More control flow simplifications. (#2673)

* More control flow and Phi param simplifications.

* Fix.

* Fix gcc error.

* Fix.

* More IR cleanup.

* Fix bug in phi param dce + ifelse simplify.

* Propagate and DCE side-effect-free functions.

* Enhance CFG simplifcation to remove loops with no side effects.

* Fix.

* Fixes.

* Fix tests. Add [__AlwaysFoldIntoUseSite] for rayPayloadLocation.

* More cleanup.

* Fixes.

* Fix.

---------

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

5 files changed, 34 insertions, 56 deletions

diff --git a/tests/cross-compile/array-of-buffers.slang.glsl b/tests/cross-compile/array-of-buffers.slang.glsl
index 1f436fad0..21961afd1 100644
--- a/tests/cross-compile/array-of-buffers.slang.glsl
+++ b/tests/cross-compile/array-of-buffers.slang.glsl
@@ -1,8 +1,6 @@
-//TEST_IGNORE_FILE:
 #version 450
 layout(row_major) uniform;
 layout(row_major) buffer;
-
 struct SLANG_ParameterGroup_C_0
 {
     uint index_0;
@@ -13,7 +11,6 @@ layout(std140) uniform _S1
 {
     SLANG_ParameterGroup_C_0 _data;
 } C_0;
-
 struct S_0
 {
     vec4 f_0;
@@ -24,31 +21,26 @@ layout(std140) uniform _S2
 {
     S_0 _data;
 } cb_0[3];
-
 layout(std430, binding = 2) readonly buffer _S3 {
     S_0 _data[];
 } sb1_0[4];
-
 layout(std430, binding = 3) buffer _S4 {
     vec4 _data[];
 } sb2_0[5];
-
 layout(std430, binding = 4) readonly buffer _S5
 {
     uint _data[];
 } bb_0[6];
-
 layout(location = 0)
 out vec4 _S6;
 
 void main()
 {
-    S_0 _S7 = ((sb1_0[C_0._data.index_0])._data[(C_0._data.index_0)]);
-    vec4 _S8 = cb_0[C_0._data.index_0]._data.f_0 + _S7.f_0;
-    vec4 _S9 = _S8 + ((sb2_0[C_0._data.index_0])._data[(C_0._data.index_0)]);
-    uint _S10 = ((bb_0[C_0._data.index_0])._data[(int(C_0._data.index_0 * 4U))/4]);
-
-    _S6 = _S9 + vec4(float(_S10));
-
+    vec4 _S7 = cb_0[C_0._data.index_0]._data.f_0;
+    S_0 _S8 = ((sb1_0[C_0._data.index_0])._data[(C_0._data.index_0)]);
+    vec4 _S9 = _S7 + _S8.f_0;
+    vec4 _S10 = _S9 + ((sb2_0[C_0._data.index_0])._data[(C_0._data.index_0)]);
+    uint _S11 = ((bb_0[C_0._data.index_0])._data[(int(C_0._data.index_0 * 4U))/4]);
+    _S6 = _S10 + vec4(float(_S11));
     return;
 }
diff --git a/tests/cross-compile/array-of-buffers.slang.hlsl b/tests/cross-compile/array-of-buffers.slang.hlsl
index 501b9c6db..960957789 100644
--- a/tests/cross-compile/array-of-buffers.slang.hlsl
+++ b/tests/cross-compile/array-of-buffers.slang.hlsl
@@ -1,5 +1,3 @@
-//TEST_IGNORE_FILE:
-
 #pragma pack_matrix(column_major)
 #ifdef SLANG_HLSL_ENABLE_NVAPI
 #include "nvHLSLExtns.h"
@@ -15,24 +13,24 @@ cbuffer C_0 : register(b0)
 {
     SLANG_ParameterGroup_C_0 C_0;
 }
-
 struct S_0
 {
     float4 f_0;
 };
 
 ConstantBuffer<S_0 >  cb_0[int(3)] : register(b1);
+
 StructuredBuffer<S_0 >  sb1_0[int(4)] : register(t0);
+
 RWStructuredBuffer<float4 >  sb2_0[int(5)] : register(u0);
-ByteAddressBuffer  bb_0[int(6)] : register(t4);
 
+ByteAddressBuffer  bb_0[int(6)] : register(t4);
 float4 main() : SV_TARGET
 {
-    S_0 _S1 = sb1_0[C_0.index_0][C_0.index_0];
-
-    float4 _S2 = cb_0[C_0.index_0].f_0 + _S1.f_0;
-    float4 _S3 = _S2 + sb2_0[C_0.index_0][C_0.index_0];
-    uint _S4 = bb_0[C_0.index_0].Load(int(C_0.index_0 * 4U));
-
-    return _S3 + (float4)float(_S4);
+    float4 _S1 = cb_0[C_0.index_0].f_0;
+    S_0 _S2 = sb1_0[C_0.index_0][C_0.index_0];
+    float4 _S3 = _S1 + _S2.f_0;
+    float4 _S4 = _S3 + sb2_0[C_0.index_0][C_0.index_0];
+    uint _S5 = bb_0[C_0.index_0].Load(int(C_0.index_0 * 4U));
+    return _S4 + (float4)float(_S5);
 }
diff --git a/tests/cross-compile/glsl-generic-in.slang.glsl b/tests/cross-compile/glsl-generic-in.slang.glsl
index c8d9b1bd1..4bf0598d0 100644
--- a/tests/cross-compile/glsl-generic-in.slang.glsl
+++ b/tests/cross-compile/glsl-generic-in.slang.glsl
@@ -1,8 +1,6 @@
-//TEST_IGNORE_FILE:
 #version 450
 layout(row_major) uniform;
 layout(row_major) buffer;
-
 struct F_0
 {
     vec4 v0_0;
@@ -14,10 +12,9 @@ float F_get_0(F_0 this_0)
     return this_0.v0_0.x + this_0.v1_0.x;
 }
 
-
 float E_get_0()
 {
-    return 1.00000000000000000000;
+    return 1.0;
 }
 
 layout(location = 0)
@@ -33,14 +30,11 @@ struct VOut_0
 {
     vec4 projPos_0;
 };
-
 void main()
 {
     F_0 _S4 = { _S2, _S3 };
     VOut_0 vout_0;
-    float _S5 = F_get_0(_S4);
-    float _S6 = E_get_0();
-    vout_0.projPos_0 = vec4(_S1, _S5 + _S6);
+    vout_0.projPos_0 = vec4(_S1, F_get_0(_S4) + E_get_0());
     gl_Position = vout_0.projPos_0;
     return;
-}
\ No newline at end of file
+}
diff --git a/tests/cross-compile/half-conversion.slang.glsl b/tests/cross-compile/half-conversion.slang.glsl
index 58d20b4fc..fb51809b4 100644
--- a/tests/cross-compile/half-conversion.slang.glsl
+++ b/tests/cross-compile/half-conversion.slang.glsl
@@ -1,8 +1,6 @@
-//half-conversion.slang.glsl
-//TEST_IGNORE_FILE:
-
 #version 450
-
+layout(row_major) uniform;
+layout(row_major) buffer;
 struct SLANG_ParameterGroup_C_0
 {
     uvec4 u_0;
@@ -13,29 +11,30 @@ layout(std140) uniform _S1
 {
     SLANG_ParameterGroup_C_0 _data;
 } C_0;
-
 vec4 f16tof32_0(uvec4 value_0)
 {
     vec4 result_0;
-    int i_0;
-    i_0 = 0;
+    int i_0 = 0;
     for(;;)
     {
-        if(i_0 < 4) {} else break;
-
-        float _S2 = (unpackHalf2x16((value_0[i_0])).x);
-        result_0[i_0] = _S2;
-        i_0 = i_0 + int(1);
+        if(i_0 < 4)
+        {
+        }
+        else
+        {
+            break;
+        }
+        result_0[i_0] = (unpackHalf2x16((value_0[i_0])).x);
+        i_0 = i_0 + 1;
     }
     return result_0;
 }
 
 layout(location = 0)
-out vec4 _S3;
+out vec4 _S2;
 
 void main()
 {
-    vec4 _S4 = f16tof32_0(C_0._data.u_0);
-    _S3 = _S4;
+    _S2 = f16tof32_0(C_0._data.u_0);
     return;
 }
diff --git a/tests/cross-compile/sign.slang.glsl b/tests/cross-compile/sign.slang.glsl
index 7a3a37c51..44c015967 100644
--- a/tests/cross-compile/sign.slang.glsl
+++ b/tests/cross-compile/sign.slang.glsl
@@ -1,17 +1,12 @@
-//TEST_IGNORE_FILE:
 #version 450
 layout(row_major) uniform;
 layout(row_major) buffer;
 
-#line 8 0
 layout(location = 0)
 out vec4 _S1;
 
-
-#line 8
 void main()
 {
-    ivec4 _S2 = ivec4(sign(vec4(1.50000000000000000000, 1.00000000000000000000, -1.50000000000000000000, -1.00000000000000000000)));
-    _S1 = vec4(_S2);
+    _S1 = vec4((ivec4(sign((vec4(1.5, 1.0, -1.5, -1.0))))));
     return;
-}
\ No newline at end of file
+}