2005-11-15 00:40:23 +00:00
|
|
|
//===-- X86ISelLowering.h - X86 DAG Lowering Interface ----------*- C++ -*-===//
|
|
|
|
//
|
|
|
|
// The LLVM Compiler Infrastructure
|
|
|
|
//
|
2007-12-29 20:36:04 +00:00
|
|
|
// This file is distributed under the University of Illinois Open Source
|
|
|
|
// License. See LICENSE.TXT for details.
|
2005-11-15 00:40:23 +00:00
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// This file defines the interfaces that X86 uses to lower LLVM code into a
|
|
|
|
// selection DAG.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
#ifndef X86ISELLOWERING_H
|
|
|
|
#define X86ISELLOWERING_H
|
|
|
|
|
2006-01-27 08:10:46 +00:00
|
|
|
#include "X86Subtarget.h"
|
2007-07-14 14:06:15 +00:00
|
|
|
#include "X86RegisterInfo.h"
|
2008-01-05 16:56:59 +00:00
|
|
|
#include "X86MachineFunctionInfo.h"
|
2005-11-15 00:40:23 +00:00
|
|
|
#include "llvm/Target/TargetLowering.h"
|
2010-01-26 19:04:47 +00:00
|
|
|
#include "llvm/Target/TargetOptions.h"
|
2008-09-03 02:54:11 +00:00
|
|
|
#include "llvm/CodeGen/FastISel.h"
|
2005-11-15 00:40:23 +00:00
|
|
|
#include "llvm/CodeGen/SelectionDAG.h"
|
2007-08-31 15:06:30 +00:00
|
|
|
#include "llvm/CodeGen/CallingConvLower.h"
|
2005-11-15 00:40:23 +00:00
|
|
|
|
|
|
|
namespace llvm {
|
|
|
|
namespace X86ISD {
|
2006-01-06 00:43:03 +00:00
|
|
|
// X86 Specific DAG Nodes
|
2005-11-15 00:40:23 +00:00
|
|
|
enum NodeType {
|
|
|
|
// Start the numbering where the builtin ops leave off.
|
2008-09-23 18:42:32 +00:00
|
|
|
FIRST_NUMBER = ISD::BUILTIN_OP_END,
|
2005-11-15 00:40:23 +00:00
|
|
|
|
2007-12-14 02:13:44 +00:00
|
|
|
/// BSF - Bit scan forward.
|
|
|
|
/// BSR - Bit scan reverse.
|
|
|
|
BSF,
|
|
|
|
BSR,
|
|
|
|
|
2006-01-09 18:33:28 +00:00
|
|
|
/// SHLD, SHRD - Double shift instructions. These correspond to
|
|
|
|
/// X86::SHLDxx and X86::SHRDxx instructions.
|
|
|
|
SHLD,
|
|
|
|
SHRD,
|
|
|
|
|
2006-01-31 03:14:29 +00:00
|
|
|
/// FAND - Bitwise logical AND of floating point values. This corresponds
|
|
|
|
/// to X86::ANDPS or X86::ANDPD.
|
|
|
|
FAND,
|
|
|
|
|
2007-01-05 07:55:56 +00:00
|
|
|
/// FOR - Bitwise logical OR of floating point values. This corresponds
|
|
|
|
/// to X86::ORPS or X86::ORPD.
|
|
|
|
FOR,
|
|
|
|
|
2006-01-31 22:28:30 +00:00
|
|
|
/// FXOR - Bitwise logical XOR of floating point values. This corresponds
|
|
|
|
/// to X86::XORPS or X86::XORPD.
|
|
|
|
FXOR,
|
|
|
|
|
2007-01-05 21:37:56 +00:00
|
|
|
/// FSRL - Bitwise logical right shift of floating point values. These
|
|
|
|
/// corresponds to X86::PSRLDQ.
|
2007-01-05 07:55:56 +00:00
|
|
|
FSRL,
|
|
|
|
|
2006-02-04 02:20:30 +00:00
|
|
|
/// FILD, FILD_FLAG - This instruction implements SINT_TO_FP with the
|
|
|
|
/// integer source in memory and FP reg result. This corresponds to the
|
|
|
|
/// X86::FILD*m instructions. It has three inputs (token chain, address,
|
|
|
|
/// and source type) and two outputs (FP value and token chain). FILD_FLAG
|
|
|
|
/// also produces a flag).
|
2006-01-12 22:54:21 +00:00
|
|
|
FILD,
|
2006-02-04 02:20:30 +00:00
|
|
|
FILD_FLAG,
|
2005-11-15 00:40:23 +00:00
|
|
|
|
|
|
|
/// FP_TO_INT*_IN_MEM - This instruction implements FP_TO_SINT with the
|
|
|
|
/// integer destination in memory and a FP reg source. This corresponds
|
|
|
|
/// to the X86::FIST*m instructions and the rounding mode change stuff. It
|
2006-10-18 18:26:48 +00:00
|
|
|
/// has two inputs (token chain and address) and two outputs (int value
|
|
|
|
/// and token chain).
|
2005-11-15 00:40:23 +00:00
|
|
|
FP_TO_INT16_IN_MEM,
|
|
|
|
FP_TO_INT32_IN_MEM,
|
|
|
|
FP_TO_INT64_IN_MEM,
|
|
|
|
|
2005-12-21 02:39:21 +00:00
|
|
|
/// FLD - This instruction implements an extending load to FP stack slots.
|
|
|
|
/// This corresponds to the X86::FLD32m / X86::FLD64m. It takes a chain
|
2005-12-23 07:31:11 +00:00
|
|
|
/// operand, ptr to load from, and a ValueType node indicating the type
|
|
|
|
/// to load to.
|
2005-12-21 02:39:21 +00:00
|
|
|
FLD,
|
|
|
|
|
2006-01-05 00:27:02 +00:00
|
|
|
/// FST - This instruction implements a truncating store to FP stack
|
|
|
|
/// slots. This corresponds to the X86::FST32m / X86::FST64m. It takes a
|
|
|
|
/// chain operand, value to store, address, and a ValueType to store it
|
|
|
|
/// as.
|
|
|
|
FST,
|
|
|
|
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
/// CALL - These operations represent an abstract X86 call
|
2005-11-15 00:40:23 +00:00
|
|
|
/// instruction, which includes a bunch of information. In particular the
|
|
|
|
/// operands of these node are:
|
|
|
|
///
|
|
|
|
/// #0 - The incoming token chain
|
|
|
|
/// #1 - The callee
|
|
|
|
/// #2 - The number of arg bytes the caller pushes on the stack.
|
|
|
|
/// #3 - The number of arg bytes the callee pops off the stack.
|
|
|
|
/// #4 - The value to pass in AL/AX/EAX (optional)
|
|
|
|
/// #5 - The value to pass in DL/DX/EDX (optional)
|
|
|
|
///
|
|
|
|
/// The result values of these nodes are:
|
|
|
|
///
|
|
|
|
/// #0 - The outgoing token chain
|
|
|
|
/// #1 - The first register result value (optional)
|
|
|
|
/// #2 - The second register result value (optional)
|
|
|
|
///
|
|
|
|
CALL,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
|
2005-11-20 21:41:10 +00:00
|
|
|
/// RDTSC_DAG - This operation implements the lowering for
|
|
|
|
/// readcyclecounter
|
|
|
|
RDTSC_DAG,
|
2005-12-17 01:21:05 +00:00
|
|
|
|
|
|
|
/// X86 compare and logical compare instructions.
|
2007-09-17 17:42:53 +00:00
|
|
|
CMP, COMI, UCOMI,
|
2005-12-17 01:21:05 +00:00
|
|
|
|
2008-12-23 22:45:23 +00:00
|
|
|
/// X86 bit-test instructions.
|
|
|
|
BT,
|
|
|
|
|
2009-03-23 15:40:10 +00:00
|
|
|
/// X86 SetCC. Operand 0 is condition code, and operand 1 is the flag
|
2005-12-21 20:21:51 +00:00
|
|
|
/// operand produced by a CMP instruction.
|
|
|
|
SETCC,
|
|
|
|
|
2009-12-15 00:53:42 +00:00
|
|
|
// Same as SETCC except it's materialized with a sbb and the value is all
|
|
|
|
// one's or all zero's.
|
|
|
|
SETCC_CARRY,
|
|
|
|
|
2009-03-12 06:46:02 +00:00
|
|
|
/// X86 conditional moves. Operand 0 and operand 1 are the two values
|
|
|
|
/// to select from. Operand 2 is the condition code, and operand 3 is the
|
|
|
|
/// flag operand produced by a CMP or TEST instruction. It also writes a
|
|
|
|
/// flag result.
|
2005-12-17 01:21:05 +00:00
|
|
|
CMOV,
|
2005-12-19 23:12:38 +00:00
|
|
|
|
2009-03-23 15:40:10 +00:00
|
|
|
/// X86 conditional branches. Operand 0 is the chain operand, operand 1
|
|
|
|
/// is the block to branch if condition is true, operand 2 is the
|
|
|
|
/// condition code, and operand 3 is the flag operand produced by a CMP
|
2005-12-21 20:21:51 +00:00
|
|
|
/// or TEST instruction.
|
2005-12-19 23:12:38 +00:00
|
|
|
BRCOND,
|
2005-12-21 02:39:21 +00:00
|
|
|
|
2009-03-23 15:40:10 +00:00
|
|
|
/// Return with a flag operand. Operand 0 is the chain operand, operand
|
|
|
|
/// 1 is the number of bytes of stack to pop.
|
2005-12-21 02:39:21 +00:00
|
|
|
RET_FLAG,
|
2006-01-11 22:15:48 +00:00
|
|
|
|
|
|
|
/// REP_STOS - Repeat fill, corresponds to X86::REP_STOSx.
|
|
|
|
REP_STOS,
|
|
|
|
|
|
|
|
/// REP_MOVS - Repeat move, corresponds to X86::REP_MOVSx.
|
|
|
|
REP_MOVS,
|
2006-01-31 22:28:30 +00:00
|
|
|
|
2006-02-18 00:15:05 +00:00
|
|
|
/// GlobalBaseReg - On Darwin, this node represents the result of the popl
|
|
|
|
/// at function entry, used for PIC code.
|
|
|
|
GlobalBaseReg,
|
2006-02-23 02:43:52 +00:00
|
|
|
|
2008-09-16 21:48:12 +00:00
|
|
|
/// Wrapper - A wrapper node for TargetConstantPool,
|
|
|
|
/// TargetExternalSymbol, and TargetGlobalAddress.
|
2006-02-23 20:41:18 +00:00
|
|
|
Wrapper,
|
2006-03-21 23:01:21 +00:00
|
|
|
|
2006-11-30 21:55:46 +00:00
|
|
|
/// WrapperRIP - Special wrapper used under X86-64 PIC mode for RIP
|
|
|
|
/// relative displacements.
|
|
|
|
WrapperRIP,
|
|
|
|
|
2010-01-24 00:05:03 +00:00
|
|
|
/// MOVQ2DQ - Copies a 64-bit value from a vector to another vector.
|
|
|
|
/// Can be used to move a vector value from a MMX register to a XMM
|
|
|
|
/// register.
|
|
|
|
MOVQ2DQ,
|
|
|
|
|
2008-02-11 04:19:36 +00:00
|
|
|
/// PEXTRB - Extract an 8-bit value from a vector and zero extend it to
|
|
|
|
/// i32, corresponds to X86::PEXTRB.
|
|
|
|
PEXTRB,
|
|
|
|
|
2006-03-31 19:22:53 +00:00
|
|
|
/// PEXTRW - Extract a 16-bit value from a vector and zero extend it to
|
2006-03-31 21:55:24 +00:00
|
|
|
/// i32, corresponds to X86::PEXTRW.
|
2006-03-31 19:22:53 +00:00
|
|
|
PEXTRW,
|
2006-03-31 21:55:24 +00:00
|
|
|
|
2008-02-11 04:19:36 +00:00
|
|
|
/// INSERTPS - Insert any element of a 4 x float vector into any element
|
|
|
|
/// of a destination 4 x floatvector.
|
|
|
|
INSERTPS,
|
|
|
|
|
|
|
|
/// PINSRB - Insert the lower 8-bits of a 32-bit value to a vector,
|
|
|
|
/// corresponds to X86::PINSRB.
|
|
|
|
PINSRB,
|
|
|
|
|
2006-03-31 21:55:24 +00:00
|
|
|
/// PINSRW - Insert the lower 16-bits of a 32-bit value to a vector,
|
|
|
|
/// corresponds to X86::PINSRW.
|
2010-02-23 02:07:48 +00:00
|
|
|
PINSRW, MMX_PINSRW,
|
2006-11-10 21:43:37 +00:00
|
|
|
|
Generate better code for v8i16 shuffles on SSE2
Generate better code for v16i8 shuffles on SSE2 (avoids stack)
Generate pshufb for v8i16 and v16i8 shuffles on SSSE3 where it is fewer uops.
Document the shuffle matching logic and add some FIXMEs for later further
cleanups.
New tests that test the above.
Examples:
New:
_shuf2:
pextrw $7, %xmm0, %eax
punpcklqdq %xmm1, %xmm0
pshuflw $128, %xmm0, %xmm0
pinsrw $2, %eax, %xmm0
Old:
_shuf2:
pextrw $2, %xmm0, %eax
pextrw $7, %xmm0, %ecx
pinsrw $2, %ecx, %xmm0
pinsrw $3, %eax, %xmm0
movd %xmm1, %eax
pinsrw $4, %eax, %xmm0
ret
=========
New:
_shuf4:
punpcklqdq %xmm1, %xmm0
pshufb LCPI1_0, %xmm0
Old:
_shuf4:
pextrw $3, %xmm0, %eax
movsd %xmm1, %xmm0
pextrw $3, %xmm1, %ecx
pinsrw $4, %ecx, %xmm0
pinsrw $5, %eax, %xmm0
========
New:
_shuf1:
pushl %ebx
pushl %edi
pushl %esi
pextrw $1, %xmm0, %eax
rolw $8, %ax
movd %xmm0, %ecx
rolw $8, %cx
pextrw $5, %xmm0, %edx
pextrw $4, %xmm0, %esi
pextrw $3, %xmm0, %edi
pextrw $2, %xmm0, %ebx
movaps %xmm0, %xmm1
pinsrw $0, %ecx, %xmm1
pinsrw $1, %eax, %xmm1
rolw $8, %bx
pinsrw $2, %ebx, %xmm1
rolw $8, %di
pinsrw $3, %edi, %xmm1
rolw $8, %si
pinsrw $4, %esi, %xmm1
rolw $8, %dx
pinsrw $5, %edx, %xmm1
pextrw $7, %xmm0, %eax
rolw $8, %ax
movaps %xmm1, %xmm0
pinsrw $7, %eax, %xmm0
popl %esi
popl %edi
popl %ebx
ret
Old:
_shuf1:
subl $252, %esp
movaps %xmm0, (%esp)
movaps %xmm0, 16(%esp)
movaps %xmm0, 32(%esp)
movaps %xmm0, 48(%esp)
movaps %xmm0, 64(%esp)
movaps %xmm0, 80(%esp)
movaps %xmm0, 96(%esp)
movaps %xmm0, 224(%esp)
movaps %xmm0, 208(%esp)
movaps %xmm0, 192(%esp)
movaps %xmm0, 176(%esp)
movaps %xmm0, 160(%esp)
movaps %xmm0, 144(%esp)
movaps %xmm0, 128(%esp)
movaps %xmm0, 112(%esp)
movzbl 14(%esp), %eax
movd %eax, %xmm1
movzbl 22(%esp), %eax
movd %eax, %xmm2
punpcklbw %xmm1, %xmm2
movzbl 42(%esp), %eax
movd %eax, %xmm1
movzbl 50(%esp), %eax
movd %eax, %xmm3
punpcklbw %xmm1, %xmm3
punpcklbw %xmm2, %xmm3
movzbl 77(%esp), %eax
movd %eax, %xmm1
movzbl 84(%esp), %eax
movd %eax, %xmm2
punpcklbw %xmm1, %xmm2
movzbl 104(%esp), %eax
movd %eax, %xmm1
punpcklbw %xmm1, %xmm0
punpcklbw %xmm2, %xmm0
movaps %xmm0, %xmm1
punpcklbw %xmm3, %xmm1
movzbl 127(%esp), %eax
movd %eax, %xmm0
movzbl 135(%esp), %eax
movd %eax, %xmm2
punpcklbw %xmm0, %xmm2
movzbl 155(%esp), %eax
movd %eax, %xmm0
movzbl 163(%esp), %eax
movd %eax, %xmm3
punpcklbw %xmm0, %xmm3
punpcklbw %xmm2, %xmm3
movzbl 188(%esp), %eax
movd %eax, %xmm0
movzbl 197(%esp), %eax
movd %eax, %xmm2
punpcklbw %xmm0, %xmm2
movzbl 217(%esp), %eax
movd %eax, %xmm4
movzbl 225(%esp), %eax
movd %eax, %xmm0
punpcklbw %xmm4, %xmm0
punpcklbw %xmm2, %xmm0
punpcklbw %xmm3, %xmm0
punpcklbw %xmm1, %xmm0
addl $252, %esp
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@65311 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-23 08:49:38 +00:00
|
|
|
/// PSHUFB - Shuffle 16 8-bit values within a vector.
|
|
|
|
PSHUFB,
|
|
|
|
|
2006-11-10 21:43:37 +00:00
|
|
|
/// FMAX, FMIN - Floating point max and min.
|
|
|
|
///
|
2007-04-20 21:38:10 +00:00
|
|
|
FMAX, FMIN,
|
2007-07-10 00:05:58 +00:00
|
|
|
|
|
|
|
/// FRSQRT, FRCP - Floating point reciprocal-sqrt and reciprocal
|
|
|
|
/// approximation. Note that these typically require refinement
|
|
|
|
/// in order to obtain suitable precision.
|
|
|
|
FRSQRT, FRCP,
|
|
|
|
|
2009-04-08 21:14:34 +00:00
|
|
|
// TLSADDR - Thread Local Storage.
|
|
|
|
TLSADDR,
|
2010-06-03 04:07:48 +00:00
|
|
|
|
|
|
|
// TLSCALL - Thread Local Storage. When calling to an OS provided
|
|
|
|
// thunk at the address from an earlier relocation.
|
|
|
|
TLSCALL,
|
2009-04-08 21:14:34 +00:00
|
|
|
|
|
|
|
// SegmentBaseAddress - The address segment:0
|
|
|
|
SegmentBaseAddress,
|
2007-07-14 14:06:15 +00:00
|
|
|
|
2008-05-08 00:57:18 +00:00
|
|
|
// EH_RETURN - Exception Handling helpers.
|
2007-10-11 19:40:01 +00:00
|
|
|
EH_RETURN,
|
|
|
|
|
2008-03-19 16:39:45 +00:00
|
|
|
/// TC_RETURN - Tail call return.
|
|
|
|
/// operand #0 chain
|
|
|
|
/// operand #1 callee (register or absolute)
|
|
|
|
/// operand #2 stack adjustment
|
|
|
|
/// operand #3 optional in flag
|
2007-11-16 01:31:51 +00:00
|
|
|
TC_RETURN,
|
|
|
|
|
2008-05-08 00:57:18 +00:00
|
|
|
// LCMPXCHG_DAG, LCMPXCHG8_DAG - Compare and swap.
|
2008-03-01 21:52:34 +00:00
|
|
|
LCMPXCHG_DAG,
|
2008-03-05 01:15:49 +00:00
|
|
|
LCMPXCHG8_DAG,
|
2008-03-01 21:52:34 +00:00
|
|
|
|
2008-05-08 00:57:18 +00:00
|
|
|
// FNSTCW16m - Store FP control world into i16 memory.
|
|
|
|
FNSTCW16m,
|
|
|
|
|
2008-05-09 21:53:03 +00:00
|
|
|
// VZEXT_MOVL - Vector move low and zero extend.
|
|
|
|
VZEXT_MOVL,
|
|
|
|
|
|
|
|
// VZEXT_LOAD - Load, scalar_to_vector, and zero extend.
|
2008-05-29 08:22:04 +00:00
|
|
|
VZEXT_LOAD,
|
|
|
|
|
|
|
|
// VSHL, VSRL - Vector logical left / right shift.
|
2008-07-17 16:51:19 +00:00
|
|
|
VSHL, VSRL,
|
2009-04-27 18:41:29 +00:00
|
|
|
|
|
|
|
// CMPPD, CMPPS - Vector double/float comparison.
|
2008-07-17 16:51:19 +00:00
|
|
|
// CMPPD, CMPPS - Vector double/float comparison.
|
|
|
|
CMPPD, CMPPS,
|
|
|
|
|
|
|
|
// PCMP* - Vector integer comparisons.
|
|
|
|
PCMPEQB, PCMPEQW, PCMPEQD, PCMPEQQ,
|
2008-12-12 00:56:36 +00:00
|
|
|
PCMPGTB, PCMPGTW, PCMPGTD, PCMPGTQ,
|
|
|
|
|
2009-03-04 19:44:21 +00:00
|
|
|
// ADD, SUB, SMUL, UMUL, etc. - Arithmetic operations with FLAGS results.
|
|
|
|
ADD, SUB, SMUL, UMUL,
|
2009-09-18 19:59:53 +00:00
|
|
|
INC, DEC, OR, XOR, AND,
|
2009-03-30 21:36:47 +00:00
|
|
|
|
|
|
|
// MUL_IMM - X86 specific multiply by immediate.
|
2009-07-29 00:28:05 +00:00
|
|
|
MUL_IMM,
|
|
|
|
|
|
|
|
// PTEST - Vector bitwise comparisons
|
2009-08-15 01:38:56 +00:00
|
|
|
PTEST,
|
|
|
|
|
|
|
|
// VASTART_SAVE_XMM_REGS - Save xmm argument registers to the stack,
|
|
|
|
// according to %al. An operator is needed so that this can be expanded
|
|
|
|
// with control flow.
|
2009-09-25 20:36:54 +00:00
|
|
|
VASTART_SAVE_XMM_REGS,
|
|
|
|
|
2010-03-06 19:32:29 +00:00
|
|
|
// MINGW_ALLOCA - MingW's __alloca call to do stack probing.
|
|
|
|
MINGW_ALLOCA,
|
|
|
|
|
2009-09-25 20:36:54 +00:00
|
|
|
// ATOMADD64_DAG, ATOMSUB64_DAG, ATOMOR64_DAG, ATOMAND64_DAG,
|
|
|
|
// ATOMXOR64_DAG, ATOMNAND64_DAG, ATOMSWAP64_DAG -
|
|
|
|
// Atomic 64-bit binary operations.
|
|
|
|
ATOMADD64_DAG = ISD::FIRST_TARGET_MEMORY_OPCODE,
|
|
|
|
ATOMSUB64_DAG,
|
|
|
|
ATOMOR64_DAG,
|
|
|
|
ATOMXOR64_DAG,
|
|
|
|
ATOMAND64_DAG,
|
|
|
|
ATOMNAND64_DAG,
|
2010-07-22 02:48:34 +00:00
|
|
|
ATOMSWAP64_DAG,
|
|
|
|
|
|
|
|
// Memory barrier
|
|
|
|
MEMBARRIER,
|
|
|
|
MFENCE,
|
|
|
|
SFENCE,
|
|
|
|
LFENCE
|
2010-03-06 19:32:29 +00:00
|
|
|
|
|
|
|
// WARNING: Do not add anything in the end unless you want the node to
|
|
|
|
// have memop! In fact, starting from ATOMADD64_DAG all opcodes will be
|
|
|
|
// thought as target memory ops!
|
2005-11-15 00:40:23 +00:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2008-01-29 19:34:22 +00:00
|
|
|
/// Define some predicates that are used for node matching.
|
|
|
|
namespace X86 {
|
|
|
|
/// isPSHUFDMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to PSHUFD.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isPSHUFDMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isPSHUFHWMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to PSHUFD.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isPSHUFHWMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isPSHUFLWMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to PSHUFD.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isPSHUFLWMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isSHUFPMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to SHUFP*.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isSHUFPMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVHLPSMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to MOVHLPS.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVHLPSMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVHLPS_v_undef_Mask - Special case of isMOVHLPSMask for canonical form
|
|
|
|
/// of vector_shuffle v, v, <2, 3, 2, 3>, i.e. vector_shuffle v, undef,
|
|
|
|
/// <2, 3, 2, 3>
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVHLPS_v_undef_Mask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVLPMask - Return true if the specified VECTOR_SHUFFLE operand
|
2009-04-27 18:41:29 +00:00
|
|
|
/// specifies a shuffle of elements that is suitable for MOVLP{S|D}.
|
|
|
|
bool isMOVLPMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVHPMask - Return true if the specified VECTOR_SHUFFLE operand
|
2009-04-27 18:41:29 +00:00
|
|
|
/// specifies a shuffle of elements that is suitable for MOVHP{S|D}.
|
2008-01-29 19:34:22 +00:00
|
|
|
/// as well as MOVLHPS.
|
2009-11-07 23:17:15 +00:00
|
|
|
bool isMOVLHPSMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isUNPCKLMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to UNPCKL.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isUNPCKLMask(ShuffleVectorSDNode *N, bool V2IsSplat = false);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isUNPCKHMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to UNPCKH.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isUNPCKHMask(ShuffleVectorSDNode *N, bool V2IsSplat = false);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isUNPCKL_v_undef_Mask - Special case of isUNPCKLMask for canonical form
|
|
|
|
/// of vector_shuffle v, v, <0, 4, 1, 5>, i.e. vector_shuffle v, undef,
|
|
|
|
/// <0, 0, 1, 1>
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isUNPCKL_v_undef_Mask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isUNPCKH_v_undef_Mask - Special case of isUNPCKHMask for canonical form
|
|
|
|
/// of vector_shuffle v, v, <2, 6, 3, 7>, i.e. vector_shuffle v, undef,
|
|
|
|
/// <2, 2, 3, 3>
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isUNPCKH_v_undef_Mask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVLMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to MOVSS,
|
|
|
|
/// MOVSD, and MOVD, i.e. setting the lowest element.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVLMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVSHDUPMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to MOVSHDUP.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVSHDUPMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
|
|
|
/// isMOVSLDUPMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to MOVSLDUP.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVSLDUPMask(ShuffleVectorSDNode *N);
|
2008-01-29 19:34:22 +00:00
|
|
|
|
2008-09-25 20:50:48 +00:00
|
|
|
/// isMOVDDUPMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to MOVDDUP.
|
2009-04-27 18:41:29 +00:00
|
|
|
bool isMOVDDUPMask(ShuffleVectorSDNode *N);
|
2008-09-25 20:50:48 +00:00
|
|
|
|
2009-10-19 02:17:23 +00:00
|
|
|
/// isPALIGNRMask - Return true if the specified VECTOR_SHUFFLE operand
|
|
|
|
/// specifies a shuffle of elements that is suitable for input to PALIGNR.
|
|
|
|
bool isPALIGNRMask(ShuffleVectorSDNode *N);
|
|
|
|
|
2008-01-29 19:34:22 +00:00
|
|
|
/// getShuffleSHUFImmediate - Return the appropriate immediate to shuffle
|
|
|
|
/// the specified isShuffleMask VECTOR_SHUFFLE mask with PSHUF* and SHUFP*
|
|
|
|
/// instructions.
|
|
|
|
unsigned getShuffleSHUFImmediate(SDNode *N);
|
|
|
|
|
|
|
|
/// getShufflePSHUFHWImmediate - Return the appropriate immediate to shuffle
|
2009-10-19 02:17:23 +00:00
|
|
|
/// the specified VECTOR_SHUFFLE mask with PSHUFHW instruction.
|
2008-01-29 19:34:22 +00:00
|
|
|
unsigned getShufflePSHUFHWImmediate(SDNode *N);
|
|
|
|
|
2009-10-19 02:17:23 +00:00
|
|
|
/// getShufflePSHUFLWImmediate - Return the appropriate immediate to shuffle
|
|
|
|
/// the specified VECTOR_SHUFFLE mask with PSHUFLW instruction.
|
2008-01-29 19:34:22 +00:00
|
|
|
unsigned getShufflePSHUFLWImmediate(SDNode *N);
|
2009-07-30 08:33:02 +00:00
|
|
|
|
2009-10-19 02:17:23 +00:00
|
|
|
/// getShufflePALIGNRImmediate - Return the appropriate immediate to shuffle
|
|
|
|
/// the specified VECTOR_SHUFFLE mask with the PALIGNR instruction.
|
|
|
|
unsigned getShufflePALIGNRImmediate(SDNode *N);
|
|
|
|
|
2009-07-30 08:33:02 +00:00
|
|
|
/// isZeroNode - Returns true if Elt is a constant zero or a floating point
|
|
|
|
/// constant +0.0.
|
|
|
|
bool isZeroNode(SDValue Elt);
|
2009-08-05 23:01:26 +00:00
|
|
|
|
|
|
|
/// isOffsetSuitableForCodeModel - Returns true of the given offset can be
|
|
|
|
/// fit into displacement field of the instruction.
|
|
|
|
bool isOffsetSuitableForCodeModel(int64_t Offset, CodeModel::Model M,
|
|
|
|
bool hasSymbolicDisplacement = true);
|
2008-01-29 19:34:22 +00:00
|
|
|
}
|
|
|
|
|
2006-10-18 18:26:48 +00:00
|
|
|
//===--------------------------------------------------------------------===//
|
2005-11-15 00:40:23 +00:00
|
|
|
// X86TargetLowering - X86 Implementation of the TargetLowering interface
|
|
|
|
class X86TargetLowering : public TargetLowering {
|
|
|
|
public:
|
2008-05-14 01:58:56 +00:00
|
|
|
explicit X86TargetLowering(X86TargetMachine &TM);
|
2005-11-15 00:40:23 +00:00
|
|
|
|
2010-01-26 06:28:43 +00:00
|
|
|
/// getPICBaseSymbol - Return the X86-32 PIC base.
|
|
|
|
MCSymbol *getPICBaseSymbol(const MachineFunction *MF, MCContext &Ctx) const;
|
|
|
|
|
2010-01-26 05:02:42 +00:00
|
|
|
virtual unsigned getJumpTableEncoding() const;
|
2010-01-25 23:38:14 +00:00
|
|
|
|
2010-01-26 05:02:42 +00:00
|
|
|
virtual const MCExpr *
|
|
|
|
LowerCustomJumpTableEntry(const MachineJumpTableInfo *MJTI,
|
|
|
|
const MachineBasicBlock *MBB, unsigned uid,
|
|
|
|
MCContext &Ctx) const;
|
|
|
|
|
Much improved pic jumptable codegen:
Then:
call "L1$pb"
"L1$pb":
popl %eax
...
LBB1_1: # entry
imull $4, %ecx, %ecx
leal LJTI1_0-"L1$pb"(%eax), %edx
addl LJTI1_0-"L1$pb"(%ecx,%eax), %edx
jmpl *%edx
.align 2
.set L1_0_set_3,LBB1_3-LJTI1_0
.set L1_0_set_2,LBB1_2-LJTI1_0
.set L1_0_set_5,LBB1_5-LJTI1_0
.set L1_0_set_4,LBB1_4-LJTI1_0
LJTI1_0:
.long L1_0_set_3
.long L1_0_set_2
Now:
call "L1$pb"
"L1$pb":
popl %eax
...
LBB1_1: # entry
addl LJTI1_0-"L1$pb"(%eax,%ecx,4), %eax
jmpl *%eax
.align 2
.set L1_0_set_3,LBB1_3-"L1$pb"
.set L1_0_set_2,LBB1_2-"L1$pb"
.set L1_0_set_5,LBB1_5-"L1$pb"
.set L1_0_set_4,LBB1_4-"L1$pb"
LJTI1_0:
.long L1_0_set_3
.long L1_0_set_2
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43924 91177308-0d34-0410-b5e6-96231b3b80d8
2007-11-09 01:32:10 +00:00
|
|
|
/// getPICJumpTableRelocaBase - Returns relocation base for the given PIC
|
|
|
|
/// jumptable.
|
2010-01-26 05:02:42 +00:00
|
|
|
virtual SDValue getPICJumpTableRelocBase(SDValue Table,
|
|
|
|
SelectionDAG &DAG) const;
|
2010-01-26 06:28:43 +00:00
|
|
|
virtual const MCExpr *
|
|
|
|
getPICJumpTableRelocBaseExpr(const MachineFunction *MF,
|
|
|
|
unsigned JTI, MCContext &Ctx) const;
|
|
|
|
|
2007-02-26 04:01:25 +00:00
|
|
|
/// getStackPtrReg - Return the stack pointer register we are using: either
|
|
|
|
/// ESP or RSP.
|
|
|
|
unsigned getStackPtrReg() const { return X86StackPtr; }
|
2008-01-23 23:17:41 +00:00
|
|
|
|
|
|
|
/// getByValTypeAlignment - Return the desired alignment for ByVal aggregate
|
|
|
|
/// function arguments in the caller parameter area. For X86, aggregates
|
|
|
|
/// that contains are placed at 16-byte boundaries while the rest are at
|
|
|
|
/// 4-byte boundaries.
|
|
|
|
virtual unsigned getByValTypeAlignment(const Type *Ty) const;
|
2008-05-15 08:39:06 +00:00
|
|
|
|
|
|
|
/// getOptimalMemOpType - Returns the target specific optimal type for load
|
2010-04-02 19:36:14 +00:00
|
|
|
/// and store operations as a result of memset, memcpy, and memmove
|
|
|
|
/// lowering. If DstAlign is zero that means it's safe to destination
|
|
|
|
/// alignment can satisfy any constraint. Similarly if SrcAlign is zero it
|
|
|
|
/// means there isn't a need to check it against alignment requirement,
|
|
|
|
/// probably because the source does not need to be loaded. If
|
|
|
|
/// 'NonScalarIntSafe' is true, that means it's safe to return a
|
|
|
|
/// non-scalar-integer type, e.g. empty string source, constant, or loaded
|
2010-04-08 07:37:57 +00:00
|
|
|
/// from memory. 'MemcpyStrSrc' indicates whether the memcpy source is
|
|
|
|
/// constant so it does not need to be loaded.
|
2010-04-16 20:11:05 +00:00
|
|
|
/// It returns EVT::Other if the type should be determined using generic
|
|
|
|
/// target-independent logic.
|
2010-04-02 19:36:14 +00:00
|
|
|
virtual EVT
|
2010-04-08 07:37:57 +00:00
|
|
|
getOptimalMemOpType(uint64_t Size, unsigned DstAlign, unsigned SrcAlign,
|
|
|
|
bool NonScalarIntSafe, bool MemcpyStrSrc,
|
2010-04-16 20:11:05 +00:00
|
|
|
MachineFunction &MF) const;
|
2009-08-15 21:21:19 +00:00
|
|
|
|
|
|
|
/// allowsUnalignedMemoryAccesses - Returns true if the target allows
|
|
|
|
/// unaligned memory accesses. of the specified type.
|
|
|
|
virtual bool allowsUnalignedMemoryAccesses(EVT VT) const {
|
|
|
|
return true;
|
|
|
|
}
|
2009-06-30 22:38:32 +00:00
|
|
|
|
2005-11-15 00:40:23 +00:00
|
|
|
/// LowerOperation - Provide custom lowering hooks for some operations.
|
|
|
|
///
|
2010-04-17 15:26:15 +00:00
|
|
|
virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const;
|
2005-11-15 00:40:23 +00:00
|
|
|
|
2008-12-01 11:39:25 +00:00
|
|
|
/// ReplaceNodeResults - Replace the results of node with an illegal result
|
|
|
|
/// type with new values built out of custom code.
|
2007-11-24 07:07:01 +00:00
|
|
|
///
|
2008-12-01 11:39:25 +00:00
|
|
|
virtual void ReplaceNodeResults(SDNode *N, SmallVectorImpl<SDValue>&Results,
|
2010-04-17 15:26:15 +00:00
|
|
|
SelectionDAG &DAG) const;
|
2007-11-24 07:07:01 +00:00
|
|
|
|
|
|
|
|
2008-07-27 21:46:04 +00:00
|
|
|
virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
|
X86 target specific DAG combine: turn build_vector (load x), (load x+4),
(load x+8), (load x+12), <0, 1, 2, 3> to a single 128-bit load (aligned and
unaligned).
e.g.
__m128 test(float a, float b, float c, float d) {
return _mm_set_ps(d, c, b, a);
}
_test:
movups 4(%esp), %xmm0
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29042 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-07 08:33:52 +00:00
|
|
|
|
2010-04-17 06:13:15 +00:00
|
|
|
/// isTypeDesirableForOp - Return true if the target has native support for
|
|
|
|
/// the specified value type and it is 'desirable' to use the type for the
|
|
|
|
/// given node type. e.g. On x86 i16 is legal, but undesirable since i16
|
|
|
|
/// instruction encodings are longer and some i16 instructions are slow.
|
|
|
|
virtual bool isTypeDesirableForOp(unsigned Opc, EVT VT) const;
|
|
|
|
|
|
|
|
/// isTypeDesirable - Return true if the target has native support for the
|
|
|
|
/// specified value type and it is 'desirable' to use the type. e.g. On x86
|
|
|
|
/// i16 is legal, but undesirable since i16 instruction encodings are longer
|
|
|
|
/// and some i16 instructions are slow.
|
|
|
|
virtual bool IsDesirableToPromoteOp(SDValue Op, EVT &PVT) const;
|
2010-04-16 06:14:10 +00:00
|
|
|
|
2010-05-01 00:01:06 +00:00
|
|
|
virtual MachineBasicBlock *
|
|
|
|
EmitInstrWithCustomInserter(MachineInstr *MI,
|
|
|
|
MachineBasicBlock *MBB) const;
|
2006-01-11 00:33:36 +00:00
|
|
|
|
2008-05-05 19:05:59 +00:00
|
|
|
|
2005-12-20 06:22:03 +00:00
|
|
|
/// getTargetNodeName - This method returns the name of a target specific
|
|
|
|
/// DAG node.
|
|
|
|
virtual const char *getTargetNodeName(unsigned Opcode) const;
|
|
|
|
|
2008-03-10 15:42:14 +00:00
|
|
|
/// getSetCCResultType - Return the ISD::SETCC ValueType
|
2009-08-11 20:47:22 +00:00
|
|
|
virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const;
|
2008-03-10 15:42:14 +00:00
|
|
|
|
2006-02-16 21:11:51 +00:00
|
|
|
/// computeMaskedBitsForTargetNode - Determine which of the bits specified
|
|
|
|
/// in Mask are known to be either zero or one and return them in the
|
|
|
|
/// KnownZero/KnownOne bitsets.
|
2008-07-27 21:46:04 +00:00
|
|
|
virtual void computeMaskedBitsForTargetNode(const SDValue Op,
|
2008-02-13 22:28:48 +00:00
|
|
|
const APInt &Mask,
|
2008-02-13 00:35:47 +00:00
|
|
|
APInt &KnownZero,
|
|
|
|
APInt &KnownOne,
|
2007-06-22 14:59:07 +00:00
|
|
|
const SelectionDAG &DAG,
|
2006-02-16 21:11:51 +00:00
|
|
|
unsigned Depth = 0) const;
|
2008-05-12 19:56:52 +00:00
|
|
|
|
|
|
|
virtual bool
|
2010-04-15 01:51:59 +00:00
|
|
|
isGAPlusOffset(SDNode *N, const GlobalValue* &GA, int64_t &Offset) const;
|
2006-02-16 21:11:51 +00:00
|
|
|
|
2010-04-17 15:26:15 +00:00
|
|
|
SDValue getReturnAddressFrameIndex(SelectionDAG &DAG) const;
|
2005-11-15 00:40:23 +00:00
|
|
|
|
2009-07-20 17:51:36 +00:00
|
|
|
virtual bool ExpandInlineAsm(CallInst *CI) const;
|
|
|
|
|
2007-03-25 02:14:49 +00:00
|
|
|
ConstraintType getConstraintType(const std::string &Constraint) const;
|
2006-07-11 02:54:03 +00:00
|
|
|
|
2006-01-31 19:43:35 +00:00
|
|
|
std::vector<unsigned>
|
2006-02-22 00:56:39 +00:00
|
|
|
getRegClassForInlineAsmConstraint(const std::string &Constraint,
|
2009-08-10 22:56:29 +00:00
|
|
|
EVT VT) const;
|
2007-08-25 00:47:38 +00:00
|
|
|
|
2009-08-10 22:56:29 +00:00
|
|
|
virtual const char *LowerXConstraint(EVT ConstraintVT) const;
|
2008-01-29 02:21:21 +00:00
|
|
|
|
2007-08-25 00:47:38 +00:00
|
|
|
/// LowerAsmOperandForConstraint - Lower the specified operand into the Ops
|
2008-09-24 00:05:32 +00:00
|
|
|
/// vector. If it is invalid, don't add anything to Ops. If hasMemory is
|
|
|
|
/// true it means one of the asm constraint of the inline asm instruction
|
|
|
|
/// being processed is 'm'.
|
2008-07-27 21:46:04 +00:00
|
|
|
virtual void LowerAsmOperandForConstraint(SDValue Op,
|
2007-08-25 00:47:38 +00:00
|
|
|
char ConstraintLetter,
|
2008-07-27 21:46:04 +00:00
|
|
|
std::vector<SDValue> &Ops,
|
2008-04-26 23:02:14 +00:00
|
|
|
SelectionDAG &DAG) const;
|
2006-10-31 20:13:11 +00:00
|
|
|
|
2006-10-18 18:26:48 +00:00
|
|
|
/// getRegForInlineAsmConstraint - Given a physical register constraint
|
|
|
|
/// (e.g. {edx}), return the register number and the register class for the
|
|
|
|
/// register. This should only be used for C_Register constraints. On
|
|
|
|
/// error, this returns a register number of 0.
|
2006-07-31 23:26:50 +00:00
|
|
|
std::pair<unsigned, const TargetRegisterClass*>
|
|
|
|
getRegForInlineAsmConstraint(const std::string &Constraint,
|
2009-08-10 22:56:29 +00:00
|
|
|
EVT VT) const;
|
2006-07-31 23:26:50 +00:00
|
|
|
|
2007-03-30 23:15:24 +00:00
|
|
|
/// isLegalAddressingMode - Return true if the addressing mode represented
|
|
|
|
/// by AM is legal for this target, for a load/store of the specified type.
|
|
|
|
virtual bool isLegalAddressingMode(const AddrMode &AM, const Type *Ty)const;
|
|
|
|
|
Loosen up iv reuse to allow reuse of the same stride but a larger type when truncating from the larger type to smaller type is free.
e.g.
Turns this loop:
LBB1_1: # entry.bb_crit_edge
xorl %ecx, %ecx
xorw %dx, %dx
movw %dx, %si
LBB1_2: # bb
movl L_X$non_lazy_ptr, %edi
movw %si, (%edi)
movl L_Y$non_lazy_ptr, %edi
movw %dx, (%edi)
addw $4, %dx
incw %si
incl %ecx
cmpl %eax, %ecx
jne LBB1_2 # bb
into
LBB1_1: # entry.bb_crit_edge
xorl %ecx, %ecx
xorw %dx, %dx
LBB1_2: # bb
movl L_X$non_lazy_ptr, %esi
movw %cx, (%esi)
movl L_Y$non_lazy_ptr, %esi
movw %dx, (%esi)
addw $4, %dx
incl %ecx
cmpl %eax, %ecx
jne LBB1_2 # bb
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43375 91177308-0d34-0410-b5e6-96231b3b80d8
2007-10-26 01:56:11 +00:00
|
|
|
/// isTruncateFree - Return true if it's free to truncate a value of
|
|
|
|
/// type Ty1 to type Ty2. e.g. On x86 it's free to truncate a i32 value in
|
|
|
|
/// register EAX to i16 by referencing its sub-register AX.
|
|
|
|
virtual bool isTruncateFree(const Type *Ty1, const Type *Ty2) const;
|
2009-08-10 22:56:29 +00:00
|
|
|
virtual bool isTruncateFree(EVT VT1, EVT VT2) const;
|
Implement support for using modeling implicit-zero-extension on x86-64
with SUBREG_TO_REG, teach SimpleRegisterCoalescing to coalesce
SUBREG_TO_REG instructions (which are similar to INSERT_SUBREG
instructions), and teach the DAGCombiner to take advantage of this on
targets which support it. This eliminates many redundant
zero-extension operations on x86-64.
This adds a new TargetLowering hook, isZExtFree. It's similar to
isTruncateFree, except it only applies to actual definitions, and not
no-op truncates which may not zero the high bits.
Also, this adds a new optimization to SimplifyDemandedBits: transform
operations like x+y into (zext (add (trunc x), (trunc y))) on targets
where all the casts are no-ops. In contexts where the high part of the
add is explicitly masked off, this allows the mask operation to be
eliminated. Fix the DAGCombiner to avoid undoing these transformations
to eliminate casts on targets where the casts are no-ops.
Also, this adds a new two-address lowering heuristic. Since
two-address lowering runs before coalescing, it helps to be able to
look through copies when deciding whether commuting and/or
three-address conversion are profitable.
Also, fix a bug in LiveInterval::MergeInClobberRanges. It didn't handle
the case that a clobber range extended both before and beyond an
existing live range. In that case, multiple live ranges need to be
added. This was exposed by the new subreg coalescing code.
Remove 2008-05-06-SpillerBug.ll. It was bugpoint-reduced, and the
spiller behavior it was looking for no longer occurrs with the new
instruction selection.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@68576 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-08 00:15:30 +00:00
|
|
|
|
|
|
|
/// isZExtFree - Return true if any actual instruction that defines a
|
|
|
|
/// value of type Ty1 implicit zero-extends the value to Ty2 in the result
|
|
|
|
/// register. This does not necessarily include registers defined in
|
|
|
|
/// unknown ways, such as incoming arguments, or copies from unknown
|
|
|
|
/// virtual registers. Also, if isTruncateFree(Ty2, Ty1) is true, this
|
|
|
|
/// does not necessarily apply to truncate instructions. e.g. on x86-64,
|
|
|
|
/// all instructions that define 32-bit values implicit zero-extend the
|
|
|
|
/// result out to 64 bits.
|
|
|
|
virtual bool isZExtFree(const Type *Ty1, const Type *Ty2) const;
|
2009-08-10 22:56:29 +00:00
|
|
|
virtual bool isZExtFree(EVT VT1, EVT VT2) const;
|
Implement support for using modeling implicit-zero-extension on x86-64
with SUBREG_TO_REG, teach SimpleRegisterCoalescing to coalesce
SUBREG_TO_REG instructions (which are similar to INSERT_SUBREG
instructions), and teach the DAGCombiner to take advantage of this on
targets which support it. This eliminates many redundant
zero-extension operations on x86-64.
This adds a new TargetLowering hook, isZExtFree. It's similar to
isTruncateFree, except it only applies to actual definitions, and not
no-op truncates which may not zero the high bits.
Also, this adds a new optimization to SimplifyDemandedBits: transform
operations like x+y into (zext (add (trunc x), (trunc y))) on targets
where all the casts are no-ops. In contexts where the high part of the
add is explicitly masked off, this allows the mask operation to be
eliminated. Fix the DAGCombiner to avoid undoing these transformations
to eliminate casts on targets where the casts are no-ops.
Also, this adds a new two-address lowering heuristic. Since
two-address lowering runs before coalescing, it helps to be able to
look through copies when deciding whether commuting and/or
three-address conversion are profitable.
Also, fix a bug in LiveInterval::MergeInClobberRanges. It didn't handle
the case that a clobber range extended both before and beyond an
existing live range. In that case, multiple live ranges need to be
added. This was exposed by the new subreg coalescing code.
Remove 2008-05-06-SpillerBug.ll. It was bugpoint-reduced, and the
spiller behavior it was looking for no longer occurrs with the new
instruction selection.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@68576 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-08 00:15:30 +00:00
|
|
|
|
2009-05-28 00:35:15 +00:00
|
|
|
/// isNarrowingProfitable - Return true if it's profitable to narrow
|
|
|
|
/// operations of type VT1 to VT2. e.g. on x86, it's profitable to narrow
|
|
|
|
/// from i32 to i8 but not from i32 to i16.
|
2009-08-10 22:56:29 +00:00
|
|
|
virtual bool isNarrowingProfitable(EVT VT1, EVT VT2) const;
|
2009-05-28 00:35:15 +00:00
|
|
|
|
2009-10-27 19:56:55 +00:00
|
|
|
/// isFPImmLegal - Returns true if the target can instruction select the
|
|
|
|
/// specified FP immediate natively. If false, the legalizer will
|
|
|
|
/// materialize the FP immediate as a load from a constant pool.
|
2009-10-28 01:43:28 +00:00
|
|
|
virtual bool isFPImmLegal(const APFloat &Imm, EVT VT) const;
|
2009-10-27 19:56:55 +00:00
|
|
|
|
2006-03-22 18:59:22 +00:00
|
|
|
/// isShuffleMaskLegal - Targets can use this to indicate that they only
|
|
|
|
/// support *some* VECTOR_SHUFFLE operations, those with specific masks.
|
2006-10-18 18:26:48 +00:00
|
|
|
/// By default, if a target supports the VECTOR_SHUFFLE node, all mask
|
|
|
|
/// values are assumed to be legal.
|
2009-04-29 05:20:52 +00:00
|
|
|
virtual bool isShuffleMaskLegal(const SmallVectorImpl<int> &Mask,
|
2009-08-10 22:56:29 +00:00
|
|
|
EVT VT) const;
|
2006-04-20 08:58:49 +00:00
|
|
|
|
|
|
|
/// isVectorClearMaskLegal - Similar to isShuffleMaskLegal. This is
|
|
|
|
/// used by Targets can use this to indicate if there is a suitable
|
|
|
|
/// VECTOR_SHUFFLE that can be used to replace a VAND with a constant
|
|
|
|
/// pool entry.
|
2009-04-29 05:20:52 +00:00
|
|
|
virtual bool isVectorClearMaskLegal(const SmallVectorImpl<int> &Mask,
|
2009-08-10 22:56:29 +00:00
|
|
|
EVT VT) const;
|
2008-03-05 01:30:59 +00:00
|
|
|
|
|
|
|
/// ShouldShrinkFPConstant - If true, then instruction selection should
|
|
|
|
/// seek to shrink the FP constant of the specified type to a smaller type
|
|
|
|
/// in order to save space and / or reduce runtime.
|
2009-08-10 22:56:29 +00:00
|
|
|
virtual bool ShouldShrinkFPConstant(EVT VT) const {
|
2008-03-05 01:30:59 +00:00
|
|
|
// Don't shrink FP constpool if SSE2 is available since cvtss2sd is more
|
|
|
|
// expensive than a straight movsd. On the other hand, it's important to
|
|
|
|
// shrink long double fp constant since fldt is very slow.
|
2009-08-11 20:47:22 +00:00
|
|
|
return !X86ScalarSSEf64 || VT == MVT::f80;
|
2008-03-05 01:30:59 +00:00
|
|
|
}
|
2007-10-11 19:40:01 +00:00
|
|
|
|
2010-05-11 16:21:03 +00:00
|
|
|
const X86Subtarget* getSubtarget() const {
|
Drop ISD::MEMSET, ISD::MEMMOVE, and ISD::MEMCPY, which are not Legal
on any current target and aren't optimized in DAGCombiner. Instead
of using intermediate nodes, expand the operations, choosing between
simple loads/stores, target-specific code, and library calls,
immediately.
Previously, the code to emit optimized code for these operations
was only used at initial SelectionDAG construction time; now it is
used at all times. This fixes some cases where rep;movs was being
used for small copies where simple loads/stores would be better.
This also cleans up code that checks for alignments less than 4;
let the targets make that decision instead of doing it in
target-independent code. This allows x86 to use rep;movs in
low-alignment cases.
Also, this fixes a bug that resulted in the use of rep;stos for
memsets of 0 with non-constant memory size when the alignment was
at least 4. It's better to use the library in this case, which
can be significantly faster when the size is large.
This also preserves more SourceValue information when memory
intrinsics are lowered into simple loads/stores.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@49572 91177308-0d34-0410-b5e6-96231b3b80d8
2008-04-12 04:36:06 +00:00
|
|
|
return Subtarget;
|
2007-11-05 23:12:20 +00:00
|
|
|
}
|
|
|
|
|
2008-01-18 06:52:41 +00:00
|
|
|
/// isScalarFPTypeInSSEReg - Return true if the specified scalar FP type is
|
|
|
|
/// computed in an SSE register, not on the X87 floating point stack.
|
2009-08-10 22:56:29 +00:00
|
|
|
bool isScalarFPTypeInSSEReg(EVT VT) const {
|
2009-08-11 20:47:22 +00:00
|
|
|
return (VT == MVT::f64 && X86ScalarSSEf64) || // f64 is when SSE2
|
|
|
|
(VT == MVT::f32 && X86ScalarSSEf32); // f32 is when SSE1
|
2008-01-18 06:52:41 +00:00
|
|
|
}
|
2008-08-19 21:32:53 +00:00
|
|
|
|
|
|
|
/// createFastISel - This method returns a target specific FastISel object,
|
|
|
|
/// or null if the target does not support "fast" ISel.
|
2010-07-07 16:29:44 +00:00
|
|
|
virtual FastISel *createFastISel(FunctionLoweringInfo &funcInfo) const;
|
2009-06-30 22:38:32 +00:00
|
|
|
|
2009-07-01 18:50:55 +00:00
|
|
|
/// getFunctionAlignment - Return the Log2 alignment of this function.
|
2009-06-30 22:38:32 +00:00
|
|
|
virtual unsigned getFunctionAlignment(const Function *F) const;
|
|
|
|
|
2010-07-24 00:39:05 +00:00
|
|
|
unsigned getRegPressureLimit(const TargetRegisterClass *RC,
|
|
|
|
MachineFunction &MF) const;
|
|
|
|
|
2010-07-06 05:18:56 +00:00
|
|
|
/// getStackCookieLocation - Return true if the target stores stack
|
|
|
|
/// protector cookies at a fixed offset in some non-standard address
|
|
|
|
/// space, and populates the address space and offset as
|
|
|
|
/// appropriate.
|
|
|
|
virtual bool getStackCookieLocation(unsigned &AddressSpace, unsigned &Offset) const;
|
|
|
|
|
2010-07-26 21:50:05 +00:00
|
|
|
protected:
|
|
|
|
std::pair<const TargetRegisterClass*, uint8_t>
|
|
|
|
findRepresentativeClass(EVT VT) const;
|
|
|
|
|
2005-11-15 00:40:23 +00:00
|
|
|
private:
|
2006-04-25 20:13:52 +00:00
|
|
|
/// Subtarget - Keep a pointer to the X86Subtarget around so that we can
|
|
|
|
/// make the right decision when generating code for different targets.
|
|
|
|
const X86Subtarget *Subtarget;
|
2008-05-14 01:58:56 +00:00
|
|
|
const X86RegisterInfo *RegInfo;
|
2008-09-09 18:22:57 +00:00
|
|
|
const TargetData *TD;
|
2006-04-25 20:13:52 +00:00
|
|
|
|
2006-09-08 06:48:29 +00:00
|
|
|
/// X86StackPtr - X86 physical register used as stack ptr.
|
|
|
|
unsigned X86StackPtr;
|
2007-10-11 19:40:01 +00:00
|
|
|
|
2007-09-23 14:52:20 +00:00
|
|
|
/// X86ScalarSSEf32, X86ScalarSSEf64 - Select between SSE or x87
|
|
|
|
/// floating point ops.
|
|
|
|
/// When SSE is available, use it for f32 operations.
|
|
|
|
/// When SSE2 is available, use it for f64 operations.
|
|
|
|
bool X86ScalarSSEf32;
|
|
|
|
bool X86ScalarSSEf64;
|
2008-01-29 19:34:22 +00:00
|
|
|
|
2009-10-27 19:56:55 +00:00
|
|
|
/// LegalFPImmediates - A list of legal fp immediates.
|
|
|
|
std::vector<APFloat> LegalFPImmediates;
|
|
|
|
|
|
|
|
/// addLegalFPImmediate - Indicate that this x86 target can instruction
|
|
|
|
/// select the specified FP immediate natively.
|
|
|
|
void addLegalFPImmediate(const APFloat& Imm) {
|
|
|
|
LegalFPImmediates.push_back(Imm);
|
|
|
|
}
|
|
|
|
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
SDValue LowerCallResult(SDValue Chain, SDValue InFlag,
|
2009-09-02 08:44:58 +00:00
|
|
|
CallingConv::ID CallConv, bool isVarArg,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::InputArg> &Ins,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG,
|
2010-04-17 15:26:15 +00:00
|
|
|
SmallVectorImpl<SDValue> &InVals) const;
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
SDValue LowerMemArgument(SDValue Chain,
|
2009-09-02 08:44:58 +00:00
|
|
|
CallingConv::ID CallConv,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::InputArg> &ArgInfo,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG,
|
|
|
|
const CCValAssign &VA, MachineFrameInfo *MFI,
|
2010-04-17 15:26:15 +00:00
|
|
|
unsigned i) const;
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
SDValue LowerMemOpCallTo(SDValue Chain, SDValue StackPtr, SDValue Arg,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG,
|
|
|
|
const CCValAssign &VA,
|
2010-04-17 15:26:15 +00:00
|
|
|
ISD::ArgFlagsTy Flags) const;
|
2007-08-31 15:06:30 +00:00
|
|
|
|
2008-01-05 16:56:59 +00:00
|
|
|
// Call lowering helpers.
|
2010-01-27 00:07:07 +00:00
|
|
|
|
|
|
|
/// IsEligibleForTailCallOptimization - Check whether the call is eligible
|
|
|
|
/// for tail call optimization. Targets which want to do tail call
|
|
|
|
/// optimization should implement this function.
|
2010-02-02 23:55:14 +00:00
|
|
|
bool IsEligibleForTailCallOptimization(SDValue Callee,
|
2010-01-27 00:07:07 +00:00
|
|
|
CallingConv::ID CalleeCC,
|
|
|
|
bool isVarArg,
|
2010-03-15 18:54:48 +00:00
|
|
|
bool isCalleeStructRet,
|
|
|
|
bool isCallerStructRet,
|
2010-01-27 06:25:16 +00:00
|
|
|
const SmallVectorImpl<ISD::OutputArg> &Outs,
|
2010-07-07 15:54:55 +00:00
|
|
|
const SmallVectorImpl<SDValue> &OutVals,
|
2010-01-27 06:25:16 +00:00
|
|
|
const SmallVectorImpl<ISD::InputArg> &Ins,
|
2010-01-27 00:07:07 +00:00
|
|
|
SelectionDAG& DAG) const;
|
2010-04-17 15:26:15 +00:00
|
|
|
bool IsCalleePop(bool isVarArg, CallingConv::ID CallConv) const;
|
2008-07-27 21:46:04 +00:00
|
|
|
SDValue EmitTailCallLoadRetAddr(SelectionDAG &DAG, SDValue &OutRetAddr,
|
|
|
|
SDValue Chain, bool IsTailCall, bool Is64Bit,
|
2010-04-17 15:26:15 +00:00
|
|
|
int FPDiff, DebugLoc dl) const;
|
2008-04-12 18:11:06 +00:00
|
|
|
|
2009-09-02 08:44:58 +00:00
|
|
|
CCAssignFn *CCAssignFnForNode(CallingConv::ID CallConv) const;
|
2010-04-17 15:26:15 +00:00
|
|
|
unsigned GetAlignedArgumentStackSize(unsigned StackSize,
|
|
|
|
SelectionDAG &DAG) const;
|
2006-01-27 08:10:46 +00:00
|
|
|
|
2009-05-23 09:59:16 +00:00
|
|
|
std::pair<SDValue,SDValue> FP_TO_INTHelper(SDValue Op, SelectionDAG &DAG,
|
2010-04-17 15:26:15 +00:00
|
|
|
bool isSigned) const;
|
Optimize splat of a scalar load into a shuffle of a vector load when it's legal. e.g.
vector_shuffle (scalar_to_vector (i32 load (ptr + 4))), undef, <0, 0, 0, 0>
=>
vector_shuffle (v4i32 load ptr), undef, <1, 1, 1, 1>
iff ptr is 16-byte aligned (or can be made into 16-byte aligned).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90984 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 21:00:30 +00:00
|
|
|
|
|
|
|
SDValue LowerAsSplatVectorLoad(SDValue SrcOp, EVT VT, DebugLoc dl,
|
2010-04-17 15:26:15 +00:00
|
|
|
SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerCONCAT_VECTORS(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerEXTRACT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerEXTRACT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerINSERT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerINSERT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerSCALAR_TO_VECTOR(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerConstantPool(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerBlockAddress(SDValue Op, SelectionDAG &DAG) const;
|
2009-02-04 20:06:27 +00:00
|
|
|
SDValue LowerGlobalAddress(const GlobalValue *GV, DebugLoc dl,
|
|
|
|
int64_t Offset, SelectionDAG &DAG) const;
|
2010-04-17 15:26:15 +00:00
|
|
|
SDValue LowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerExternalSymbol(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerShift(SDValue Op, SelectionDAG &DAG) const;
|
2009-08-10 22:56:29 +00:00
|
|
|
SDValue BuildFILD(SDValue Op, EVT SrcVT, SDValue Chain, SDValue StackSlot,
|
2010-04-17 15:26:15 +00:00
|
|
|
SelectionDAG &DAG) const;
|
2010-05-21 00:52:33 +00:00
|
|
|
SDValue LowerBIT_CONVERT(SDValue op, SelectionDAG &DAG) const;
|
2010-04-17 15:26:15 +00:00
|
|
|
SDValue LowerSINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerUINT_TO_FP_i32(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFP_TO_SINT(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFP_TO_UINT(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFABS(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFNEG(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) const;
|
2010-04-21 01:47:12 +00:00
|
|
|
SDValue LowerToBT(SDValue And, ISD::CondCode CC,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG) const;
|
2010-04-17 15:26:15 +00:00
|
|
|
SDValue LowerSETCC(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerVSETCC(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerSELECT(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerBRCOND(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerMEMSET(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerJumpTable(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerVASTART(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerVAARG(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerVACOPY(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerRETURNADDR(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFRAME_TO_ARGS_OFFSET(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerEH_RETURN(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerTRAMPOLINE(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerFLT_ROUNDS_(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerCTLZ(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerCTTZ(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerMUL_V2I64(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerXALUO(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
|
|
|
|
SDValue LowerCMP_SWAP(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerLOAD_SUB(SDValue Op, SelectionDAG &DAG) const;
|
|
|
|
SDValue LowerREADCYCLECOUNTER(SDValue Op, SelectionDAG &DAG) const;
|
2010-07-22 02:48:34 +00:00
|
|
|
SDValue LowerMEMBARRIER(SDValue Op, SelectionDAG &DAG) const;
|
2008-12-01 11:39:25 +00:00
|
|
|
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
virtual SDValue
|
|
|
|
LowerFormalArguments(SDValue Chain,
|
2009-09-02 08:44:58 +00:00
|
|
|
CallingConv::ID CallConv, bool isVarArg,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::InputArg> &Ins,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG,
|
2010-04-17 15:26:15 +00:00
|
|
|
SmallVectorImpl<SDValue> &InVals) const;
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
virtual SDValue
|
2010-02-02 23:55:14 +00:00
|
|
|
LowerCall(SDValue Chain, SDValue Callee,
|
2010-01-27 00:07:07 +00:00
|
|
|
CallingConv::ID CallConv, bool isVarArg, bool &isTailCall,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::OutputArg> &Outs,
|
2010-07-07 15:54:55 +00:00
|
|
|
const SmallVectorImpl<SDValue> &OutVals,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::InputArg> &Ins,
|
|
|
|
DebugLoc dl, SelectionDAG &DAG,
|
2010-04-17 15:26:15 +00:00
|
|
|
SmallVectorImpl<SDValue> &InVals) const;
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
|
|
|
|
virtual SDValue
|
|
|
|
LowerReturn(SDValue Chain,
|
2009-09-02 08:44:58 +00:00
|
|
|
CallingConv::ID CallConv, bool isVarArg,
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
const SmallVectorImpl<ISD::OutputArg> &Outs,
|
2010-07-07 15:54:55 +00:00
|
|
|
const SmallVectorImpl<SDValue> &OutVals,
|
2010-04-17 15:26:15 +00:00
|
|
|
DebugLoc dl, SelectionDAG &DAG) const;
|
Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-05 01:29:28 +00:00
|
|
|
|
2009-11-07 02:11:54 +00:00
|
|
|
virtual bool
|
|
|
|
CanLowerReturn(CallingConv::ID CallConv, bool isVarArg,
|
2010-07-10 09:00:22 +00:00
|
|
|
const SmallVectorImpl<ISD::OutputArg> &Outs,
|
2010-07-06 22:19:37 +00:00
|
|
|
LLVMContext &Context) const;
|
2009-11-07 02:11:54 +00:00
|
|
|
|
2008-12-01 11:39:25 +00:00
|
|
|
void ReplaceATOMIC_BINARY_64(SDNode *N, SmallVectorImpl<SDValue> &Results,
|
2010-04-17 15:26:15 +00:00
|
|
|
SelectionDAG &DAG, unsigned NewOp) const;
|
2008-12-01 11:39:25 +00:00
|
|
|
|
2009-08-18 22:50:32 +00:00
|
|
|
/// Utility function to emit string processing sse4.2 instructions
|
|
|
|
/// that return in xmm0.
|
2009-09-19 10:09:15 +00:00
|
|
|
/// This takes the instruction to expand, the associated machine basic
|
|
|
|
/// block, the number of args, and whether or not the second arg is
|
|
|
|
/// in memory or not.
|
2009-08-18 22:50:32 +00:00
|
|
|
MachineBasicBlock *EmitPCMP(MachineInstr *BInstr, MachineBasicBlock *BB,
|
Reapply address space patch after fixing an issue in MemCopyOptimizer.
Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100304 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-04 03:10:48 +00:00
|
|
|
unsigned argNum, bool inMem) const;
|
2009-08-18 22:50:32 +00:00
|
|
|
|
2008-05-05 19:05:59 +00:00
|
|
|
/// Utility function to emit atomic bitwise operations (and, or, xor).
|
2009-09-19 10:09:15 +00:00
|
|
|
/// It takes the bitwise instruction to expand, the associated machine basic
|
|
|
|
/// block, and the associated X86 opcodes for reg/reg and reg/imm.
|
2008-05-05 19:05:59 +00:00
|
|
|
MachineBasicBlock *EmitAtomicBitwiseWithCustomInserter(
|
|
|
|
MachineInstr *BInstr,
|
|
|
|
MachineBasicBlock *BB,
|
|
|
|
unsigned regOpc,
|
2008-06-14 05:48:15 +00:00
|
|
|
unsigned immOpc,
|
2008-08-19 18:47:28 +00:00
|
|
|
unsigned loadOpc,
|
|
|
|
unsigned cxchgOpc,
|
|
|
|
unsigned notOpc,
|
|
|
|
unsigned EAXreg,
|
|
|
|
TargetRegisterClass *RC,
|
2009-02-07 16:15:20 +00:00
|
|
|
bool invSrc = false) const;
|
2008-10-02 18:53:47 +00:00
|
|
|
|
|
|
|
MachineBasicBlock *EmitAtomicBit6432WithCustomInserter(
|
|
|
|
MachineInstr *BInstr,
|
|
|
|
MachineBasicBlock *BB,
|
|
|
|
unsigned regOpcL,
|
|
|
|
unsigned regOpcH,
|
|
|
|
unsigned immOpcL,
|
|
|
|
unsigned immOpcH,
|
2009-02-07 16:15:20 +00:00
|
|
|
bool invSrc = false) const;
|
2008-05-05 19:05:59 +00:00
|
|
|
|
|
|
|
/// Utility function to emit atomic min and max. It takes the min/max
|
2009-03-26 01:46:56 +00:00
|
|
|
/// instruction to expand, the associated basic block, and the associated
|
|
|
|
/// cmov opcode for moving the min or max value.
|
2008-05-05 19:05:59 +00:00
|
|
|
MachineBasicBlock *EmitAtomicMinMaxWithCustomInserter(MachineInstr *BInstr,
|
|
|
|
MachineBasicBlock *BB,
|
2009-02-07 16:15:20 +00:00
|
|
|
unsigned cmovOpc) const;
|
2009-03-04 19:44:21 +00:00
|
|
|
|
2009-08-15 01:38:56 +00:00
|
|
|
/// Utility function to emit the xmm reg save portion of va_start.
|
|
|
|
MachineBasicBlock *EmitVAStartSaveXMMRegsWithCustomInserter(
|
|
|
|
MachineInstr *BInstr,
|
|
|
|
MachineBasicBlock *BB) const;
|
|
|
|
|
2009-09-02 05:57:00 +00:00
|
|
|
MachineBasicBlock *EmitLoweredSelect(MachineInstr *I,
|
2010-05-01 00:01:06 +00:00
|
|
|
MachineBasicBlock *BB) const;
|
2010-03-06 19:32:29 +00:00
|
|
|
|
|
|
|
MachineBasicBlock *EmitLoweredMingwAlloca(MachineInstr *MI,
|
2010-05-01 00:01:06 +00:00
|
|
|
MachineBasicBlock *BB) const;
|
2010-06-03 04:07:48 +00:00
|
|
|
|
|
|
|
MachineBasicBlock *EmitLoweredTLSCall(MachineInstr *MI,
|
|
|
|
MachineBasicBlock *BB) const;
|
2010-03-06 19:32:29 +00:00
|
|
|
|
2009-03-04 19:44:21 +00:00
|
|
|
/// Emit nodes that will be selected as "test Op0,Op0", or something
|
2009-03-07 01:58:32 +00:00
|
|
|
/// equivalent, for use with the given x86 condition code.
|
2010-04-26 19:06:11 +00:00
|
|
|
SDValue EmitTest(SDValue Op0, unsigned X86CC, SelectionDAG &DAG) const;
|
2009-03-04 19:44:21 +00:00
|
|
|
|
|
|
|
/// Emit nodes that will be selected as "cmp Op0,Op1", or something
|
2009-03-07 01:58:32 +00:00
|
|
|
/// equivalent, for use with the given x86 condition code.
|
2010-04-26 19:06:11 +00:00
|
|
|
SDValue EmitCmp(SDValue Op0, SDValue Op1, unsigned X86CC,
|
2010-04-17 15:26:15 +00:00
|
|
|
SelectionDAG &DAG) const;
|
2005-11-15 00:40:23 +00:00
|
|
|
};
|
2008-09-03 00:03:49 +00:00
|
|
|
|
|
|
|
namespace X86 {
|
2010-07-07 16:29:44 +00:00
|
|
|
FastISel *createFastISel(FunctionLoweringInfo &funcInfo);
|
2008-09-03 00:03:49 +00:00
|
|
|
}
|
2005-11-15 00:40:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
#endif // X86ISELLOWERING_H
|