Add matrix select intrinsic (#7566)

* Add matrix select intrinsic

* Fix hlsl test

* Restrict matrix select to HLSL

* Better test for HLSL side

* Select route for GLSL/SPIRV

* Exclude matrices from select legalization

* Exclude CUDA from select test

* Inline and move

* format code

---------

Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>

2 files changed, 49 insertions, 5 deletions

diff --git a/tests/hlsl-intrinsic/matrix-cast-to-vector.slang b/tests/hlsl-intrinsic/matrix-cast-to-vector.slang
index 522f3ce11..74df140d3 100644
--- a/tests/hlsl-intrinsic/matrix-cast-to-vector.slang
+++ b/tests/hlsl-intrinsic/matrix-cast-to-vector.slang
@@ -17,10 +17,9 @@ void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
     float2x2 matrix2x2_2 = (float2x2)vector4_2;
 
     outputBuffer[0] = uint(true
-            && all(vector4_1 == float4(1, 2, 3, 4))
-            
-            && all(matrix2x2_2[0] == float2(1,2))
-            && all(matrix2x2_2[1] == float2(3,4))
-        );
+        && all(vector4_1 == float4(1, 2, 3, 4))
+        && all(matrix2x2_2[0] == float2(1,2))
+        && all(matrix2x2_2[1] == float2(3,4))
+    );
     //BUF: 1
 }
diff --git a/tests/language-feature/matrix-select.slang b/tests/language-feature/matrix-select.slang
new file mode 100644
index 000000000..a3cab6906
--- /dev/null
+++ b/tests/language-feature/matrix-select.slang
@@ -0,0 +1,45 @@
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-dx12 -use-dxil -compute -shaderobj -output-using-type -xslang -matrix-layout-column-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-dx12 -use-dxil -compute -shaderobj -output-using-type -xslang -matrix-layout-row-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-vk -compute -shaderobj -output-using-type -xslang -matrix-layout-column-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-vk -compute -shaderobj -output-using-type -xslang -matrix-layout-row-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-mtl -compute -output-using-type -xslang -matrix-layout-column-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-mtl -compute -output-using-type -xslang -matrix-layout-row-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-wgpu -compute -output-using-type -xslang -matrix-layout-column-major
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-wgpu -compute -output-using-type -xslang -matrix-layout-row-major
+
+//TEST_INPUT:ubuffer(data=[0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+int selectDims<int N, int M>(bool cond)
+{
+    return select(
+        matrix<bool, N, M>(cond),
+        matrix<int, N, M>(1),
+        matrix<int, N, M>(0)
+    )[0][0];
+}
+
+int selectDimsDigit<int N, int M, int D>(int x)
+{
+    return selectDims<N, M>(((x >> D) & 0b1) == 0b1) << D;
+}
+
+[numthreads(1, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    int x = 324;
+
+    int s = 0;
+    s += selectDimsDigit<2, 2, 0>(x);
+    s += selectDimsDigit<2, 3, 1>(x);
+    s += selectDimsDigit<2, 4, 2>(x);
+    s += selectDimsDigit<3, 2, 3>(x);
+    s += selectDimsDigit<3, 3, 4>(x);
+    s += selectDimsDigit<3, 4, 5>(x);
+    s += selectDimsDigit<4, 2, 6>(x);
+    s += selectDimsDigit<4, 3, 7>(x);
+    s += selectDimsDigit<4, 4, 8>(x);
+
+    // CHK: 324
+    outputBuffer[0] = s;
+} 
\ No newline at end of file