Merge pull request #1143 from KhronosGroup/interlock

Merge SPV_EXT_fragment_shader_interlock with fixes

reference/opt/shaders-hlsl/frag/pixel-interlock-ordered.sm51.fxconly.frag

@@ -0,0 +1,24 @@
+RWByteAddressBuffer _9 : register(u6, space0);
+globallycoherent RasterizerOrderedByteAddressBuffer _42 : register(u3, space0);
+RasterizerOrderedByteAddressBuffer _52 : register(u4, space0);
+RWTexture2D<unorm float4> img4 : register(u5, space0);
+RasterizerOrderedTexture2D<unorm float4> img : register(u0, space0);
+RasterizerOrderedTexture2D<unorm float4> img3 : register(u2, space0);
+RasterizerOrderedTexture2D<uint> img2 : register(u1, space0);
+
+void frag_main()
+{
+    _9.Store(0, uint(0));
+    img4[int2(1, 1)] = float4(1.0f, 0.0f, 0.0f, 1.0f);
+    img[int2(0, 0)] = img3[int2(0, 0)];
+    uint _39;
+    InterlockedAdd(img2[int2(0, 0)], 1u, _39);
+    _42.Store(0, uint(int(_42.Load(0)) + 42));
+    uint _55;
+    _42.InterlockedAnd(4, _52.Load(0), _55);
+}
+
+void main()
+{
+    frag_main();
+}

reference/opt/shaders-msl/frag/pixel-interlock-ordered.msl2.argument.frag

@@ -0,0 +1,43 @@
+#pragma clang diagnostic ignored "-Wunused-variable"
+
+#include <metal_stdlib>
+#include <simd/simd.h>
+#include <metal_atomic>
+
+using namespace metal;
+
+struct Buffer3
+{
+    int baz;
+};
+
+struct Buffer
+{
+    int foo;
+    uint bar;
+};
+
+struct Buffer2
+{
+    uint quux;
+};
+
+struct spvDescriptorSetBuffer0
+{
+    device Buffer3* m_9 [[id(0)]];
+    texture2d<float, access::write> img4 [[id(1)]];
+    texture2d<float, access::write> img [[id(2), raster_order_group(0)]];
+    texture2d<float> img3 [[id(3), raster_order_group(0)]];
+    volatile device Buffer* m_34 [[id(4), raster_order_group(0)]];
+    device Buffer2* m_44 [[id(5), raster_order_group(0)]];
+};
+
+fragment void main0(constant spvDescriptorSetBuffer0& spvDescriptorSet0 [[buffer(0)]])
+{
+    (*spvDescriptorSet0.m_9).baz = 0;
+    spvDescriptorSet0.img4.write(float4(1.0, 0.0, 0.0, 1.0), uint2(int2(1)));
+    spvDescriptorSet0.img.write(spvDescriptorSet0.img3.read(uint2(int2(0))), uint2(int2(0)));
+    (*spvDescriptorSet0.m_34).foo += 42;
+    uint _49 = atomic_fetch_and_explicit((volatile device atomic_uint*)&(*spvDescriptorSet0.m_34).bar, (*spvDescriptorSet0.m_44).quux, memory_order_relaxed);
+}
+

reference/opt/shaders-msl/frag/pixel-interlock-ordered.msl2.frag

@@ -0,0 +1,33 @@
+#pragma clang diagnostic ignored "-Wunused-variable"
+
+#include <metal_stdlib>
+#include <simd/simd.h>
+#include <metal_atomic>
+
+using namespace metal;
+
+struct Buffer3
+{
+    int baz;
+};
+
+struct Buffer
+{
+    int foo;
+    uint bar;
+};
+
+struct Buffer2
+{
+    uint quux;
+};
+
+fragment void main0(device Buffer3& _9 [[buffer(0)]], volatile device Buffer& _34 [[buffer(1), raster_order_group(0)]], device Buffer2& _44 [[buffer(2), raster_order_group(0)]], texture2d<float, access::write> img4 [[texture(0)]], texture2d<float, access::write> img [[texture(1), raster_order_group(0)]], texture2d<float> img3 [[texture(2), raster_order_group(0)]])
+{
+    _9.baz = 0;
+    img4.write(float4(1.0, 0.0, 0.0, 1.0), uint2(int2(1)));
+    img.write(img3.read(uint2(int2(0))), uint2(int2(0)));
+    _34.foo += 42;
+    uint _49 = atomic_fetch_and_explicit((volatile device atomic_uint*)&_34.bar, _44.quux, memory_order_relaxed);
+}
+

reference/opt/shaders/frag/pixel-interlock-ordered.frag

@@ -0,0 +1,23 @@
+#version 450
+#extension GL_ARB_fragment_shader_interlock : require
+layout(pixel_interlock_ordered) in;
+
+layout(binding = 2, std430) coherent buffer Buffer
+{
+    int foo;
+    uint bar;
+} _30;
+
+layout(binding = 0, rgba8) uniform writeonly image2D img;
+layout(binding = 1, r32ui) uniform uimage2D img2;
+
+void main()
+{
+    beginInvocationInterlockARB();
+    imageStore(img, ivec2(0), vec4(1.0, 0.0, 0.0, 1.0));
+    uint _27 = imageAtomicAdd(img2, ivec2(0), 1u);
+    _30.foo += 42;
+    uint _41 = atomicAnd(_30.bar, 255u);
+    endInvocationInterlockARB();
+}
+

reference/opt/shaders/frag/pixel-interlock-unordered.frag

@@ -0,0 +1,23 @@
+#version 450
+#extension GL_ARB_fragment_shader_interlock : require
+layout(pixel_interlock_unordered) in;
+
+layout(binding = 2, std430) coherent buffer Buffer
+{
+    int foo;
+    uint bar;
+} _30;
+
+layout(binding = 0, rgba8) uniform writeonly image2D img;
+layout(binding = 1, r32ui) uniform uimage2D img2;
+
+void main()
+{
+    beginInvocationInterlockARB();
+    imageStore(img, ivec2(0), vec4(1.0, 0.0, 0.0, 1.0));
+    uint _27 = imageAtomicAdd(img2, ivec2(0), 1u);
+    _30.foo += 42;
+    uint _41 = atomicAnd(_30.bar, 255u);
+    endInvocationInterlockARB();
+}
+

reference/opt/shaders/frag/sample-interlock-ordered.frag

@@ -0,0 +1,23 @@
+#version 450
+#extension GL_ARB_fragment_shader_interlock : require
+layout(sample_interlock_ordered) in;
+
+layout(binding = 2, std430) coherent buffer Buffer
+{
+    int foo;
+    uint bar;
+} _30;
+
+layout(binding = 0, rgba8) uniform writeonly image2D img;
+layout(binding = 1, r32ui) uniform uimage2D img2;
+
+void main()
+{
+    beginInvocationInterlockARB();
+    imageStore(img, ivec2(0), vec4(1.0, 0.0, 0.0, 1.0));
+    uint _27 = imageAtomicAdd(img2, ivec2(0), 1u);
+    _30.foo += 42;
+    uint _47 = atomicAnd(_30.bar, uint(gl_SampleMaskIn[0]));
+    endInvocationInterlockARB();
+}
+

reference/opt/shaders/frag/sample-interlock-unordered.frag

@@ -0,0 +1,23 @@
+#version 450
+#extension GL_ARB_fragment_shader_interlock : require
+layout(sample_interlock_unordered) in;
+
+layout(binding = 2, std430) coherent buffer Buffer
+{
+    int foo;
+    uint bar;
+} _30;
+
+layout(binding = 0, rgba8) uniform writeonly image2D img;
+layout(binding = 1, r32ui) uniform uimage2D img2;
+
+void main()
+{
+    beginInvocationInterlockARB();
+    imageStore(img, ivec2(0), vec4(1.0, 0.0, 0.0, 1.0));
+    uint _27 = imageAtomicAdd(img2, ivec2(0), 1u);
+    _30.foo += 42;
+    uint _41 = atomicAnd(_30.bar, 255u);
+    endInvocationInterlockARB();
+}
+

reference/shaders-hlsl-no-opt/asm/frag/pixel-interlock-callstack.sm51.fxconly.asm.frag

@@ -0,0 +1,32 @@
+RasterizerOrderedByteAddressBuffer _7 : register(u1, space0);
+RWByteAddressBuffer _9 : register(u0, space0);
+
+static float4 gl_FragCoord;
+struct SPIRV_Cross_Input
+{
+    float4 gl_FragCoord : SV_Position;
+};
+
+void callee2()
+{
+    int _31 = int(gl_FragCoord.x);
+    _7.Store(_31 * 4 + 0, _7.Load(_31 * 4 + 0) + 1u);
+}
+
+void callee()
+{
+    int _39 = int(gl_FragCoord.x);
+    _9.Store(_39 * 4 + 0, _9.Load(_39 * 4 + 0) + 1u);
+    callee2();
+}
+
+void frag_main()
+{
+    callee();
+}
+
+void main(SPIRV_Cross_Input stage_input)
+{
+    gl_FragCoord = stage_input.gl_FragCoord;
+    frag_main();
+}

reference/shaders-hlsl-no-opt/asm/frag/pixel-interlock-control-flow.sm51.fxconly.asm.frag

@@ -0,0 +1,42 @@
+RasterizerOrderedByteAddressBuffer _7 : register(u1, space0);
+RWByteAddressBuffer _13 : register(u2, space0);
+RasterizerOrderedByteAddressBuffer _9 : register(u0, space0);
+
+static float4 gl_FragCoord;
+struct SPIRV_Cross_Input
+{
+    float4 gl_FragCoord : SV_Position;
+};
+
+void callee2()
+{
+    int _44 = int(gl_FragCoord.x);
+    _7.Store(_44 * 4 + 0, _7.Load(_44 * 4 + 0) + 1u);
+}
+
+void callee()
+{
+    int _52 = int(gl_FragCoord.x);
+    _9.Store(_52 * 4 + 0, _9.Load(_52 * 4 + 0) + 1u);
+    callee2();
+    if (true)
+    {
+    }
+}
+
+void _35()
+{
+    _13.Store(int(gl_FragCoord.x) * 4 + 0, 4u);
+}
+
+void frag_main()
+{
+    callee();
+    _35();
+}
+
+void main(SPIRV_Cross_Input stage_input)
+{
+    gl_FragCoord = stage_input.gl_FragCoord;
+    frag_main();
+}

reference/shaders-hlsl-no-opt/asm/frag/pixel-interlock-split-functions.sm51.fxconly.asm.frag

@@ -0,0 +1,42 @@
+RasterizerOrderedByteAddressBuffer _7 : register(u1, space0);
+RasterizerOrderedByteAddressBuffer _9 : register(u0, space0);
+
+static float4 gl_FragCoord;
+struct SPIRV_Cross_Input
+{
+    float4 gl_FragCoord : SV_Position;
+};
+
+void callee2()
+{
+    int _37 = int(gl_FragCoord.x);
+    _7.Store(_37 * 4 + 0, _7.Load(_37 * 4 + 0) + 1u);
+}
+
+void callee()
+{
+    int _45 = int(gl_FragCoord.x);
+    _9.Store(_45 * 4 + 0, _9.Load(_45 * 4 + 0) + 1u);
+    callee2();
+}
+
+void _29()
+{
+}
+
+void _31()
+{
+}
+
+void frag_main()
+{
+    callee();
+    _29();
+    _31();
+}
+
+void main(SPIRV_Cross_Input stage_input)
+{
+    gl_FragCoord = stage_input.gl_FragCoord;
+    frag_main();
+}

reference/shaders-hlsl-no-opt/frag/pixel-interlock-simple-callstack.sm51.fxconly.frag

@@ -0,0 +1,32 @@
+RasterizerOrderedByteAddressBuffer _14 : register(u1, space0);
+RasterizerOrderedByteAddressBuffer _35 : register(u0, space0);
+
+static float4 gl_FragCoord;
+struct SPIRV_Cross_Input
+{
+    float4 gl_FragCoord : SV_Position;
+};
+
+void callee2()
+{
+    int _25 = int(gl_FragCoord.x);
+    _14.Store(_25 * 4 + 0, _14.Load(_25 * 4 + 0) + 1u);
+}
+
+void callee()
+{
+    int _38 = int(gl_FragCoord.x);
+    _35.Store(_38 * 4 + 0, _35.Load(_38 * 4 + 0) + 1u);
+    callee2();
+}
+
+void frag_main()
+{
+    callee();
+}
+
+void main(SPIRV_Cross_Input stage_input)
+{
+    gl_FragCoord = stage_input.gl_FragCoord;
+    frag_main();
+}

reference/shaders-hlsl/frag/pixel-interlock-ordered.sm51.fxconly.frag

@@ -0,0 +1,24 @@
+RWByteAddressBuffer _9 : register(u6, space0);
+globallycoherent RasterizerOrderedByteAddressBuffer _42 : register(u3, space0);
+RasterizerOrderedByteAddressBuffer _52 : register(u4, space0);
+RWTexture2D<unorm float4> img4 : register(u5, space0);
+RasterizerOrderedTexture2D<unorm float4> img : register(u0, space0);
+RasterizerOrderedTexture2D<unorm float4> img3 : register(u2, space0);
+RasterizerOrderedTexture2D<uint> img2 : register(u1, space0);
+
+void frag_main()
+{
+    _9.Store(0, uint(0));
+    img4[int2(1, 1)] = float4(1.0f, 0.0f, 0.0f, 1.0f);
+    img[int2(0, 0)] = img3[int2(0, 0)];
+    uint _39;
+    InterlockedAdd(img2[int2(0, 0)], 1u, _39);
+    _42.Store(0, uint(int(_42.Load(0)) + 42));
+    uint _55;
+    _42.InterlockedAnd(4, _52.Load(0), _55);
+}
+
+void main()
+{
+    frag_main();
+}