def SDT_X86SEG_ALLOCA : SDTypeProfile<1, 1, [SDTCisVT<0, iPTR>, SDTCisVT<1, iPTR>]>;
-def SDT_X86WIN_FTOL : SDTypeProfile<0, 1, [SDTCisFP<0>]>;
-
def SDT_X86EHRET : SDTypeProfile<0, 1, [SDTCisInt<0>]>;
def SDT_X86TCRET : SDTypeProfile<0, 2, [SDTCisPtrTy<0>, SDTCisVT<1, i32>]>;
def X86TLSCall : SDNode<"X86ISD::TLSCALL", SDT_X86TLSCALL,
[SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>;
-def X86WinFTOL : SDNode<"X86ISD::WIN_FTOL", SDT_X86WIN_FTOL,
- [SDNPHasChain, SDNPOutGlue]>;
-
//===----------------------------------------------------------------------===//
// X86 Operand Definitions.
//
def vz32mem : X86VMemOperand<VR512, "printi32mem", X86MemVZ32Operand>;
def vz64mem : X86VMemOperand<VR512, "printi64mem", X86MemVZ64Operand>;
-// A version of i8mem for use on x86-64 that uses GR64_NOREX instead of
-// plain GR64, so that it doesn't potentially require a REX prefix.
-def i8mem_NOREX : Operand<i64> {
+// A version of i8mem for use on x86-64 and x32 that uses a NOREX GPR instead
+// of a plain GPR, so that it doesn't potentially require a REX prefix.
+def ptr_rc_norex : PointerLikeRegClass<2>;
+def ptr_rc_norex_nosp : PointerLikeRegClass<3>;
+
+def i8mem_NOREX : Operand<iPTR> {
let PrintMethod = "printi8mem";
- let MIOperandInfo = (ops GR64_NOREX, i8imm, GR64_NOREX_NOSP, i32imm, i8imm);
+ let MIOperandInfo = (ops ptr_rc_norex, i8imm, ptr_rc_norex_nosp, i32imm, i8imm);
let ParserMatchClass = X86Mem8AsmOperand;
let OperandType = "OPERAND_MEMORY";
}
// GPRs available for tailcall.
// It represents GR32_TC, GR64_TC or GR64_TCW64.
-def ptr_rc_tailcall : PointerLikeRegClass<2>;
+def ptr_rc_tailcall : PointerLikeRegClass<4>;
// Special i32mem for addresses of load folding tail calls. These are not
// allowed to use callee-saved registers since they must be scheduled
// X86 Complex Pattern Definitions.
//
-// Define X86 specific addressing mode.
-def addr : ComplexPattern<iPTR, 5, "SelectAddr", [], [SDNPWantParent]>;
-def lea32addr : ComplexPattern<i32, 5, "SelectLEAAddr",
+// Define X86-specific addressing mode.
+def addr : ComplexPattern<iPTR, 5, "selectAddr", [], [SDNPWantParent]>;
+def lea32addr : ComplexPattern<i32, 5, "selectLEAAddr",
[add, sub, mul, X86mul_imm, shl, or, frameindex],
[]>;
// In 64-bit mode 32-bit LEAs can use RIP-relative addressing.
-def lea64_32addr : ComplexPattern<i32, 5, "SelectLEA64_32Addr",
+def lea64_32addr : ComplexPattern<i32, 5, "selectLEA64_32Addr",
[add, sub, mul, X86mul_imm, shl, or,
frameindex, X86WrapperRIP],
[]>;
-def tls32addr : ComplexPattern<i32, 5, "SelectTLSADDRAddr",
+def tls32addr : ComplexPattern<i32, 5, "selectTLSADDRAddr",
[tglobaltlsaddr], []>;
-def tls32baseaddr : ComplexPattern<i32, 5, "SelectTLSADDRAddr",
+def tls32baseaddr : ComplexPattern<i32, 5, "selectTLSADDRAddr",
[tglobaltlsaddr], []>;
-def lea64addr : ComplexPattern<i64, 5, "SelectLEAAddr",
+def lea64addr : ComplexPattern<i64, 5, "selectLEAAddr",
[add, sub, mul, X86mul_imm, shl, or, frameindex,
X86WrapperRIP], []>;
-def tls64addr : ComplexPattern<i64, 5, "SelectTLSADDRAddr",
+def tls64addr : ComplexPattern<i64, 5, "selectTLSADDRAddr",
[tglobaltlsaddr], []>;
-def tls64baseaddr : ComplexPattern<i64, 5, "SelectTLSADDRAddr",
+def tls64baseaddr : ComplexPattern<i64, 5, "selectTLSADDRAddr",
[tglobaltlsaddr], []>;
-def vectoraddr : ComplexPattern<iPTR, 5, "SelectVectorAddr", [],[SDNPWantParent]>;
+def vectoraddr : ComplexPattern<iPTR, 5, "selectVectorAddr", [],[SDNPWantParent]>;
//===----------------------------------------------------------------------===//
// X86 Instruction Predicate Definitions.
def HasPOPCNT : Predicate<"Subtarget->hasPOPCNT()">;
def HasAES : Predicate<"Subtarget->hasAES()">;
+def HasFXSR : Predicate<"Subtarget->hasFXSR()">;
+def HasXSAVE : Predicate<"Subtarget->hasXSAVE()">;
+def HasXSAVEOPT : Predicate<"Subtarget->hasXSAVEOPT()">;
+def HasXSAVEC : Predicate<"Subtarget->hasXSAVEC()">;
+def HasXSAVES : Predicate<"Subtarget->hasXSAVES()">;
def HasPCLMUL : Predicate<"Subtarget->hasPCLMUL()">;
def HasFMA : Predicate<"Subtarget->hasFMA()">;
def UseFMAOnAVX : Predicate<"Subtarget->hasFMA() && !Subtarget->hasAVX512()">;
}]>;
-def i16immSExt8 : ImmLeaf<i16, [{ return Imm == (int8_t)Imm; }]>;
-def i32immSExt8 : ImmLeaf<i32, [{ return Imm == (int8_t)Imm; }]>;
-def i64immSExt8 : ImmLeaf<i64, [{ return Imm == (int8_t)Imm; }]>;
+def i16immSExt8 : ImmLeaf<i16, [{ return isInt<8>(Imm); }]>;
+def i32immSExt8 : ImmLeaf<i32, [{ return isInt<8>(Imm); }]>;
+def i64immSExt8 : ImmLeaf<i64, [{ return isInt<8>(Imm); }]>;
+
+// If we have multiple users of an immediate, it's much smaller to reuse
+// the register, rather than encode the immediate in every instruction.
+// This has the risk of increasing register pressure from stretched live
+// ranges, however, the immediates should be trivial to rematerialize by
+// the RA in the event of high register pressure.
+// TODO : This is currently enabled for stores and binary ops. There are more
+// cases for which this can be enabled, though this catches the bulk of the
+// issues.
+// TODO2 : This should really also be enabled under O2, but there's currently
+// an issue with RA where we don't pull the constants into their users
+// when we rematerialize them. I'll follow-up on enabling O2 after we fix that
+// issue.
+// TODO3 : This is currently limited to single basic blocks (DAG creation
+// pulls block immediates to the top and merges them if necessary).
+// Eventually, it would be nice to allow ConstantHoisting to merge constants
+// globally for potentially added savings.
+//
+def imm8_su : PatLeaf<(i8 imm), [{
+ return !shouldAvoidImmediateInstFormsForSize(N);
+}]>;
+def imm16_su : PatLeaf<(i16 imm), [{
+ return !shouldAvoidImmediateInstFormsForSize(N);
+}]>;
+def imm32_su : PatLeaf<(i32 imm), [{
+ return !shouldAvoidImmediateInstFormsForSize(N);
+}]>;
+
+def i16immSExt8_su : PatLeaf<(i16immSExt8), [{
+ return !shouldAvoidImmediateInstFormsForSize(N);
+}]>;
+def i32immSExt8_su : PatLeaf<(i32immSExt8), [{
+ return !shouldAvoidImmediateInstFormsForSize(N);
+}]>;
-def i64immSExt32 : ImmLeaf<i64, [{ return Imm == (int32_t)Imm; }]>;
+def i64immSExt32 : ImmLeaf<i64, [{ return isInt<32>(Imm); }]>;
// i64immZExt32 predicate - True if the 64-bit immediate fits in a 32-bit
// unsigned field.
-def i64immZExt32 : ImmLeaf<i64, [{ return (uint64_t)Imm == (uint32_t)Imm; }]>;
+def i64immZExt32 : ImmLeaf<i64, [{ return isUInt<32>(Imm); }]>;
def i64immZExt32SExt8 : ImmLeaf<i64, [{
- return (uint64_t)Imm == (uint32_t)Imm && (int32_t)Imm == (int8_t)Imm;
+ return isUInt<32>(Imm) && isInt<8>(static_cast<int32_t>(Imm));
}]>;
// Helper fragments for loads.
let SchedRW = [WriteStore] in {
def MOV8mi : Ii8 <0xC6, MRM0m, (outs), (ins i8mem :$dst, i8imm :$src),
"mov{b}\t{$src, $dst|$dst, $src}",
- [(store (i8 imm:$src), addr:$dst)], IIC_MOV_MEM>;
+ [(store (i8 imm8_su:$src), addr:$dst)], IIC_MOV_MEM>;
def MOV16mi : Ii16<0xC7, MRM0m, (outs), (ins i16mem:$dst, i16imm:$src),
"mov{w}\t{$src, $dst|$dst, $src}",
- [(store (i16 imm:$src), addr:$dst)], IIC_MOV_MEM>, OpSize16;
+ [(store (i16 imm16_su:$src), addr:$dst)], IIC_MOV_MEM>, OpSize16;
def MOV32mi : Ii32<0xC7, MRM0m, (outs), (ins i32mem:$dst, i32imm:$src),
"mov{l}\t{$src, $dst|$dst, $src}",
- [(store (i32 imm:$src), addr:$dst)], IIC_MOV_MEM>, OpSize32;
+ [(store (i32 imm32_su:$src), addr:$dst)], IIC_MOV_MEM>, OpSize32;
def MOV64mi32 : RIi32S<0xC7, MRM0m, (outs), (ins i64mem:$dst, i64i32imm:$src),
"mov{q}\t{$src, $dst|$dst, $src}",
[(store i64immSExt32:$src, addr:$dst)], IIC_MOV_MEM>;
def : MnemonicAlias<"xsaveq", "xsave64", "att">;
def : MnemonicAlias<"xrstorq", "xrstor64", "att">;
def : MnemonicAlias<"xsaveoptq", "xsaveopt64", "att">;
-
+def : MnemonicAlias<"xrstorsq", "xrstors64", "att">;
+def : MnemonicAlias<"xsavecq", "xsavec64", "att">;
+def : MnemonicAlias<"xsavesq", "xsaves64", "att">;
class CondCodeAlias<string Prefix,string Suffix, string OldCond, string NewCond,
string VariantName>
//===----------------------------------------------------------------------===//
// aad/aam default to base 10 if no operand is specified.
-def : InstAlias<"aad", (AAD8i8 10)>;
-def : InstAlias<"aam", (AAM8i8 10)>;
+def : InstAlias<"aad", (AAD8i8 10)>, Requires<[Not64BitMode]>;
+def : InstAlias<"aam", (AAM8i8 10)>, Requires<[Not64BitMode]>;
// Disambiguate the mem/imm form of bt-without-a-suffix as btl.
// Likewise for btc/btr/bts.
def : InstAlias<"inb\t{%dx|dx}", (IN8rr), 0>;
def : InstAlias<"inw\t{%dx|dx}", (IN16rr), 0>;
def : InstAlias<"inl\t{%dx|dx}", (IN32rr), 0>;
-def : InstAlias<"inb\t$port", (IN8ri i8imm:$port), 0>;
-def : InstAlias<"inw\t$port", (IN16ri i8imm:$port), 0>;
-def : InstAlias<"inl\t$port", (IN32ri i8imm:$port), 0>;
+def : InstAlias<"inb\t$port", (IN8ri u8imm:$port), 0>;
+def : InstAlias<"inw\t$port", (IN16ri u8imm:$port), 0>;
+def : InstAlias<"inl\t$port", (IN32ri u8imm:$port), 0>;
// jmp and call aliases for lcall and ljmp. jmp $42,$5 -> ljmp
def : InstAlias<"outb\t{%dx|dx}", (OUT8rr), 0>;
def : InstAlias<"outw\t{%dx|dx}", (OUT16rr), 0>;
def : InstAlias<"outl\t{%dx|dx}", (OUT32rr), 0>;
-def : InstAlias<"outb\t$port", (OUT8ir i8imm:$port), 0>;
-def : InstAlias<"outw\t$port", (OUT16ir i8imm:$port), 0>;
-def : InstAlias<"outl\t$port", (OUT32ir i8imm:$port), 0>;
+def : InstAlias<"outb\t$port", (OUT8ir u8imm:$port), 0>;
+def : InstAlias<"outw\t$port", (OUT16ir u8imm:$port), 0>;
+def : InstAlias<"outl\t$port", (OUT32ir u8imm:$port), 0>;
// 'sldt <mem>' can be encoded with either sldtw or sldtq with the same
// effect (both store to a 16-bit mem). Force to sldtw to avoid ambiguity
def : InstAlias<"xchg{l}\t{%eax, $src|$src, eax}",
(XCHG32ar64 GR32_NOAX:$src), 0>, Requires<[In64BitMode]>;
def : InstAlias<"xchg{q}\t{%rax, $src|$src, rax}", (XCHG64ar GR64:$src), 0>;
+
+// These aliases exist to get the parser to prioritize matching 8-bit
+// immediate encodings over matching the implicit ax/eax/rax encodings. By
+// explicitly mentioning the A register here, these entries will be ordered
+// first due to the more explicit immediate type.
+def : InstAlias<"adc{w}\t{$imm, %ax|ax, $imm}", (ADC16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"add{w}\t{$imm, %ax|ax, $imm}", (ADD16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"and{w}\t{$imm, %ax|ax, $imm}", (AND16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"cmp{w}\t{$imm, %ax|ax, $imm}", (CMP16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"or{w}\t{$imm, %ax|ax, $imm}", (OR16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"sbb{w}\t{$imm, %ax|ax, $imm}", (SBB16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"sub{w}\t{$imm, %ax|ax, $imm}", (SUB16ri8 AX, i16i8imm:$imm), 0>;
+def : InstAlias<"xor{w}\t{$imm, %ax|ax, $imm}", (XOR16ri8 AX, i16i8imm:$imm), 0>;
+
+def : InstAlias<"adc{l}\t{$imm, %eax|eax, $imm}", (ADC32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"add{l}\t{$imm, %eax|eax, $imm}", (ADD32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"and{l}\t{$imm, %eax|eax, $imm}", (AND32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"cmp{l}\t{$imm, %eax|eax, $imm}", (CMP32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"or{l}\t{$imm, %eax|eax, $imm}", (OR32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"sbb{l}\t{$imm, %eax|eax, $imm}", (SBB32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"sub{l}\t{$imm, %eax|eax, $imm}", (SUB32ri8 EAX, i32i8imm:$imm), 0>;
+def : InstAlias<"xor{l}\t{$imm, %eax|eax, $imm}", (XOR32ri8 EAX, i32i8imm:$imm), 0>;
+
+def : InstAlias<"adc{q}\t{$imm, %rax|rax, $imm}", (ADC64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"add{q}\t{$imm, %rax|rax, $imm}", (ADD64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"and{q}\t{$imm, %rax|rax, $imm}", (AND64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"cmp{q}\t{$imm, %rax|rax, $imm}", (CMP64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"or{q}\t{$imm, %rax|rax, $imm}", (OR64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"sbb{q}\t{$imm, %rax|rax, $imm}", (SBB64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"sub{q}\t{$imm, %rax|rax, $imm}", (SUB64ri8 RAX, i64i8imm:$imm), 0>;
+def : InstAlias<"xor{q}\t{$imm, %rax|rax, $imm}", (XOR64ri8 RAX, i64i8imm:$imm), 0>;