of performing 1/sqrt(x).
Note that the Vulkan spec states that inversesqrt()'s precision is 2
ULP, and sqrt()'s precision is inherited from 1.0 / inversesqrt();
however, our rr::Sqrt is implemented in terms of x86's sqrt intrinsic on
x86, or as calls to sqrt from Math.h.

Bug: b/169760262
Change-Id: I65ba9a64d1db934c523dda11c1a2c186059d220b
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/51268
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: 's avatarNicolas Capens <nicolascapens@google.com>
Tested-by: 's avatarAntonio Maiorano <amaiorano@google.com>
    • Optimize reciprocal operation · d1561871
      Antonio Maiorano authored 
      This change deprecates rr::Rcp_pp with rr::Rcp, which makes sure to
correctly compute the reciprocal using the Newton-Rhapson refinement
only if the current target supports the required instrinsic, otherwise
using 1 / x. Currently, only LLVM on Intel will use NR. Note that
passing in Precision::Relaxed will produce a faster, but less precise
reciprocal.

Also removed PixelProgram::linearToSRGB as it's unused.

Bug: b/169760262
Bug: b/149574741
Change-Id: I4a2f943aa60116c4397d7a8ae18583a260824788
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/50648Reviewed-by: 's avatarAlexis Hétu <sugoi@google.com>
Reviewed-by: 's avatarNicolas Capens <nicolascapens@google.com>
Tested-by: 's avatarAntonio Maiorano <amaiorano@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
  29. 15 Dec, 2020 1 commit
  31. 14 Dec, 2020 1 commit