Improved SCCP, inlining and resource specialization passes, legalize `ImageSubscript` for GLSL (#2146)

10 files changed, 149 insertions, 24 deletions

diff --git a/tests/bugs/sccp-switch-case-removal.slang b/tests/bugs/sccp-switch-case-removal.slang
new file mode 100644
index 000000000..cf8d36ef7
--- /dev/null
+++ b/tests/bugs/sccp-switch-case-removal.slang
@@ -0,0 +1,25 @@
+//TEST(compute,vulkan):COMPARE_COMPUTE_EX:-vk -slang -compute -shaderobj
+
+//TEST_INPUT:ubuffer(data=[9 9 9 9], stride=4):out,name=outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+static const uint kConstant = 5;
+
+[numthreads(4, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    float x = 1.0;
+    switch(kConstant)
+    {
+    case 5:
+        int tid = int(dispatchThreadID.x);
+	    outputBuffer[tid] = 0;
+        break;
+    case 1:
+        ddx(x); // this should trigger glslang error if it doesn't get eliminated by sccp.
+        break;
+    default:
+        ddy(x); // this should trigger glslang error if it doesn't get eliminated by sccp.
+        break;
+    }
+}
\ No newline at end of file
diff --git a/tests/bugs/sccp-switch-case-removal.slang.expected.txt b/tests/bugs/sccp-switch-case-removal.slang.expected.txt
new file mode 100644
index 000000000..ae25f7400
--- /dev/null
+++ b/tests/bugs/sccp-switch-case-removal.slang.expected.txt
@@ -0,0 +1,4 @@
+0
+0
+0
+0
\ No newline at end of file
diff --git a/tests/bugs/vk-image-write.slang b/tests/bugs/vk-image-write.slang
new file mode 100644
index 000000000..10141384c
--- /dev/null
+++ b/tests/bugs/vk-image-write.slang
@@ -0,0 +1,16 @@
+//TEST:CROSS_COMPILE: -profile ps_5_0 -entry main -target spirv-assembly
+
+// Ensure that we can lower to `imageStore` correctly.
+
+RWTexture2D<float4> t;
+
+void writeColor(float3 v)
+{
+    t[uint2(0,0)].xyz += v;
+}
+
+float4 main() : SV_Target
+{
+	writeColor(float3(1.0));
+	return float4(0);
+}
diff --git a/tests/bugs/vk-image-write.slang.glsl b/tests/bugs/vk-image-write.slang.glsl
new file mode 100644
index 000000000..7b9bcebf9
--- /dev/null
+++ b/tests/bugs/vk-image-write.slang.glsl
@@ -0,0 +1,40 @@
+//TEST_IGNORE_FILE:
+
+#version 450
+layout(row_major) uniform;
+layout(row_major) buffer;
+
+layout(rgba32f)
+layout(binding = 0)
+uniform image2D t_0;
+
+void writeColor_0(vec3 v_0)
+{
+    const uvec2 _S1 = uvec2(0U, 0U);
+
+    vec4 _S2 = (imageLoad((t_0), ivec2((_S1))));
+
+    vec3 _S3 = _S2.xyz + v_0;
+
+    ivec2 _S4 = ivec2(_S1);
+
+    vec4 _S5 = imageLoad(t_0,_S4);
+
+    vec4 _S6 = _S5;
+    _S6.xyz = _S3;
+
+    imageStore(t_0,_S4,_S6);
+    return;
+}
+
+
+layout(location = 0)
+out vec4 _S7;
+
+void main()
+{
+    writeColor_0(vec3(1.00000000000000000000));
+    _S7 = vec4(0);
+    return;
+}
+
diff --git a/tests/bugs/vk-structured-buffer-load.hlsl.glsl b/tests/bugs/vk-structured-buffer-load.hlsl.glsl
index 1c7ec8043..05c8de193 100644
--- a/tests/bugs/vk-structured-buffer-load.hlsl.glsl
+++ b/tests/bugs/vk-structured-buffer-load.hlsl.glsl
@@ -43,13 +43,12 @@ void main()
     float HitT_0 = (gl_RayTmaxNV);
     RayData.PackedHitInfoA_0.x = HitT_0;
 
-    const uint use_rcp_0 = uint(0);
 
-    float offsfloat_0 = ((gParamBlock_sbuf_0)._data[(int(uint(0)))]);
+    float offsfloat_0 = ((gParamBlock_sbuf_0)._data[(0)]);
 
-    uint use_rcp_1 = use_rcp_0|uint(HitT_0 > 0.00000000000000000000);
+    uint use_rcp_0 = 0U | uint(HitT_0 > 0.00000000000000000000);
 
-    if(bool(use_rcp_1))
+    if(bool(use_rcp_0))
     {
 
         float tmpA = rcp_0(offsfloat_0);
@@ -60,7 +59,7 @@ void main()
     else
     {
 
-        if(use_rcp_1 > uint(0)&&offsfloat_0 == 0.00000000000000000000)
+        if(use_rcp_0 > 0U&&offsfloat_0 == 0.00000000000000000000)
         {
 
             float tmpB = (inversesqrt((offsfloat_0 + 1.00000000000000000000)));
diff --git a/tests/optimization/func-resource-result/func-resource-result-complex.slang b/tests/optimization/func-resource-result/func-resource-result-complex.slang
new file mode 100644
index 000000000..a5585ff4c
--- /dev/null
+++ b/tests/optimization/func-resource-result/func-resource-result-complex.slang
@@ -0,0 +1,45 @@
+// func-resource-result-simple.slang
+
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj
+
+// Test that a function that returns a resource type can be
+// compiled for targets that don't natively support resource
+// return values.
+
+//TEST_INPUT:set textures=[Texture2D(size=4, content = zero), Texture2D(size=4, content = one)]
+Texture2D textures[2];
+
+//TEST_INPUT:set sampler=Sampler
+SamplerState sampler;
+
+Texture2D getTex(int index)
+{
+    Texture2D result;
+    // Note: `index` here will need to be a compile time constant in order to generate
+    // valid GLSL. If constant folding and function inlining are all done correctly
+    // we should be able to compile this.
+    if (index == 0)
+        result = textures[1];
+    else
+        result = textures[0];
+    return result;
+}
+
+int test(int val)
+{
+    // Make sure index is a compile-time constant.
+	return getTex(int(0.0) + 1*2 < 5 ? 0 : 1).SampleLevel(sampler, float2(0,0), 0).x == 0.0 ? 0 : 1;
+}
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+[numthreads(4, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    uint tid = dispatchThreadID.x;
+    int inVal = tid;
+    int outVal = test(inVal);
+    outputBuffer[tid] = outVal;
+}
diff --git a/tests/optimization/func-resource-result/func-resource-result-complex.slang.expected.txt b/tests/optimization/func-resource-result/func-resource-result-complex.slang.expected.txt
new file mode 100644
index 000000000..ef529012e
--- /dev/null
+++ b/tests/optimization/func-resource-result/func-resource-result-complex.slang.expected.txt
@@ -0,0 +1,4 @@
+1
+1
+1
+1
\ No newline at end of file
diff --git a/tests/pipeline/rasterization/conservative-rasterization/inner-coverage.slang.glsl b/tests/pipeline/rasterization/conservative-rasterization/inner-coverage.slang.glsl
index 1454d493f..59fecd544 100644
--- a/tests/pipeline/rasterization/conservative-rasterization/inner-coverage.slang.glsl
+++ b/tests/pipeline/rasterization/conservative-rasterization/inner-coverage.slang.glsl
@@ -1,14 +1,14 @@
+//TEST_IGNORE_FILE:
 #version 450
-
 #extension GL_NV_conservative_raster_underestimation : require
+layout(row_major) uniform;
+layout(row_major) buffer;
 
 layout(location = 0)
 out vec4 _S1;
 
 void main()
 {
-    vec4 _S2;
-    _S2 = vec4(uint(gl_FragFullyCoveredNV));
-    _S1 = _S2;
+    _S1 = vec4(uint(gl_FragFullyCoveredNV));
     return;
 }
diff --git a/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl b/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
index a12b9827b..1818b7789 100644
--- a/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
+++ b/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
@@ -2,6 +2,8 @@
 
 #version 450
 #extension GL_ARB_fragment_shader_interlock : require
+layout(row_major) uniform;
+layout(row_major) buffer;
 
 layout(rgba32f)
 layout(binding = 0)
@@ -15,13 +17,9 @@ out vec4 _S2;
 
 void main()
 {
-    vec4 _S3;
-
     beginInvocationInterlockARB();
 
-    vec4 _S4 = (imageLoad((entryPointParams_texture_0), ivec2((uvec2(_S1.xy)))));
-
-    _S3 = _S4;
+    vec4 _S3 = (imageLoad((entryPointParams_texture_0), ivec2((uvec2(_S1.xy)))));
     imageStore((entryPointParams_texture_0), ivec2((uvec2(_S1.xy))), _S3 + _S1);
 
     endInvocationInterlockARB();
diff --git a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
index c07e9b61c..1da5f4f8a 100644
--- a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
+++ b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
@@ -2,24 +2,18 @@
 //TEST_IGNORE_FILE:
 
 #version 450
-
 #extension GL_NV_fragment_shader_barycentric : require
+layout(row_major) uniform;
+layout(row_major) buffer;
 
 pervertexNV layout(location = 0)
-in vec4 _S1[3];
+in vec4  _S1[3];
 
 layout(location = 0)
 out vec4 _S2;
 
 void main()
 {
-    vec4 _S3;
-
-    _S3 = gl_BaryCoordNV.x * _S1[0]
-        + gl_BaryCoordNV.y * _S1[1]
-        + gl_BaryCoordNV.z * _S1[2];
-
-    _S2 = _S3;
-
+    _S2 = gl_BaryCoordNV.x * ((_S1)[(0U)]) + gl_BaryCoordNV.y * ((_S1)[(1U)]) + gl_BaryCoordNV.z * ((_S1)[(2U)]);
     return;
 }