Our lookup key for generated sampling functions was only using the
sampler and image view descriptors into account. The same descriptors
can be used in various different sampling instructions, so we need to
also make that part of the lookup key.

Bug: b/129523279
Change-Id: I325aa5571191e4ffc69214c2d13d431d21316fed
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31949
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Z was only being interpolated when the Position builtin is used (which
isn't valid for fragment shaders) and/or when there's a depth buffer
present. Instead interpolate it when FragCoord is being used.

Bug: b/133325267
Tests: dEQP-VK.draw.inverted_depth_ranges.*
Change-Id: I0567084b6f7bb054bd864545fa34acc52c8ac3d6
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31848
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

Change-Id: Ia954596bcd7e6c206240fbf6cef7dd30731c3a99
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31847
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Texture dimensions larger than SHRT_MAX (32767)
were not supported by SwiftShader. Used integers
instead of shorts to store texture sizes to
increase the maximum texture size to INT_MAX
(2147483647).

Tests: dEQP-VK.memory.pipeline_barrier.host_write_uniform_texel_buffer.*
Tests: dEQP-VK.memory.pipeline_barrier.transfer_dst_uniform_texel_buffer.*

Bug b/133429305

Change-Id: I7913f5f4c1f933889e74c57851a837ad7982f87a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31989Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

vk::destroy was crashing for vk::Fence after:
https://swiftshader-review.googlesource.com/c/SwiftShader/+/31988

Reverted that change and fixed it in VkObject.cpp to make sure the
result of the cast is always equal to the allocated memory.

Bug: b/117835459
Change-Id: I831625418fc83503329e833d48d9765b75d83b80
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31990Tested-by: Alexis Hétu <sugoi@google.com>
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Fence has a second base class with virtual functions, but its first base does not.
In "static VkResult Create(...), when assigning the handle via:

*outObject = *object;

since this is code run in the context of the first base class, the address returned
is offset by sizeof the vtable pointer from the second base class (8 bytes on x64).

Problem is that the Cast overload for Fence does a simple reinterpret_cast of this
stored pointer back to Fence*, without the offset being taken into account. The
fix is to simply first cast the stored pointer back to the first base class type, and
then static_cast it to the derived Fence type.

Bug: b/117835459
Change-Id: Ieb2856320e996a5ad5d9e69aad5e01cfaee1cd9e
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31988Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

As requested here:
https://swiftshader-review.googlesource.com/c/SwiftShader/+/31841/1#message-24da63454b89e5d545f3b686f2470c8d2392ee61

Change-Id: I76500807933b565fb13f30268c47c7caa4be7c69
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31843Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

We were assuming the condition was the same width as the lhs and rhs types - it can also be a scalar even if the result is not.

Bug: b/133440380
Change-Id: Ib0e971af426b71030a097838295b07bb6e4261e0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31841Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

The BUILD.gn files used to ignore the platform entirely. All that was
necessary was to point them at the proper include directory for Mac.

Bug chromium:966543

Change-Id: I755a432cc7d01ce094d7191fd48d0702d8a1ae3e
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31889
Presubmit-Ready: Sean Risser <srisser@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Sean Risser <srisser@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I8017db3f860c111143835e8ccb73c712489ede0c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31842
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

"Buffer updates performed with vkCmdUpdateBuffer first copy the data into command buffer memory when the command is recorded (which requires additional storage and may incur an additional allocation), and then copy the data from the command buffer into dstBuffer when the command is executed on a device."

Bug: b/133127573
Bug: b/118619338
Bug: b/118383648
Change-Id: I8ba9bcdd6bf9e5b4f7e181c6fc379595a1d32e2a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31839
Presubmit-Ready: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

I had made a bad assumption earlier that these would choose the correct operation based on signedness. That's not the case -- LLVM has separate Max/UMax and Min/UMin RMW operations.

Test: dEQP-VK.glsl.atomic_operations.*
Change-Id: I92a9e56011c4ef8ac658a826c480b16f65d0704d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31909
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>

Test: dEQP-VK.glsl.440.linkage.varying.component.vert_in.ivec2.as_int_int_unused
Change-Id: I6e8da5863cfcee722dd4cf202810ddc505f76a65
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31890Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Tests: dEQP-VK.draw.basic_draw.*
Change-Id: I609e22601c4aae1bdebaca6d42d33754cfaa57bb
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31840
Presubmit-Ready: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Fixes a whole lot of leak-detector warnings.

Bug: b/133399620
Change-Id: I4f35150712dd2035a84c16d06ca75a4eba2c3c69
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31836Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

SpirvRoutine holds reactor fields which must not outlive the scope of the reactor mutex.

While the routine was not directly referenced outside of ComputeProgram::generate(), the destructor of the SpirvRoutine's rr::Variables was touching reactor, leading to data races.

Bug: b/133127573
Change-Id: I854e079efdc0ef2d5b1ae67f014fe7148ad8f892
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31838
Presubmit-Ready: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

We were only passing them to the compiler. These should go to the linker as well.

Bug: b/133127573
Change-Id: Ifc02a2c17a164f30c80311fd17d2a6311e3d5cba
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31817Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Hide all the error-prone synchronization logic in the class, and expose a documented API.

This should have no impact on behavior - it was authored to make it harder to break things in the future.
That said, it appears to fix a whole bunch of flakes with the dEQP query tests.

Bug: b/133127573
Change-Id: I5c30b79b9b1cd36dba1fa2d3c34af0f5bd62772a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31816
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Change-Id: I767fe366bb3484b2f21db61c9fd964172c5b0537
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31908Reviewed-by: Saman Sami <samans@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Presubmit-Ready: Saman Sami <samans@chromium.org>
Tested-by: Saman Sami <samans@chromium.org>

Resolve must occur in linear space.

Bug b/119620965

Change-Id: I5c7c7c45a1448f44c702969e34e580dc299df70a
Test: dEQP-VK.renderpass.suballocation.multisample_resolve.*
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31871Tested-by: Alexis Hétu <sugoi@google.com>
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

Bug: b/133127573
Change-Id: I0915cd5f90b75d2bb97152a6b1083755ad14c8b3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31815Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

We call the constructor, so make sure we also call the destructor.
Prevents non-POD inner fields from being deallocated without a destructor call before being freed.

Change-Id: Ibf1291619f0ab4c7f7761a80a045d605fa9cbd53
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31813Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug: b/133127573
Change-Id: I834939c708dcb1a075b397a919fa63c5aec652f5
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31593Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Replace sw::MutexLock with std::mutex.
Replace use of sw::Thread with std::thread.

helgrind seems to take objection to the init Event.

AtomicInt remains as a wrapper around std::atomic<int> as it provides an 'operator=' and
'std::atomic& operator=( const std::atomic& )' is deleted.

This change is to reduce the amount of code I have to verify while looking for synchronization issues.

Bug: b/133127573
Change-Id: I8cbe38c918298ffafaf59fe25b0ff258670f9a99
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31812Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

If we don't do this then calls to StorePhi() will assume the instruction belongs to the wrong block.

Bug: b/133212545
Change-Id: If8436dcc677847a86fd629d6354e437ef46ba048
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31833
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

In this particular case, I do not see the issue described in b/133135427 to ever happen here - however helgrind seems to find this questionable, and unless this is causing a noticable performance impact I think we should err on the side of caution.

Bug: b/133135427
Change-Id: I1e5fa02ed28b6146f8a6044999f92b141e001e13
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31810Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug: b/133135427
Change-Id: Iadb691c7b0b957a494b57a8a5046a58a6b42180c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31809Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

As we destruct the Queue now, there's no need to have a shared_ptr indirection.

This makes Renderer part of the vk::Queue allocation.

Bug: b/133127573
Change-Id: I9ea0d427752689e73346e0da950144832d908a2e
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31685Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

We were never destructing VkQueues, but were freeing the underlying memory.

Bug: b/133127573
Change-Id: Id6d9f615428073adad7267dd1bd48ced17f1b433
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31592Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Don't (ab)use vk::Fence for its start() finish() events.
sw::WaitGroup shares the same interface, so use that.

Allows vk::Fence to have its destructor deleted again.

Bug: b/133127573
Change-Id: I747ea6965bf85343c720235083b309d8b855a950
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31684Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Bug: b/133127573
Change-Id: I0ce0164de401524014eb295547b899f24f437e21
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31683Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This is implemented by sw::WaitGroup, and vk::Fence.

The goal here is to abstract the underlying implementation of the task tracker. We currently (ab)use vk::Fence for queue tracking when a sw::WaitGroup could be used.

This also allows us to implement a task logger in the future that can show what's inflight at any time.

Bug: b/133127573
Change-Id: I0177b26e7577ffd5cb687a5105a69a22070a7ac6
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31682Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This change fixes the ASAN issue as described in b/133135427.

Reproduction case:
./build/vk-unittests --gtest_repeat=-1 --gtest_filter=ComputeParams/SwiftShaderVulkanBufferToBufferComputeTest.Memcpy/0

Bug: b/133127573
Bug: b/133135427
Bug: swiftshader:130
Change-Id: I06fbf10ab042160e8ca481f2afaa30d4f676dc75
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31681
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

A synchronization primitive borrowed from golang:
https://golang.org/pkg/sync/#WaitGroup

Bug: b/133127573
Change-Id: Id3ec093c0de478cfe16fad4d064cf2c26288acf7
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31680Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Simplify to a single implementation across all platforms.
Document the class and members.
Also add:
• Constructor with event type and initial state
• clear() method.
• wait() method with timeout.
• operator bool().

Bug: b/133127573
Change-Id: I612dc8a1011aaa2cfa50f40c59849323348b3054
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31679
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Change-Id: I372351c769f8da843acc097927ed2abf872259b0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31835
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Error happened while moving code from android tree to
swiftshader tree. Whoops!

Change-Id: Ie7085f8e6d54b2b11547bb6a38348c20140beae9
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31868Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Hernan Liatis <hliatis@google.com>

Using 'volatile' does not make anything thread safe.

Shuts up a bunch of TSAN errors.

Bug: b/133127573
Change-Id: I90b708ac53824de91c316f61f137ecca34087047
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31830
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

Don't bake the string into the printf format, use a '%s' and pass the string as a parameter.

Change-Id: I47be05383637caaa81a93d19623ff39c47988df3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31429Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Tests: dEQP-VK.draw.basic_draw.draw_indirect.*
Tests: dEQP-VK.draw.basic_draw.draw_indexed_indirect.*
Bug: b/133127573
Change-Id: Ie64ae58d352d0fab23a0ce8811ce95c024e59367
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31687Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>