Wave intrinsics for Vector and Matrix types (#1262)

* Update slang-binaries to verison with SPIR-V version support.

* Support vec and matrix Wave intrinsics on vk.
Added wave-vector.slang test
Add wave-diverge.slang test
Add support for more wave intrinsics to vk.

* Test out Wave intrinsic support for matrices.

* Remove matrix glsl intrinsics -> not available.
Fix some typo.

8 files changed, 114 insertions, 2 deletions

diff --git a/tests/hlsl-intrinsic/wave-active-product.slang b/tests/hlsl-intrinsic/wave-active-product.slang
index 351df1635..ca3fdcb77 100644
--- a/tests/hlsl-intrinsic/wave-active-product.slang
+++ b/tests/hlsl-intrinsic/wave-active-product.slang
@@ -2,7 +2,7 @@
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
 //TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
-//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-cuda -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0], stride=4):out,name outputBuffer
 RWStructuredBuffer<int> outputBuffer;
diff --git a/tests/hlsl-intrinsic/wave-diverge.slang b/tests/hlsl-intrinsic/wave-diverge.slang
new file mode 100644
index 000000000..ab83a1553
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-diverge.slang
@@ -0,0 +1,26 @@
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute 
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
+
+//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)
+{
+    int idx = int(dispatchThreadID.x);
+    
+    int value = 0;
+    
+    if (idx == 2)
+    {
+        // diverge
+        return;
+    }
+    
+    value = WaveActiveMin(idx + 1);
+    
+    outputBuffer[idx] = value;
+}
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/wave-diverge.slang.expected.txt b/tests/hlsl-intrinsic/wave-diverge.slang.expected.txt
new file mode 100644
index 000000000..68b8a88e2
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-diverge.slang.expected.txt
@@ -0,0 +1,4 @@
+1
+1
+0
+1
diff --git a/tests/hlsl-intrinsic/wave-matrix.slang b/tests/hlsl-intrinsic/wave-matrix.slang
new file mode 100644
index 000000000..022182164
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-matrix.slang
@@ -0,0 +1,37 @@
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute 
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
+//DISABLE_TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0], stride=4):out,name outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+[numthreads(8, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    const int idx = int(dispatchThreadID.x);
+    
+    // NOTE! dxc only supports bit ops on uint and associated types NOT int
+    // Also GLSL does not have built in support for int matrices. So we'll just try with float for now    
+    // GLSL does not support matrix types for Wave like intrinsics   
+       
+    matrix<int, 2, 2> v0 = matrix<int, 2, 2>(idx + 1, idx + 2, idx + 3, idx + 4);   
+    matrix<float, 2, 2> v1 = matrix<float, 2, 2>(v0) + matrix<float, 2, 2>(1, 1, 1, 1);
+    
+    
+    matrix<uint, 2, 2> uv0 = matrix<uint, 2, 2>(v0[0][0], v0[0][1], v0[1][0], v0[0][1]);
+    
+    matrix<int, 2, 2> r0 = WaveActiveSum(v0);
+    matrix<float, 2, 2> r1 = WaveActiveSum(v1);
+    matrix<uint, 2, 2> r2 = WaveActiveBitXor(uv0);
+    matrix<uint, 2, 2> r3 = WaveActiveBitOr(uv0);
+    matrix<uint, 2, 2> r4 = WaveActiveBitAnd(uv0);
+    
+    matrix<uint, 2, 2> r5 = r2 + r3 + r4;
+    matrix<int, 2, 2> r6 = matrix<int, 2, 2>(r5[0][0], r5[0][1], r5[1][0], r5[1][1]); 
+    
+    matrix<int, 2, 2> r = r0 + matrix<int, 2, 2>(r1) + r6;
+   
+    outputBuffer[idx] = r[0][0] + r[0][1] + r[1][0] + r[1][1];
+}
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/wave-matrix.slang.expected.txt b/tests/hlsl-intrinsic/wave-matrix.slang.expected.txt
new file mode 100644
index 000000000..23f9285c3
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-matrix.slang.expected.txt
@@ -0,0 +1,8 @@
+1EC
+1EC
+1EC
+1EC
+1EC
+1EC
+1EC
+1EC
diff --git a/tests/hlsl-intrinsic/wave-vector.slang b/tests/hlsl-intrinsic/wave-vector.slang
new file mode 100644
index 000000000..808f0c5f6
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-vector.slang
@@ -0,0 +1,29 @@
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute 
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0 0 0 0], stride=4):out,name outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+[numthreads(8, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    const int idx = int(dispatchThreadID.x);
+       
+    int2 v0 = int2(idx + 1, idx + 2);   
+    float2 v1 = float2(idx + 2, idx + 3);
+    // NOTE! dxc only supports bit ops on uint and associated types NOT int
+    uint2 uv0 = v0;
+    
+    int2 r0 = WaveActiveSum(v0);
+    float2 r1 = WaveActiveSum(v1);
+    int2 r2 = WaveActiveBitXor(uv0);
+    int2 r3 = WaveActiveBitOr(uv0);
+    int2 r4 = WaveActiveBitAnd(uv0);
+    
+    int2 r = r0 + int2(r1) + r2 + r3 + r4;
+   
+    outputBuffer[idx] = r.x + r.y;
+}
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/wave-vector.slang.expected.txt b/tests/hlsl-intrinsic/wave-vector.slang.expected.txt
new file mode 100644
index 000000000..eb6984bb6
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-vector.slang.expected.txt
@@ -0,0 +1,8 @@
+D6
+D6
+D6
+D6
+D6
+D6
+D6
+D6
diff --git a/tests/hlsl-intrinsic/wave.slang b/tests/hlsl-intrinsic/wave.slang
index 9fc9dc26d..d8273080c 100644
--- a/tests/hlsl-intrinsic/wave.slang
+++ b/tests/hlsl-intrinsic/wave.slang
@@ -2,7 +2,7 @@
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
 //TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
-//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-cuda -compute
+//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name outputBuffer
 RWStructuredBuffer<int> outputBuffer;