// MMX vector types are always returned in MM0. If the target doesn't have
// MM0, it doesn't support these vector types.
- CCIfType<[v8i8, v4i16, v2i32, v1i64], CCAssignToReg<[MM0]>>,
+ CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32], CCAssignToReg<[MM0]>>,
// Long double types are always returned in ST0 (even with SSE).
CCIfType<[f80], CCAssignToReg<[ST0, ST1]>>
// The X86-64 calling convention always returns FP values in XMM0.
CCIfType<[f32], CCAssignToReg<[XMM0, XMM1]>>,
CCIfType<[f64], CCAssignToReg<[XMM0, XMM1]>>,
+
+ // MMX vector types are always returned in XMM0.
+ CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32], CCAssignToReg<[XMM0, XMM1]>>,
CCDelegateTo<RetCC_X86Common>
]>;
// The first 8 MMX (except for v1i64) vector arguments are passed in XMM
// registers on Darwin.
- CCIfType<[v8i8, v4i16, v2i32],
+ CCIfType<[v8i8, v4i16, v2i32, v2f32],
CCIfSubtarget<"isTargetDarwin()",
CCIfSubtarget<"hasSSE2()",
CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>>>,
CCIfType<[v16i8, v8i16, v4i32, v2i64, v4f32, v2f64], CCAssignToStack<16, 16>>,
// __m64 vectors get 8-byte stack slots that are 8-byte aligned.
- CCIfType<[v8i8, v4i16, v2i32, v1i64], CCAssignToStack<8, 8>>
+ CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32], CCAssignToStack<8, 8>>
]>;
// Calling convention used on Win64
[RCX , RDX , R8 , R9 ]>>,
// The first 4 MMX vector arguments are passed in GPRs.
- CCIfType<[v8i8, v4i16, v2i32, v1i64],
+ CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32],
CCAssignToRegWithShadow<[RCX , RDX , R8 , R9 ],
[XMM0, XMM1, XMM2, XMM3]>>,
// The first 8 MMX (except for v1i64) vector arguments are passed in XMM
// registers on Darwin.
- CCIfType<[v8i8, v4i16, v2i32],
+ CCIfType<[v8i8, v4i16, v2i32, v2f32],
CCIfSubtarget<"isTargetDarwin()",
CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>>,
// The first 3 __m64 (except for v1i64) vector arguments are passed in mmx
// registers if the call is not a vararg call.
- CCIfNotVarArg<CCIfType<[v8i8, v4i16, v2i32],
+ CCIfNotVarArg<CCIfType<[v8i8, v4i16, v2i32, v2f32],
CCAssignToReg<[MM0, MM1, MM2]>>>,
// Integer/Float values get stored in stack slots that are 4 bytes in
addRegisterClass(MVT::v8i8, X86::VR64RegisterClass);
addRegisterClass(MVT::v4i16, X86::VR64RegisterClass);
addRegisterClass(MVT::v2i32, X86::VR64RegisterClass);
+ addRegisterClass(MVT::v2f32, X86::VR64RegisterClass);
addRegisterClass(MVT::v1i64, X86::VR64RegisterClass);
// FIXME: add MMX packed arithmetics
AddPromotedToType (ISD::LOAD, MVT::v4i16, MVT::v1i64);
setOperationAction(ISD::LOAD, MVT::v2i32, Promote);
AddPromotedToType (ISD::LOAD, MVT::v2i32, MVT::v1i64);
+ setOperationAction(ISD::LOAD, MVT::v2f32, Promote);
+ AddPromotedToType (ISD::LOAD, MVT::v2f32, MVT::v1i64);
setOperationAction(ISD::LOAD, MVT::v1i64, Legal);
setOperationAction(ISD::BUILD_VECTOR, MVT::v8i8, Custom);
setOperationAction(ISD::BUILD_VECTOR, MVT::v4i16, Custom);
setOperationAction(ISD::BUILD_VECTOR, MVT::v2i32, Custom);
+ setOperationAction(ISD::BUILD_VECTOR, MVT::v2f32, Custom);
setOperationAction(ISD::BUILD_VECTOR, MVT::v1i64, Custom);
setOperationAction(ISD::VECTOR_SHUFFLE, MVT::v8i8, Custom);
// Don't emit a copytoreg.
continue;
}
-
+
Chain = DAG.getCopyToReg(Chain, VA.getLocReg(), ValToCopy, Flag);
Flag = Chain.getValue(1);
}
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
def : Pat<(store (v2i32 VR64:$src), addr:$dst),
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
+def : Pat<(store (v2f32 VR64:$src), addr:$dst),
+ (MMX_MOVQ64mr addr:$dst, VR64:$src)>;
def : Pat<(store (v1i64 VR64:$src), addr:$dst),
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
// Bit convert.
def : Pat<(v8i8 (bitconvert (v1i64 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v8i8 (bitconvert (v2i32 VR64:$src))), (v8i8 VR64:$src)>;
+def : Pat<(v8i8 (bitconvert (v2f32 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v8i8 (bitconvert (v4i16 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v1i64 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v2i32 VR64:$src))), (v4i16 VR64:$src)>;
+def : Pat<(v4i16 (bitconvert (v2f32 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v8i8 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v1i64 VR64:$src))), (v2i32 VR64:$src)>;
+def : Pat<(v2i32 (bitconvert (v2f32 VR64:$src))), (v2i32 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v4i16 VR64:$src))), (v2i32 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v8i8 VR64:$src))), (v2i32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v1i64 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v2i32 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v4i16 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v8i8 VR64:$src))), (v2f32 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v2i32 VR64:$src))), (v1i64 VR64:$src)>;
+def : Pat<(v1i64 (bitconvert (v2f32 VR64:$src))), (v1i64 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v4i16 VR64:$src))), (v1i64 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v8i8 VR64:$src))), (v1i64 VR64:$src)>;
(MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v2i32 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64to64rr GR64:$src)>;
+def : Pat<(v2f32 (bitconvert (i64 GR64:$src))),
+ (MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v4i16 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v8i8 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64from64rr VR64:$src)>;
def : Pat<(i64 (bitconvert (v2i32 VR64:$src))),
(MMX_MOVD64from64rr VR64:$src)>;
+def : Pat<(i64 (bitconvert (v2f32 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
def : Pat<(i64 (bitconvert (v4i16 VR64:$src))),
(MMX_MOVD64from64rr VR64:$src)>;
def : Pat<(i64 (bitconvert (v8i8 VR64:$src))),
}
// Generic vector registers: VR64 and VR128.
-def VR64 : RegisterClass<"X86", [v8i8, v4i16, v2i32, v1i64], 64,
+def VR64 : RegisterClass<"X86", [v8i8, v4i16, v2i32, v1i64, v2f32], 64,
[MM0, MM1, MM2, MM3, MM4, MM5, MM6, MM7]>;
def VR128 : RegisterClass<"X86", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],128,
[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,