def SDTX86CwdStore : SDTypeProfile<0, 1, [SDTCisPtrTy<0>]>;
def X86fld : SDNode<"X86ISD::FLD", SDTX86Fld,
- [SDNPHasChain, SDNPMayLoad]>;
+ [SDNPHasChain, SDNPMayLoad, SDNPMemOperand]>;
def X86fst : SDNode<"X86ISD::FST", SDTX86Fst,
- [SDNPHasChain, SDNPInFlag, SDNPMayStore]>;
+ [SDNPHasChain, SDNPInGlue, SDNPMayStore,
+ SDNPMemOperand]>;
def X86fild : SDNode<"X86ISD::FILD", SDTX86Fild,
- [SDNPHasChain, SDNPMayLoad]>;
+ [SDNPHasChain, SDNPMayLoad, SDNPMemOperand]>;
def X86fildflag : SDNode<"X86ISD::FILD_FLAG", SDTX86Fild,
- [SDNPHasChain, SDNPOutFlag, SDNPMayLoad]>;
+ [SDNPHasChain, SDNPOutGlue, SDNPMayLoad,
+ SDNPMemOperand]>;
def X86fp_to_i16mem : SDNode<"X86ISD::FP_TO_INT16_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain, SDNPMayStore]>;
+ [SDNPHasChain, SDNPMayStore, SDNPMemOperand]>;
def X86fp_to_i32mem : SDNode<"X86ISD::FP_TO_INT32_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain, SDNPMayStore]>;
+ [SDNPHasChain, SDNPMayStore, SDNPMemOperand]>;
def X86fp_to_i64mem : SDNode<"X86ISD::FP_TO_INT64_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain, SDNPMayStore]>;
+ [SDNPHasChain, SDNPMayStore, SDNPMemOperand]>;
def X86fp_cwd_get16 : SDNode<"X86ISD::FNSTCW16m", SDTX86CwdStore,
- [SDNPHasChain, SDNPMayStore, SDNPSideEffect]>;
+ [SDNPHasChain, SDNPMayStore, SDNPSideEffect,
+ SDNPMemOperand]>;
//===----------------------------------------------------------------------===//
// FPStack pattern fragments
// Some 'special' instructions
let usesCustomInserter = 1 in { // Expanded after instruction selection.
- def FP32_TO_INT16_IN_MEM : I<0, Pseudo,
- (outs), (ins i16mem:$dst, RFP32:$src),
- "##FP32_TO_INT16_IN_MEM PSEUDO!",
+ def FP32_TO_INT16_IN_MEM : PseudoI<(outs), (ins i16mem:$dst, RFP32:$src),
[(X86fp_to_i16mem RFP32:$src, addr:$dst)]>;
- def FP32_TO_INT32_IN_MEM : I<0, Pseudo,
- (outs), (ins i32mem:$dst, RFP32:$src),
- "##FP32_TO_INT32_IN_MEM PSEUDO!",
+ def FP32_TO_INT32_IN_MEM : PseudoI<(outs), (ins i32mem:$dst, RFP32:$src),
[(X86fp_to_i32mem RFP32:$src, addr:$dst)]>;
- def FP32_TO_INT64_IN_MEM : I<0, Pseudo,
- (outs), (ins i64mem:$dst, RFP32:$src),
- "##FP32_TO_INT64_IN_MEM PSEUDO!",
+ def FP32_TO_INT64_IN_MEM : PseudoI<(outs), (ins i64mem:$dst, RFP32:$src),
[(X86fp_to_i64mem RFP32:$src, addr:$dst)]>;
- def FP64_TO_INT16_IN_MEM : I<0, Pseudo,
- (outs), (ins i16mem:$dst, RFP64:$src),
- "##FP64_TO_INT16_IN_MEM PSEUDO!",
+ def FP64_TO_INT16_IN_MEM : PseudoI<(outs), (ins i16mem:$dst, RFP64:$src),
[(X86fp_to_i16mem RFP64:$src, addr:$dst)]>;
- def FP64_TO_INT32_IN_MEM : I<0, Pseudo,
- (outs), (ins i32mem:$dst, RFP64:$src),
- "##FP64_TO_INT32_IN_MEM PSEUDO!",
+ def FP64_TO_INT32_IN_MEM : PseudoI<(outs), (ins i32mem:$dst, RFP64:$src),
[(X86fp_to_i32mem RFP64:$src, addr:$dst)]>;
- def FP64_TO_INT64_IN_MEM : I<0, Pseudo,
- (outs), (ins i64mem:$dst, RFP64:$src),
- "##FP64_TO_INT64_IN_MEM PSEUDO!",
+ def FP64_TO_INT64_IN_MEM : PseudoI<(outs), (ins i64mem:$dst, RFP64:$src),
[(X86fp_to_i64mem RFP64:$src, addr:$dst)]>;
- def FP80_TO_INT16_IN_MEM : I<0, Pseudo,
- (outs), (ins i16mem:$dst, RFP80:$src),
- "##FP80_TO_INT16_IN_MEM PSEUDO!",
+ def FP80_TO_INT16_IN_MEM : PseudoI<(outs), (ins i16mem:$dst, RFP80:$src),
[(X86fp_to_i16mem RFP80:$src, addr:$dst)]>;
- def FP80_TO_INT32_IN_MEM : I<0, Pseudo,
- (outs), (ins i32mem:$dst, RFP80:$src),
- "##FP80_TO_INT32_IN_MEM PSEUDO!",
+ def FP80_TO_INT32_IN_MEM : PseudoI<(outs), (ins i32mem:$dst, RFP80:$src),
[(X86fp_to_i32mem RFP80:$src, addr:$dst)]>;
- def FP80_TO_INT64_IN_MEM : I<0, Pseudo,
- (outs), (ins i64mem:$dst, RFP80:$src),
- "##FP80_TO_INT64_IN_MEM PSEUDO!",
+ def FP80_TO_INT64_IN_MEM : PseudoI<(outs), (ins i64mem:$dst, RFP80:$src),
[(X86fp_to_i64mem RFP80:$src, addr:$dst)]>;
}
-let isTerminator = 1 in
- let Defs = [FP0, FP1, FP2, FP3, FP4, FP5, FP6] in
- def FP_REG_KILL : I<0, Pseudo, (outs), (ins), "##FP_REG_KILL", []>;
-
// All FP Stack operations are represented with four instructions here. The
// first three instructions, generated by the instruction selector, use "RFP32"
// "RFP64" or "RFP80" registers: traditional register files to reference 32-bit,
// a pattern) and the FPI instruction should have emission info (e.g. opcode
// encoding and asm printing info).
-// Pseudo Instructions for FP stack return values.
-def FpGET_ST0_32 : FpI_<(outs RFP32:$dst), (ins), SpecialFP, []>; // FPR = ST(0)
-def FpGET_ST0_64 : FpI_<(outs RFP64:$dst), (ins), SpecialFP, []>; // FPR = ST(0)
-def FpGET_ST0_80 : FpI_<(outs RFP80:$dst), (ins), SpecialFP, []>; // FPR = ST(0)
-
-// FpGET_ST1* should only be issued *after* an FpGET_ST0* has been issued when
-// there are two values live out on the stack from a call or inlineasm. This
-// magic is handled by the stackifier. It is not valid to emit FpGET_ST1* and
-// then FpGET_ST0*. In addition, it is invalid for any FP-using operations to
-// occur between them.
-def FpGET_ST1_32 : FpI_<(outs RFP32:$dst), (ins), SpecialFP, []>; // FPR = ST(1)
-def FpGET_ST1_64 : FpI_<(outs RFP64:$dst), (ins), SpecialFP, []>; // FPR = ST(1)
-def FpGET_ST1_80 : FpI_<(outs RFP80:$dst), (ins), SpecialFP, []>; // FPR = ST(1)
-
-let Defs = [ST0] in {
-def FpSET_ST0_32 : FpI_<(outs), (ins RFP32:$src), SpecialFP, []>; // ST(0) = FPR
-def FpSET_ST0_64 : FpI_<(outs), (ins RFP64:$src), SpecialFP, []>; // ST(0) = FPR
-def FpSET_ST0_80 : FpI_<(outs), (ins RFP80:$src), SpecialFP, []>; // ST(0) = FPR
-}
-
-let Defs = [ST1] in {
-def FpSET_ST1_32 : FpI_<(outs), (ins RFP32:$src), SpecialFP, []>; // ST(1) = FPR
-def FpSET_ST1_64 : FpI_<(outs), (ins RFP64:$src), SpecialFP, []>; // ST(1) = FPR
-def FpSET_ST1_80 : FpI_<(outs), (ins RFP80:$src), SpecialFP, []>; // ST(1) = FPR
-}
+// Pseudo Instruction for FP stack return values.
+def FpPOP_RETVAL : FpI_<(outs RFP80:$dst), (ins), SpecialFP, []>;
-// FpIf32, FpIf64 - Floating Point Psuedo Instruction template.
+// FpIf32, FpIf64 - Floating Point Pseudo Instruction template.
// f32 instructions can use SSE1 and are predicated on FPStackf32 == !SSE1.
// f64 instructions can use SSE2 and are predicated on FPStackf64 == !SSE2.
// f80 instructions cannot use SSE and use neither of these.
class FpIf64<dag outs, dag ins, FPFormat fp, list<dag> pattern> :
FpI_<outs, ins, fp, pattern>, Requires<[FPStackf64]>;
-// Register copies. Just copies, the shortening ones do not truncate.
-let neverHasSideEffects = 1 in {
- def MOV_Fp3232 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src), SpecialFP, []>;
- def MOV_Fp3264 : FpIf32<(outs RFP64:$dst), (ins RFP32:$src), SpecialFP, []>;
- def MOV_Fp6432 : FpIf32<(outs RFP32:$dst), (ins RFP64:$src), SpecialFP, []>;
- def MOV_Fp6464 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src), SpecialFP, []>;
- def MOV_Fp8032 : FpIf32<(outs RFP32:$dst), (ins RFP80:$src), SpecialFP, []>;
- def MOV_Fp3280 : FpIf32<(outs RFP80:$dst), (ins RFP32:$src), SpecialFP, []>;
- def MOV_Fp8064 : FpIf64<(outs RFP64:$dst), (ins RFP80:$src), SpecialFP, []>;
- def MOV_Fp6480 : FpIf64<(outs RFP80:$dst), (ins RFP64:$src), SpecialFP, []>;
- def MOV_Fp8080 : FpI_ <(outs RFP80:$dst), (ins RFP80:$src), SpecialFP, []>;
-}
-
// Factoring for arithmetic.
multiclass FPBinary_rr<SDNode OpNode> {
// Register op register -> register
[(set RFP80:$dst,
(OpNode RFP80:$src1, (f80 (extloadf64 addr:$src2))))]>;
def _F32m : FPI<0xD8, fp, (outs), (ins f32mem:$src),
- !strconcat("f", !strconcat(asmstring, "{s}\t$src"))> {
+ !strconcat("f", asmstring, "{s}\t$src")> {
let mayLoad = 1;
}
def _F64m : FPI<0xDC, fp, (outs), (ins f64mem:$src),
- !strconcat("f", !strconcat(asmstring, "{l}\t$src"))> {
+ !strconcat("f", asmstring, "{l}\t$src")> {
let mayLoad = 1;
}
// ST(0) = ST(0) + [memint]
[(set RFP80:$dst, (OpNode RFP80:$src1,
(X86fild addr:$src2, i32)))]>;
def _FI16m : FPI<0xDE, fp, (outs), (ins i16mem:$src),
- !strconcat("fi", !strconcat(asmstring, "{s}\t$src"))> {
+ !strconcat("fi", asmstring, "{s}\t$src")> {
let mayLoad = 1;
}
def _FI32m : FPI<0xDA, fp, (outs), (ins i32mem:$src),
- !strconcat("fi", !strconcat(asmstring, "{l}\t$src"))> {
+ !strconcat("fi", asmstring, "{l}\t$src")> {
let mayLoad = 1;
}
}
// Versions of FP instructions that take a single memory operand. Added for the
// disassembler; remove as they are included with patterns elsewhere.
-def FCOM32m : FPI<0xD8, MRM2m, (outs), (ins f32mem:$src), "fcom{l}\t$src">;
-def FCOMP32m : FPI<0xD8, MRM3m, (outs), (ins f32mem:$src), "fcomp{l}\t$src">;
+def FCOM32m : FPI<0xD8, MRM2m, (outs), (ins f32mem:$src), "fcom{s}\t$src">;
+def FCOMP32m : FPI<0xD8, MRM3m, (outs), (ins f32mem:$src), "fcomp{s}\t$src">;
def FLDENVm : FPI<0xD9, MRM4m, (outs), (ins f32mem:$src), "fldenv\t$src">;
def FSTENVm : FPI<0xD9, MRM6m, (outs f32mem:$dst), (ins), "fnstenv\t$dst">;
def FICOM32m : FPI<0xDA, MRM2m, (outs), (ins i32mem:$src), "ficom{l}\t$src">;
def FICOMP32m: FPI<0xDA, MRM3m, (outs), (ins i32mem:$src), "ficomp{l}\t$src">;
-def FCOM64m : FPI<0xDC, MRM2m, (outs), (ins f64mem:$src), "fcom{ll}\t$src">;
-def FCOMP64m : FPI<0xDC, MRM3m, (outs), (ins f64mem:$src), "fcomp{ll}\t$src">;
+def FCOM64m : FPI<0xDC, MRM2m, (outs), (ins f64mem:$src), "fcom{l}\t$src">;
+def FCOMP64m : FPI<0xDC, MRM3m, (outs), (ins f64mem:$src), "fcomp{l}\t$src">;
def FRSTORm : FPI<0xDD, MRM4m, (outs f32mem:$dst), (ins), "frstor\t$dst">;
def FSAVEm : FPI<0xDD, MRM6m, (outs f32mem:$dst), (ins), "fnsave\t$dst">;
def FNSTSWm : FPI<0xDD, MRM7m, (outs f32mem:$dst), (ins), "fnstsw\t$dst">;
-def FICOM16m : FPI<0xDE, MRM2m, (outs), (ins i16mem:$src), "ficom{w}\t$src">;
-def FICOMP16m: FPI<0xDE, MRM3m, (outs), (ins i16mem:$src), "ficomp{w}\t$src">;
+def FICOM16m : FPI<0xDE, MRM2m, (outs), (ins i16mem:$src), "ficom{s}\t$src">;
+def FICOMP16m: FPI<0xDE, MRM3m, (outs), (ins i16mem:$src), "ficomp{s}\t$src">;
def FBLDm : FPI<0xDF, MRM4m, (outs), (ins f32mem:$src), "fbld\t$src">;
def FBSTPm : FPI<0xDF, MRM6m, (outs f32mem:$dst), (ins), "fbstp\t$dst">;
// Floating point cmovs.
+class FpIf32CMov<dag outs, dag ins, FPFormat fp, list<dag> pattern> :
+ FpI_<outs, ins, fp, pattern>, Requires<[FPStackf32, HasCMov]>;
+class FpIf64CMov<dag outs, dag ins, FPFormat fp, list<dag> pattern> :
+ FpI_<outs, ins, fp, pattern>, Requires<[FPStackf64, HasCMov]>;
+
multiclass FPCMov<PatLeaf cc> {
- def _Fp32 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src1, RFP32:$src2),
+ def _Fp32 : FpIf32CMov<(outs RFP32:$dst), (ins RFP32:$src1, RFP32:$src2),
CondMovFP,
[(set RFP32:$dst, (X86cmov RFP32:$src1, RFP32:$src2,
cc, EFLAGS))]>;
- def _Fp64 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src1, RFP64:$src2),
+ def _Fp64 : FpIf64CMov<(outs RFP64:$dst), (ins RFP64:$src1, RFP64:$src2),
CondMovFP,
[(set RFP64:$dst, (X86cmov RFP64:$src1, RFP64:$src2,
cc, EFLAGS))]>;
def _Fp80 : FpI_<(outs RFP80:$dst), (ins RFP80:$src1, RFP80:$src2),
CondMovFP,
[(set RFP80:$dst, (X86cmov RFP80:$src1, RFP80:$src2,
- cc, EFLAGS))]>;
+ cc, EFLAGS))]>,
+ Requires<[HasCMov]>;
}
-let Uses = [EFLAGS], isTwoAddress = 1 in {
+
+let Uses = [EFLAGS], Constraints = "$src1 = $dst" in {
defm CMOVB : FPCMov<X86_COND_B>;
defm CMOVBE : FPCMov<X86_COND_BE>;
defm CMOVE : FPCMov<X86_COND_E>;
defm CMOVNBE: FPCMov<X86_COND_A>;
defm CMOVNE : FPCMov<X86_COND_NE>;
defm CMOVNP : FPCMov<X86_COND_NP>;
-}
+} // Uses = [EFLAGS], Constraints = "$src1 = $dst"
+let Predicates = [HasCMov] in {
// These are not factored because there's no clean way to pass DA/DB.
def CMOVB_F : FPI<0xC0, AddRegFrm, (outs RST:$op), (ins),
"fcmovb\t{$op, %st(0)|%ST(0), $op}">, DA;
"fcmovne\t{$op, %st(0)|%ST(0), $op}">, DB;
def CMOVNP_F : FPI<0xD8, AddRegFrm, (outs RST:$op), (ins),
"fcmovnu\t{$op, %st(0)|%ST(0), $op}">, DB;
+} // Predicates = [HasCMov]
// Floating point loads & stores.
let canFoldAsLoad = 1 in {
def LD_Fp32m : FpIf32<(outs RFP32:$dst), (ins f32mem:$src), ZeroArgFP,
[(set RFP32:$dst, (loadf32 addr:$src))]>;
-let isReMaterializable = 1, mayHaveSideEffects = 1 in
+let isReMaterializable = 1 in
def LD_Fp64m : FpIf64<(outs RFP64:$dst), (ins f64mem:$src), ZeroArgFP,
[(set RFP64:$dst, (loadf64 addr:$src))]>;
def LD_Fp80m : FpI_<(outs RFP80:$dst), (ins f80mem:$src), ZeroArgFP,
def UCOM_Fpr80 : FpI_ <(outs), (ins RFP80:$lhs, RFP80:$rhs), CompareFP,
[]>; // FPSW = cmp ST(0) with ST(i)
+// CC = ST(0) cmp ST(i)
def UCOM_FpIr32: FpIf32<(outs), (ins RFP32:$lhs, RFP32:$rhs), CompareFP,
- [(X86cmp RFP32:$lhs, RFP32:$rhs),
- (implicit EFLAGS)]>; // CC = ST(0) cmp ST(i)
+ [(set EFLAGS, (X86cmp RFP32:$lhs, RFP32:$rhs))]>;
def UCOM_FpIr64: FpIf64<(outs), (ins RFP64:$lhs, RFP64:$rhs), CompareFP,
- [(X86cmp RFP64:$lhs, RFP64:$rhs),
- (implicit EFLAGS)]>; // CC = ST(0) cmp ST(i)
+ [(set EFLAGS, (X86cmp RFP64:$lhs, RFP64:$rhs))]>;
def UCOM_FpIr80: FpI_<(outs), (ins RFP80:$lhs, RFP80:$rhs), CompareFP,
- [(X86cmp RFP80:$lhs, RFP80:$rhs),
- (implicit EFLAGS)]>; // CC = ST(0) cmp ST(i)
+ [(set EFLAGS, (X86cmp RFP80:$lhs, RFP80:$rhs))]>;
}
let Defs = [EFLAGS], Uses = [ST0] in {
def UCOM_FIr : FPI<0xE8, AddRegFrm, // CC = cmp ST(0) with ST(i)
(outs), (ins RST:$reg),
- "fucomi\t{$reg, %st(0)|%ST(0), $reg}">, DB;
+ "fucomi\t$reg">, DB;
def UCOM_FIPr : FPI<0xE8, AddRegFrm, // CC = cmp ST(0) with ST(i), pop
(outs), (ins RST:$reg),
- "fucomip\t{$reg, %st(0)|%ST(0), $reg}">, DF;
+ "fucompi\t$reg">, DF;
}
def COM_FIr : FPI<0xF0, AddRegFrm, (outs), (ins RST:$reg),
- "fcomi\t{$reg, %st(0)|%ST(0), $reg}">, DB;
+ "fcomi\t$reg">, DB;
def COM_FIPr : FPI<0xF0, AddRegFrm, (outs), (ins RST:$reg),
- "fcomip\t{$reg, %st(0)|%ST(0), $reg}">, DF;
+ "fcompi\t$reg">, DF;
// Floating point flag ops.
let Defs = [AX] in
def FLDCW16m : I<0xD9, MRM5m, // X87 control world = [mem16]
(outs), (ins i16mem:$dst), "fldcw\t$dst", []>;
-// Register free
-
+// FPU control instructions
+def FNINIT : I<0xE3, RawFrm, (outs), (ins), "fninit", []>, DB;
def FFREE : FPI<0xC0, AddRegFrm, (outs), (ins RST:$reg),
"ffree\t$reg">, DD;
def FNCLEX : I<0xE2, RawFrm, (outs), (ins), "fnclex", []>, DB;
-// Operandless floating-point instructions for the disassembler
+// Operandless floating-point instructions for the disassembler.
+def WAIT : I<0x9B, RawFrm, (outs), (ins), "wait", []>;
def FNOP : I<0xD0, RawFrm, (outs), (ins), "fnop", []>, D9;
def FXAM : I<0xE5, RawFrm, (outs), (ins), "fxam", []>, D9;
def FXSAVE : I<0xAE, MRM0m, (outs opaque512mem:$dst), (ins),
"fxsave\t$dst", []>, TB;
+def FXSAVE64 : I<0xAE, MRM0m, (outs opaque512mem:$dst), (ins),
+ "fxsaveq\t$dst", []>, TB, REX_W, Requires<[In64BitMode]>;
def FXRSTOR : I<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
"fxrstor\t$src", []>, TB;
+def FXRSTOR64 : I<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
+ "fxrstorq\t$src", []>, TB, REX_W, Requires<[In64BitMode]>;
//===----------------------------------------------------------------------===//
// Non-Instruction Patterns
// FP extensions map onto simple pseudo-value conversions if they are to/from
// the FP stack.
-def : Pat<(f64 (fextend RFP32:$src)), (MOV_Fp3264 RFP32:$src)>,
+def : Pat<(f64 (fextend RFP32:$src)), (COPY_TO_REGCLASS RFP32:$src, RFP64)>,
Requires<[FPStackf32]>;
-def : Pat<(f80 (fextend RFP32:$src)), (MOV_Fp3280 RFP32:$src)>,
+def : Pat<(f80 (fextend RFP32:$src)), (COPY_TO_REGCLASS RFP32:$src, RFP80)>,
Requires<[FPStackf32]>;
-def : Pat<(f80 (fextend RFP64:$src)), (MOV_Fp6480 RFP64:$src)>,
+def : Pat<(f80 (fextend RFP64:$src)), (COPY_TO_REGCLASS RFP64:$src, RFP80)>,
Requires<[FPStackf64]>;
// FP truncations map onto simple pseudo-value conversions if they are to/from
// the FP stack. We have validated that only value-preserving truncations make
// it through isel.
-def : Pat<(f32 (fround RFP64:$src)), (MOV_Fp6432 RFP64:$src)>,
+def : Pat<(f32 (fround RFP64:$src)), (COPY_TO_REGCLASS RFP64:$src, RFP32)>,
Requires<[FPStackf32]>;
-def : Pat<(f32 (fround RFP80:$src)), (MOV_Fp8032 RFP80:$src)>,
+def : Pat<(f32 (fround RFP80:$src)), (COPY_TO_REGCLASS RFP80:$src, RFP32)>,
Requires<[FPStackf32]>;
-def : Pat<(f64 (fround RFP80:$src)), (MOV_Fp8064 RFP80:$src)>,
+def : Pat<(f64 (fround RFP80:$src)), (COPY_TO_REGCLASS RFP80:$src, RFP64)>,
Requires<[FPStackf64]>;
projects / oota-llvm.git / blobdiff