Full address insts elimination for backward autodiff. (#2604)

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

14 files changed, 159 insertions, 87 deletions

diff --git a/tests/autodiff/reverse-struct-multi-write.slang b/tests/autodiff/reverse-struct-multi-write.slang
new file mode 100644
index 000000000..dd12c7d3d
--- /dev/null
+++ b/tests/autodiff/reverse-struct-multi-write.slang
@@ -0,0 +1,48 @@
+
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj -output-using-type
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj -output-using-type
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<float> outputBuffer;
+
+struct A : IDifferentiable
+{
+    float x;
+    float y;
+};
+
+[BackwardDifferentiable]
+A f(A a)
+{
+    // Read/writes to local struct variables won't be SSA'd out by default.
+    // The backward diff preparation pass will kick in to create temp vars for them.
+    A aout;
+    aout.y = 2 * a.x;
+    aout.y = aout.y + 2 * a.x;
+    aout.x = aout.y + 5 * a.x;
+
+    // The result should be equivalent to:
+    /*
+    A aout;
+    var tmp = 2 * a.x;
+    tmp = tmp + 2 * a.x;
+    aout.y = tmp;
+    aout.x = tmp + 5 * a.x;
+    */
+    return aout;
+
+}
+
+[numthreads(1, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    A a = {1.0, 2.0};
+
+    var dpa = diffPair(a);
+
+    A.Differential dout = {1.0, 1.0};
+    
+    __bwd_diff(f)(dpa, dout);
+    outputBuffer[0] = dpa.d.x; // Expect: 13
+    outputBuffer[1] = dpa.d.y; // Expect: 0
+}
diff --git a/tests/autodiff/reverse-struct-multi-write.slang.expected.txt b/tests/autodiff/reverse-struct-multi-write.slang.expected.txt
new file mode 100644
index 000000000..403f2ffd4
--- /dev/null
+++ b/tests/autodiff/reverse-struct-multi-write.slang.expected.txt
@@ -0,0 +1,6 @@
+type: float
+13.000000
+0.000000
+0.000000
+0.000000
+0.000000
diff --git a/tests/compute/half-texture.slang.glsl b/tests/compute/half-texture.slang.glsl
index 88f585378..0eccccaaf 100644
--- a/tests/compute/half-texture.slang.glsl
+++ b/tests/compute/half-texture.slang.glsl
@@ -21,20 +21,23 @@ layout(std430, binding = 0) buffer _S1 {
     int _data[];
 } outputBuffer_0;
 
-layout(local_size_x = 4, local_size_y = 4, local_size_z = 1) in;void main()
+layout(local_size_x = 4, local_size_y = 4, local_size_z = 1) in;
+void main()
 {
     ivec2 pos_0 = ivec2(gl_GlobalInvocationID.xy);
     const float _S2 = 1.00000000000000000000 / 3.00000000000000000000;
-    ivec2 pos2_0 = ivec2(3 - pos_0.y, 3 - pos_0.x);
+    int _S3 = pos_0.y;
+    int _S4 = pos_0.x;
+    ivec2 pos2_0 = ivec2(3 - _S3, 3 - _S4);
 
     float16_t h_0 = (float16_t(imageLoad((halfTexture_0), ivec2((uvec2(pos2_0)))).x));
     f16vec2 h2_0 = (f16vec2(imageLoad((halfTexture2_0), ivec2((uvec2(pos2_0)))).xy));
     f16vec4 h4_0 = (f16vec4(imageLoad((halfTexture4_0), ivec2((uvec2(pos2_0))))));
-	imageStore((halfTexture_0), ivec2((uvec2(pos_0))), f16vec4(h2_0.x + h2_0.y, float16_t(0), float16_t(0), float16_t(0)));
-	imageStore((halfTexture2_0), ivec2((uvec2(pos_0))), f16vec4(h4_0.xy, float16_t(0), float16_t(0)));
-	imageStore((halfTexture4_0), ivec2((uvec2(pos_0))), f16vec4(h2_0, h_0, h_0));
+    imageStore((halfTexture_0), ivec2((uvec2(pos_0))), f16vec4(h2_0.x + h2_0.y, float16_t(0), float16_t(0), float16_t(0)));
+    imageStore((halfTexture2_0), ivec2((uvec2(pos_0))), f16vec4(h4_0.xy, float16_t(0), float16_t(0)));
+    imageStore((halfTexture4_0), ivec2((uvec2(pos_0))), f16vec4(h2_0, h_0, h_0));
 
-    int index_0 = pos_0.x + pos_0.y * 4;
+    int index_0 = _S4 + _S3 * 4;
     ((outputBuffer_0)._data[(uint(index_0))]) = index_0;
 
     return;
diff --git a/tests/compute/half-texture.slang.hlsl b/tests/compute/half-texture.slang.hlsl
index c606703a4..2d04ee17f 100644
--- a/tests/compute/half-texture.slang.hlsl
+++ b/tests/compute/half-texture.slang.hlsl
@@ -8,19 +8,21 @@ RWStructuredBuffer<int > outputBuffer_0 : register(u0);
 [shader("compute")][numthreads(4, 4, 1)]
 void computeMain(uint3 dispatchThreadID_0 : SV_DISPATCHTHREADID)
 {
-    int2 pos_0 = (int2) dispatchThreadID_0.xy;
+    int2 pos_0 = int2(dispatchThreadID_0.xy);
     float _S1 = 1.00000000000000000000 / 3.00000000000000000000;
-    int2 pos2_0 = int2(int(3) - pos_0.y, int(3) - pos_0.x);
+    int _S2 = pos_0.y;
+    int _S3 = pos_0.x;
+    int2 pos2_0 = int2(int(3) - _S2, int(3) - _S3);
 
-    half h_0 = halfTexture_0[(uint2) pos2_0];
-    vector<half,2> h2_0 = halfTexture2_0[(uint2) pos2_0];
-    vector<half,4> h4_0 = halfTexture4_0[(uint2) pos2_0];
+    half h_0 = halfTexture_0[uint2(pos2_0)];
+    vector<half, 2> h2_0 = halfTexture2_0[uint2(pos2_0)];
+    vector<half, 4> h4_0 = halfTexture4_0[uint2(pos2_0)];
 
-    halfTexture_0[(uint2) pos_0] = h2_0.x + h2_0.y;
-    halfTexture2_0[(uint2) pos_0] = h4_0.xy;
-    halfTexture4_0[(uint2) pos_0] = vector<half,4>(h2_0, h_0, h_0);
+    halfTexture_0[uint2(pos_0)] = h2_0.x + h2_0.y;
+    halfTexture2_0[uint2(pos_0)] = h4_0.xy;
+    halfTexture4_0[uint2(pos_0)] = vector<half, 4>(h2_0, h_0, h_0);
 
-    int index_0 = pos_0.x + pos_0.y * int(4);
-    outputBuffer_0[(uint) index_0] = index_0;
+    int index_0 = _S3 + _S2 * int(4);
+    outputBuffer_0[uint(index_0)] = index_0;
     return;
 }
diff --git a/tests/cross-compile/precise-keyword.slang.glsl b/tests/cross-compile/precise-keyword.slang.glsl
index 17fed739e..027a8eb3b 100644
--- a/tests/cross-compile/precise-keyword.slang.glsl
+++ b/tests/cross-compile/precise-keyword.slang.glsl
@@ -11,15 +11,18 @@ in vec2 _S2;
 
 void main()
 {
+    float _S3 = _S2.x;
+
     precise float z_0;
 
-    if(_S2.x > float(0))
+    if(_S3 > 0.00000000000000000000)
     {
-        z_0 = _S2.x * _S2.y + _S2.x;
+        z_0 = _S3 * _S2.y + _S3;
     }
     else
     {
-        z_0 = _S2.y * _S2.x + _S2.y;
+        float _S4 = _S2.y;
+        z_0 = _S4 * _S3 + _S4;
     }
     _S1 = vec4(z_0);
     return;
diff --git a/tests/cross-compile/precise-keyword.slang.hlsl b/tests/cross-compile/precise-keyword.slang.hlsl
index 54017868b..7a07fdc5e 100644
--- a/tests/cross-compile/precise-keyword.slang.hlsl
+++ b/tests/cross-compile/precise-keyword.slang.hlsl
@@ -3,15 +3,17 @@
 
 float4 main(float2 v_0 : V) : SV_TARGET
 {
+    float _S1 = v_0.x;
     precise float z_0;
 
-    if(v_0.x > (float) 0)
+    if (_S1 > 0.00000000000000000000)
     {
-        z_0 = v_0.x * v_0.y + v_0.x;
+        z_0 = _S1 * v_0.y + _S1;
     }
     else
     {
-        z_0 = v_0.y * v_0.x + v_0.y;
+        float _S2 = v_0.y;
+        z_0 = _S2 * _S1 + _S2;
     }
 
     return (float4) z_0;
diff --git a/tests/experimental/liveness/liveness-6.slang.expected b/tests/experimental/liveness/liveness-6.slang.expected
index ac1894f95..26a537330 100644
--- a/tests/experimental/liveness/liveness-6.slang.expected
+++ b/tests/experimental/liveness/liveness-6.slang.expected
@@ -60,15 +60,16 @@ int calcThing_0(int offset_0)
             i_0 = i_0 + 1;
         }
         livenessEnd_0(i_0, 0);
-        int _S3 = another_0[k_0 & 1];
-        int _S4 = total_0;
+        int _S3 = k_0 & 1;
+        int _S4 = another_0[_S3];
+        int _S5 = total_0;
         livenessEnd_0(total_0, 0);
-        int total_1 = _S4 + _S3;
-        int _S5 = arr_0[k_0 & 1];
+        int total_1 = _S5 + _S4;
+        int _S6 = arr_0[_S3];
         livenessEnd_1(arr_0, 0);
-        int total_2 = total_1 + _S5;
-        int _S6 = (k_0 + 7) % 5;
-        if(_S6 == 4)
+        int total_2 = total_1 + _S6;
+        int _S7 = (k_0 + 7) % 5;
+        if(_S7 == 4)
         {
             livenessEnd_0(k_0, 0);
             livenessEnd_1(another_0, 0);
@@ -83,32 +84,32 @@ int calcThing_0(int offset_0)
     int total_3;
     if(total_0 > 4)
     {
-        int _S7 = total_0;
+        int _S8 = total_0;
         livenessEnd_0(total_0, 0);
-        int _S8 = - _S7;
+        int _S9 = - _S8;
         livenessStart_1(total_3, 0);
-        total_3 = _S8;
+        total_3 = _S9;
     }
     else
     {
-        int _S9 = total_0;
+        int _S10 = total_0;
         livenessEnd_0(total_0, 0);
         livenessStart_1(total_3, 0);
-        total_3 = _S9;
+        total_3 = _S10;
     }
     return total_3;
 }
 
-layout(std430, binding = 0) buffer _S10 {
+layout(std430, binding = 0) buffer _S11 {
     int _data[];
 } outputBuffer_0;
 layout(local_size_x = 4, local_size_y = 1, local_size_z = 1) in;
 void main()
 {
     int index_0 = int(gl_GlobalInvocationID.x);
-    uint _S11 = uint(index_0);
-    int _S12 = calcThing_0(index_0);
-    ((outputBuffer_0)._data[(_S11)]) = _S12;
+    uint _S12 = uint(index_0);
+    int _S13 = calcThing_0(index_0);
+    ((outputBuffer_0)._data[(_S12)]) = _S13;
     return;
 }
 
diff --git a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-make-hit.slang.1.expected b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-make-hit.slang.1.expected
index 8fc391feb..15221b921 100644
--- a/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-make-hit.slang.1.expected
+++ b/tests/hlsl-intrinsic/shader-execution-reordering/hit-object-make-hit.slang.1.expected
@@ -57,15 +57,16 @@ uint calcValue_0(hitObjectNV hit_0)
         uint hitKind_0 = (hitObjectGetHitKindNV((hit_0)));
         uint r_1 = 0U + hitKind_0 + instanceIndex_0 + instanceID_0 + geometryIndex_0 + primitiveIndex_0;
         RayDesc_0 ray_1 = HitObject_GetRayDesc_0(hit_0);
-        uint r_2 = r_1 + uint(ray_1.TMin_0 > 0.00000000000000000000) + uint(ray_1.TMax_0 < ray_1.TMin_0);
+        float _S6 = ray_1.TMin_0;
+        uint r_2 = r_1 + uint(_S6 > 0.00000000000000000000) + uint(ray_1.TMax_0 < _S6);
         SomeValues_0 objSomeValues_0 = HitObject_GetAttributes_0(hit_0);
         r_0 = r_2 + uint(objSomeValues_0.a_0);
     }
     else
     {
-        bool _S6 = (hitObjectIsMissNV((hit_0)));
+        bool _S7 = (hitObjectIsMissNV((hit_0)));
         uint r_3;
-        if(_S6)
+        if(_S7)
         {
             r_3 = 1U;
         }
@@ -78,29 +79,29 @@ uint calcValue_0(hitObjectNV hit_0)
     return r_0;
 }
 
-layout(std430, binding = 1) buffer _S7 {
+layout(std430, binding = 1) buffer _S8 {
     uint _data[];
 } outputBuffer_0;
 void main()
 {
-    uvec3 _S8 = ((gl_LaunchIDEXT));
-    ivec2 launchID_0 = ivec2(_S8.xy);
-    uvec3 _S9 = ((gl_LaunchSizeEXT));
+    uvec3 _S9 = ((gl_LaunchIDEXT));
+    ivec2 launchID_0 = ivec2(_S9.xy);
+    uvec3 _S10 = ((gl_LaunchSizeEXT));
     int idx_0 = launchID_0.x;
     RayDesc_0 ray_2;
     ray_2.Origin_0 = vec3(float(idx_0), 0.00000000000000000000, 0.00000000000000000000);
     ray_2.TMin_0 = 0.00999999977648258209;
     ray_2.Direction_0 = vec3(0.00000000000000000000, 1.00000000000000000000, 0.00000000000000000000);
     ray_2.TMax_0 = 10000.00000000000000000000;
-    RayDesc_0 _S10 = ray_2;
+    RayDesc_0 _S11 = ray_2;
     hitObjectNV hitObj_0;
-    hitObjectRecordHitWithIndexNV(hitObj_0, scene_0, int(uint(idx_0)), int(uint(idx_0 * 2)), int(uint(idx_0 * 3)), 0U, 0U, _S10.Origin_0, _S10.TMin_0, _S10.Direction_0, _S10.TMax_0, (0));
+    hitObjectRecordHitWithIndexNV(hitObj_0, scene_0, int(uint(idx_0)), int(uint(idx_0 * 2)), int(uint(idx_0 * 3)), 0U, 0U, _S11.Origin_0, _S11.TMin_0, _S11.Direction_0, _S11.TMax_0, (0));
     uint r_4 = calcValue_0(hitObj_0);
-    RayDesc_0 _S11 = ray_2;
+    RayDesc_0 _S12 = ray_2;
     hitObjectNV hitObj_1;
-    hitObjectRecordHitNV(hitObj_1, scene_0, int(uint(idx_0)), int(uint(idx_0 * 3)), int(uint(idx_0 * 2)), 0U, 0U, 4U, _S11.Origin_0, _S11.TMin_0, _S11.Direction_0, _S11.TMax_0, (0));
-    uint _S12 = calcValue_0(hitObj_1);
-    uint r_5 = r_4 + _S12;
+    hitObjectRecordHitNV(hitObj_1, scene_0, int(uint(idx_0)), int(uint(idx_0 * 3)), int(uint(idx_0 * 2)), 0U, 0U, 4U, _S12.Origin_0, _S12.TMin_0, _S12.Direction_0, _S12.TMax_0, (0));
+    uint _S13 = calcValue_0(hitObj_1);
+    uint r_5 = r_4 + _S13;
     ((outputBuffer_0)._data[(uint(idx_0))]) = r_5;
     return;
 }
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 90223115b..f250c1c92 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
@@ -79,36 +79,37 @@ void main()
     ivec2 launchID_0 = ivec2(_S3.xy);
     uvec3 _S4 = ((gl_LaunchSizeEXT));
     int idx_0 = launchID_0.x;
-    SomeValues_0 someValues_0 = { idx_0, float(idx_0) * 2.00000000000000000000 };
+    float _S5 = float(idx_0);
+    SomeValues_0 someValues_0 = { idx_0, _S5 * 2.00000000000000000000 };
     RayDesc_0 ray_0;
-    ray_0.Origin_0 = vec3(float(idx_0), 0.00000000000000000000, 0.00000000000000000000);
+    ray_0.Origin_0 = vec3(_S5, 0.00000000000000000000, 0.00000000000000000000);
     ray_0.TMin_0 = 0.00999999977648258209;
     ray_0.Direction_0 = vec3(0.00000000000000000000, 1.00000000000000000000, 0.00000000000000000000);
     ray_0.TMax_0 = 10000.00000000000000000000;
-    RayDesc_0 _S5 = ray_0;
+    RayDesc_0 _S6 = ray_0;
     p_0 = someValues_0;
     hitObjectNV hitObj_0;
-    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S5.Origin_0, _S5.TMin_0, _S5.Direction_0, _S5.TMax_0, (0));
+    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S6.Origin_0, _S6.TMin_0, _S6.Direction_0, _S6.TMax_0, (0));
     uint r_1 = calcValue_0(hitObj_0);
     reorderThreadNV(hitObj_0);
     SomeValues_0 otherValues_0;
-    SomeValues_0 _S6 = { idx_0 * -1, float(idx_0) * 4.00000000000000000000 };
-    otherValues_0 = _S6;
+    SomeValues_0 _S7 = { idx_0 * -1, _S5 * 4.00000000000000000000 };
+    otherValues_0 = _S7;
     HitObject_Invoke_0(scene_0, hitObj_0, otherValues_0);
-    uint _S7 = calcValue_0(hitObj_0);
-    uint r_2 = r_1 + _S7;
+    uint _S8 = calcValue_0(hitObj_0);
+    uint r_2 = r_1 + _S8;
     reorderThreadNV(hitObj_0, uint(idx_0 & 3), 2U);
-    SomeValues_0 _S8 = { idx_0 * -2, float(idx_0) * 8.00000000000000000000 };
-    otherValues_0 = _S8;
+    SomeValues_0 _S9 = { idx_0 * -2, _S5 * 8.00000000000000000000 };
+    otherValues_0 = _S9;
     HitObject_Invoke_0(scene_0, hitObj_0, otherValues_0);
-    uint _S9 = calcValue_0(hitObj_0);
-    uint r_3 = r_2 + _S9;
+    uint _S10 = calcValue_0(hitObj_0);
+    uint r_3 = r_2 + _S10;
     reorderThreadNV(uint(idx_0 & 1), 1U);
-    SomeValues_0 _S10 = { idx_0 * -4, float(idx_0) * 16.00000000000000000000 };
-    otherValues_0 = _S10;
+    SomeValues_0 _S11 = { idx_0 * -4, _S5 * 16.00000000000000000000 };
+    otherValues_0 = _S11;
     HitObject_Invoke_0(scene_0, hitObj_0, otherValues_0);
-    uint _S11 = calcValue_0(hitObj_0);
-    uint r_4 = r_3 + _S11;
+    uint _S12 = calcValue_0(hitObj_0);
+    uint r_4 = r_3 + _S12;
     ((outputBuffer_0)._data[(uint(idx_0))]) = r_4;
     return;
 }
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 a86dc6aa7..f6f6f132d 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
@@ -70,19 +70,20 @@ void main()
     int idx_0 = launchID_0.x;
     int _S5 = idx_0 / 4;
     float currentTime_0 = float(_S5);
-    SomeValues_0 someValues_0 = { idx_0, float(idx_0) * 2.00000000000000000000 };
+    float _S6 = float(idx_0);
+    SomeValues_0 someValues_0 = { idx_0, _S6 * 2.00000000000000000000 };
     RayDesc_0 ray_0;
-    ray_0.Origin_0 = vec3(float(idx_0), 0.00000000000000000000, 0.00000000000000000000);
+    ray_0.Origin_0 = vec3(_S6, 0.00000000000000000000, 0.00000000000000000000);
     ray_0.TMin_0 = 0.00999999977648258209;
     ray_0.Direction_0 = vec3(0.00000000000000000000, 1.00000000000000000000, 0.00000000000000000000);
     ray_0.TMax_0 = 10000.00000000000000000000;
-    RayDesc_0 _S6 = ray_0;
+    RayDesc_0 _S7 = ray_0;
     p_0 = someValues_0;
     hitObjectNV hitObj_0;
-    hitObjectTraceRayMotionNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S6.Origin_0, _S6.TMin_0, _S6.Direction_0, _S6.TMax_0, currentTime_0, (0));
-    uint _S7 = uint(idx_0);
-    uint _S8 = calcValue_0(hitObj_0);
-    ((outputBuffer_0)._data[(_S7)]) = _S8;
+    hitObjectTraceRayMotionNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S7.Origin_0, _S7.TMin_0, _S7.Direction_0, _S7.TMax_0, currentTime_0, (0));
+    uint _S8 = uint(idx_0);
+    uint _S9 = calcValue_0(hitObj_0);
+    ((outputBuffer_0)._data[(_S8)]) = _S9;
     return;
 }
 
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 38ddbf233..16099b5e2 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
@@ -67,19 +67,20 @@ void main()
     ivec2 launchID_0 = ivec2(_S3.xy);
     uvec3 _S4 = ((gl_LaunchSizeEXT));
     int idx_0 = launchID_0.x;
-    SomeValues_0 someValues_0 = { idx_0, float(idx_0) * 2.00000000000000000000 };
+    float _S5 = float(idx_0);
+    SomeValues_0 someValues_0 = { idx_0, _S5 * 2.00000000000000000000 };
     RayDesc_0 ray_0;
-    ray_0.Origin_0 = vec3(float(idx_0), 0.00000000000000000000, 0.00000000000000000000);
+    ray_0.Origin_0 = vec3(_S5, 0.00000000000000000000, 0.00000000000000000000);
     ray_0.TMin_0 = 0.00999999977648258209;
     ray_0.Direction_0 = vec3(0.00000000000000000000, 1.00000000000000000000, 0.00000000000000000000);
     ray_0.TMax_0 = 10000.00000000000000000000;
-    RayDesc_0 _S5 = ray_0;
+    RayDesc_0 _S6 = ray_0;
     p_0 = someValues_0;
     hitObjectNV hitObj_0;
-    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S5.Origin_0, _S5.TMin_0, _S5.Direction_0, _S5.TMax_0, (0));
-    uint _S6 = uint(idx_0);
-    uint _S7 = calcValue_0(hitObj_0);
-    ((outputBuffer_0)._data[(_S6)]) = _S7;
+    hitObjectTraceRayNV(hitObj_0, scene_0, 20U, 255U, 0U, 4U, 0U, _S6.Origin_0, _S6.TMin_0, _S6.Direction_0, _S6.TMax_0, (0));
+    uint _S7 = uint(idx_0);
+    uint _S8 = calcValue_0(hitObj_0);
+    ((outputBuffer_0)._data[(_S7)]) = _S8;
     return;
 }
 
diff --git a/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl b/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
index 1818b7789..84eba46f0 100644
--- a/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
+++ b/tests/pipeline/rasterization/fragment-shader-interlock.slang.glsl
@@ -19,11 +19,13 @@ void main()
 {
     beginInvocationInterlockARB();
 
-    vec4 _S3 = (imageLoad((entryPointParams_texture_0), ivec2((uvec2(_S1.xy)))));
-    imageStore((entryPointParams_texture_0), ivec2((uvec2(_S1.xy))), _S3 + _S1);
+    vec2 _S3 = _S1.xy;
+
+    vec4 _S4 = (imageLoad((entryPointParams_texture_0), ivec2((uvec2(_S3)))));
+    imageStore((entryPointParams_texture_0), ivec2((uvec2(_S3))), _S4 + _S1);
 
     endInvocationInterlockARB();
 
-    _S2 = _S3;
+    _S2 = _S4;
     return;
 }
diff --git a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
index 1da5f4f8a..864f44eb3 100644
--- a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
+++ b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.glsl
@@ -14,6 +14,7 @@ out vec4 _S2;
 
 void main()
 {
+    uvec2 _S3 = uvec2(0U, 0U);
     _S2 = gl_BaryCoordNV.x * ((_S1)[(0U)]) + gl_BaryCoordNV.y * ((_S1)[(1U)]) + gl_BaryCoordNV.z * ((_S1)[(2U)]);
     return;
 }
diff --git a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.hlsl b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.hlsl
index 257b334bf..ce23492c9 100644
--- a/tests/pipeline/rasterization/get-attribute-at-vertex.slang.hlsl
+++ b/tests/pipeline/rasterization/get-attribute-at-vertex.slang.hlsl
@@ -8,7 +8,7 @@ void main(
     vector<float,3> bary_0 : SV_BARYCENTRICS,
     out vector<float,4> result_0 : SV_TARGET)
 {
-    result_0 = bary_0.x * GetAttributeAtVertex(color_0, (uint) int(0))
-             + bary_0.y * GetAttributeAtVertex(color_0, (uint) int(1))
-             + bary_0.z * GetAttributeAtVertex(color_0, (uint) int(2));
+    result_0 = bary_0.x * GetAttributeAtVertex(color_0, 0U)
+             + bary_0.y * GetAttributeAtVertex(color_0, 1U)
+             + bary_0.z * GetAttributeAtVertex(color_0, 2U);
 }