The LLVM Target-Independent Code Generator

Machine instructions are created by using the BuildMI functions, located in the include/llvm/CodeGen/MachineInstrBuilder.h file. The BuildMI functions make it easy to build arbitrary machine instructions. Usage of the BuildMI functions look like this:

// Create a 'DestReg = mov 42' (rendered in X86 assembly as 'mov DestReg, 42')
// instruction.  The '1' specifies how many operands will be added.
MachineInstr *MI = BuildMI(X86::MOV32ri, 1, DestReg).addImm(42);

// Create the same instr, but insert it at the end of a basic block.
MachineBasicBlock &MBB = ...
BuildMI(MBB, X86::MOV32ri, 1, DestReg).addImm(42);

// Create the same instr, but insert it before a specified iterator point.
MachineBasicBlock::iterator MBBI = ...
BuildMI(MBB, MBBI, X86::MOV32ri, 1, DestReg).addImm(42);

// Create a 'cmp Reg, 0' instruction, no destination reg.
MI = BuildMI(X86::CMP32ri, 2).addReg(Reg).addImm(0);
// Create an 'sahf' instruction which takes no operands and stores nothing.
MI = BuildMI(X86::SAHF, 0);

// Create a self looping branch instruction.
BuildMI(MBB, X86::JNE, 1).addMBB(&MBB);

The key thing to remember with the BuildMI functions is that you have to specify the number of operands that the machine instruction will take. This allows for efficient memory allocation. You also need to specify if operands default to be uses of values, not definitions. If you need to add a definition operand (other than the optional destination register), you must explicitly mark it as such:

MI.addReg(Reg, RegState::Define);

One important issue that the code generator needs to be aware of is the presence of fixed registers. In particular, there are often places in the instruction stream where the register allocator must arrange for a particular value to be in a particular register. This can occur due to limitations of the instruction set (e.g., the X86 can only do a 32-bit divide with the EAX/EDX registers), or external factors like calling conventions. In any case, the instruction selector should emit code that copies a virtual register into or out of a physical register when needed.

For example, consider this simple LLVM example:

define i32 @test(i32 %X, i32 %Y) {
  %Z = udiv i32 %X, %Y
  ret i32 %Z
}

The X86 instruction selector produces this machine code for the div and ret (use "llc X.bc -march=x86 -print-machineinstrs" to get this):

;; Start of div
%EAX = mov %reg1024           ;; Copy X (in reg1024) into EAX
%reg1027 = sar %reg1024, 31
%EDX = mov %reg1027           ;; Sign extend X into EDX
idiv %reg1025                 ;; Divide by Y (in reg1025)
%reg1026 = mov %EAX           ;; Read the result (Z) out of EAX

;; Start of ret
%EAX = mov %reg1026           ;; 32-bit return value goes in EAX
ret

By the end of code generation, the register allocator has coalesced the registers and deleted the resultant identity moves producing the following code:

;; X is in EAX, Y is in ECX
mov %EAX, %EDX
sar %EDX, 31
idiv %ECX
ret 

This approach is extremely general (if it can handle the X86 architecture, it can handle anything!) and allows all of the target specific knowledge about the instruction stream to be isolated in the instruction selector. Note that physical registers should have a short lifetime for good code generation, and all physical registers are assumed dead on entry to and exit from basic blocks (before register allocation). Thus, if you need a value to be live across basic block boundaries, it must live in a virtual register.

Some machine instructions, like calls, clobber a large number of physical registers. Rather than adding <def,dead> operands for all of them, it is possible to use an MO_RegisterMask operand instead. The register mask operand holds a bit mask of preserved registers, and everything else is considered to be clobbered by the instruction.

MachineInstr's are initially selected in SSA-form, and are maintained in SSA-form until register allocation happens. For the most part, this is trivially simple since LLVM is already in SSA form; LLVM PHI nodes become machine code PHI nodes, and virtual registers are only allowed to have a single definition.

After register allocation, machine code is no longer in SSA-form because there are no virtual registers left in the code.

The SelectionDAG provides an abstraction for code representation in a way that is amenable to instruction selection using automatic techniques (e.g. dynamic-programming based optimal pattern matching selectors). It is also well-suited to other phases of code generation; in particular, instruction scheduling (SelectionDAG's are very close to scheduling DAGs post-selection). Additionally, the SelectionDAG provides a host representation where a large variety of very-low-level (but target-independent) optimizations may be performed; ones which require extensive information about the instructions efficiently supported by the target.

The SelectionDAG is a Directed-Acyclic-Graph whose nodes are instances of the SDNode class. The primary payload of the SDNode is its operation code (Opcode) that indicates what operation the node performs and the operands to the operation. The various operation node types are described at the top of the include/llvm/CodeGen/SelectionDAGNodes.h file.

Although most operations define a single value, each node in the graph may define multiple values. For example, a combined div/rem operation will define both the dividend and the remainder. Many other situations require multiple values as well. Each node also has some number of operands, which are edges to the node defining the used value. Because nodes may define multiple values, edges are represented by instances of the SDValue class, which is a <SDNode, unsigned> pair, indicating the node and result value being used, respectively. Each value produced by an SDNode has an associated MVT (Machine Value Type) indicating what the type of the value is.

SelectionDAGs contain two different kinds of values: those that represent data flow and those that represent control flow dependencies. Data values are simple edges with an integer or floating point value type. Control edges are represented as "chain" edges which are of type MVT::Other. These edges provide an ordering between nodes that have side effects (such as loads, stores, calls, returns, etc). All nodes that have side effects should take a token chain as input and produce a new one as output. By convention, token chain inputs are always operand #0, and chain results are always the last value produced by an operation.

A SelectionDAG has designated "Entry" and "Root" nodes. The Entry node is always a marker node with an Opcode of ISD::EntryToken. The Root node is the final side-effecting node in the token chain. For example, in a single basic block function it would be the return node.

One important concept for SelectionDAGs is the notion of a "legal" vs. "illegal" DAG. A legal DAG for a target is one that only uses supported operations and supported types. On a 32-bit PowerPC, for example, a DAG with a value of type i1, i8, i16, or i64 would be illegal, as would a DAG that uses a SREM or UREM operation. The legalize types and legalize operations phases are responsible for turning an illegal DAG into a legal DAG.

SelectionDAG-based instruction selection consists of the following steps:

1. Build initial DAG — This stage performs a simple translation from the input LLVM code to an illegal SelectionDAG.
2. Optimize SelectionDAG — This stage performs simple optimizations on the SelectionDAG to simplify it, and recognize meta instructions (like rotates and div/rem pairs) for targets that support these meta operations. This makes the resultant code more efficient and the select instructions from DAG phase (below) simpler.
3. Legalize SelectionDAG Types — This stage transforms SelectionDAG nodes to eliminate any types that are unsupported on the target.
4. Optimize SelectionDAG — The SelectionDAG optimizer is run to clean up redundancies exposed by type legalization.
5. Legalize SelectionDAG Ops — This stage transforms SelectionDAG nodes to eliminate any operations that are unsupported on the target.
6. Optimize SelectionDAG — The SelectionDAG optimizer is run to eliminate inefficiencies introduced by operation legalization.
7. Select instructions from DAG — Finally, the target instruction selector matches the DAG operations to target instructions. This process translates the target-independent input DAG into another DAG of target instructions.
8. SelectionDAG Scheduling and Formation — The last phase assigns a linear order to the instructions in the target-instruction DAG and emits them into the MachineFunction being compiled. This step uses traditional prepass scheduling techniques.

After all of these steps are complete, the SelectionDAG is destroyed and the rest of the code generation passes are run.

One great way to visualize what is going on here is to take advantage of a few LLC command line options. The following options pop up a window displaying the SelectionDAG at specific times (if you only get errors printed to the console while using this, you probably need to configure your system to add support for it).

• -view-dag-combine1-dags displays the DAG after being built, before the first optimization pass.
• -view-legalize-dags displays the DAG before Legalization.
• -view-dag-combine2-dags displays the DAG before the second optimization pass.
• -view-isel-dags displays the DAG before the Select phase.
• -view-sched-dags displays the DAG before Scheduling.

The -view-sunit-dags displays the Scheduler's dependency graph. This graph is based on the final SelectionDAG, with nodes that must be scheduled together bundled into a single scheduling-unit node, and with immediate operands and other nodes that aren't relevant for scheduling omitted.

The initial SelectionDAG is naïvely peephole expanded from the LLVM input by the SelectionDAGLowering class in the lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp file. The intent of this pass is to expose as much low-level, target-specific details to the SelectionDAG as possible. This pass is mostly hard-coded (e.g. an LLVM add turns into an SDNode add while a getelementptr is expanded into the obvious arithmetic). This pass requires target-specific hooks to lower calls, returns, varargs, etc. For these features, the TargetLowering interface is used.

The Legalize phase is in charge of converting a DAG to only use the types that are natively supported by the target.

There are two main ways of converting values of unsupported scalar types to values of supported types: converting small types to larger types ("promoting"), and breaking up large integer types into smaller ones ("expanding"). For example, a target might require that all f32 values are promoted to f64 and that all i1/i8/i16 values are promoted to i32. The same target might require that all i64 values be expanded into pairs of i32 values. These changes can insert sign and zero extensions as needed to make sure that the final code has the same behavior as the input.

There are two main ways of converting values of unsupported vector types to value of supported types: splitting vector types, multiple times if necessary, until a legal type is found, and extending vector types by adding elements to the end to round them out to legal types ("widening"). If a vector gets split all the way down to single-element parts with no supported vector type being found, the elements are converted to scalars ("scalarizing").

A target implementation tells the legalizer which types are supported (and which register class to use for them) by calling the addRegisterClass method in its TargetLowering constructor.

The Legalize phase is in charge of converting a DAG to only use the operations that are natively supported by the target.

Targets often have weird constraints, such as not supporting every operation on every supported datatype (e.g. X86 does not support byte conditional moves and PowerPC does not support sign-extending loads from a 16-bit memory location). Legalize takes care of this by open-coding another sequence of operations to emulate the operation ("expansion"), by promoting one type to a larger type that supports the operation ("promotion"), or by using a target-specific hook to implement the legalization ("custom").

A target implementation tells the legalizer which operations are not supported (and which of the above three actions to take) by calling the setOperationAction method in its TargetLowering constructor.

Prior to the existence of the Legalize passes, we required that every target selector supported and handled every operator and type even if they are not natively supported. The introduction of the Legalize phases allows all of the canonicalization patterns to be shared across targets, and makes it very easy to optimize the canonicalized code because it is still in the form of a DAG.

The SelectionDAG optimization phase is run multiple times for code generation, immediately after the DAG is built and once after each legalization. The first run of the pass allows the initial code to be cleaned up (e.g. performing optimizations that depend on knowing that the operators have restricted type inputs). Subsequent runs of the pass clean up the messy code generated by the Legalize passes, which allows Legalize to be very simple (it can focus on making code legal instead of focusing on generating good and legal code).

One important class of optimizations performed is optimizing inserted sign and zero extension instructions. We currently use ad-hoc techniques, but could move to more rigorous techniques in the future. Here are some good papers on the subject:

"Widening integer arithmetic"
Kevin Redwine and Norman Ramsey
International Conference on Compiler Construction (CC) 2004

"Effective sign extension elimination"
Motohiro Kawahito, Hideaki Komatsu, and Toshio Nakatani
Proceedings of the ACM SIGPLAN 2002 Conference on Programming Language Design and Implementation.

The Select phase is the bulk of the target-specific code for instruction selection. This phase takes a legal SelectionDAG as input, pattern matches the instructions supported by the target to this DAG, and produces a new DAG of target code. For example, consider the following LLVM fragment:

%t1 = fadd float %W, %X
%t2 = fmul float %t1, %Y
%t3 = fadd float %t2, %Z

This LLVM code corresponds to a SelectionDAG that looks basically like this:

(fadd:f32 (fmul:f32 (fadd:f32 W, X), Y), Z)

If a target supports floating point multiply-and-add (FMA) operations, one of the adds can be merged with the multiply. On the PowerPC, for example, the output of the instruction selector might look like this DAG:

(FMADDS (FADDS W, X), Y, Z)

The FMADDS instruction is a ternary instruction that multiplies its first two operands and adds the third (as single-precision floating-point numbers). The FADDS instruction is a simple binary single-precision add instruction. To perform this pattern match, the PowerPC backend includes the following instruction definitions:

def FMADDS : AForm_1<59, 29,
                    (ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB),
                    "fmadds $FRT, $FRA, $FRC, $FRB",
                    [(set F4RC:$FRT, (fadd (fmul F4RC:$FRA, F4RC:$FRC),
                                           F4RC:$FRB))]>;
def FADDS : AForm_2<59, 21,
                    (ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB),
                    "fadds $FRT, $FRA, $FRB",
                    [(set F4RC:$FRT, (fadd F4RC:$FRA, F4RC:$FRB))]>;

The portion of the instruction definition in bold indicates the pattern used to match the instruction. The DAG operators (like fmul/fadd) are defined in the include/llvm/Target/TargetSelectionDAG.td file. " F4RC" is the register class of the input and result values.

The TableGen DAG instruction selector generator reads the instruction patterns in the .td file and automatically builds parts of the pattern matching code for your target. It has the following strengths:

• At compiler-compiler time, it analyzes your instruction patterns and tells you if your patterns make sense or not.
• It can handle arbitrary constraints on operands for the pattern match. In particular, it is straight-forward to say things like "match any immediate that is a 13-bit sign-extended value". For examples, see the immSExt16 and related tblgen classes in the PowerPC backend.
• It knows several important identities for the patterns defined. For example, it knows that addition is commutative, so it allows the FMADDS pattern above to match "(fadd X, (fmul Y, Z))" as well as "(fadd (fmul X, Y), Z)", without the target author having to specially handle this case.
• It has a full-featured type-inferencing system. In particular, you should rarely have to explicitly tell the system what type parts of your patterns are. In the FMADDS case above, we didn't have to tell tblgen that all of the nodes in the pattern are of type 'f32'. It was able to infer and propagate this knowledge from the fact that F4RC has type 'f32'.
• Targets can define their own (and rely on built-in) "pattern fragments". Pattern fragments are chunks of reusable patterns that get inlined into your patterns during compiler-compiler time. For example, the integer "(not x)" operation is actually defined as a pattern fragment that expands as "(xor x, -1)", since the SelectionDAG does not have a native 'not' operation. Targets can define their own short-hand fragments as they see fit. See the definition of 'not' and 'ineg' for examples.
• In addition to instructions, targets can specify arbitrary patterns that map to one or more instructions using the 'Pat' class. For example, the PowerPC has no way to load an arbitrary integer immediate into a register in one instruction. To tell tblgen how to do this, it defines:
  
  

  // Arbitrary immediate support.  Implement in terms of LIS/ORI.
  def : Pat<(i32 imm:$imm),
            (ORI (LIS (HI16 imm:$imm)), (LO16 imm:$imm))>;
  

  
  If none of the single-instruction patterns for loading an immediate into a register match, this will be used. This rule says "match an arbitrary i32 immediate, turning it into an ORI ('or a 16-bit immediate') and an LIS ('load 16-bit immediate, where the immediate is shifted to the left 16 bits') instruction". To make this work, the LO16/HI16 node transformations are used to manipulate the input immediate (in this case, take the high or low 16-bits of the immediate).
• While the system does automate a lot, it still allows you to write custom C++ code to match special cases if there is something that is hard to express.

While it has many strengths, the system currently has some limitations, primarily because it is a work in progress and is not yet finished:

• Overall, there is no way to define or match SelectionDAG nodes that define multiple values (e.g. SMUL_LOHI, LOAD, CALL, etc). This is the biggest reason that you currently still have to write custom C++ code for your instruction selector.
• There is no great way to support matching complex addressing modes yet. In the future, we will extend pattern fragments to allow them to define multiple values (e.g. the four operands of the X86 addressing mode, which are currently matched with custom C++ code). In addition, we'll extend fragments so that a fragment can match multiple different patterns.
• We don't automatically infer flags like isStore/isLoad yet.
• We don't automatically generate the set of supported registers and operations for the Legalizer yet.
• We don't have a way of tying in custom legalized nodes yet.

Despite these limitations, the instruction selector generator is still quite useful for most of the binary and logical operations in typical instruction sets. If you run into any problems or can't figure out how to do something, please let Chris know!

The scheduling phase takes the DAG of target instructions from the selection phase and assigns an order. The scheduler can pick an order depending on various constraints of the machines (i.e. order for minimal register pressure or try to cover instruction latencies). Once an order is established, the DAG is converted to a list of MachineInstrs and the SelectionDAG is destroyed.

Note that this phase is logically separate from the instruction selection phase, but is tied to it closely in the code because it operates on SelectionDAGs.

1. Optional function-at-a-time selection.
2. Auto-generate entire selector from .td file.

The first step in determining the live intervals of variables is to calculate the set of registers that are immediately dead after the instruction (i.e., the instruction calculates the value, but it is never used) and the set of registers that are used by the instruction, but are never used after the instruction (i.e., they are killed). Live variable information is computed for each virtual register and register allocatable physical register in the function. This is done in a very efficient manner because it uses SSA to sparsely compute lifetime information for virtual registers (which are in SSA form) and only has to track physical registers within a block. Before register allocation, LLVM can assume that physical registers are only live within a single basic block. This allows it to do a single, local analysis to resolve physical register lifetimes within each basic block. If a physical register is not register allocatable (e.g., a stack pointer or condition codes), it is not tracked.

Physical registers may be live in to or out of a function. Live in values are typically arguments in registers. Live out values are typically return values in registers. Live in values are marked as such, and are given a dummy "defining" instruction during live intervals analysis. If the last basic block of a function is a return, then it's marked as using all live out values in the function.

PHI nodes need to be handled specially, because the calculation of the live variable information from a depth first traversal of the CFG of the function won't guarantee that a virtual register used by the PHI node is defined before it's used. When a PHI node is encountered, only the definition is handled, because the uses will be handled in other basic blocks.

For each PHI node of the current basic block, we simulate an assignment at the end of the current basic block and traverse the successor basic blocks. If a successor basic block has a PHI node and one of the PHI node's operands is coming from the current basic block, then the variable is marked as alive within the current basic block and all of its predecessor basic blocks, until the basic block with the defining instruction is encountered.

We now have the information available to perform the live intervals analysis and build the live intervals themselves. We start off by numbering the basic blocks and machine instructions. We then handle the "live-in" values. These are in physical registers, so the physical register is assumed to be killed by the end of the basic block. Live intervals for virtual registers are computed for some ordering of the machine instructions [1, N]. A live interval is an interval [i, j), where 1 <= i <= j < N, for which a variable is live.

More to come...

In LLVM, physical registers are denoted by integer numbers that normally range from 1 to 1023. To see how this numbering is defined for a particular architecture, you can read the GenRegisterNames.inc file for that architecture. For instance, by inspecting lib/Target/X86/X86GenRegisterNames.inc we see that the 32-bit register EAX is denoted by 15, and the MMX register MM0 is mapped to 48.

Some architectures contain registers that share the same physical location. A notable example is the X86 platform. For instance, in the X86 architecture, the registers EAX, AX and AL share the first eight bits. These physical registers are marked as aliased in LLVM. Given a particular architecture, you can check which registers are aliased by inspecting its RegisterInfo.td file. Moreover, the method TargetRegisterInfo::getAliasSet(p_reg) returns an array containing all the physical registers aliased to the register p_reg.

Physical registers, in LLVM, are grouped in Register Classes. Elements in the same register class are functionally equivalent, and can be interchangeably used. Each virtual register can only be mapped to physical registers of a particular class. For instance, in the X86 architecture, some virtuals can only be allocated to 8 bit registers. A register class is described by TargetRegisterClass objects. To discover if a virtual register is compatible with a given physical, this code can be used:

bool RegMapping_Fer::compatible_class(MachineFunction &mf,
                                      unsigned v_reg,
                                      unsigned p_reg) {
  assert(TargetRegisterInfo::isPhysicalRegister(p_reg) &&
         "Target register must be physical");
  const TargetRegisterClass *trc = mf.getRegInfo().getRegClass(v_reg);
  return trc->contains(p_reg);
}

Sometimes, mostly for debugging purposes, it is useful to change the number of physical registers available in the target architecture. This must be done statically, inside the TargetRegsterInfo.td file. Just grep for RegisterClass, the last parameter of which is a list of registers. Just commenting some out is one simple way to avoid them being used. A more polite way is to explicitly exclude some registers from the allocation order. See the definition of the GR8 register class in lib/Target/X86/X86RegisterInfo.td for an example of this.

Virtual registers are also denoted by integer numbers. Contrary to physical registers, different virtual registers never share the same number. Whereas physical registers are statically defined in a TargetRegisterInfo.td file and cannot be created by the application developer, that is not the case with virtual registers. In order to create new virtual registers, use the method MachineRegisterInfo::createVirtualRegister(). This method will return a new virtual register. Use an IndexedMap<Foo, VirtReg2IndexFunctor> to hold information per virtual register. If you need to enumerate all virtual registers, use the function TargetRegisterInfo::index2VirtReg() to find the virtual register numbers:

  for (unsigned i = 0, e = MRI->getNumVirtRegs(); i != e; ++i) {
    unsigned VirtReg = TargetRegisterInfo::index2VirtReg(i);
    stuff(VirtReg);
  }

Before register allocation, the operands of an instruction are mostly virtual registers, although physical registers may also be used. In order to check if a given machine operand is a register, use the boolean function MachineOperand::isRegister(). To obtain the integer code of a register, use MachineOperand::getReg(). An instruction may define or use a register. For instance, ADD reg:1026 := reg:1025 reg:1024 defines the registers 1024, and uses registers 1025 and 1026. Given a register operand, the method MachineOperand::isUse() informs if that register is being used by the instruction. The method MachineOperand::isDef() informs if that registers is being defined.

We will call physical registers present in the LLVM bitcode before register allocation pre-colored registers. Pre-colored registers are used in many different situations, for instance, to pass parameters of functions calls, and to store results of particular instructions. There are two types of pre-colored registers: the ones implicitly defined, and those explicitly defined. Explicitly defined registers are normal operands, and can be accessed with MachineInstr::getOperand(int)::getReg(). In order to check which registers are implicitly defined by an instruction, use the TargetInstrInfo::get(opcode)::ImplicitDefs, where opcode is the opcode of the target instruction. One important difference between explicit and implicit physical registers is that the latter are defined statically for each instruction, whereas the former may vary depending on the program being compiled. For example, an instruction that represents a function call will always implicitly define or use the same set of physical registers. To read the registers implicitly used by an instruction, use TargetInstrInfo::get(opcode)::ImplicitUses. Pre-colored registers impose constraints on any register allocation algorithm. The register allocator must make sure that none of them are overwritten by the values of virtual registers while still alive.

There are two ways to map virtual registers to physical registers (or to memory slots). The first way, that we will call direct mapping, is based on the use of methods of the classes TargetRegisterInfo, and MachineOperand. The second way, that we will call indirect mapping, relies on the VirtRegMap class in order to insert loads and stores sending and getting values to and from memory.

The direct mapping provides more flexibility to the developer of the register allocator; however, it is more error prone, and demands more implementation work. Basically, the programmer will have to specify where load and store instructions should be inserted in the target function being compiled in order to get and store values in memory. To assign a physical register to a virtual register present in a given operand, use MachineOperand::setReg(p_reg). To insert a store instruction, use TargetInstrInfo::storeRegToStackSlot(...), and to insert a load instruction, use TargetInstrInfo::loadRegFromStackSlot.

The indirect mapping shields the application developer from the complexities of inserting load and store instructions. In order to map a virtual register to a physical one, use VirtRegMap::assignVirt2Phys(vreg, preg). In order to map a certain virtual register to memory, use VirtRegMap::assignVirt2StackSlot(vreg). This method will return the stack slot where vreg's value will be located. If it is necessary to map another virtual register to the same stack slot, use VirtRegMap::assignVirt2StackSlot(vreg, stack_location). One important point to consider when using the indirect mapping, is that even if a virtual register is mapped to memory, it still needs to be mapped to a physical register. This physical register is the location where the virtual register is supposed to be found before being stored or after being reloaded.

If the indirect strategy is used, after all the virtual registers have been mapped to physical registers or stack slots, it is necessary to use a spiller object to place load and store instructions in the code. Every virtual that has been mapped to a stack slot will be stored to memory after been defined and will be loaded before being used. The implementation of the spiller tries to recycle load/store instructions, avoiding unnecessary instructions. For an example of how to invoke the spiller, see RegAllocLinearScan::runOnMachineFunction in lib/CodeGen/RegAllocLinearScan.cpp.

With very rare exceptions (e.g., function calls), the LLVM machine code instructions are three address instructions. That is, each instruction is expected to define at most one register, and to use at most two registers. However, some architectures use two address instructions. In this case, the defined register is also one of the used register. For instance, an instruction such as ADD %EAX, %EBX, in X86 is actually equivalent to %EAX = %EAX + %EBX.

In order to produce correct code, LLVM must convert three address instructions that represent two address instructions into true two address instructions. LLVM provides the pass TwoAddressInstructionPass for this specific purpose. It must be run before register allocation takes place. After its execution, the resulting code may no longer be in SSA form. This happens, for instance, in situations where an instruction such as %a = ADD %b %c is converted to two instructions such as:

%a = MOVE %b
%a = ADD %a %c

Notice that, internally, the second instruction is represented as ADD %a[def/use] %c. I.e., the register operand %a is both used and defined by the instruction.

An important transformation that happens during register allocation is called the SSA Deconstruction Phase. The SSA form simplifies many analyses that are performed on the control flow graph of programs. However, traditional instruction sets do not implement PHI instructions. Thus, in order to generate executable code, compilers must replace PHI instructions with other instructions that preserve their semantics.

There are many ways in which PHI instructions can safely be removed from the target code. The most traditional PHI deconstruction algorithm replaces PHI instructions with copy instructions. That is the strategy adopted by LLVM. The SSA deconstruction algorithm is implemented in lib/CodeGen/PHIElimination.cpp. In order to invoke this pass, the identifier PHIEliminationID must be marked as required in the code of the register allocator.

Instruction folding is an optimization performed during register allocation that removes unnecessary copy instructions. For instance, a sequence of instructions such as:

%EBX = LOAD %mem_address
%EAX = COPY %EBX

can be safely substituted by the single instruction:

%EAX = LOAD %mem_address

Instructions can be folded with the TargetRegisterInfo::foldMemoryOperand(...) method. Care must be taken when folding instructions; a folded instruction can be quite different from the original instruction. See LiveIntervals::addIntervalsForSpills in lib/CodeGen/LiveIntervalAnalysis.cpp for an example of its use.

The LLVM infrastructure provides the application developer with three different register allocators:

• Fast — This register allocator is the default for debug builds. It allocates registers on a basic block level, attempting to keep values in registers and reusing registers as appropriate.
• Basic — This is an incremental approach to register allocation. Live ranges are assigned to registers one at a time in an order that is driven by heuristics. Since code can be rewritten on-the-fly during allocation, this framework allows interesting allocators to be developed as extensions. It is not itself a production register allocator but is a potentially useful stand-alone mode for triaging bugs and as a performance baseline.
• Greedy — The default allocator. This is a highly tuned implementation of the Basic allocator that incorporates global live range splitting. This allocator works hard to minimize the cost of spill code.
• PBQP — A Partitioned Boolean Quadratic Programming (PBQP) based register allocator. This allocator works by constructing a PBQP problem representing the register allocation problem under consideration, solving this using a PBQP solver, and mapping the solution back to a register assignment.

The type of register allocator used in llc can be chosen with the command line option -regalloc=...:

$ llc -regalloc=linearscan file.bc -o ln.s;
$ llc -regalloc=fast file.bc -o fa.s;
$ llc -regalloc=pbqp file.bc -o pbqp.s;

Throwing an exception requires unwinding out of a function. The information on how to unwind a given function is traditionally expressed in DWARF unwind (a.k.a. frame) info. But that format was originally developed for debuggers to backtrace, and each Frame Description Entry (FDE) requires ~20-30 bytes per function. There is also the cost of mapping from an address in a function to the corresponding FDE at runtime. An alternative unwind encoding is called compact unwind and requires just 4-bytes per function.

The compact unwind encoding is a 32-bit value, which is encoded in an architecture-specific way. It specifies which registers to restore and from where, and how to unwind out of the function. When the linker creates a final linked image, it will create a __TEXT,__unwind_info section. This section is a small and fast way for the runtime to access unwind info for any given function. If we emit compact unwind info for the function, that compact unwind info will be encoded in the __TEXT,__unwind_info section. If we emit DWARF unwind info, the __TEXT,__unwind_info section will contain the offset of the FDE in the __TEXT,__eh_frame section in the final linked image.

For X86, there are three modes for the compact unwind encoding:

Function with a Frame Pointer (EBP or RBP)

EBP/RBP-based frame, where EBP/RBP is pushed onto the stack immediately after the return address, then ESP/RSP is moved to EBP/RBP. Thus to unwind, ESP/RSP is restored with the current EBP/RBP value, then EBP/RBP is restored by popping the stack, and the return is done by popping the stack once more into the PC. All non-volatile registers that need to be restored must have been saved in a small range on the stack that starts EBP-4 to EBP-1020 (RBP-8 to RBP-1020). The offset (divided by 4 in 32-bit mode and 8 in 64-bit mode) is encoded in bits 16-23 (mask: 0x00FF0000). The registers saved are encoded in bits 0-14 (mask: 0x00007FFF) as five 3-bit entries from the following table:

Compact Number	i386 Register	x86-64 Regiser
1	EBX	RBX
2	ECX	R12
3	EDX	R13
4	EDI	R14
5	ESI	R15
6	EBP	RBP

Frameless with a Small Constant Stack Size (EBP or RBP is not used as a frame pointer)

To return, a constant (encoded in the compact unwind encoding) is added to the ESP/RSP. Then the return is done by popping the stack into the PC. All non-volatile registers that need to be restored must have been saved on the stack immediately after the return address. The stack size (divided by 4 in 32-bit mode and 8 in 64-bit mode) is encoded in bits 16-23 (mask: 0x00FF0000). There is a maximum stack size of 1024 bytes in 32-bit mode and 2048 in 64-bit mode. The number of registers saved is encoded in bits 9-12 (mask: 0x00001C00). Bits 0-9 (mask: 0x000003FF) contain which registers were saved and their order. (See the encodeCompactUnwindRegistersWithoutFrame() function in lib/Target/X86FrameLowering.cpp for the encoding algorithm.)

Frameless with a Large Constant Stack Size (EBP or RBP is not used as a frame pointer)

This case is like the "Frameless with a Small Constant Stack Size" case, but the stack size is too large to encode in the compact unwind encoding. Instead it requires that the function contains "subl $nnnnnn, %esp" in its prolog. The compact encoding contains the offset to the $nnnnnn value in the function in bits 9-12 (mask: 0x00001C00).

Instructions in a VLIW target can typically be mapped to multiple functional units. During the process of packetizing, the compiler must be able to reason about whether an instruction can be added to a packet. This decision can be complex since the compiler has to examine all possible mappings of instructions to functional units. Therefore to alleviate compilation-time complexity, the VLIW packetizer parses the instruction classes of a target and generates tables at compiler build time. These tables can then be queried by the provided machine-independent API to determine if an instruction can be accommodated in a packet.

The packetizer reads instruction classes from a target's itineraries and creates a deterministic finite automaton (DFA) to represent the state of a packet. A DFA consists of three major elements: inputs, states, and transitions. The set of inputs for the generated DFA represents the instruction being added to a packet. The states represent the possible consumption of functional units by instructions in a packet. In the DFA, transitions from one state to another occur on the addition of an instruction to an existing packet. If there is a legal mapping of functional units to instructions, then the DFA contains a corresponding transition. The absence of a transition indicates that a legal mapping does not exist and that the instruction cannot be added to the packet.

To generate tables for a VLIW target, add TargetGenDFAPacketizer.inc as a target to the Makefile in the target directory. The exported API provides three functions: DFAPacketizer::clearResources(), DFAPacketizer::reserveResources(MachineInstr *MI), and DFAPacketizer::canReserveResources(MachineInstr *MI). These functions allow a target packetizer to add an instruction to an existing packet and to check whether an instruction can be added to a packet. See llvm/CodeGen/DFAPacketizer.h for more information.

The first phase of alias processing is simple instruction mnemonic remapping for classes of instructions which are allowed with two different mnemonics. This phase is a simple and unconditionally remapping from one input mnemonic to one output mnemonic. It isn't possible for this form of alias to look at the operands at all, so the remapping must apply for all forms of a given mnemonic. Mnemonic aliases are defined simply, for example X86 has:

def : MnemonicAlias<"cbw",     "cbtw">;
def : MnemonicAlias<"smovq",   "movsq">;
def : MnemonicAlias<"fldcww",  "fldcw">;
def : MnemonicAlias<"fucompi", "fucomip">;
def : MnemonicAlias<"ud2a",    "ud2">;

... and many others. With a MnemonicAlias definition, the mnemonic is remapped simply and directly. Though MnemonicAlias's can't look at any aspect of the instruction (such as the operands) they can depend on global modes (the same ones supported by the matcher), through a Requires clause:

def : MnemonicAlias<"pushf", "pushfq">, Requires<[In64BitMode]>;
def : MnemonicAlias<"pushf", "pushfl">, Requires<[In32BitMode]>;

In this example, the mnemonic gets mapped into different a new one depending on the current instruction set.

The most general phase of alias processing occurs while matching is happening: it provides new forms for the matcher to match along with a specific instruction to generate. An instruction alias has two parts: the string to match and the instruction to generate. For example:

def : InstAlias<"movsx $src, $dst", (MOVSX16rr8W GR16:$dst, GR8  :$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX16rm8W GR16:$dst, i8mem:$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX32rr8  GR32:$dst, GR8  :$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX32rr16 GR32:$dst, GR16 :$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX64rr8  GR64:$dst, GR8  :$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX64rr16 GR64:$dst, GR16 :$src)>;
def : InstAlias<"movsx $src, $dst", (MOVSX64rr32 GR64:$dst, GR32 :$src)>;

This shows a powerful example of the instruction aliases, matching the same mnemonic in multiple different ways depending on what operands are present in the assembly. The result of instruction aliases can include operands in a different order than the destination instruction, and can use an input multiple times, for example:

def : InstAlias<"clrb $reg", (XOR8rr  GR8 :$reg, GR8 :$reg)>;
def : InstAlias<"clrw $reg", (XOR16rr GR16:$reg, GR16:$reg)>;
def : InstAlias<"clrl $reg", (XOR32rr GR32:$reg, GR32:$reg)>;
def : InstAlias<"clrq $reg", (XOR64rr GR64:$reg, GR64:$reg)>;

This example also shows that tied operands are only listed once. In the X86 backend, XOR8rr has two input GR8's and one output GR8 (where an input is tied to the output). InstAliases take a flattened operand list without duplicates for tied operands. The result of an instruction alias can also use immediates and fixed physical registers which are added as simple immediate operands in the result, for example:

// Fixed Immediate operand.
def : InstAlias<"aad", (AAD8i8 10)>;

// Fixed register operand.
def : InstAlias<"fcomi", (COM_FIr ST1)>;

// Simple alias.
def : InstAlias<"fcomi $reg", (COM_FIr RST:$reg)>;

Instruction aliases can also have a Requires clause to make them subtarget specific.

If the back-end supports it, the instruction printer can automatically emit the alias rather than what's being aliased. It typically leads to better, more readable code. If it's better to print out what's being aliased, then pass a '0' as the third parameter to the InstAlias definition.

This box indicates whether the target is considered to be production quality. This indicates that the target has been used as a static compiler to compile large amounts of code by a variety of different people and is in continuous use.

This box indicates whether the target supports parsing target specific .s files by implementing the MCAsmParser interface. This is required for llvm-mc to be able to act as a native assembler and is required for inline assembly support in the native .o file writer.

This box indicates whether the target supports the MCDisassembler API for disassembling machine opcode bytes into MCInst's.

This box indicates whether the target supports most popular inline assembly constraints and modifiers.

This box indicates whether the target supports the JIT compiler through the ExecutionEngine interface.

The ARM backend has basic support for integer code in ARM codegen mode, but lacks NEON and full Thumb support.

This box indicates whether the target supports writing .o files (e.g. MachO, ELF, and/or COFF) files directly from the target. Note that the target also must include an assembly parser and general inline assembly support for full inline assembly support in the .o writer.

Targets that don't support this feature can obviously still write out .o files, they just rely on having an external assembler to translate from a .s file to a .o file (as is the case for many C compilers).

This box indicates whether the target supports guaranteed tail calls. These are calls marked "tail" and use the fastcc calling convention. Please see the tail call section more more details.

This box indicates whether the target supports segmented stacks. This replaces the traditional large C stack with many linked segments. It is compatible with the gcc implementation used by the Go front end.

Basic support exists on the X86 backend. Currently vararg doesn't work and the object files are not marked the way the gold linker expects, but simple Go programs can be built by dragonegg.

The following are the known target triples that are supported by the X86 backend. This is not an exhaustive list, and it would be useful to add those that people test.

• i686-pc-linux-gnu — Linux
• i386-unknown-freebsd5.3 — FreeBSD 5.3
• i686-pc-cygwin — Cygwin on Win32
• i686-pc-mingw32 — MingW on Win32
• i386-pc-mingw32msvc — MingW crosscompiler on Linux
• i686-apple-darwin* — Apple Darwin on X86
• x86_64-unknown-linux-gnu — Linux

The following target-specific calling conventions are known to backend:

• x86_StdCall — stdcall calling convention seen on Microsoft Windows platform (CC ID = 64).
• x86_FastCall — fastcall calling convention seen on Microsoft Windows platform (CC ID = 65).
• x86_ThisCall — Similar to X86_StdCall. Passes first argument in ECX, others via stack. Callee is responsible for stack cleaning. This convention is used by MSVC by default for methods in its ABI (CC ID = 70).

The x86 has a very flexible way of accessing memory. It is capable of forming memory addresses of the following expression directly in integer instructions (which use ModR/M addressing):

SegmentReg: Base + [1,2,4,8] * IndexReg + Disp32

In order to represent this, LLVM tracks no less than 5 operands for each memory operand of this form. This means that the "load" form of 'mov' has the following MachineOperands in this order:

Index:        0     |    1        2       3           4          5
Meaning:   DestReg, | BaseReg,  Scale, IndexReg, Displacement Segment
OperandTy: VirtReg, | VirtReg, UnsImm, VirtReg,   SignExtImm  PhysReg

Stores, and all other instructions, treat the four memory operands in the same way and in the same order. If the segment register is unspecified (regno = 0), then no segment override is generated. "Lea" operations do not have a segment register specified, so they only have 4 operands for their memory reference.

x86 has a feature which provides the ability to perform loads and stores to different address spaces via the x86 segment registers. A segment override prefix byte on an instruction causes the instruction's memory access to go to the specified segment. LLVM address space 0 is the default address space, which includes the stack, and any unqualified memory accesses in a program. Address spaces 1-255 are currently reserved for user-defined code. The GS-segment is represented by address space 256, while the FS-segment is represented by address space 257. Other x86 segments have yet to be allocated address space numbers.

While these address spaces may seem similar to TLS via the thread_local keyword, and often use the same underlying hardware, there are some fundamental differences.

The thread_local keyword applies to global variables and specifies that they are to be allocated in thread-local memory. There are no type qualifiers involved, and these variables can be pointed to with normal pointers and accessed with normal loads and stores. The thread_local keyword is target-independent at the LLVM IR level (though LLVM doesn't yet have implementations of it for some configurations).

Special address spaces, in contrast, apply to static types. Every load and store has a particular address space in its address operand type, and this is what determines which address space is accessed. LLVM ignores these special address space qualifiers on global variables, and does not provide a way to directly allocate storage in them. At the LLVM IR level, the behavior of these special address spaces depends in part on the underlying OS or runtime environment, and they are specific to x86 (and LLVM doesn't yet handle them correctly in some cases).

Some operating systems and runtime environments use (or may in the future use) the FS/GS-segment registers for various low-level purposes, so care should be taken when considering them.

An instruction name consists of the base name, a default operand size, and a a character per operand with an optional special size. For example:

ADD8rr      -> add, 8-bit register, 8-bit register
IMUL16rmi   -> imul, 16-bit register, 16-bit memory, 16-bit immediate
IMUL16rmi8  -> imul, 16-bit register, 16-bit memory, 8-bit immediate
MOVSX32rm16 -> movsx, 32-bit register, 16-bit memory

LLVM follows the AIX PowerPC ABI, with two deviations. LLVM uses a PC relative (PIC) or static addressing for accessing global values, so no TOC (r2) is used. Second, r31 is used as a frame pointer to allow dynamic growth of a stack frame. LLVM takes advantage of having no TOC to provide space to save the frame pointer in the PowerPC linkage area of the caller frame. Other details of PowerPC ABI can be found at PowerPC ABI. Note: This link describes the 32 bit ABI. The 64 bit ABI is similar except space for GPRs are 8 bytes wide (not 4) and r13 is reserved for system use.

The size of a PowerPC frame is usually fixed for the duration of a function's invocation. Since the frame is fixed size, all references into the frame can be accessed via fixed offsets from the stack pointer. The exception to this is when dynamic alloca or variable sized arrays are present, then a base pointer (r31) is used as a proxy for the stack pointer and stack pointer is free to grow or shrink. A base pointer is also used if llvm-gcc is not passed the -fomit-frame-pointer flag. The stack pointer is always aligned to 16 bytes, so that space allocated for altivec vectors will be properly aligned.

An invocation frame is laid out as follows (low memory at top);

The linkage area is used by a callee to save special registers prior to allocating its own frame. Only three entries are relevant to LLVM. The first entry is the previous stack pointer (sp), aka link. This allows probing tools like gdb or exception handlers to quickly scan the frames in the stack. A function epilog can also use the link to pop the frame from the stack. The third entry in the linkage area is used to save the return address from the lr register. Finally, as mentioned above, the last entry is used to save the previous frame pointer (r31.) The entries in the linkage area are the size of a GPR, thus the linkage area is 24 bytes long in 32 bit mode and 48 bytes in 64 bit mode.

32 bit linkage area

64 bit linkage area

The parameter area is used to store arguments being passed to a callee function. Following the PowerPC ABI, the first few arguments are actually passed in registers, with the space in the parameter area unused. However, if there are not enough registers or the callee is a thunk or vararg function, these register arguments can be spilled into the parameter area. Thus, the parameter area must be large enough to store all the parameters for the largest call sequence made by the caller. The size must also be minimally large enough to spill registers r3-r10. This allows callees blind to the call signature, such as thunks and vararg functions, enough space to cache the argument registers. Therefore, the parameter area is minimally 32 bytes (64 bytes in 64 bit mode.) Also note that since the parameter area is a fixed offset from the top of the frame, that a callee can access its spilt arguments using fixed offsets from the stack pointer (or base pointer.)

Combining the information about the linkage, parameter areas and alignment. A stack frame is minimally 64 bytes in 32 bit mode and 128 bytes in 64 bit mode.

The dynamic area starts out as size zero. If a function uses dynamic alloca then space is added to the stack, the linkage and parameter areas are shifted to top of stack, and the new space is available immediately below the linkage and parameter areas. The cost of shifting the linkage and parameter areas is minor since only the link value needs to be copied. The link value can be easily fetched by adding the original frame size to the base pointer. Note that allocations in the dynamic space need to observe 16 byte alignment.

The locals area is where the llvm compiler reserves space for local variables.

The saved registers area is where the llvm compiler spills callee saved registers on entry to the callee.

The llvm prolog and epilog are the same as described in the PowerPC ABI, with the following exceptions. Callee saved registers are spilled after the frame is created. This allows the llvm epilog/prolog support to be common with other targets. The base pointer callee saved register r31 is saved in the TOC slot of linkage area. This simplifies allocation of space for the base pointer and makes it convenient to locate programatically and during debugging.

TODO - More to come.