Allow CPP/CUDA/Metal to lower/legalize buffer-elements to support column_major/row_major. (#4653)

* Allow CPP/CUDA/Metal to legalize their buffer-elements.

Fixes: #4537

Changes:
1. Matrix inputs require legalization (pack/unpack) to ensure consistent row_major/column_major throughout entire shader, the following enabled legalization pass fixes this.
2. Added missing CUDA intrinsic so CUDA can run more tests.
3. Added a memory packing test since this still fails for cpp/cuda/metal (due to having no memory packing enforcement).

* change memory packing tests to run for targets without packing

---------

Co-authored-by: Yong He <yonghe@outlook.com>

1 files changed, 118 insertions, 0 deletions

diff --git a/tests/compute/constant-buffer-memory-packing.slang b/tests/compute/constant-buffer-memory-packing.slang
new file mode 100644
index 000000000..5246c4d33
--- /dev/null
+++ b/tests/compute/constant-buffer-memory-packing.slang
@@ -0,0 +1,118 @@
+// column-major-with-row-major-operations.slang
+
+// Metal/CPP/CUDA do not deal with packing currently, different results will occur.
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-cpu -compute -xslang -DTARGET_WITHOUT_PACKING
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-cuda -compute -xslang -DTARGET_WITHOUT_PACKING
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-mtl -compute -xslang -DTARGET_WITHOUT_PACKING
+
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-slang -compute -dx12
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-vk -compute
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX(filecheck-buffer=BUF):-vk -compute -emit-spirv-via-glsl
+
+// CPP/Metal/CUDA due to not having memory packing will recieve the following ROW matrix:
+// {1,2,3}
+// {0,4,5}
+// {6,0,7}
+
+// GLSL/SPIRV/HLSL due to having memory packing will recieve the following ROW/COL matrix:
+// {1,2,3}
+// {0,4,5}
+// {6,0,7}
+
+//TEST_INPUT:cbuffer(data=[1.0 2.0 3.0 0.0  4.0 5.0 6.0 0.0  7.0 8.0 9.0 0]):name matrixTestCBuf1
+ConstantBuffer<row_major float3x3> matrixTestCBuf1;
+
+// CPP/Metal/CUDA due to not having memory packing will recieve the following COL matrix post-transpose:
+// {1,0,8}
+// {4,2,0}
+// {7,5,3}
+
+//TEST_INPUT:cbuffer(data=[1.0 4.0 7.0 0.0  2.0 5.0 8.0 0.0  3.0 6.0 9.0 0.0]):name matrixTestCBuf2
+ConstantBuffer<column_major float3x3> matrixTestCBuf2;
+
+//TEST_INPUT:cbuffer(data=[1.0 2.0 3.0 0.0  4.0 5.0 6.0 0.0]):name NeedsPadding
+cbuffer NeedsPadding
+{
+    float3 data1;
+    float3 data2;
+};
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name output
+RWStructuredBuffer<uint> output;
+
+bool floatCheck(float data, float valueToCheckFor)
+{
+    return data < (valueToCheckFor + 0.001) && data > valueToCheckFor - 0.001;
+}
+
+[numthreads(1, 1, 1)]
+void computeMain(uint3 tid : SV_DispatchThreadID)
+{
+    float3x3 matrixTest1;
+    matrixTest1 = matrixTestCBuf1;
+
+    float3x3 matrixTest2;
+    matrixTest2 = matrixTestCBuf2;
+
+    output[0] = uint(true
+#ifndef TARGET_WITHOUT_PACKING
+            && floatCheck(matrixTest1[0][0], 1)
+            && floatCheck(matrixTest1[0][1], 2)
+            && floatCheck(matrixTest1[0][2], 3)
+            && floatCheck(matrixTest1[1][0], 4)
+            && floatCheck(matrixTest1[1][1], 5)
+            && floatCheck(matrixTest1[1][2], 6)
+            && floatCheck(matrixTest1[2][0], 7)
+            && floatCheck(matrixTest1[2][1], 8)
+            && floatCheck(matrixTest1[2][2], 9)
+
+            && floatCheck(matrixTest2[0][0], 1)
+            && floatCheck(matrixTest2[0][1], 2)
+            && floatCheck(matrixTest2[0][2], 3)
+            && floatCheck(matrixTest2[1][0], 4)
+            && floatCheck(matrixTest2[1][1], 5)
+            && floatCheck(matrixTest2[1][2], 6)
+            && floatCheck(matrixTest2[2][0], 7)
+            && floatCheck(matrixTest2[2][1], 8)
+            && floatCheck(matrixTest2[2][2], 9)
+
+            && floatCheck(data1[0], 1)
+            && floatCheck(data1[1], 2)
+            && floatCheck(data1[2], 3)
+            && floatCheck(data2[0], 4)
+            && floatCheck(data2[1], 5)
+            && floatCheck(data2[2], 6)
+#else
+            && floatCheck(matrixTest1[0][0], 1)
+            && floatCheck(matrixTest1[0][1], 2)
+            && floatCheck(matrixTest1[0][2], 3)
+            && floatCheck(matrixTest1[1][0], 0)
+            && floatCheck(matrixTest1[1][1], 4)
+            && floatCheck(matrixTest1[1][2], 5)
+            && floatCheck(matrixTest1[2][0], 6)
+            && floatCheck(matrixTest1[2][1], 0)
+            && floatCheck(matrixTest1[2][2], 7)
+
+            && floatCheck(matrixTest2[0][0], 1)
+            && floatCheck(matrixTest2[0][1], 0)
+            && floatCheck(matrixTest2[0][2], 8)
+            && floatCheck(matrixTest2[1][0], 4)
+            && floatCheck(matrixTest2[1][1], 2)
+            && floatCheck(matrixTest2[1][2], 0)
+            && floatCheck(matrixTest2[2][0], 7)
+            && floatCheck(matrixTest2[2][1], 5)
+            && floatCheck(matrixTest2[2][2], 3)
+
+            && floatCheck(data1[0], 1)
+            && floatCheck(data1[1], 2)
+            && floatCheck(data1[2], 3)
+            && floatCheck(data2[0], 0)
+            && floatCheck(data2[1], 4)
+            && floatCheck(data2[2], 5)
+#endif
+        );
+    output[1] = (uint)matrixTest2[0][0];
+    output[2] = (uint)matrixTest2[0][1];
+    output[3] = (uint)matrixTest2[0][2];
+    //BUF: 1
+}