WaveBroadcastAt/WaveShuffle (#1299)

* Support for WaveReadLaneAt with dynamic (but uniform across Wave) on Vk by enabling VK1.4.
Fixed wave-lane-at.slang test to test with laneId that is uniform across the Wave.

* Added WaveShuffle intrinsic.
Test for WaveShuffle intrinsic.

* Added some documentation on WaveShuffle

* Fix that version required for subgroupBroadcast to be non constexpr is actually 1.5

* Added WaveBroadcastLaneAt
Documented WaveShuffle/BroadcastLaneAt/ReadLaneAt

* Update docs around WaveBroadcast/Read/Shuffle.
Use '_waveShuffle` as name in CUDA prelude to better describe it's more flexible behavior.

9 files changed, 78 insertions, 3 deletions

diff --git a/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang b/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang
new file mode 100644
index 000000000..3c746476a
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang
@@ -0,0 +1,28 @@
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -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;
+    
+    // Scalar
+
+    value += WaveBroadcastLaneAt(idx, 1);
+    
+    // vector
+    
+    {
+        float2 v = float2(idx + 1, idx + 2);
+        float2 readValue = WaveBroadcastLaneAt(v, 4 & 3);
+        
+        value += int(readValue[0] + readValue[1]);
+    }            
+   
+    outputBuffer[idx] = value;
+}
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang.expected.txt b/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang.expected.txt
new file mode 100644
index 000000000..e785149d2
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-broadcast-lane-at-vk.slang.expected.txt
@@ -0,0 +1,4 @@
+4
+4
+4
+4
diff --git a/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang b/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang
new file mode 100644
index 000000000..b6f5d3847
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang
@@ -0,0 +1,41 @@
+//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
+// Disabled on VK because glsl can't do WaveReadLaneAt on matrix. 
+//DISABLE_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;
+    
+    // Scalar
+    
+    value += WaveBroadcastLaneAt(idx, 1);
+    
+    // vector
+    
+    {
+        float2 v = float2(idx + 1, idx + 2);
+        float2 readValue = WaveBroadcastLaneAt(v, 2);
+        
+        value += int(readValue[0] + readValue[1]);
+    }
+        
+    // matrix
+    {
+        matrix<int, 2, 2> v = matrix<int, 2, 2>(idx, idx - 1, idx * 3, idx - 2);
+        
+        matrix<int, 2, 2> readValue = WaveBroadcastLaneAt(v, 3);
+        
+        value += int(readValue[0][0] + readValue[0][1] + readValue[1][0] + readValue[1][1]);
+    }
+    
+    outputBuffer[idx] = value;
+}
\ No newline at end of file
diff --git a/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang.expected.txt b/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang.expected.txt
new file mode 100644
index 000000000..5ce1f8639
--- /dev/null
+++ b/tests/hlsl-intrinsic/wave-broadcast-lane-at.slang.expected.txt
@@ -0,0 +1,4 @@
+17
+17
+17
+17
diff --git a/tests/hlsl-intrinsic/wave-lane-at-vk.slang b/tests/hlsl-intrinsic/wave-read-lane-at-vk.slang
index 0d52f781e..3bd6b36b8 100644
--- a/tests/hlsl-intrinsic/wave-lane-at-vk.slang
+++ b/tests/hlsl-intrinsic/wave-read-lane-at-vk.slang
@@ -2,8 +2,7 @@
 // We have this 'simple' test, because we can't do matrix (or imat) operations on GLSL/Vk target
 
 //TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -use-dxil -profile cs_6_0
-// TODO(JS): Disabled for now, as requires upgraded glslang
-//DISABLE_TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name outputBuffer
 RWStructuredBuffer<int> outputBuffer;
diff --git a/tests/hlsl-intrinsic/wave-lane-at-vk.slang.expected.txt b/tests/hlsl-intrinsic/wave-read-lane-at-vk.slang.expected.txt
index 4e98888c6..4e98888c6 100644
--- a/tests/hlsl-intrinsic/wave-lane-at-vk.slang.expected.txt
+++ b/tests/hlsl-intrinsic/wave-read-lane-at-vk.slang.expected.txt
diff --git a/tests/hlsl-intrinsic/wave-lane-at.slang b/tests/hlsl-intrinsic/wave-read-lane-at.slang
index c3caaa4e8..c3caaa4e8 100644
--- a/tests/hlsl-intrinsic/wave-lane-at.slang
+++ b/tests/hlsl-intrinsic/wave-read-lane-at.slang
diff --git a/tests/hlsl-intrinsic/wave-lane-at.slang.expected.txt b/tests/hlsl-intrinsic/wave-read-lane-at.slang.expected.txt
index c6167dbae..c6167dbae 100644
--- a/tests/hlsl-intrinsic/wave-lane-at.slang.expected.txt
+++ b/tests/hlsl-intrinsic/wave-read-lane-at.slang.expected.txt
diff --git a/tests/hlsl-intrinsic/wave-shuffle-vk.slang b/tests/hlsl-intrinsic/wave-shuffle-vk.slang
index 01fb59155..75aa392ea 100644
--- a/tests/hlsl-intrinsic/wave-shuffle-vk.slang
+++ b/tests/hlsl-intrinsic/wave-shuffle-vk.slang
@@ -2,7 +2,6 @@
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute 
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
 //DISABLE_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