[Backend] Improve dot support to target FMA (#4516)

This PR:
- Refactors FMA dot implementation
- Supports dot3d in FMA path
- Fixes several issues in operand offset computation
- Enables small dot operands for AMD backend

Author: binarman

include/triton/Conversion/TritonGPUToLLVM/Utility.h

@@ -335,12 +335,18 @@ SmallVector<Value> delinearize(RewriterBase &rewriter, Location loc,
 SmallVector<Value> delinearize(RewriterBase &rewriter, Location loc,
                                Value linear, ArrayRef<unsigned> shape);
 
+SmallVector<unsigned> delinearize(unsigned linear, ArrayRef<unsigned> shape,
+                                  ArrayRef<unsigned> order);
+
 Value linearize(RewriterBase &rewriter, Location loc, ArrayRef<Value> multiDim,
                 ArrayRef<unsigned> shape, ArrayRef<unsigned> order);
 
 Value linearize(RewriterBase &rewriter, Location loc, ArrayRef<Value> multiDim,
                 ArrayRef<unsigned> shape);
 
+size_t linearize(ArrayRef<unsigned> multiDim, ArrayRef<unsigned> shape,
+                 ArrayRef<unsigned> order);
+
 Value addStringToModule(Location loc, RewriterBase &rewriter, StringRef key,
                         StringRef content);
 
@@ -495,6 +501,24 @@ inline Value dot(RewriterBase &rewriter, Location loc, ArrayRef<Value> offsets,
   return ret;
 }
 
+/// Extend 2d shared object to 3d.
+///
+/// If tensor has 3 dimensions, returns original shared object.
+/// If tensor shape is [M, N], return shared object describing shape [1, M, N]
+///
+/// This Function is used to simplify processing of 2d and 3d dot operands,
+/// particularly in the conversion of local_load operation.
+///
+/// \param rewriter
+/// \param loc
+/// \param smemObj
+/// \param shape shape of a tensor represented by smemObj
+/// \returns shared object describing 3d tensor
+SharedMemoryObject
+getExpandedSharedMemoryObject(ConversionPatternRewriter &rewriter, Location loc,
+                              SharedMemoryObject smemObj,
+                              ArrayRef<int64_t> shape);
+
 // -----------------------------------------------------------------------
 // Blocked layout indices
 // -----------------------------------------------------------------------

include/triton/Dialect/TritonGPU/IR/Dialect.h

@@ -234,6 +234,12 @@ void dumpHWLayout(RankedTensorType tensorType);
 // Return a string representation of the layout of the tensor.
 std::string getLayoutStr(RankedTensorType tensorType, bool useHWPointOfView);
 
+template <typename T>
+llvm::SmallVector<T> expandMatrixShapeWithBatch(llvm::ArrayRef<T> s);
+
+llvm::SmallVector<unsigned>
+expandMatrixOrderWithBatch(llvm::ArrayRef<unsigned> o);
+
 } // namespace gpu
 } // namespace triton
 } // namespace mlir