176 lines · plain
1===============================2How To Use Instruction Mappings3===============================4 5.. contents::6 :local:7 8Introduction9============10 11This document contains information about adding instruction mapping support12for a target. The motivation behind this feature comes from the need to switch13between different instruction formats during various optimizations. One approach14could be to use switch cases which list all the instructions along with formats15they can transition to. However, it has large maintenance overhead16because of the hardcoded instruction names. Also, whenever a new instruction is17added in the .td files, all the relevant switch cases should be modified18accordingly. Instead, the same functionality could be achieved with TableGen and19some support from the .td files for a fraction of maintenance cost.20 21``InstrMapping`` Class Overview22===============================23 24TableGen uses relationship models to map instructions with each other. These25models are described using ``InstrMapping`` class as a base. Each model sets26various fields of the ``InstrMapping`` class such that they can uniquely27describe all the instructions using that model. TableGen parses all the relation28models and uses the information to construct relation tables which relate29instructions with each other. These tables are emitted in the30``XXXInstrInfo.inc`` file along with the functions to query them. Following31is the definition of ``InstrMapping`` class defined in Target.td file:32 33.. code-block:: text34 35 class InstrMapping {36 // Used to reduce search space only to the instructions using this37 // relation model.38 string FilterClass;39 40 // List of fields/attributes that should be same for all the instructions in41 // a row of the relation table. Think of this as a set of properties shared42 // by all the instructions related by this relationship.43 list<string> RowFields = [];44 45 // List of fields/attributes that are same for all the instructions46 // in a column of the relation table.47 list<string> ColFields = [];48 49 // Values for the fields/attributes listed in 'ColFields' corresponding to50 // the key instruction. This is the instruction that will be transformed51 // using this relation model.52 list<string> KeyCol = [];53 54 // List of values for the fields/attributes listed in 'ColFields', one for55 // each column in the relation table. These are the instructions a key56 // instruction will be transformed into.57 list<list<string> > ValueCols = [];58 }59 60Sample Example61--------------62 63Let's say that we want to have a function64``int getPredOpcode(uint16_t Opcode, enum PredSense inPredSense)`` which65takes a non-predicated instruction and returns its predicated true or false form66depending on some input flag, ``inPredSense``. The first step in the process is67to define a relationship model that relates predicated instructions to their68non-predicated form by assigning appropriate values to the ``InstrMapping``69fields. For this relationship, non-predicated instructions are treated as key70instruction since they are the ones used to query the interface function.71 72.. code-block:: text73 74 def getPredOpcode : InstrMapping {75 // Choose a FilterClass that is used as a base class for all the76 // instructions modeling this relationship. This is done to reduce the77 // search space only to these set of instructions.78 let FilterClass = "PredRel";79 80 // Instructions with same values for all the fields in RowFields form a81 // row in the resulting relation table.82 // For example, if we want to relate 'ADD' (non-predicated) with 'Add_pt'83 // (predicated true) and 'Add_pf' (predicated false), then all 384 // instructions need to have same value for BaseOpcode field. It can be any85 // unique value (Ex: XYZ) and should not be shared with any other86 // instruction not related to 'add'.87 let RowFields = ["BaseOpcode"];88 89 // List of attributes that can be used to define key and column instructions90 // for a relation. Key instruction is passed as an argument91 // to the function used for querying relation tables. Column instructions92 // are the instructions they (key) can transform into.93 //94 // Here, we choose 'PredSense' as ColFields since this is the unique95 // attribute of the key (non-predicated) and column (true/false)96 // instructions involved in this relationship model.97 let ColFields = ["PredSense"];98 99 // The key column contains non-predicated instructions.100 let KeyCol = ["none"];101 102 // Two value columns - first column contains instructions with103 // PredSense=true while second column has instructions with PredSense=false.104 let ValueCols = [["true"], ["false"]];105 }106 107TableGen uses the above relationship model to emit relation table that maps108non-predicated instructions with their predicated forms. It also outputs the109interface function110``int getPredOpcode(uint16_t Opcode, enum PredSense inPredSense)`` to query111the table. Here, Function ``getPredOpcode`` takes two arguments, opcode of the112current instruction and PredSense of the desired instruction, and returns113predicated form of the instruction, if found in the relation table.114In order for an instruction to be added into the relation table, it needs115to include relevant information in its definition. For example, consider116following to be the current definitions of ADD, ADD_pt (true) and ADD_pf (false)117instructions:118 119.. code-block:: text120 121 def ADD : ALU32_rr<(outs IntRegs:$dst), (ins IntRegs:$a, IntRegs:$b),122 "$dst = add($a, $b)",123 [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$a),124 (i32 IntRegs:$b)))]>;125 126 def ADD_Pt : ALU32_rr<(outs IntRegs:$dst),127 (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),128 "if ($p) $dst = add($a, $b)",129 []>;130 131 def ADD_Pf : ALU32_rr<(outs IntRegs:$dst),132 (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),133 "if (!$p) $dst = add($a, $b)",134 []>;135 136In this step, we modify these instructions to include the information137required by the relationship model, <tt>getPredOpcode</tt>, so that they can138be related.139 140.. code-block:: text141 142 def ADD : PredRel, ALU32_rr<(outs IntRegs:$dst), (ins IntRegs:$a, IntRegs:$b),143 "$dst = add($a, $b)",144 [(set (i32 IntRegs:$dst), (add (i32 IntRegs:$a),145 (i32 IntRegs:$b)))]> {146 let BaseOpcode = "ADD";147 let PredSense = "none";148 }149 150 def ADD_Pt : PredRel, ALU32_rr<(outs IntRegs:$dst),151 (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),152 "if ($p) $dst = add($a, $b)",153 []> {154 let BaseOpcode = "ADD";155 let PredSense = "true";156 }157 158 def ADD_Pf : PredRel, ALU32_rr<(outs IntRegs:$dst),159 (ins PredRegs:$p, IntRegs:$a, IntRegs:$b),160 "if (!$p) $dst = add($a, $b)",161 []> {162 let BaseOpcode = "ADD";163 let PredSense = "false";164 }165 166Please note that all the above instructions use ``PredRel`` as a base class.167This is extremely important since TableGen uses it as a filter for selecting168instructions for ``getPredOpcode`` model. Any instruction not derived from169``PredRel`` is excluded from the analysis. ``BaseOpcode`` is another important170field. Since it's selected as a ``RowFields`` of the model, it is required171to have the same value for all 3 instructions in order to be related. Next,172``PredSense`` is used to determine their column positions by comparing its value173with ``KeyCol`` and ``ValueCols``. If an instruction sets its ``PredSense``174value to something not used in the relation model, it will not be assigned175a column in the relation table.176