//===- AsmMatcherEmitter.cpp - Generate an assembly matcher ---------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This tablegen backend emits a target specifier matcher for converting parsed
// assembly operands in the MCInst structures.
//
// The input to the target specific matcher is a list of literal tokens and
// operands. The target specific parser should generally eliminate any syntax
// which is not relevant for matching; for example, comma tokens should have
// already been consumed and eliminated by the parser. Most instructions will
// end up with a single literal token (the instruction name) and some number of
// operands.
//
// Some example inputs, for X86:
// 'addl' (immediate ...) (register ...)
// 'add' (immediate ...) (memory ...)
// 'call' '*' %epc
//
// The assembly matcher is responsible for converting this input into a precise
// machine instruction (i.e., an instruction with a well defined encoding). This
// mapping has several properties which complicate matching:
//
// - It may be ambiguous; many architectures can legally encode particular
// variants of an instruction in different ways (for example, using a smaller
// encoding for small immediates). Such ambiguities should never be
// arbitrarily resolved by the assembler, the assembler is always responsible
// for choosing the "best" available instruction.
//
// - It may depend on the subtarget or the assembler context. Instructions
// which are invalid for the current mode, but otherwise unambiguous (e.g.,
// an SSE instruction in a file being assembled for i486) should be accepted
// and rejected by the assembler front end. However, if the proper encoding
// for an instruction is dependent
