Overhaul global inst deduplication and cpp/cuda backend. (#2654)

* Overhaul global inst deduplication and cpp/cuda backend.

* Update IR documentation.

---------

Co-authored-by: Yong He <yhe@nvidia.com>

6 files changed, 64 insertions, 57 deletions

diff --git a/tests/hlsl-intrinsic/f16tof32.slang b/tests/hlsl-intrinsic/f16tof32.slang
index 78c5fdae6..d45eab00b 100644
--- a/tests/hlsl-intrinsic/f16tof32.slang
+++ b/tests/hlsl-intrinsic/f16tof32.slang
@@ -2,7 +2,7 @@
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12
 //TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
-//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -render-features half
+//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -render-features half
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0  0 0 0 0], stride=4):out,name outputBuffer
 RWStructuredBuffer<float> outputBuffer;
diff --git a/tests/hlsl-intrinsic/f32tof16.slang b/tests/hlsl-intrinsic/f32tof16.slang
index ebcb6b40a..ad8e8e5df 100644
--- a/tests/hlsl-intrinsic/f32tof16.slang
+++ b/tests/hlsl-intrinsic/f32tof16.slang
@@ -2,7 +2,7 @@
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12
 //TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
-//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -render-features half 
+//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -render-features half 
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0], stride=4):out,name outputBuffer
 RWStructuredBuffer<uint> outputBuffer;
diff --git a/tests/hlsl-intrinsic/matrix-double.slang b/tests/hlsl-intrinsic/matrix-double.slang
index b9e36bef8..08bd78cee 100644
--- a/tests/hlsl-intrinsic/matrix-double.slang
+++ b/tests/hlsl-intrinsic/matrix-double.slang
@@ -35,8 +35,7 @@ Float calcTotal(FloatMatrix v)
 
 FloatMatrix makeFloatMatrix(Float f)
 {
-    FloatMatrix m = { { f, f }, { f, f } };
-    return m;
+    return FloatMatrix(f);
 }
 
 IntMatrix makeIntMatrix(int v)
@@ -45,68 +44,58 @@ IntMatrix makeIntMatrix(int v)
     return m;
 }
 
-[numthreads(4, 1, 1)]
-void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+void test1(inout FloatMatrix ft, inout FloatMatrix f, int idx)
 {
-    int idx = int(dispatchThreadID.x);
-
-    Float scalarF = idx * (1.0f / (4.0f));
-
-    FloatMatrix ft = {}; 
-
-    FloatMatrix f = { { scalarF + 0.01, scalarF + 0.02}, { scalarF + 0.011, scalarF + 0.022}};
-    
     // fmod
     ft += FloatMatrix(IntMatrix(((f % makeFloatMatrix(0.11f)) * makeFloatMatrix(100)) + makeFloatMatrix(0.5)));
-    
+
     ft += sin(f);
-      
+
     // Lets try some matrix/matrix
     ft = f * ft;
-    
+
     // Lets try some vector matrix
-    
+
     {
-        FloatMatrix r = {mul(f[0], ft), mul(ft, f[1])};
+        FloatMatrix r = { mul(f[0], ft), mul(ft, f[1]) };
         ft += r;
     }
-    
+
     // Back to the transcendentals
-   
+
     ft += cos(f);
     ft += tan(f);
-    
+
     ft += asin(f);
     ft += acos(f);
     ft += atan(f);
-    
-    ft += atan2(f, makeFloatMatrix(2)); 
 
+    ft += atan2(f, makeFloatMatrix(2));
     {
         FloatMatrix sf, cf;
         sincos(f, sf, cf);
-        
+
         ft += sf;
         ft += cf;
     }
-    
+
     ft += rcp(makeFloatMatrix(1.0) + f);
     ft += FloatMatrix(sign(f - makeFloatMatrix(0.5)));
-   
+
     ft += saturate(f * makeFloatMatrix(4) - makeFloatMatrix(2.0));
-    
+
     ft += sqrt(f);
     ft += rsqrt(makeFloatMatrix(1.0f) + f);
-    
+
     ft += exp2(f);
     ft += exp(f);
-                
+
     ft += frac(f * makeFloatMatrix(3));
     ft += ceil(f * makeFloatMatrix(5) - makeFloatMatrix(3));
-    
+
     ft += floor(f * makeFloatMatrix(10) - makeFloatMatrix(7));
     ft += trunc(f * makeFloatMatrix(7));
-     
+
     ft += log(f + makeFloatMatrix(10.0));
     ft += log2(f * makeFloatMatrix(3) + makeFloatMatrix(2));
 
@@ -114,12 +103,15 @@ void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
         float scalarVs[] = { 1, 10, 100, 1000 };
         ft += FloatMatrix(IntMatrix(log10(makeFloatMatrix(scalarVs[idx])) + makeFloatMatrix(0.5f)));
     }
-       
+
     ft += abs(f * makeFloatMatrix(4) - makeFloatMatrix(2.0f));
-    
+
     ft += min(makeFloatMatrix(0.5), f);
     ft += max(f, makeFloatMatrix(0.75));
+}
 
+void test2(inout FloatMatrix ft, inout FloatMatrix f)
+{
     ft += pow(makeFloatMatrix(0.5), f);
 
     ft += smoothstep(makeFloatMatrix(0.2), makeFloatMatrix(0.7), f);
@@ -135,7 +127,22 @@ void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
     
     UIntMatrix vu = asuint(f);
     ft += asfloat(vu);
-#endif   
-   
+#endif
+}
+
+[numthreads(4, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    int idx = int(dispatchThreadID.x);
+
+    Float scalarF = idx * (1.0f / (4.0f));
+
+    FloatMatrix ft = {}; 
+
+    FloatMatrix f = { { scalarF + 0.01, scalarF + 0.02}, { scalarF + 0.011, scalarF + 0.022}};
+
+    test1(ft, f, idx);
+    test2(ft, f);
+    
     outputBuffer[idx] = calcTotal(ft);
 }
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-reorder-thread.slang.1.expected b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-reorder-thread.slang.1.expected
index 58f06cfec..6e3e5d5d8 100644
--- a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-reorder-thread.slang.1.expected
+++ b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-reorder-thread.slang.1.expected
@@ -17,13 +17,17 @@ layout(location = 0)
 rayPayloadEXT
 SomeValues_0 p_0;
 
-layout(binding = 0)
-uniform accelerationStructureEXT scene_0;
-
 layout(location = 0)
 hitObjectAttributeNV
 SomeValues_0 t_0;
 
+layout(location = 1)
+rayPayloadEXT
+SomeValues_0 p_1;
+
+layout(binding = 0)
+uniform accelerationStructureEXT scene_0;
+
 SomeValues_0 HitObject_GetAttributes_0(hitObjectNV this_0)
 {
     hitObjectGetAttributesNV((this_0), ((0)));
@@ -50,15 +54,11 @@ uint calcValue_0(hitObjectNV hit_0)
     return r_0;
 }
 
-layout(location = 1)
-rayPayloadEXT
-SomeValues_0 p_1;
-
 void HitObject_Invoke_0(accelerationStructureEXT AccelerationStructure_0, hitObjectNV HitOrMiss_0, inout SomeValues_0 Payload_0)
 {
-    p_1 = Payload_0;
-    hitObjectExecuteShaderNV(HitOrMiss_0, (1));
-    Payload_0 = p_1;
+    p_0 = Payload_0;
+    hitObjectExecuteShaderNV(HitOrMiss_0, (0));
+    Payload_0 = p_0;
     return;
 }
 
@@ -87,10 +87,10 @@ void main()
     ray_0.Direction_0 = vec3(0.0, 1.0, 0.0);
     ray_0.TMax_0 = 10000.0;
     RayDesc_0 _S7 = ray_0;
-    p_0.a_0 = idx_0;
-    p_0.b_0 = _S6;
+    p_1.a_0 = idx_0;
+    p_1.b_0 = _S6;
     hitObjectNV hitObj_0;
-    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S7.Origin_0, _S7.TMin_0, _S7.Direction_0, _S7.TMax_0, (0));
+    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S7.Origin_0, _S7.TMin_0, _S7.Direction_0, _S7.TMax_0, (1));
     uint r_1 = calcValue_0(hitObj_0);
     reorderThreadNV(hitObj_0);
     float _S8 = _S5 * 4.0;
diff --git a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-motion-ray.slang.1.expected b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-motion-ray.slang.1.expected
index e6d31a3c8..12025efe0 100644
--- a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-motion-ray.slang.1.expected
+++ b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-motion-ray.slang.1.expected
@@ -15,16 +15,16 @@ struct SomeValues_0
 };
 
 layout(location = 0)
+hitObjectAttributeNV
+SomeValues_0 t_0;
+
+layout(location = 0)
 rayPayloadEXT
 SomeValues_0 p_0;
 
 layout(binding = 0)
 uniform accelerationStructureEXT scene_0;
 
-layout(location = 0)
-hitObjectAttributeNV
-SomeValues_0 t_0;
-
 SomeValues_0 HitObject_GetAttributes_0(hitObjectNV this_0)
 {
     hitObjectGetAttributesNV((this_0), ((0)));
diff --git a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-ray.slang.1.expected b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-ray.slang.1.expected
index 9b5a4d193..ac4d78234 100644
--- a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-ray.slang.1.expected
+++ b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-trace-ray.slang.1.expected
@@ -14,16 +14,16 @@ struct SomeValues_0
 };
 
 layout(location = 0)
+hitObjectAttributeNV
+SomeValues_0 t_0;
+
+layout(location = 0)
 rayPayloadEXT
 SomeValues_0 p_0;
 
 layout(binding = 0)
 uniform accelerationStructureEXT scene_0;
 
-layout(location = 0)
-hitObjectAttributeNV
-SomeValues_0 t_0;
-
 SomeValues_0 HitObject_GetAttributes_0(hitObjectNV this_0)
 {
     hitObjectGetAttributesNV((this_0), ((0)));