[SM6.10] Update DXIL op allowed shaders (#8102)

Sets the allowed shader stages for the LinAlg DXIL ops.

This is another commit extracted from the defunct branch that needed to
be merged in before it was lost.

Real changes are in `utils/hct/hctdb.py` the rest is generated code

Author: V-FEXrt

lib/DXIL/DxilOperations.cpp

@@ -3916,22 +3916,31 @@ void OP::GetMinShaderModelAndMask(OpCode C, bool bWithTranslation,
   // RayQuery_CommittedClusterID=2147483653,
   // RayQuery_CandidateTriangleObjectPosition=2147483656,
   // RayQuery_CommittedTriangleObjectPosition=2147483657,
-  // CreateMatrix=2147483659, FillMatrix=2147483660,
-  // CopyConvertMatrix=2147483661, MatrixLoadFromDescriptor=2147483662,
+  // CreateMatrix=2147483659, MatrixLoadFromDescriptor=2147483662,
+  // MatrixQueryAccumulatorLayout=2147483670, MatrixVecMul=2147483673,
+  // MatrixVecMulAdd=2147483674, MatrixAccumulateToDescriptor=2147483675,
+  // MatrixOuterProduct=2147483677
+  if ((305 <= op && op <= 308) || op == 2147483648 ||
+      (2147483652 <= op && op <= 2147483653) ||
+      (2147483656 <= op && op <= 2147483657) || op == 2147483659 ||
+      op == 2147483662 || op == 2147483670 ||
+      (2147483673 <= op && op <= 2147483675) || op == 2147483677) {
+    major = 6;
+    minor = 10;
+    return;
+  }
+  // Instructions: FillMatrix=2147483660, CopyConvertMatrix=2147483661,
   // MatrixLoadFromMemory=2147483663, MatrixLength=2147483664,
   // MatrixGetCoordinate=2147483665, MatrixGetElement=2147483666,
   // MatrixSetElement=2147483667, MatrixStoreToDescriptor=2147483668,
-  // MatrixStoreToMemory=2147483669, MatrixQueryAccumulatorLayout=2147483670,
-  // MatrixMulOp=2147483671, MatrixAccumulate=2147483672,
-  // MatrixVecMul=2147483673, MatrixVecMulAdd=2147483674,
-  // MatrixAccumulateToDescriptor=2147483675,
-  // MatrixAccumulateToMemory=2147483676, MatrixOuterProduct=2147483677
-  if ((305 <= op && op <= 308) || op == 2147483648 ||
-      (2147483652 <= op && op <= 2147483653) ||
-      (2147483656 <= op && op <= 2147483657) ||
-      (2147483659 <= op && op <= 2147483677)) {
+  // MatrixStoreToMemory=2147483669, MatrixMulOp=2147483671,
+  // MatrixAccumulate=2147483672, MatrixAccumulateToMemory=2147483676
+  if ((2147483660 <= op && op <= 2147483661) ||
+      (2147483663 <= op && op <= 2147483669) ||
+      (2147483671 <= op && op <= 2147483672) || op == 2147483676) {
     major = 6;
     minor = 10;
+    mask = SFLAG(Compute) | SFLAG(Mesh) | SFLAG(Amplification);
     return;
   }
   // Instructions: GetGroupWaveIndex=2147483649, GetGroupWaveCount=2147483650

utils/hct/hctdb.py

@@ -1159,18 +1159,30 @@ def populate_categories_and_models_ExperimentalOps(self):
                 "miss",
             )
 
+        # Thread/Wave/ThreadGroup scope operations
         for i in insts(
-            "CreateMatrix,FillMatrix,CopyConvertMatrix,"
-            + "MatrixLoadFromDescriptor,MatrixLoadFromMemory,"
+            "CreateMatrix,MatrixQueryAccumulatorLayout,"
+            + "MatrixLoadFromDescriptor,MatrixAccumulateToDescriptor,"
+            + "MatrixVecMul,MatrixVecMulAdd,MatrixOuterProduct"
+        ):
+            i.category = "Linear Algebra Operations"
+            i.shader_model = experimental_sm
+
+        # Wave/ThreadGroup scope operations
+        for i in insts(
+            "FillMatrix,CopyConvertMatrix,"
             + "MatrixLength,MatrixGetCoordinate,MatrixGetElement,MatrixSetElement,"
-            + "MatrixStoreToDescriptor,MatrixStoreToMemory,"
-            + "MatrixQueryAccumulatorLayout,MatrixMulOp,MatrixAccumulate,"
-            + "MatrixVecMul,MatrixVecMulAdd,"
-            + "MatrixAccumulateToDescriptor,MatrixAccumulateToMemory,"
-            + "MatrixOuterProduct"
+            + "MatrixStoreToDescriptor,"
+            + "MatrixLoadFromMemory,MatrixStoreToMemory,MatrixAccumulateToMemory,"
+            + "MatrixMulOp,MatrixAccumulate"
         ):
             i.category = "Linear Algebra Operations"
             i.shader_model = experimental_sm
+            i.shader_stages = (
+                "compute",
+                "mesh",
+                "amplification",
+            )
 
     def populate_llvm_instructions(self):
         # Add instructions that map to LLVM instructions.