From 010496c6a759fd8d8ee76e0af90b331106894233 Mon Sep 17 00:00:00 2001 From: Chris Lattner Date: Tue, 5 Oct 2010 06:22:35 +0000 Subject: [PATCH] move the atomic pseudo instructions out to X86InstrCompiler.td git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@115599 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86Instr64bit.td | 153 ------------- lib/Target/X86/X86InstrCompiler.td | 348 ++++++++++++++++++++++++++++- lib/Target/X86/X86InstrInfo.td | 272 +++------------------- lib/Target/X86/X86InstrSystem.td | 8 + 4 files changed, 388 insertions(+), 393 deletions(-) diff --git a/lib/Target/X86/X86Instr64bit.td b/lib/Target/X86/X86Instr64bit.td index 73a9153724a..36cda076c7b 100644 --- a/lib/Target/X86/X86Instr64bit.td +++ b/lib/Target/X86/X86Instr64bit.td @@ -1386,159 +1386,6 @@ def SETB_C64r : RI<0x19, MRMInitReg, (outs GR64:$dst), (ins), "", def : Pat<(i64 (anyext (i8 (X86setcc_c X86_COND_B, EFLAGS)))), (SETB_C64r)>; -//===----------------------------------------------------------------------===// -// Descriptor-table support instructions - -// LLDT is not interpreted specially in 64-bit mode because there is no sign -// extension. -def SLDT64r : RI<0x00, MRM0r, (outs GR64:$dst), (ins), - "sldt{q}\t$dst", []>, TB; -def SLDT64m : RI<0x00, MRM0m, (outs i16mem:$dst), (ins), - "sldt{q}\t$dst", []>, TB; - -//===----------------------------------------------------------------------===// -// Alias Instructions -//===----------------------------------------------------------------------===// - -// We want to rewrite MOV64r0 in terms of MOV32r0, because it's sometimes a -// smaller encoding, but doing so at isel time interferes with rematerialization -// in the current register allocator. For now, this is rewritten when the -// instruction is lowered to an MCInst. -// FIXME: AddedComplexity gives this a higher priority than MOV64ri32. Remove -// when we have a better way to specify isel priority. -let Defs = [EFLAGS], - AddedComplexity = 1, isReMaterializable = 1, isAsCheapAsAMove = 1 in -def MOV64r0 : I<0x31, MRMInitReg, (outs GR64:$dst), (ins), "", - [(set GR64:$dst, 0)]>; - -// Materialize i64 constant where top 32-bits are zero. This could theoretically -// use MOV32ri with a SUBREG_TO_REG to represent the zero-extension, however -// that would make it more difficult to rematerialize. -let AddedComplexity = 1, isReMaterializable = 1, isAsCheapAsAMove = 1 in -def MOV64ri64i32 : Ii32<0xB8, AddRegFrm, (outs GR64:$dst), (ins i64i32imm:$src), - "", [(set GR64:$dst, i64immZExt32:$src)]>; - -//===----------------------------------------------------------------------===// -// Atomic Instructions -//===----------------------------------------------------------------------===// - -// TODO: Get this to fold the constant into the instruction. -let hasSideEffects = 1, Defs = [ESP] in -def Int_MemBarrierNoSSE64 : RI<0x09, MRM1r, (outs), (ins GR64:$zero), - "lock\n\t" - "or{q}\t{$zero, (%rsp)|(%rsp), $zero}", - [(X86MemBarrierNoSSE GR64:$zero)]>, - Requires<[In64BitMode]>, LOCK; - -let Defs = [RAX, EFLAGS], Uses = [RAX] in { -def LCMPXCHG64 : RI<0xB1, MRMDestMem, (outs), (ins i64mem:$ptr, GR64:$swap), - "lock\n\t" - "cmpxchgq\t$swap,$ptr", - [(X86cas addr:$ptr, GR64:$swap, 8)]>, TB, LOCK; -} - -let Constraints = "$val = $dst" in { -let Defs = [EFLAGS] in -def LXADD64 : RI<0xC1, MRMSrcMem, (outs GR64:$dst), (ins GR64:$val,i64mem:$ptr), - "lock\n\t" - "xadd\t$val, $ptr", - [(set GR64:$dst, (atomic_load_add_64 addr:$ptr, GR64:$val))]>, - TB, LOCK; - -def XCHG64rm : RI<0x87, MRMSrcMem, (outs GR64:$dst), - (ins GR64:$val,i64mem:$ptr), - "xchg{q}\t{$val, $ptr|$ptr, $val}", - [(set GR64:$dst, (atomic_swap_64 addr:$ptr, GR64:$val))]>; - -def XCHG64rr : RI<0x87, MRMSrcReg, (outs GR64:$dst), (ins GR64:$val,GR64:$src), - "xchg{q}\t{$val, $src|$src, $val}", []>; -} - -def XADD64rr : RI<0xC1, MRMDestReg, (outs GR64:$dst), (ins GR64:$src), - "xadd{q}\t{$src, $dst|$dst, $src}", []>, TB; -let mayLoad = 1, mayStore = 1 in -def XADD64rm : RI<0xC1, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), - "xadd{q}\t{$src, $dst|$dst, $src}", []>, TB; - -def CMPXCHG64rr : RI<0xB1, MRMDestReg, (outs GR64:$dst), (ins GR64:$src), - "cmpxchg{q}\t{$src, $dst|$dst, $src}", []>, TB; -let mayLoad = 1, mayStore = 1 in -def CMPXCHG64rm : RI<0xB1, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), - "cmpxchg{q}\t{$src, $dst|$dst, $src}", []>, TB; - -let Defs = [RAX, RDX, EFLAGS], Uses = [RAX, RBX, RCX, RDX] in -def CMPXCHG16B : RI<0xC7, MRM1m, (outs), (ins i128mem:$dst), - "cmpxchg16b\t$dst", []>, TB; - -def XCHG64ar : RI<0x90, AddRegFrm, (outs), (ins GR64:$src), - "xchg{q}\t{$src, %rax|%rax, $src}", []>; - -// Optimized codegen when the non-memory output is not used. -let Defs = [EFLAGS], mayLoad = 1, mayStore = 1 in { -// FIXME: Use normal add / sub instructions and add lock prefix dynamically. -def LOCK_ADD64mr : RI<0x01, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2), - "lock\n\t" - "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD64mi8 : RIi8<0x83, MRM0m, (outs), - (ins i64mem:$dst, i64i8imm :$src2), - "lock\n\t" - "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD64mi32 : RIi32<0x81, MRM0m, (outs), - (ins i64mem:$dst, i64i32imm :$src2), - "lock\n\t" - "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB64mr : RI<0x29, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2), - "lock\n\t" - "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB64mi8 : RIi8<0x83, MRM5m, (outs), - (ins i64mem:$dst, i64i8imm :$src2), - "lock\n\t" - "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB64mi32 : RIi32<0x81, MRM5m, (outs), - (ins i64mem:$dst, i64i32imm:$src2), - "lock\n\t" - "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_INC64m : RI<0xFF, MRM0m, (outs), (ins i64mem:$dst), - "lock\n\t" - "inc{q}\t$dst", []>, LOCK; -def LOCK_DEC64m : RI<0xFF, MRM1m, (outs), (ins i64mem:$dst), - "lock\n\t" - "dec{q}\t$dst", []>, LOCK; -} -// Atomic exchange, and, or, xor -let Constraints = "$val = $dst", Defs = [EFLAGS], - usesCustomInserter = 1 in { -def ATOMAND64 : I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMAND64 PSEUDO!", - [(set GR64:$dst, (atomic_load_and_64 addr:$ptr, GR64:$val))]>; -def ATOMOR64 : I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMOR64 PSEUDO!", - [(set GR64:$dst, (atomic_load_or_64 addr:$ptr, GR64:$val))]>; -def ATOMXOR64 : I<0, Pseudo,(outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMXOR64 PSEUDO!", - [(set GR64:$dst, (atomic_load_xor_64 addr:$ptr, GR64:$val))]>; -def ATOMNAND64 : I<0, Pseudo,(outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMNAND64 PSEUDO!", - [(set GR64:$dst, (atomic_load_nand_64 addr:$ptr, GR64:$val))]>; -def ATOMMIN64: I<0, Pseudo, (outs GR64:$dst), (ins i64mem:$ptr, GR64:$val), - "#ATOMMIN64 PSEUDO!", - [(set GR64:$dst, (atomic_load_min_64 addr:$ptr, GR64:$val))]>; -def ATOMMAX64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMMAX64 PSEUDO!", - [(set GR64:$dst, (atomic_load_max_64 addr:$ptr, GR64:$val))]>; -def ATOMUMIN64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMUMIN64 PSEUDO!", - [(set GR64:$dst, (atomic_load_umin_64 addr:$ptr, GR64:$val))]>; -def ATOMUMAX64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), - "#ATOMUMAX64 PSEUDO!", - [(set GR64:$dst, (atomic_load_umax_64 addr:$ptr, GR64:$val))]>; -} - - -// String manipulation instructions - -def LODSQ : RI<0xAD, RawFrm, (outs), (ins), "lodsq", []>; - //===----------------------------------------------------------------------===// // X86-64 SSE Instructions diff --git a/lib/Target/X86/X86InstrCompiler.td b/lib/Target/X86/X86InstrCompiler.td index 3046a69add4..7955a916d15 100644 --- a/lib/Target/X86/X86InstrCompiler.td +++ b/lib/Target/X86/X86InstrCompiler.td @@ -127,6 +127,27 @@ def MOV32r0 : I<0x31, MRMInitReg, (outs GR32:$dst), (ins), "", [(set GR32:$dst, 0)]>; } +// We want to rewrite MOV64r0 in terms of MOV32r0, because it's sometimes a +// smaller encoding, but doing so at isel time interferes with rematerialization +// in the current register allocator. For now, this is rewritten when the +// instruction is lowered to an MCInst. +// FIXME: AddedComplexity gives this a higher priority than MOV64ri32. Remove +// when we have a better way to specify isel priority. +let Defs = [EFLAGS], + AddedComplexity = 1, isReMaterializable = 1, isAsCheapAsAMove = 1 in +def MOV64r0 : I<0x31, MRMInitReg, (outs GR64:$dst), (ins), "", + [(set GR64:$dst, 0)]>; + +// Materialize i64 constant where top 32-bits are zero. This could theoretically +// use MOV32ri with a SUBREG_TO_REG to represent the zero-extension, however +// that would make it more difficult to rematerialize. +let AddedComplexity = 1, isReMaterializable = 1, isAsCheapAsAMove = 1 in +def MOV64ri64i32 : Ii32<0xB8, AddRegFrm, (outs GR64:$dst), (ins i64i32imm:$src), + "", [(set GR64:$dst, i64immZExt32:$src)]>; + + + + //===----------------------------------------------------------------------===// // Thread Local Storage Instructions // @@ -187,7 +208,330 @@ def TLSCall_64 : I<0, Pseudo, (outs), (ins i64mem:$sym), Requires<[In64BitMode]>; //===----------------------------------------------------------------------===// -// Non-Instruction Patterns +// Atomic Instruction Pseudo Instructions +//===----------------------------------------------------------------------===// + +// Atomic exchange, and, or, xor +let Constraints = "$val = $dst", Defs = [EFLAGS], + usesCustomInserter = 1 in { + +def ATOMAND8 : I<0, Pseudo, (outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), + "#ATOMAND8 PSEUDO!", + [(set GR8:$dst, (atomic_load_and_8 addr:$ptr, GR8:$val))]>; +def ATOMOR8 : I<0, Pseudo, (outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), + "#ATOMOR8 PSEUDO!", + [(set GR8:$dst, (atomic_load_or_8 addr:$ptr, GR8:$val))]>; +def ATOMXOR8 : I<0, Pseudo,(outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), + "#ATOMXOR8 PSEUDO!", + [(set GR8:$dst, (atomic_load_xor_8 addr:$ptr, GR8:$val))]>; +def ATOMNAND8 : I<0, Pseudo,(outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), + "#ATOMNAND8 PSEUDO!", + [(set GR8:$dst, (atomic_load_nand_8 addr:$ptr, GR8:$val))]>; + +def ATOMAND16 : I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMAND16 PSEUDO!", + [(set GR16:$dst, (atomic_load_and_16 addr:$ptr, GR16:$val))]>; +def ATOMOR16 : I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMOR16 PSEUDO!", + [(set GR16:$dst, (atomic_load_or_16 addr:$ptr, GR16:$val))]>; +def ATOMXOR16 : I<0, Pseudo,(outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMXOR16 PSEUDO!", + [(set GR16:$dst, (atomic_load_xor_16 addr:$ptr, GR16:$val))]>; +def ATOMNAND16 : I<0, Pseudo,(outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMNAND16 PSEUDO!", + [(set GR16:$dst, (atomic_load_nand_16 addr:$ptr, GR16:$val))]>; +def ATOMMIN16: I<0, Pseudo, (outs GR16:$dst), (ins i16mem:$ptr, GR16:$val), + "#ATOMMIN16 PSEUDO!", + [(set GR16:$dst, (atomic_load_min_16 addr:$ptr, GR16:$val))]>; +def ATOMMAX16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMMAX16 PSEUDO!", + [(set GR16:$dst, (atomic_load_max_16 addr:$ptr, GR16:$val))]>; +def ATOMUMIN16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMUMIN16 PSEUDO!", + [(set GR16:$dst, (atomic_load_umin_16 addr:$ptr, GR16:$val))]>; +def ATOMUMAX16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), + "#ATOMUMAX16 PSEUDO!", + [(set GR16:$dst, (atomic_load_umax_16 addr:$ptr, GR16:$val))]>; + + +def ATOMAND32 : I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMAND32 PSEUDO!", + [(set GR32:$dst, (atomic_load_and_32 addr:$ptr, GR32:$val))]>; +def ATOMOR32 : I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMOR32 PSEUDO!", + [(set GR32:$dst, (atomic_load_or_32 addr:$ptr, GR32:$val))]>; +def ATOMXOR32 : I<0, Pseudo,(outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMXOR32 PSEUDO!", + [(set GR32:$dst, (atomic_load_xor_32 addr:$ptr, GR32:$val))]>; +def ATOMNAND32 : I<0, Pseudo,(outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMNAND32 PSEUDO!", + [(set GR32:$dst, (atomic_load_nand_32 addr:$ptr, GR32:$val))]>; +def ATOMMIN32: I<0, Pseudo, (outs GR32:$dst), (ins i32mem:$ptr, GR32:$val), + "#ATOMMIN32 PSEUDO!", + [(set GR32:$dst, (atomic_load_min_32 addr:$ptr, GR32:$val))]>; +def ATOMMAX32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMMAX32 PSEUDO!", + [(set GR32:$dst, (atomic_load_max_32 addr:$ptr, GR32:$val))]>; +def ATOMUMIN32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMUMIN32 PSEUDO!", + [(set GR32:$dst, (atomic_load_umin_32 addr:$ptr, GR32:$val))]>; +def ATOMUMAX32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), + "#ATOMUMAX32 PSEUDO!", + [(set GR32:$dst, (atomic_load_umax_32 addr:$ptr, GR32:$val))]>; + + + +def ATOMAND64 : I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMAND64 PSEUDO!", + [(set GR64:$dst, (atomic_load_and_64 addr:$ptr, GR64:$val))]>; +def ATOMOR64 : I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMOR64 PSEUDO!", + [(set GR64:$dst, (atomic_load_or_64 addr:$ptr, GR64:$val))]>; +def ATOMXOR64 : I<0, Pseudo,(outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMXOR64 PSEUDO!", + [(set GR64:$dst, (atomic_load_xor_64 addr:$ptr, GR64:$val))]>; +def ATOMNAND64 : I<0, Pseudo,(outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMNAND64 PSEUDO!", + [(set GR64:$dst, (atomic_load_nand_64 addr:$ptr, GR64:$val))]>; +def ATOMMIN64: I<0, Pseudo, (outs GR64:$dst), (ins i64mem:$ptr, GR64:$val), + "#ATOMMIN64 PSEUDO!", + [(set GR64:$dst, (atomic_load_min_64 addr:$ptr, GR64:$val))]>; +def ATOMMAX64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMMAX64 PSEUDO!", + [(set GR64:$dst, (atomic_load_max_64 addr:$ptr, GR64:$val))]>; +def ATOMUMIN64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMUMIN64 PSEUDO!", + [(set GR64:$dst, (atomic_load_umin_64 addr:$ptr, GR64:$val))]>; +def ATOMUMAX64: I<0, Pseudo, (outs GR64:$dst),(ins i64mem:$ptr, GR64:$val), + "#ATOMUMAX64 PSEUDO!", + [(set GR64:$dst, (atomic_load_umax_64 addr:$ptr, GR64:$val))]>; +} + +let Constraints = "$val1 = $dst1, $val2 = $dst2", + Defs = [EFLAGS, EAX, EBX, ECX, EDX], + Uses = [EAX, EBX, ECX, EDX], + mayLoad = 1, mayStore = 1, + usesCustomInserter = 1 in { +def ATOMAND6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMAND6432 PSEUDO!", []>; +def ATOMOR6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMOR6432 PSEUDO!", []>; +def ATOMXOR6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMXOR6432 PSEUDO!", []>; +def ATOMNAND6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMNAND6432 PSEUDO!", []>; +def ATOMADD6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMADD6432 PSEUDO!", []>; +def ATOMSUB6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMSUB6432 PSEUDO!", []>; +def ATOMSWAP6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), + (ins i64mem:$ptr, GR32:$val1, GR32:$val2), + "#ATOMSWAP6432 PSEUDO!", []>; +} + +//===----------------------------------------------------------------------===// +// Normal-Instructions-With-Lock-Prefix Pseudo Instructions +//===----------------------------------------------------------------------===// + +// FIXME: Use normal instructions and add lock prefix dynamically. + +// Memory barriers + +// TODO: Get this to fold the constant into the instruction. +def OR32mrLocked : I<0x09, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$zero), + "lock\n\t" + "or{l}\t{$zero, $dst|$dst, $zero}", + []>, Requires<[In32BitMode]>, LOCK; + +let hasSideEffects = 1 in +def Int_MemBarrier : I<0, Pseudo, (outs), (ins), + "#MEMBARRIER", + [(X86MemBarrier)]>, Requires<[HasSSE2]>; + +// TODO: Get this to fold the constant into the instruction. +let hasSideEffects = 1, Defs = [ESP] in +def Int_MemBarrierNoSSE64 : RI<0x09, MRM1r, (outs), (ins GR64:$zero), + "lock\n\t" + "or{q}\t{$zero, (%rsp)|(%rsp), $zero}", + [(X86MemBarrierNoSSE GR64:$zero)]>, + Requires<[In64BitMode]>, LOCK; + + +// Optimized codegen when the non-memory output is not used. +let Defs = [EFLAGS], mayLoad = 1, mayStore = 1 in { +def LOCK_ADD8mr : I<0x00, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src2), + "lock\n\t" + "add{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD16mr : I<0x01, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), + "lock\n\t" + "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; +def LOCK_ADD32mr : I<0x01, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), + "lock\n\t" + "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD64mr : RI<0x01, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2), + "lock\n\t" + "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + +def LOCK_ADD8mi : Ii8<0x80, MRM0m, (outs), (ins i8mem :$dst, i8imm :$src2), + "lock\n\t" + "add{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD16mi : Ii16<0x81, MRM0m, (outs), (ins i16mem:$dst, i16imm:$src2), + "lock\n\t" + "add{w}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD32mi : Ii32<0x81, MRM0m, (outs), (ins i32mem:$dst, i32imm:$src2), + "lock\n\t" + "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD64mi32 : RIi32<0x81, MRM0m, (outs), + (ins i64mem:$dst, i64i32imm :$src2), + "lock\n\t" + "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + +def LOCK_ADD16mi8 : Ii8<0x83, MRM0m, (outs), (ins i16mem:$dst, i16i8imm :$src2), + "lock\n\t" + "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; +def LOCK_ADD32mi8 : Ii8<0x83, MRM0m, (outs), (ins i32mem:$dst, i32i8imm :$src2), + "lock\n\t" + "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_ADD64mi8 : RIi8<0x83, MRM0m, (outs), + (ins i64mem:$dst, i64i8imm :$src2), + "lock\n\t" + "add{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + +def LOCK_SUB8mr : I<0x28, MRMDestMem, (outs), (ins i8mem :$dst, GR8 :$src2), + "lock\n\t" + "sub{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_SUB16mr : I<0x29, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), + "lock\n\t" + "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; +def LOCK_SUB32mr : I<0x29, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), + "lock\n\t" + "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_SUB64mr : RI<0x29, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2), + "lock\n\t" + "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + + +def LOCK_SUB8mi : Ii8<0x80, MRM5m, (outs), (ins i8mem :$dst, i8imm:$src2), + "lock\n\t" + "sub{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_SUB16mi : Ii16<0x81, MRM5m, (outs), (ins i16mem:$dst, i16imm:$src2), + "lock\n\t" + "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; +def LOCK_SUB32mi : Ii32<0x81, MRM5m, (outs), (ins i32mem:$dst, i32imm:$src2), + "lock\n\t" + "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_SUB64mi32 : RIi32<0x81, MRM5m, (outs), + (ins i64mem:$dst, i64i32imm:$src2), + "lock\n\t" + "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + + +def LOCK_SUB16mi8 : Ii8<0x83, MRM5m, (outs), (ins i16mem:$dst, i16i8imm :$src2), + "lock\n\t" + "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; +def LOCK_SUB32mi8 : Ii8<0x83, MRM5m, (outs), (ins i32mem:$dst, i32i8imm :$src2), + "lock\n\t" + "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; +def LOCK_SUB64mi8 : RIi8<0x83, MRM5m, (outs), + (ins i64mem:$dst, i64i8imm :$src2), + "lock\n\t" + "sub{q}\t{$src2, $dst|$dst, $src2}", []>, LOCK; + +def LOCK_INC8m : I<0xFE, MRM0m, (outs), (ins i8mem :$dst), + "lock\n\t" + "inc{b}\t$dst", []>, LOCK; +def LOCK_INC16m : I<0xFF, MRM0m, (outs), (ins i16mem:$dst), + "lock\n\t" + "inc{w}\t$dst", []>, OpSize, LOCK; +def LOCK_INC32m : I<0xFF, MRM0m, (outs), (ins i32mem:$dst), + "lock\n\t" + "inc{l}\t$dst", []>, LOCK; +def LOCK_INC64m : RI<0xFF, MRM0m, (outs), (ins i64mem:$dst), + "lock\n\t" + "inc{q}\t$dst", []>, LOCK; + +def LOCK_DEC8m : I<0xFE, MRM1m, (outs), (ins i8mem :$dst), + "lock\n\t" + "dec{b}\t$dst", []>, LOCK; +def LOCK_DEC16m : I<0xFF, MRM1m, (outs), (ins i16mem:$dst), + "lock\n\t" + "dec{w}\t$dst", []>, OpSize, LOCK; +def LOCK_DEC32m : I<0xFF, MRM1m, (outs), (ins i32mem:$dst), + "lock\n\t" + "dec{l}\t$dst", []>, LOCK; +def LOCK_DEC64m : RI<0xFF, MRM1m, (outs), (ins i64mem:$dst), + "lock\n\t" + "dec{q}\t$dst", []>, LOCK; +} + +// Atomic compare and swap. +let Defs = [EAX, EDX, EFLAGS], Uses = [EAX, EBX, ECX, EDX] in { +def LCMPXCHG8B : I<0xC7, MRM1m, (outs), (ins i64mem:$ptr), + "lock\n\t" + "cmpxchg8b\t$ptr", + [(X86cas8 addr:$ptr)]>, TB, LOCK; +} +let Defs = [AL, EFLAGS], Uses = [AL] in { +def LCMPXCHG8 : I<0xB0, MRMDestMem, (outs), (ins i8mem:$ptr, GR8:$swap), + "lock\n\t" + "cmpxchg{b}\t{$swap, $ptr|$ptr, $swap}", + [(X86cas addr:$ptr, GR8:$swap, 1)]>, TB, LOCK; +} + +let Defs = [AX, EFLAGS], Uses = [AX] in { +def LCMPXCHG16 : I<0xB1, MRMDestMem, (outs), (ins i16mem:$ptr, GR16:$swap), + "lock\n\t" + "cmpxchg{w}\t{$swap, $ptr|$ptr, $swap}", + [(X86cas addr:$ptr, GR16:$swap, 2)]>, TB, OpSize, LOCK; +} + +let Defs = [EAX, EFLAGS], Uses = [EAX] in { +def LCMPXCHG32 : I<0xB1, MRMDestMem, (outs), (ins i32mem:$ptr, GR32:$swap), + "lock\n\t" + "cmpxchg{l}\t{$swap, $ptr|$ptr, $swap}", + [(X86cas addr:$ptr, GR32:$swap, 4)]>, TB, LOCK; +} + +let Defs = [RAX, EFLAGS], Uses = [RAX] in { +def LCMPXCHG64 : RI<0xB1, MRMDestMem, (outs), (ins i64mem:$ptr, GR64:$swap), + "lock\n\t" + "cmpxchgq\t$swap,$ptr", + [(X86cas addr:$ptr, GR64:$swap, 8)]>, TB, LOCK; +} + +// Atomic exchange and add +let Constraints = "$val = $dst", Defs = [EFLAGS] in { +def LXADD8 : I<0xC0, MRMSrcMem, (outs GR8:$dst), (ins GR8:$val, i8mem:$ptr), + "lock\n\t" + "xadd{b}\t{$val, $ptr|$ptr, $val}", + [(set GR8:$dst, (atomic_load_add_8 addr:$ptr, GR8:$val))]>, + TB, LOCK; +def LXADD16 : I<0xC1, MRMSrcMem, (outs GR16:$dst), (ins GR16:$val, i16mem:$ptr), + "lock\n\t" + "xadd{w}\t{$val, $ptr|$ptr, $val}", + [(set GR16:$dst, (atomic_load_add_16 addr:$ptr, GR16:$val))]>, + TB, OpSize, LOCK; +def LXADD32 : I<0xC1, MRMSrcMem, (outs GR32:$dst), (ins GR32:$val, i32mem:$ptr), + "lock\n\t" + "xadd{l}\t{$val, $ptr|$ptr, $val}", + [(set GR32:$dst, (atomic_load_add_32 addr:$ptr, GR32:$val))]>, + TB, LOCK; +def LXADD64 : RI<0xC1, MRMSrcMem, (outs GR64:$dst), (ins GR64:$val,i64mem:$ptr), + "lock\n\t" + "xadd\t$val, $ptr", + [(set GR64:$dst, (atomic_load_add_64 addr:$ptr, GR64:$val))]>, + TB, LOCK; +} + + +//===----------------------------------------------------------------------===// +// DAG Pattern Matching Rules //===----------------------------------------------------------------------===// // ConstantPool GlobalAddress, ExternalSymbol, and JumpTable @@ -1109,6 +1453,4 @@ def : Pat<(and GR64:$src1, i64immSExt8:$src2), (AND64ri8 GR64:$src1, i64immSExt8:$src2)>; def : Pat<(and GR64:$src1, i64immSExt32:$src2), (AND64ri32 GR64:$src1, i64immSExt32:$src2)>; - - \ No newline at end of file diff --git a/lib/Target/X86/X86InstrInfo.td b/lib/Target/X86/X86InstrInfo.td index 81c6dd08311..8b5823d03f3 100644 --- a/lib/Target/X86/X86InstrInfo.td +++ b/lib/Target/X86/X86InstrInfo.td @@ -3495,94 +3495,45 @@ let neverHasSideEffects = 1 in { // Atomic support // -// Memory barriers - -// TODO: Get this to fold the constant into the instruction. -def OR32mrLocked : I<0x09, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$zero), - "lock\n\t" - "or{l}\t{$zero, $dst|$dst, $zero}", - []>, Requires<[In32BitMode]>, LOCK; - -let hasSideEffects = 1 in { -def Int_MemBarrier : I<0, Pseudo, (outs), (ins), - "#MEMBARRIER", - [(X86MemBarrier)]>, Requires<[HasSSE2]>; -} // Atomic swap. These are just normal xchg instructions. But since a memory // operand is referenced, the atomicity is ensured. let Constraints = "$val = $dst" in { -def XCHG32rm : I<0x87, MRMSrcMem, (outs GR32:$dst), - (ins GR32:$val, i32mem:$ptr), - "xchg{l}\t{$val, $ptr|$ptr, $val}", - [(set GR32:$dst, (atomic_swap_32 addr:$ptr, GR32:$val))]>; +def XCHG8rm : I<0x86, MRMSrcMem, (outs GR8:$dst), (ins GR8:$val, i8mem:$ptr), + "xchg{b}\t{$val, $ptr|$ptr, $val}", + [(set GR8:$dst, (atomic_swap_8 addr:$ptr, GR8:$val))]>; def XCHG16rm : I<0x87, MRMSrcMem, (outs GR16:$dst), (ins GR16:$val, i16mem:$ptr), "xchg{w}\t{$val, $ptr|$ptr, $val}", [(set GR16:$dst, (atomic_swap_16 addr:$ptr, GR16:$val))]>, OpSize; -def XCHG8rm : I<0x86, MRMSrcMem, (outs GR8:$dst), (ins GR8:$val, i8mem:$ptr), - "xchg{b}\t{$val, $ptr|$ptr, $val}", - [(set GR8:$dst, (atomic_swap_8 addr:$ptr, GR8:$val))]>; +def XCHG32rm : I<0x87, MRMSrcMem, (outs GR32:$dst), + (ins GR32:$val, i32mem:$ptr), + "xchg{l}\t{$val, $ptr|$ptr, $val}", + [(set GR32:$dst, (atomic_swap_32 addr:$ptr, GR32:$val))]>; +def XCHG64rm : RI<0x87, MRMSrcMem, (outs GR64:$dst), + (ins GR64:$val,i64mem:$ptr), + "xchg{q}\t{$val, $ptr|$ptr, $val}", + [(set GR64:$dst, (atomic_swap_64 addr:$ptr, GR64:$val))]>; -def XCHG32rr : I<0x87, MRMSrcReg, (outs GR32:$dst), (ins GR32:$val, GR32:$src), - "xchg{l}\t{$val, $src|$src, $val}", []>; -def XCHG16rr : I<0x87, MRMSrcReg, (outs GR16:$dst), (ins GR16:$val, GR16:$src), - "xchg{w}\t{$val, $src|$src, $val}", []>, OpSize; def XCHG8rr : I<0x86, MRMSrcReg, (outs GR8:$dst), (ins GR8:$val, GR8:$src), "xchg{b}\t{$val, $src|$src, $val}", []>; +def XCHG16rr : I<0x87, MRMSrcReg, (outs GR16:$dst), (ins GR16:$val, GR16:$src), + "xchg{w}\t{$val, $src|$src, $val}", []>, OpSize; +def XCHG32rr : I<0x87, MRMSrcReg, (outs GR32:$dst), (ins GR32:$val, GR32:$src), + "xchg{l}\t{$val, $src|$src, $val}", []>; +def XCHG64rr : RI<0x87, MRMSrcReg, (outs GR64:$dst), (ins GR64:$val,GR64:$src), + "xchg{q}\t{$val, $src|$src, $val}", []>; } def XCHG16ar : I<0x90, AddRegFrm, (outs), (ins GR16:$src), "xchg{w}\t{$src, %ax|%ax, $src}", []>, OpSize; def XCHG32ar : I<0x90, AddRegFrm, (outs), (ins GR32:$src), "xchg{l}\t{$src, %eax|%eax, $src}", []>; +def XCHG64ar : RI<0x90, AddRegFrm, (outs), (ins GR64:$src), + "xchg{q}\t{$src, %rax|%rax, $src}", []>; + -// Atomic compare and swap. -let Defs = [EAX, EFLAGS], Uses = [EAX] in { -def LCMPXCHG32 : I<0xB1, MRMDestMem, (outs), (ins i32mem:$ptr, GR32:$swap), - "lock\n\t" - "cmpxchg{l}\t{$swap, $ptr|$ptr, $swap}", - [(X86cas addr:$ptr, GR32:$swap, 4)]>, TB, LOCK; -} -let Defs = [EAX, EDX, EFLAGS], Uses = [EAX, EBX, ECX, EDX] in { -def LCMPXCHG8B : I<0xC7, MRM1m, (outs), (ins i64mem:$ptr), - "lock\n\t" - "cmpxchg8b\t$ptr", - [(X86cas8 addr:$ptr)]>, TB, LOCK; -} - -let Defs = [AX, EFLAGS], Uses = [AX] in { -def LCMPXCHG16 : I<0xB1, MRMDestMem, (outs), (ins i16mem:$ptr, GR16:$swap), - "lock\n\t" - "cmpxchg{w}\t{$swap, $ptr|$ptr, $swap}", - [(X86cas addr:$ptr, GR16:$swap, 2)]>, TB, OpSize, LOCK; -} -let Defs = [AL, EFLAGS], Uses = [AL] in { -def LCMPXCHG8 : I<0xB0, MRMDestMem, (outs), (ins i8mem:$ptr, GR8:$swap), - "lock\n\t" - "cmpxchg{b}\t{$swap, $ptr|$ptr, $swap}", - [(X86cas addr:$ptr, GR8:$swap, 1)]>, TB, LOCK; -} - -// Atomic exchange and add -let Constraints = "$val = $dst", Defs = [EFLAGS] in { -def LXADD32 : I<0xC1, MRMSrcMem, (outs GR32:$dst), (ins GR32:$val, i32mem:$ptr), - "lock\n\t" - "xadd{l}\t{$val, $ptr|$ptr, $val}", - [(set GR32:$dst, (atomic_load_add_32 addr:$ptr, GR32:$val))]>, - TB, LOCK; -def LXADD16 : I<0xC1, MRMSrcMem, (outs GR16:$dst), (ins GR16:$val, i16mem:$ptr), - "lock\n\t" - "xadd{w}\t{$val, $ptr|$ptr, $val}", - [(set GR16:$dst, (atomic_load_add_16 addr:$ptr, GR16:$val))]>, - TB, OpSize, LOCK; -def LXADD8 : I<0xC0, MRMSrcMem, (outs GR8:$dst), (ins GR8:$val, i8mem:$ptr), - "lock\n\t" - "xadd{b}\t{$val, $ptr|$ptr, $val}", - [(set GR8:$dst, (atomic_load_add_8 addr:$ptr, GR8:$val))]>, - TB, LOCK; -} def XADD8rr : I<0xC0, MRMDestReg, (outs GR8:$dst), (ins GR8:$src), "xadd{b}\t{$src, $dst|$dst, $src}", []>, TB; @@ -3590,6 +3541,8 @@ def XADD16rr : I<0xC1, MRMDestReg, (outs GR16:$dst), (ins GR16:$src), "xadd{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize; def XADD32rr : I<0xC1, MRMDestReg, (outs GR32:$dst), (ins GR32:$src), "xadd{l}\t{$src, $dst|$dst, $src}", []>, TB; +def XADD64rr : RI<0xC1, MRMDestReg, (outs GR64:$dst), (ins GR64:$src), + "xadd{q}\t{$src, $dst|$dst, $src}", []>, TB; let mayLoad = 1, mayStore = 1 in { def XADD8rm : I<0xC0, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src), @@ -3598,6 +3551,9 @@ def XADD16rm : I<0xC1, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src), "xadd{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize; def XADD32rm : I<0xC1, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src), "xadd{l}\t{$src, $dst|$dst, $src}", []>, TB; +def XADD64rm : RI<0xC1, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), + "xadd{q}\t{$src, $dst|$dst, $src}", []>, TB; + } def CMPXCHG8rr : I<0xB0, MRMDestReg, (outs GR8:$dst), (ins GR8:$src), @@ -3606,6 +3562,8 @@ def CMPXCHG16rr : I<0xB1, MRMDestReg, (outs GR16:$dst), (ins GR16:$src), "cmpxchg{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize; def CMPXCHG32rr : I<0xB1, MRMDestReg, (outs GR32:$dst), (ins GR32:$src), "cmpxchg{l}\t{$src, $dst|$dst, $src}", []>, TB; +def CMPXCHG64rr : RI<0xB1, MRMDestReg, (outs GR64:$dst), (ins GR64:$src), + "cmpxchg{q}\t{$src, $dst|$dst, $src}", []>, TB; let mayLoad = 1, mayStore = 1 in { def CMPXCHG8rm : I<0xB0, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src), @@ -3614,180 +3572,19 @@ def CMPXCHG16rm : I<0xB1, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src), "cmpxchg{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize; def CMPXCHG32rm : I<0xB1, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src), "cmpxchg{l}\t{$src, $dst|$dst, $src}", []>, TB; +def CMPXCHG64rm : RI<0xB1, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), + "cmpxchg{q}\t{$src, $dst|$dst, $src}", []>, TB; } let Defs = [EAX, EDX, EFLAGS], Uses = [EAX, EBX, ECX, EDX] in def CMPXCHG8B : I<0xC7, MRM1m, (outs), (ins i64mem:$dst), "cmpxchg8b\t$dst", []>, TB; -// Optimized codegen when the non-memory output is not used. -// FIXME: Use normal add / sub instructions and add lock prefix dynamically. -let Defs = [EFLAGS], mayLoad = 1, mayStore = 1 in { -def LOCK_ADD8mr : I<0x00, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src2), - "lock\n\t" - "add{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD16mr : I<0x01, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), - "lock\n\t" - "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; -def LOCK_ADD32mr : I<0x01, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), - "lock\n\t" - "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD8mi : Ii8<0x80, MRM0m, (outs), (ins i8mem :$dst, i8imm :$src2), - "lock\n\t" - "add{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD16mi : Ii16<0x81, MRM0m, (outs), (ins i16mem:$dst, i16imm:$src2), - "lock\n\t" - "add{w}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD32mi : Ii32<0x81, MRM0m, (outs), (ins i32mem:$dst, i32imm:$src2), - "lock\n\t" - "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_ADD16mi8 : Ii8<0x83, MRM0m, (outs), (ins i16mem:$dst, i16i8imm :$src2), - "lock\n\t" - "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; -def LOCK_ADD32mi8 : Ii8<0x83, MRM0m, (outs), (ins i32mem:$dst, i32i8imm :$src2), - "lock\n\t" - "add{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; - -def LOCK_INC8m : I<0xFE, MRM0m, (outs), (ins i8mem :$dst), - "lock\n\t" - "inc{b}\t$dst", []>, LOCK; -def LOCK_INC16m : I<0xFF, MRM0m, (outs), (ins i16mem:$dst), - "lock\n\t" - "inc{w}\t$dst", []>, OpSize, LOCK; -def LOCK_INC32m : I<0xFF, MRM0m, (outs), (ins i32mem:$dst), - "lock\n\t" - "inc{l}\t$dst", []>, LOCK; - -def LOCK_SUB8mr : I<0x28, MRMDestMem, (outs), (ins i8mem :$dst, GR8 :$src2), - "lock\n\t" - "sub{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB16mr : I<0x29, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), - "lock\n\t" - "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; -def LOCK_SUB32mr : I<0x29, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), - "lock\n\t" - "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB8mi : Ii8<0x80, MRM5m, (outs), (ins i8mem :$dst, i8imm:$src2), - "lock\n\t" - "sub{b}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB16mi : Ii16<0x81, MRM5m, (outs), (ins i16mem:$dst, i16imm:$src2), - "lock\n\t" - "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; -def LOCK_SUB32mi : Ii32<0x81, MRM5m, (outs), (ins i32mem:$dst, i32imm:$src2), - "lock\n\t" - "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; -def LOCK_SUB16mi8 : Ii8<0x83, MRM5m, (outs), (ins i16mem:$dst, i16i8imm :$src2), - "lock\n\t" - "sub{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize, LOCK; -def LOCK_SUB32mi8 : Ii8<0x83, MRM5m, (outs), (ins i32mem:$dst, i32i8imm :$src2), - "lock\n\t" - "sub{l}\t{$src2, $dst|$dst, $src2}", []>, LOCK; - -def LOCK_DEC8m : I<0xFE, MRM1m, (outs), (ins i8mem :$dst), - "lock\n\t" - "dec{b}\t$dst", []>, LOCK; -def LOCK_DEC16m : I<0xFF, MRM1m, (outs), (ins i16mem:$dst), - "lock\n\t" - "dec{w}\t$dst", []>, OpSize, LOCK; -def LOCK_DEC32m : I<0xFF, MRM1m, (outs), (ins i32mem:$dst), - "lock\n\t" - "dec{l}\t$dst", []>, LOCK; -} - -// Atomic exchange, and, or, xor -let Constraints = "$val = $dst", Defs = [EFLAGS], - usesCustomInserter = 1 in { -def ATOMAND32 : I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMAND32 PSEUDO!", - [(set GR32:$dst, (atomic_load_and_32 addr:$ptr, GR32:$val))]>; -def ATOMOR32 : I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMOR32 PSEUDO!", - [(set GR32:$dst, (atomic_load_or_32 addr:$ptr, GR32:$val))]>; -def ATOMXOR32 : I<0, Pseudo,(outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMXOR32 PSEUDO!", - [(set GR32:$dst, (atomic_load_xor_32 addr:$ptr, GR32:$val))]>; -def ATOMNAND32 : I<0, Pseudo,(outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMNAND32 PSEUDO!", - [(set GR32:$dst, (atomic_load_nand_32 addr:$ptr, GR32:$val))]>; -def ATOMMIN32: I<0, Pseudo, (outs GR32:$dst), (ins i32mem:$ptr, GR32:$val), - "#ATOMMIN32 PSEUDO!", - [(set GR32:$dst, (atomic_load_min_32 addr:$ptr, GR32:$val))]>; -def ATOMMAX32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMMAX32 PSEUDO!", - [(set GR32:$dst, (atomic_load_max_32 addr:$ptr, GR32:$val))]>; -def ATOMUMIN32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMUMIN32 PSEUDO!", - [(set GR32:$dst, (atomic_load_umin_32 addr:$ptr, GR32:$val))]>; -def ATOMUMAX32: I<0, Pseudo, (outs GR32:$dst),(ins i32mem:$ptr, GR32:$val), - "#ATOMUMAX32 PSEUDO!", - [(set GR32:$dst, (atomic_load_umax_32 addr:$ptr, GR32:$val))]>; - -def ATOMAND16 : I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMAND16 PSEUDO!", - [(set GR16:$dst, (atomic_load_and_16 addr:$ptr, GR16:$val))]>; -def ATOMOR16 : I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMOR16 PSEUDO!", - [(set GR16:$dst, (atomic_load_or_16 addr:$ptr, GR16:$val))]>; -def ATOMXOR16 : I<0, Pseudo,(outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMXOR16 PSEUDO!", - [(set GR16:$dst, (atomic_load_xor_16 addr:$ptr, GR16:$val))]>; -def ATOMNAND16 : I<0, Pseudo,(outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMNAND16 PSEUDO!", - [(set GR16:$dst, (atomic_load_nand_16 addr:$ptr, GR16:$val))]>; -def ATOMMIN16: I<0, Pseudo, (outs GR16:$dst), (ins i16mem:$ptr, GR16:$val), - "#ATOMMIN16 PSEUDO!", - [(set GR16:$dst, (atomic_load_min_16 addr:$ptr, GR16:$val))]>; -def ATOMMAX16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMMAX16 PSEUDO!", - [(set GR16:$dst, (atomic_load_max_16 addr:$ptr, GR16:$val))]>; -def ATOMUMIN16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMUMIN16 PSEUDO!", - [(set GR16:$dst, (atomic_load_umin_16 addr:$ptr, GR16:$val))]>; -def ATOMUMAX16: I<0, Pseudo, (outs GR16:$dst),(ins i16mem:$ptr, GR16:$val), - "#ATOMUMAX16 PSEUDO!", - [(set GR16:$dst, (atomic_load_umax_16 addr:$ptr, GR16:$val))]>; - -def ATOMAND8 : I<0, Pseudo, (outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), - "#ATOMAND8 PSEUDO!", - [(set GR8:$dst, (atomic_load_and_8 addr:$ptr, GR8:$val))]>; -def ATOMOR8 : I<0, Pseudo, (outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), - "#ATOMOR8 PSEUDO!", - [(set GR8:$dst, (atomic_load_or_8 addr:$ptr, GR8:$val))]>; -def ATOMXOR8 : I<0, Pseudo,(outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), - "#ATOMXOR8 PSEUDO!", - [(set GR8:$dst, (atomic_load_xor_8 addr:$ptr, GR8:$val))]>; -def ATOMNAND8 : I<0, Pseudo,(outs GR8:$dst),(ins i8mem:$ptr, GR8:$val), - "#ATOMNAND8 PSEUDO!", - [(set GR8:$dst, (atomic_load_nand_8 addr:$ptr, GR8:$val))]>; -} - -let Constraints = "$val1 = $dst1, $val2 = $dst2", - Defs = [EFLAGS, EAX, EBX, ECX, EDX], - Uses = [EAX, EBX, ECX, EDX], - mayLoad = 1, mayStore = 1, - usesCustomInserter = 1 in { -def ATOMAND6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMAND6432 PSEUDO!", []>; -def ATOMOR6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMOR6432 PSEUDO!", []>; -def ATOMXOR6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMXOR6432 PSEUDO!", []>; -def ATOMNAND6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMNAND6432 PSEUDO!", []>; -def ATOMADD6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMADD6432 PSEUDO!", []>; -def ATOMSUB6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMSUB6432 PSEUDO!", []>; -def ATOMSWAP6432 : I<0, Pseudo, (outs GR32:$dst1, GR32:$dst2), - (ins i64mem:$ptr, GR32:$val1, GR32:$val2), - "#ATOMSWAP6432 PSEUDO!", []>; -} +let Defs = [RAX, RDX, EFLAGS], Uses = [RAX, RBX, RCX, RDX] in +def CMPXCHG16B : RI<0xC7, MRM1m, (outs), (ins i128mem:$dst), + "cmpxchg16b\t$dst", []>, TB; + + // Lock instruction prefix def LOCK_PREFIX : I<0xF0, RawFrm, (outs), (ins), "lock", []>; @@ -3807,6 +3604,7 @@ def REPNE_PREFIX : I<0xF2, RawFrm, (outs), (ins), "repne", []>; def LODSB : I<0xAC, RawFrm, (outs), (ins), "lodsb", []>; def LODSW : I<0xAD, RawFrm, (outs), (ins), "lodsw", []>, OpSize; def LODSD : I<0xAD, RawFrm, (outs), (ins), "lods{l|d}", []>; +def LODSQ : RI<0xAD, RawFrm, (outs), (ins), "lodsq", []>; def OUTSB : I<0x6E, RawFrm, (outs), (ins), "outsb", []>; def OUTSW : I<0x6F, RawFrm, (outs), (ins), "outsw", []>, OpSize; diff --git a/lib/Target/X86/X86InstrSystem.td b/lib/Target/X86/X86InstrSystem.td index 68f348ae239..2d27209f5fa 100644 --- a/lib/Target/X86/X86InstrSystem.td +++ b/lib/Target/X86/X86InstrSystem.td @@ -331,6 +331,14 @@ def SLDT16m : I<0x00, MRM0m, (outs i16mem:$dst), (ins), "sldt{w}\t$dst", []>, TB; def SLDT32r : I<0x00, MRM0r, (outs GR32:$dst), (ins), "sldt{l}\t$dst", []>, TB; + +// LLDT is not interpreted specially in 64-bit mode because there is no sign +// extension. +def SLDT64r : RI<0x00, MRM0r, (outs GR64:$dst), (ins), + "sldt{q}\t$dst", []>, TB; +def SLDT64m : RI<0x00, MRM0m, (outs i16mem:$dst), (ins), + "sldt{q}\t$dst", []>, TB; + def LGDTm : I<0x01, MRM2m, (outs), (ins opaque48mem:$src), "lgdt\t$src", []>, TB; def LIDTm : I<0x01, MRM3m, (outs), (ins opaque48mem:$src), -- 2.34.1