Change-Id: I05b0a9fbad99116a7163a6cb4ad43e7eccb060f3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41808
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

To temporarily work around build issues related to Google3 integration,
make ASTC support optional and only enable it for CMake, Chromium, and
Android builds.

Bug: b/150600814
Change-Id: Iac6a404322c65a6f2727807ea8af2edfb219a5a2
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41828Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Sean Risser <srisser@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

It's valid in Vulkan for a shader to have specialization constants
and the pipeline to be lacking specialization constant data. This
could lead to a key match between the two even though there should
be a different shader payload. From the spec:

> It is legal for a SPIR-V module with specializations to be
> compiled into a pipeline where no specialization info was provided.
> SPIR-V specialization constants contain default values such that
> if a specialization is not provided.

Fix this by adding in the required checks. Should not affect perf.

Bug: b/150449637
Bug: angleproject:4426
Change-Id: I5f384420bdc1f03a36852db0b20393ae56452b62
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41728Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: Jamie Madill <jmadill@chromium.org>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

This cl adds both LDR and HDR support for ASTC compressed
textures. Only LDR formats are exposed in the
PhysicalDevice's properties, but HDR support can be added
trivially later by adding the HDR formats and exposing
"VK_EXT_texture_compression_astc_hdr".

Pulled from: https://github.com/ARM-software/astc-encoder
Git hash: 81a5e50741b4c8302cf7d78f314a53e44ee68e1f

The astc-encode git repo was added to third-party, with a
few minor modifications:
1) All encoding related code has been ripped out, only
   decoding related code remains
2) Replaced ASTC_CODEC_INTERNAL_ERROR() with UNREACHABLE()
   in a switch statement in astc_color_unquantize.cpp
3) Some functions were using a lot of stack memory, so I
   added a unique_ptr to allocate the same objects on the
   heap, to avoid potential issues.

LDR ASTC is decoded to 8bit unsigned RGBA.
HDR ASTC is decoded to 32b floating point.

Tests: dEQP-VK.*astc*
Bug: b/150130101
Change-Id: I6b03fed6e1f326a95c7aefe9f9a9d0a89cf24428
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41568Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

EGL_ANGLE_iosurface_client_buffer added a new attribute to the config:
EGL_BIND_TO_TEXTURE_TARGET_ANGLE, which needed to be properly exposed
in the EGL frontend.

Bug: chromium:1015454
Change-Id: I58881df18d711ed7974a0a06e53fed42b1f28638
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41709
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Set squash uploads to false. This should make git cl upload work
properly from SwiftShader for those who prefer that over 'git push'
to a ref.

Bug: None
Change-Id: Ib80fdbd28c2a7a5808ab73ceec8ddb98f8962e0b
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41708Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

For the System V ABI, variadic calls must store the number of floating
point arguments into RAX. This was mostly working by accident for
our calls to printf since RAX is used as a scratch register, and was
often non-zero, which is all that's really needed for printf with float
args to work; however, when RAX would become 0, printf would print the
wrong thing.

Bug: b/149913889
Change-Id: Id4b519c5416927d537fca078109c0dc850f08359
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41668Tested-by: Antonio Maiorano <amaiorano@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

The code in VkPipelineCache was correctly comparing all the attributes
of the specialization constants and then was also doing a faulty shared
pointer comparison. This was causing major drag on draw calls as every
draw call in ANGLE was recompiling the shader.

Bug: angleproject:4426
Change-Id: Ia0466738eb7de78c0c3b53e61e2f2bc6831bbcca
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41688
Kokoro-Presubmit: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

If the stack size is > 4K, emit chkstk, which probes the stack to commit
the pages required to support the large stack.

Bug: swiftshader:25
Change-Id: I6b9f09218736ffb641cb1dbf95a1de7149633ef8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41608
Presubmit-Ready: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

The CMake build was already using -fvisibility=protected, which has the
same effect but uses duplicate symbol entries for internal/external use.

Bug: b/148240133
Change-Id: I8069a1b293a6ec0604be32179ed591ffaa645ed7
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41648
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>

Referencing //third_party/SPIRV-Tools directly only works when building
inside a Chromium checkout.

Bug: angleproject:4425
Change-Id: I70ace58145a93b747a67d98b6efd4af7e3cfd370
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41628Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Bug: angleproject:4425
Change-Id: I2785b9f5c6f96b2b12eeacf64ff5316355f1ae54
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41609Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

Currently we produce a warning every time an unsupported extension is
used. The original bug (b/139528538) called for logging these structs
silently unless a debugger's attached.

So I've replaced all of our warnings for unsupported structs with
calls to LOG_TRAP. This is an update to TRACE_ASSERT. LOG_TRAP will
never emit to the debug log, instead it will only print to a file if
writing to files is enabled for logv.

Bug: b/148415347
Change-Id: Ib4ad2b20b3dffce4fac597c891b2f5ee23e032c4
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41348Tested-by: Sean Risser <srisser@google.com>
Presubmit-Ready: Sean Risser <srisser@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Change-Id: Id5607e9329a01bcc137cd2f9c25db798dab15a4a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41488
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

And remove external SPIRV-Headers dependency. We use an internal copy.

Bug: angleproject:4425
Change-Id: I36c8c836197d6642d4c696520d61bdf7790fce6c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41589Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Will be used by SpirvShaderEnumNames.cpp in https://swiftshader-review.googlesource.com/c/SwiftShader/+/41510

Bug: angleproject:4425
Change-Id: I631c593432c12a6ccf64edcb83b2e11ac87104c0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41588Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Also applies a fix for `spvOpcodeString` being moved to the public spirv-tools interface.

Changes:
    8910ea5f1 Fix Wrange-loop-analysis warnings in SPIRV-Tools. (#3201)
    c316fb15f Add missing dependencies when generating spvtools_core_tables (#3199)
    03c9effb5 Brief guide to writing a spirv-fuzz fuzzer pass (#3190)
    dea1040fa Fix ignored const qualifier warning in static_cast (#3197)
    4a80497a8 Make spvOpcodeString part of the public API (#3174)
    03794b8f5 Fix static libraries linking order (#3189)
    79f8caf91 vscode: Add missing fields to schema.Opcode (#3169)
    4d912f4e6 spirv-val: Add support for SPV_AMD_shader_image_load_store_lod (#3186)
    77fefe765 spirvfuzz: Fix type-related bug, change undef to zero, and add assert (#3188)
    6c218ec60 spirv-fuzz: Fuzzer pass that adds access chains (#3182)
    77fb303e5 spirv-fuzz: Fuzzer pass to add function calls (#3178)
    3d4a0dd48 spirv-fuzz: Ensure that donated variables are always initialized (#3181)
    fe10239f9 spirv-fuzz: Add fuzzer passes to add loads/stores (#3176)
    1f03ac102 spirv-fuzz: Fuzzer passes to add local and global variables (#3175)
    9e52bc0d0 utils/vscode: LSP - Support OpExtInst (#3140)
    a3b5ad909 Fix typos in opt's help. Update environment version. (#3170)
    da5457f6b Start SPIRV-Tools v2020.2
    8dd174809 Finalize SPIRV-Tools v2020.1
    76616bab6 Update CHANGES
    bb56e892f spirv-fuzz: Fuzzer pass to add composite types (#3171)
    fdd0c8776 Update script that checks copyright years. (#3173)
    b7e0998e3 spirv-fuzz: Disallow copying of null and undefined pointers (#3172)
    a9624b4d5 Handle TimeAMD in AmdExtensionToKhrPass. (#3168)
    ddcc11763 Update CHANGES
    0265a9d4d Implement constant folding for many transcendentals (#3166)
    7a2d408de Fix typo in comment. (#3163)
    1fc7a9ec7 spirv-fuzz: Arbitrary variable facts (#3165)
    dd3f8d492 spirv-fuzz: Add outlining test (#3164)
    521223b70 spirv-fuzz: Make functions "livesafe" during donation (#3146)
    97f1d485b Dead branch elim fix (#3160)
    1b3441036 Fix chromium build (#3152)
    5d786f6cc Clarify mapping of target env to SPIR-V version (#3150)
    e7afeb060 Use dummy switch instead of dummy loop in MergeReturn pass. (#3151)
    6729c4a7c Fix structured exit validation (#3141)
    bb236c326 Add spvParseVulkanEnv (#3142)
    dd37d73c5 Handle conflict between debug info and existing validation rule (#3104)
    b97057e7f Use spirv-headers' BUILD.gn (#3148)
    6a1a1b326 Roll external/spirv-headers/ af64a9e82..dc77030ac (4 commits) (#3147)
    ab7ac60f1 spirv-fuzz: Refactoring and type-related fixes (#3144)
    323a81fc5 Validate Vulkan 1.2 capabilities
    d46a1b048 Add basic support for Vulkan 1.2: SPV_ENV_VULKAN_1_2
    f8d7df760 Fix OpLine bug of merge-blocks pass (#3130)
    8013d477a spirv-fuzz: add dead blocks (#3135)
    18b3b9456 Remove names and decorations of imported symbols (#3081)
    fb2e81981 Fix GN build for OpenCL.DebugInfo.100 update (#3134)
    c86dc0bcd Fix bad parameter names in error message (#3129)
    899735f45 Revert PR #3093 (#3131)
    a466b99db Disallow forward references in arrays (#3093)
    31acc7882 spirv-fuzz: Add fuzzer pass to perform module donation (#3117)
    c8bf14393 GetOperandConstants operand can be const (#3126)

Commands:
    git subtree pull --prefix third_party/SPIRV-Tools https://github.com/KhronosGroup/SPIRV-Tools master --squash

Bug: b/123642959
Bug: angleproject:4425
Change-Id: I73076a22b474893d900a017d94a5d122061381bc

8910ea5f1 Fix Wrange-loop-analysis warnings in SPIRV-Tools. (#3201)
c316fb15f Add missing dependencies when generating spvtools_core_tables (#3199)
03c9effb5 Brief guide to writing a spirv-fuzz fuzzer pass (#3190)
dea1040fa Fix ignored const qualifier warning in static_cast (#3197)
4a80497a8 Make spvOpcodeString part of the public API (#3174)
03794b8f5 Fix static libraries linking order (#3189)
79f8caf91 vscode: Add missing fields to schema.Opcode (#3169)
4d912f4e6 spirv-val: Add support for SPV_AMD_shader_image_load_store_lod (#3186)
77fefe765 spirvfuzz: Fix type-related bug, change undef to zero, and add assert (#3188)
6c218ec60 spirv-fuzz: Fuzzer pass that adds access chains (#3182)
77fb303e5 spirv-fuzz: Fuzzer pass to add function calls (#3178)
3d4a0dd48 spirv-fuzz: Ensure that donated variables are always initialized (#3181)
fe10239f9 spirv-fuzz: Add fuzzer passes to add loads/stores (#3176)
1f03ac102 spirv-fuzz: Fuzzer passes to add local and global variables (#3175)
9e52bc0d0 utils/vscode: LSP - Support OpExtInst (#3140)
a3b5ad909 Fix typos in opt's help. Update environment version. (#3170)
da5457f6b Start SPIRV-Tools v2020.2
8dd174809 Finalize SPIRV-Tools v2020.1
76616bab6 Update CHANGES
bb56e892f spirv-fuzz: Fuzzer pass to add composite types (#3171)
fdd0c8776 Update script that checks copyright years. (#3173)
b7e0998e3 spirv-fuzz: Disallow copying of null and undefined pointers (#3172)
a9624b4d5 Handle TimeAMD in AmdExtensionToKhrPass. (#3168)
ddcc11763 Update CHANGES
0265a9d4d Implement constant folding for many transcendentals (#3166)
7a2d408de Fix typo in comment. (#3163)
1fc7a9ec7 spirv-fuzz: Arbitrary variable facts (#3165)
dd3f8d492 spirv-fuzz: Add outlining test (#3164)
521223b70 spirv-fuzz: Make functions "livesafe" during donation (#3146)
97f1d485b Dead branch elim fix (#3160)
1b3441036 Fix chromium build (#3152)
5d786f6cc Clarify mapping of target env to SPIR-V version (#3150)
e7afeb060 Use dummy switch instead of dummy loop in MergeReturn pass. (#3151)
6729c4a7c Fix structured exit validation (#3141)
bb236c326 Add spvParseVulkanEnv (#3142)
dd37d73c5 Handle conflict between debug info and existing validation rule (#3104)
b97057e7f Use spirv-headers' BUILD.gn (#3148)
6a1a1b326 Roll external/spirv-headers/ af64a9e82..dc77030ac (4 commits) (#3147)
ab7ac60f1 spirv-fuzz: Refactoring and type-related fixes (#3144)
323a81fc5 Validate Vulkan 1.2 capabilities
d46a1b048 Add basic support for Vulkan 1.2: SPV_ENV_VULKAN_1_2
f8d7df760 Fix OpLine bug of merge-blocks pass (#3130)
8013d477a spirv-fuzz: add dead blocks (#3135)
18b3b9456 Remove names and decorations of imported symbols (#3081)
fb2e81981 Fix GN build for OpenCL.DebugInfo.100 update (#3134)
c86dc0bcd Fix bad parameter names in error message (#3129)
899735f45 Revert PR #3093 (#3131)
a466b99db Disallow forward references in arrays (#3093)
31acc7882 spirv-fuzz: Add fuzzer pass to perform module donation (#3117)
c8bf14393 GetOperandConstants operand can be const (#3126)

git-subtree-dir: third_party/SPIRV-Tools
git-subtree-split: 8910ea5f1c7bc38f79a8b70b265cd9d1571f4b56

Bug: chromium:1039697
Change-Id: I5af47f44696a0eb1da7516ea57f3ae766651cae3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41548
Presubmit-Ready: Nico Weber <thakis@chromium.org>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

A recent change was made that failed when building nacl, because despite
the target_cpu being set to x86, current_cpu is x64 for nacl, as it is
cross-compiled to x64 using a different toolchain. As such, we were
building and linking the x86 version of Subzero with the x64 version of
nacl, resulting in a linker error.

See the following doc for more info:
https://chromium.googlesource.com/dart/dartium/src/+/refs/heads/master/tools/gn/docs/cross_compiles.md#as-a-build_gn-author

Bug: b/150008309
Change-Id: I581538cb7bb090e5b30819052eb7e245bc90ab12
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41528Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

This reverts commit f50cc82b.

Reason for revert: Skia roll broken. See:
https://chromium-review.googlesource.com/c/chromium/src/+/2067664
"lld-link: error: undefined symbol: _chkstk"

Change-Id: I2ac508c9560e3a49e8a856906c3a5f6141c45207
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41489Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

If the stack size is > 4K, emit chkstk, which probes the stack to commit
the pages required to support the large stack.

Bug: swiftshader:25
Change-Id: Iac1850c528460f4244f6f14f851ef47ee5a737d0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41388Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

VK_EXT_shader_stencil_export is a simple extension to implement in
SwiftShader. Added PixelRoutine::stencilReplaceRef(), which can be
trivially modified to output Byte4 instead of Byte8, should the
stencil code be refactored.

Verified that these tests still pass with SwANGLE
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth32f_stencil8_basic
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth32f_stencil8_stencil_only
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth24_stencil8_basic
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth24_stencil8_stencil_only

Change-Id: Ia5829489ccba75ee7c41e365ca2d3d586c987c2d
Tests: dEQP-VK.pipeline.shader_stencil_export.op_replace
Bug: b/148175198
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41469
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

sw::Thread and rr::Thread uses have been replaced by std::thread.

Note the latter does not support controlling the stack size, but it
appears to default to 2 MiB on Linux, 1 MiB on Windows, and 512 KiB on
macOS.

Bug: b/149829034
Change-Id: Ie4ad8af12c111e951761248b69f8206dd143af24
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41368Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

On platforms where the calling thread has a very small stack, LLVM can
overflow the stack. Work around it by creating a temporary thread for
performing LLVM optimizations and codegen.

Bug: b/149829034
Change-Id: I59edd0f7d02b4724ea30f9413ea3efbb4d200290
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41389Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>

This reverts commit f9881ba0.

Reason for revert: Causes failures. See:
https://ci.chromium.org/p/chromium/builders/try/linux_optional_gpu_tests_rel/36449
https://ci.chromium.org/p/chromium/builders/try/mac_optional_gpu_tests_rel/36978
https://ci.chromium.org/p/chromium/builders/try/win_optional_gpu_tests_rel/37221

Change-Id: Ib95459cefea22001fa43e07bc20a15002976a9da
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41468Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

VK_EXT_shader_stencil_export is a simple extension to implement in
SwiftShader. Added PixelRoutine::stencilReplaceRef(), which can be
trivially modified to output Byte4 instead of Byte8, should the
stencil code be refactored.

Verified that these tests still pass with SwANGLE
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth32f_stencil8_basic
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth32f_stencil8_stencil_only
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth24_stencil8_basic
dEQP-GLES3.functional.fbo.blit.depth_stencil.depth24_stencil8_stencil_only

Tests: dEQP-VK.pipeline.shader_stencil_export.op_replace
Bug: b/148175198
Change-Id: Ib98e9917a42aa487d2ee6ffed8b8e317b4fec5d3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41208
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

When available, the decorations for the array member of the structure
used in EmitArrayLength should be used to find the array stride and
use it in the length computation.

This cl should fix these tests with SwANGLE:
dEQP-GLES31.functional.compute.basic.write_multiple_unsized_arr_*

Bug: b/148459131
Change-Id: I5722ebc761b81faf75bb9bab7c60f240afa50446
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41271
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Bug: dawn:283
Change-Id: I8abd042ce1cb1a278d77b6639aeda650630963a8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41328Tested-by: Corentin Wallez <cwallez@google.com>
Kokoro-Presubmit: Corentin Wallez <cwallez@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

When sampling an sRGB texture format, we have a vector of 16-bit
elements where 0xFF00 represents 1.0, but we were using a lookup table
which produced results in a 0xFFFF base.

Also on linear to sRGB conversion, precision was lost due to truncating
to 12-bit. Adding a rounding factor before truncating preserves
precision.

Bug: b/149574741
Change-Id: I19218c958867c4b5e97dc3b62b4ea83f37161134
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41268
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

This works well as our use-case is to lookup the same addresses multiple
times (loops, helper functions, etc.). With a simple timing of
ReactorUnitTests.exe, with this optimization enabled, total time dropped
from 16.72 seconds to 2.24 seconds (~7.5X faster).

This optimization is especially important for Mac where resolving stack
addresses takes much longer than on other platforms. With this
optimization, running a specific gles-unittests test, it was able to
output the callstack trace in about 30 seconds; without, it was still
emitting jit code after 10 minutes at which point I gave up.

Bug: b/131425026
Change-Id: I97e76256220f73fba0a1fc4099532e833f561551
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/32369
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

Adds support for emitting print locations when using the Subzero
backend. Very useful for debugging.

Bug: b/149572931
Change-Id: I32fc14486cd7a021f424fb926dc776072374b931
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41288Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

I find myself copy and pasting the commands out of the previous change, looking in the squash change for the SHAs and descriptions, and amending the top merge change with all this info. It's tedious.

This script does all of this with a one-liner.

Bug: b/140546382
Change-Id: I2ecfd8074b053cbf6de4001a56bca2ffbaccba42
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41230
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Factored out most of the rr::Print code in LLVMReactor.cpp to
Reactor.cpp, and rewritten in terms of Nucleus. Added a couple of new
Nucleus functions to support this.

Bug: b/149477527
Change-Id: I0a28626f1aa6133a37f9e75abc08544f3de15a45
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41188
Presubmit-Ready: Antonio Maiorano <amaiorano@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

When I (hack) fixed the fact that Subzero fails to load from a constant
source because it interprets it as an offset, I did so by routing all
loads into sz::createLoad, and in there, making sure to bitcast constant
sources to variables. I thought this covered all loads, but it turns out
that one of our optimization passes optimizes stores by removing
intermediate stores. This would sometimes replace a variable source with
a constant one, so I applied a similar hack here to avoid doing replace
in this case.

Bug: b/148272103
Change-Id: I809688e2c79fa1b62918d3b484a8e2a9e601bd90
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41269
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Change-Id: I1b54ba532578106fc4906b1d1b2bf284d7b27002
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41231
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolas.capens@gmail.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Bug: b/149328074
Change-Id: Ia5af4622f2a024ea08395f7b7412e086fdb34c39
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/41270Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

From VK_EXT_external_memory_host, there is a type of VkDeviceMemory
that takes a pointer to host mapped foreign memory, where the
underlying memory is owned by the host, not the device.
This is used by ANGLE on Mac when creating a VkDeviceMemory given an
IOSurface.

Tests: dEQP-VK.*
Bug: chromium:1965434
Change-Id: I944a5d3cd23ad8b193f58981829a8847419a4b0d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/39968
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Jonah Ryan-Davis <jonahr@google.com>

Fixes another hairy bug in condition-variable timeouts, adds docs

Changes:
    a047dd0bb Include benchmarks on project's README.md
    4c702da52 Scheduler: Fix lock state on Fiber::wait timeout.
    b4e305525 Docs: Add documentation for marl::Scheduler.
    5f18ac0e0 ConditionVariable: Delete copy and move constructors
    f78eb441f Scheduler: document requirement to unbind() before destruction
    9f9f6d32e Defer benchmark - avoid benchmark::DoNotOptimize()
    3b610e902 Fix compiler warnings with MARL_FIBERS_USE_UCONTEXT
    0dbab1184 Scheduler: Delete copy and move constructors / assignment ops.
    cbef55d58 Kokoro: Build benchmarks
    e923c3d96 Rework the 'hello task' example to be more idiomatic

Commands:
    git subtree pull --prefix third_party/marl https://github.com/google/marl master --squash

Bug: b/140546382
Change-Id: I3412287ec95bdf761d84fd9f0dffae8f08bcec3c

a047dd0bb Include benchmarks on project's README.md
4c702da52 Scheduler: Fix lock state on Fiber::wait timeout.
b4e305525 Docs: Add documentation for marl::Scheduler.
5f18ac0e0 ConditionVariable: Delete copy and move constructors
f78eb441f Scheduler: document requirement to unbind() before destruction
9f9f6d32e Defer benchmark - avoid benchmark::DoNotOptimize()
3b610e902 Fix compiler warnings with MARL_FIBERS_USE_UCONTEXT
0dbab1184 Scheduler: Delete copy and move constructors / assignment ops.
cbef55d58 Kokoro: Build benchmarks
e923c3d96 Rework the 'hello task' example to be more idiomatic

git-subtree-dir: third_party/marl
git-subtree-split: a047dd0bbbd6a65ee4d03d0ceb4fedfa56da02a5