// FPStack specific DAG Nodes.
//===----------------------------------------------------------------------===//
-def SDTX86FpGet2 : SDTypeProfile<2, 0, [SDTCisVT<0, f80>,
+def SDTX86FpGet2 : SDTypeProfile<2, 0, [SDTCisVT<0, f80>,
SDTCisVT<1, f80>]>;
def SDTX86Fld : SDTypeProfile<1, 2, [SDTCisFP<0>,
- SDTCisPtrTy<1>,
+ SDTCisPtrTy<1>,
SDTCisVT<2, OtherVT>]>;
def SDTX86Fst : SDTypeProfile<0, 3, [SDTCisFP<0>,
- SDTCisPtrTy<1>,
+ SDTCisPtrTy<1>,
SDTCisVT<2, OtherVT>]>;
def SDTX86Fild : SDTypeProfile<1, 2, [SDTCisFP<0>, SDTCisPtrTy<1>,
SDTCisVT<2, OtherVT>]>;
// 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,
-// 64-bit or 80-bit floating point values. These sizes apply to the values,
+// 64-bit or 80-bit floating point values. These sizes apply to the values,
// not the registers, which are always 80 bits; RFP32, RFP64 and RFP80 can be
// copied to each other without losing information. These instructions are all
// pseudo instructions and use the "_Fp" suffix.
// The second instruction is defined with FPI, which is the actual instruction
// emitted by the assembler. These use "RST" registers, although frequently
// the actual register(s) used are implicit. These are always 80 bits.
-// The FP stackifier pass converts one to the other after register allocation
+// The FP stackifier pass converts one to the other after register allocation
// occurs.
//
// Note that the FpI instruction should have instruction selection info (e.g.
// The FopST0 series are not included here because of the irregularities
// in where the 'r' goes in assembly output.
// These instructions cannot address 80-bit memory.
-multiclass FPBinary<SDNode OpNode, Format fp, string asmstring> {
+multiclass FPBinary<SDNode OpNode, Format fp, string asmstring,
+ bit Forward = 1> {
// ST(0) = ST(0) + [mem]
-def _Fp32m : FpIf32<(outs RFP32:$dst),
+def _Fp32m : FpIf32<(outs RFP32:$dst),
(ins RFP32:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP32:$dst,
- (OpNode RFP32:$src1, (loadf32 addr:$src2)))]>;
-def _Fp64m : FpIf64<(outs RFP64:$dst),
+ [!if(Forward,
+ (set RFP32:$dst,
+ (OpNode RFP32:$src1, (loadf32 addr:$src2))),
+ (set RFP32:$dst,
+ (OpNode (loadf32 addr:$src2), RFP32:$src1)))]>;
+def _Fp64m : FpIf64<(outs RFP64:$dst),
(ins RFP64:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP64:$dst,
- (OpNode RFP64:$src1, (loadf64 addr:$src2)))]>;
-def _Fp64m32: FpIf64<(outs RFP64:$dst),
+ [!if(Forward,
+ (set RFP64:$dst,
+ (OpNode RFP64:$src1, (loadf64 addr:$src2))),
+ (set RFP64:$dst,
+ (OpNode (loadf64 addr:$src2), RFP64:$src1)))]>;
+def _Fp64m32: FpIf64<(outs RFP64:$dst),
(ins RFP64:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP64:$dst,
- (OpNode RFP64:$src1, (f64 (extloadf32 addr:$src2))))]>;
-def _Fp80m32: FpI_<(outs RFP80:$dst),
+ [!if(Forward,
+ (set RFP64:$dst,
+ (OpNode RFP64:$src1, (f64 (extloadf32 addr:$src2)))),
+ (set RFP64:$dst,
+ (OpNode (f64 (extloadf32 addr:$src2)), RFP64:$src1)))]>;
+def _Fp80m32: FpI_<(outs RFP80:$dst),
(ins RFP80:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP80:$dst,
- (OpNode RFP80:$src1, (f80 (extloadf32 addr:$src2))))]>;
-def _Fp80m64: FpI_<(outs RFP80:$dst),
+ [!if(Forward,
+ (set RFP80:$dst,
+ (OpNode RFP80:$src1, (f80 (extloadf32 addr:$src2)))),
+ (set RFP80:$dst,
+ (OpNode (f80 (extloadf32 addr:$src2)), RFP80:$src1)))]>;
+def _Fp80m64: FpI_<(outs RFP80:$dst),
(ins RFP80:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP80:$dst,
- (OpNode RFP80:$src1, (f80 (extloadf64 addr:$src2))))]>;
-def _F32m : FPI<0xD8, fp, (outs), (ins f32mem:$src),
- !strconcat("f", asmstring, "{s}\t$src")> {
- let mayLoad = 1;
-}
-def _F64m : FPI<0xDC, fp, (outs), (ins f64mem:$src),
- !strconcat("f", asmstring, "{l}\t$src")> {
- let mayLoad = 1;
-}
+ [!if(Forward,
+ (set RFP80:$dst,
+ (OpNode RFP80:$src1, (f80 (extloadf64 addr:$src2)))),
+ (set RFP80:$dst,
+ (OpNode (f80 (extloadf64 addr:$src2)), RFP80:$src1)))]>;
+let mayLoad = 1 in
+def _F32m : FPI<0xD8, fp, (outs), (ins f32mem:$src),
+ !strconcat("f", asmstring, "{s}\t$src")>;
+let mayLoad = 1 in
+def _F64m : FPI<0xDC, fp, (outs), (ins f64mem:$src),
+ !strconcat("f", asmstring, "{l}\t$src")>;
// ST(0) = ST(0) + [memint]
-def _FpI16m32 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src1, i16mem:$src2),
- OneArgFPRW,
- [(set RFP32:$dst, (OpNode RFP32:$src1,
- (X86fild addr:$src2, i16)))]>;
-def _FpI32m32 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src1, i32mem:$src2),
+def _FpI16m32 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src1, i16mem:$src2),
OneArgFPRW,
- [(set RFP32:$dst, (OpNode RFP32:$src1,
- (X86fild addr:$src2, i32)))]>;
-def _FpI16m64 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src1, i16mem:$src2),
+ [!if(Forward,
+ (set RFP32:$dst,
+ (OpNode RFP32:$src1, (X86fild addr:$src2, i16))),
+ (set RFP32:$dst,
+ (OpNode (X86fild addr:$src2, i16), RFP32:$src1)))]>;
+def _FpI32m32 : FpIf32<(outs RFP32:$dst), (ins RFP32:$src1, i32mem:$src2),
OneArgFPRW,
- [(set RFP64:$dst, (OpNode RFP64:$src1,
- (X86fild addr:$src2, i16)))]>;
-def _FpI32m64 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src1, i32mem:$src2),
+ [!if(Forward,
+ (set RFP32:$dst,
+ (OpNode RFP32:$src1, (X86fild addr:$src2, i32))),
+ (set RFP32:$dst,
+ (OpNode (X86fild addr:$src2, i32), RFP32:$src1)))]>;
+def _FpI16m64 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src1, i16mem:$src2),
OneArgFPRW,
- [(set RFP64:$dst, (OpNode RFP64:$src1,
- (X86fild addr:$src2, i32)))]>;
-def _FpI16m80 : FpI_<(outs RFP80:$dst), (ins RFP80:$src1, i16mem:$src2),
+ [!if(Forward,
+ (set RFP64:$dst,
+ (OpNode RFP64:$src1, (X86fild addr:$src2, i16))),
+ (set RFP64:$dst,
+ (OpNode (X86fild addr:$src2, i16), RFP64:$src1)))]>;
+def _FpI32m64 : FpIf64<(outs RFP64:$dst), (ins RFP64:$src1, i32mem:$src2),
OneArgFPRW,
- [(set RFP80:$dst, (OpNode RFP80:$src1,
- (X86fild addr:$src2, i16)))]>;
-def _FpI32m80 : FpI_<(outs RFP80:$dst), (ins RFP80:$src1, i32mem:$src2),
- OneArgFPRW,
- [(set RFP80:$dst, (OpNode RFP80:$src1,
- (X86fild addr:$src2, i32)))]>;
-def _FI16m : FPI<0xDE, fp, (outs), (ins i16mem:$src),
- !strconcat("fi", asmstring, "{s}\t$src")> {
- let mayLoad = 1;
-}
-def _FI32m : FPI<0xDA, fp, (outs), (ins i32mem:$src),
- !strconcat("fi", asmstring, "{l}\t$src")> {
- let mayLoad = 1;
-}
+ [!if(Forward,
+ (set RFP64:$dst,
+ (OpNode RFP64:$src1, (X86fild addr:$src2, i32))),
+ (set RFP64:$dst,
+ (OpNode (X86fild addr:$src2, i32), RFP64:$src1)))]>;
+def _FpI16m80 : FpI_<(outs RFP80:$dst), (ins RFP80:$src1, i16mem:$src2),
+ OneArgFPRW,
+ [!if(Forward,
+ (set RFP80:$dst,
+ (OpNode RFP80:$src1, (X86fild addr:$src2, i16))),
+ (set RFP80:$dst,
+ (OpNode (X86fild addr:$src2, i16), RFP80:$src1)))]>;
+def _FpI32m80 : FpI_<(outs RFP80:$dst), (ins RFP80:$src1, i32mem:$src2),
+ OneArgFPRW,
+ [!if(Forward,
+ (set RFP80:$dst,
+ (OpNode RFP80:$src1, (X86fild addr:$src2, i32))),
+ (set RFP80:$dst,
+ (OpNode (X86fild addr:$src2, i32), RFP80:$src1)))]>;
+let mayLoad = 1 in
+def _FI16m : FPI<0xDE, fp, (outs), (ins i16mem:$src),
+ !strconcat("fi", asmstring, "{s}\t$src")>;
+let mayLoad = 1 in
+def _FI32m : FPI<0xDA, fp, (outs), (ins i32mem:$src),
+ !strconcat("fi", asmstring, "{l}\t$src")>;
}
let Defs = [FPSW] in {
let SchedRW = [WriteFAddLd] in {
defm ADD : FPBinary<fadd, MRM0m, "add">;
defm SUB : FPBinary<fsub, MRM4m, "sub">;
-defm SUBR: FPBinary<fsub ,MRM5m, "subr">;
+defm SUBR: FPBinary<fsub ,MRM5m, "subr", 0>;
}
let SchedRW = [WriteFMulLd] in {
defm MUL : FPBinary<fmul, MRM1m, "mul">;
}
let SchedRW = [WriteFDivLd] in {
defm DIV : FPBinary<fdiv, MRM6m, "div">;
-defm DIVR: FPBinary<fdiv, MRM7m, "divr">;
+defm DIVR: FPBinary<fdiv, MRM7m, "divr", 0>;
}
}
def FRSTORm : FPI<0xDD, MRM4m, (outs f32mem:$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$dst">;
+def FNSTSWm : FPI<0xDD, MRM7m, (outs i16mem:$dst), (ins), "fnstsw\t$dst">;
def FICOM16m : FPI<0xDE, MRM2m, (outs), (ins i16mem:$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$dst">;
+def FBLDm : FPI<0xDF, MRM4m, (outs), (ins f80mem:$src), "fbld\t$src">;
+def FBSTPm : FPI<0xDF, MRM6m, (outs f80mem:$dst), (ins), "fbstp\t$dst">;
// Floating point cmovs.
class FpIf32CMov<dag outs, dag ins, FPFormat fp, list<dag> pattern> :
let Predicates = [HasCMov] in {
// These are not factored because there's no clean way to pass DA/DB.
-def CMOVB_F : FPI<0xDA, MRM0r, (outs RST:$op), (ins),
+def CMOVB_F : FPI<0xDA, MRM0r, (outs), (ins RST:$op),
"fcmovb\t{$op, %st(0)|st(0), $op}">;
-def CMOVBE_F : FPI<0xDA, MRM2r, (outs RST:$op), (ins),
+def CMOVBE_F : FPI<0xDA, MRM2r, (outs), (ins RST:$op),
"fcmovbe\t{$op, %st(0)|st(0), $op}">;
-def CMOVE_F : FPI<0xDA, MRM1r, (outs RST:$op), (ins),
+def CMOVE_F : FPI<0xDA, MRM1r, (outs), (ins RST:$op),
"fcmove\t{$op, %st(0)|st(0), $op}">;
-def CMOVP_F : FPI<0xDA, MRM3r, (outs RST:$op), (ins),
+def CMOVP_F : FPI<0xDA, MRM3r, (outs), (ins RST:$op),
"fcmovu\t{$op, %st(0)|st(0), $op}">;
-def CMOVNB_F : FPI<0xDB, MRM0r, (outs RST:$op), (ins),
+def CMOVNB_F : FPI<0xDB, MRM0r, (outs), (ins RST:$op),
"fcmovnb\t{$op, %st(0)|st(0), $op}">;
-def CMOVNBE_F: FPI<0xDB, MRM2r, (outs RST:$op), (ins),
+def CMOVNBE_F: FPI<0xDB, MRM2r, (outs), (ins RST:$op),
"fcmovnbe\t{$op, %st(0)|st(0), $op}">;
-def CMOVNE_F : FPI<0xDB, MRM1r, (outs RST:$op), (ins),
+def CMOVNE_F : FPI<0xDB, MRM1r, (outs), (ins RST:$op),
"fcmovne\t{$op, %st(0)|st(0), $op}">;
-def CMOVNP_F : FPI<0xDB, MRM3r, (outs RST:$op), (ins),
+def CMOVNP_F : FPI<0xDB, MRM3r, (outs), (ins RST:$op),
"fcmovnu\t{$op, %st(0)|st(0), $op}">;
} // Predicates = [HasCMov]
IIC_FST>;
def ISTT_FP32m : FPI<0xDB, MRM1m, (outs), (ins i32mem:$dst), "fisttp{l}\t$dst",
IIC_FST>;
-def ISTT_FP64m : FPI<0xDD, MRM1m, (outs), (ins i64mem:$dst),
+def ISTT_FP64m : FPI<0xDD, MRM1m, (outs), (ins i64mem:$dst),
"fisttp{ll}\t$dst", IIC_FST>;
}
def FSCALE : I<0xD9, MRM_FD, (outs), (ins), "fscale", [], IIC_FSCALE>;
def FCOMPP : I<0xDE, MRM_D9, (outs), (ins), "fcompp", [], IIC_FCOMPP>;
-def FXSAVE : I<0xAE, MRM0m, (outs opaque512mem:$dst), (ins),
- "fxsave\t$dst", [], IIC_FXSAVE>, TB;
-def FXSAVE64 : RI<0xAE, MRM0m, (outs opaque512mem:$dst), (ins),
- "fxsave{q|64}\t$dst", [], IIC_FXSAVE>, TB,
- Requires<[In64BitMode]>;
-def FXRSTOR : I<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
- "fxrstor\t$src", [], IIC_FXRSTOR>, TB;
-def FXRSTOR64 : RI<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
- "fxrstor{q|64}\t$src", [], IIC_FXRSTOR>, TB,
- Requires<[In64BitMode]>;
+let Predicates = [HasFXSR] in {
+ def FXSAVE : I<0xAE, MRM0m, (outs), (ins opaque512mem:$dst),
+ "fxsave\t$dst", [(int_x86_fxsave addr:$dst)], IIC_FXSAVE>, TB;
+ def FXSAVE64 : RI<0xAE, MRM0m, (outs), (ins opaque512mem:$dst),
+ "fxsave64\t$dst", [(int_x86_fxsave64 addr:$dst)],
+ IIC_FXSAVE>, TB, Requires<[In64BitMode]>;
+ def FXRSTOR : I<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
+ "fxrstor\t$src", [(int_x86_fxrstor addr:$src)], IIC_FXRSTOR>, TB;
+ def FXRSTOR64 : RI<0xAE, MRM1m, (outs), (ins opaque512mem:$src),
+ "fxrstor64\t$src", [(int_x86_fxrstor64 addr:$src)],
+ IIC_FXRSTOR>, TB, Requires<[In64BitMode]>;
+} // Predicates = [FeatureFXSR]
} // SchedRW
//===----------------------------------------------------------------------===//
// Required for CALL which return f32 / f64 / f80 values.
def : Pat<(X86fst RFP32:$src, addr:$op, f32), (ST_Fp32m addr:$op, RFP32:$src)>;
-def : Pat<(X86fst RFP64:$src, addr:$op, f32), (ST_Fp64m32 addr:$op,
+def : Pat<(X86fst RFP64:$src, addr:$op, f32), (ST_Fp64m32 addr:$op,
RFP64:$src)>;
def : Pat<(X86fst RFP64:$src, addr:$op, f64), (ST_Fp64m addr:$op, RFP64:$src)>;
-def : Pat<(X86fst RFP80:$src, addr:$op, f32), (ST_Fp80m32 addr:$op,
+def : Pat<(X86fst RFP80:$src, addr:$op, f32), (ST_Fp80m32 addr:$op,
RFP80:$src)>;
-def : Pat<(X86fst RFP80:$src, addr:$op, f64), (ST_Fp80m64 addr:$op,
+def : Pat<(X86fst RFP80:$src, addr:$op, f64), (ST_Fp80m64 addr:$op,
RFP80:$src)>;
def : Pat<(X86fst RFP80:$src, addr:$op, f80), (ST_FpP80m addr:$op,
RFP80:$src)>;