// Return-value conventions common to all X86 CC's.
def RetCC_X86Common : CallingConv<[
- // Scalar values are returned in AX first, then DX.
- CCIfType<[i8] , CCAssignToReg<[AL]>>,
+ // Scalar values are returned in AX first, then DX. For i8, the ABI
+ // requires the values to be in AL and AH, however this code uses AL and DL
+ // instead. This is because using AH for the second register conflicts with
+ // the way LLVM does multiple return values -- a return of {i16,i8} would end
+ // up in AX and AH, which overlap. Front-ends wishing to conform to the ABI
+ // for functions that return two i8 values are currently expected to pack the
+ // values into an i16 (which uses AX, and thus AL:AH).
+ CCIfType<[i8] , CCAssignToReg<[AL, DL]>>,
CCIfType<[i16], CCAssignToReg<[AX, DX]>>,
CCIfType<[i32], CCAssignToReg<[EAX, EDX]>>,
CCIfType<[i64], CCAssignToReg<[RAX, RDX]>>,
// weirdly; this is really the sse-regparm calling convention) in which
// case they use XMM0, otherwise it is the same as the common X86 calling
// conv.
- CCIfInReg<CCIfSubtarget<"hasSSE2()",
+ CCIfInReg<CCIfSubtarget<"hasSSE2()",
CCIfType<[f32, f64], CCAssignToReg<[XMM0,XMM1,XMM2]>>>>,
CCIfType<[f32,f64], CCAssignToReg<[ST0, ST1]>>,
CCDelegateTo<RetCC_X86Common>
// X86-32 FastCC return-value convention.
def RetCC_X86_32_Fast : CallingConv<[
// The X86-32 fastcc returns 1, 2, or 3 FP values in XMM0-2 if the target has
- // SSE2, otherwise it is the the C calling conventions.
+ // SSE2.
// This can happen when a float, 2 x float, or 3 x float vector is split by
// target lowering, and is returned in 1-3 sse regs.
CCIfType<[f32], CCIfSubtarget<"hasSSE2()", CCAssignToReg<[XMM0,XMM1,XMM2]>>>,
CCIfType<[f64], CCIfSubtarget<"hasSSE2()", CCAssignToReg<[XMM0,XMM1,XMM2]>>>,
+
+ // For integers, ECX can be used as an extra return register
+ CCIfType<[i8], CCAssignToReg<[AL, DL, CL]>>,
+ CCIfType<[i16], CCAssignToReg<[AX, DX, CX]>>,
+ CCIfType<[i32], CCAssignToReg<[EAX, EDX, ECX]>>,
+
+ // Otherwise, it is the same as the common X86 calling convention.
CCDelegateTo<RetCC_X86Common>
]>;
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]>>,
+ // MMX vector types are always returned in XMM0 except for v1i64 which is
+ // returned in RAX. This disagrees with ABI documentation but is bug
+ // compatible with gcc.
+ CCIfType<[v1i64], CCAssignToReg<[RAX]>>,
+ CCIfType<[v8i8, v4i16, v2i32, v2f32], CCAssignToReg<[XMM0, XMM1]>>,
CCDelegateTo<RetCC_X86Common>
]>;
// X86-Win64 C return-value convention.
def RetCC_X86_Win64_C : CallingConv<[
// The X86-Win64 calling convention always returns __m64 values in RAX.
- CCIfType<[v8i8, v4i16, v2i32, v1i64], CCAssignToReg<[RAX]>>,
+ CCIfType<[v8i8, v4i16, v2i32, v1i64], CCBitConvertToType<i64>>,
// And FP in XMM0 only.
CCIfType<[f32], CCAssignToReg<[XMM0]>>,
// The 'nest' parameter, if any, is passed in R10.
CCIfNest<CCAssignToReg<[R10]>>,
+ // The first 6 v1i64 vector arguments are passed in GPRs on Darwin.
+ CCIfType<[v1i64],
+ CCIfSubtarget<"isTargetDarwin()",
+ CCBitConvertToType<i64>>>,
+
// The first 6 integer arguments are passed in integer registers.
CCIfType<[i32], CCAssignToReg<[EDI, ESI, EDX, ECX, R8D, R9D]>>,
CCIfType<[i64], CCAssignToReg<[RDI, RSI, RDX, RCX, R8 , R9 ]>>,
-
- // The first 8 FP/Vector arguments are passed in XMM registers.
- CCIfType<[f32, f64, v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
- CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>,
// The first 8 MMX (except for v1i64) vector arguments are passed in XMM
// registers on Darwin.
CCIfType<[v8i8, v4i16, v2i32, v2f32],
CCIfSubtarget<"isTargetDarwin()",
CCIfSubtarget<"hasSSE2()",
- CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>>>,
+ CCPromoteToType<v2i64>>>>,
- // The first 8 v1i64 vector arguments are passed in GPRs on Darwin.
- CCIfType<[v1i64],
- CCIfSubtarget<"isTargetDarwin()",
- CCAssignToReg<[RDI, RSI, RDX, RCX, R8]>>>,
+ // The first 8 FP/Vector arguments are passed in XMM registers.
+ CCIfType<[f32, f64, v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
+ CCIfSubtarget<"hasSSE1()",
+ CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>>,
// Integer/FP values get stored in stack slots that are 8 bytes in size and
// 8-byte aligned if there are no more registers to hold them.
// The 'nest' parameter, if any, is passed in R10.
CCIfNest<CCAssignToReg<[R10]>>,
+ // 128 bit vectors are passed by pointer
+ CCIfType<[v16i8, v8i16, v4i32, v2i64, v4f32, v2f64], CCPassIndirect<i64>>,
+
+ // The first 4 MMX vector arguments are passed in GPRs.
+ CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32],
+ CCBitConvertToType<i64>>,
+
// The first 4 integer arguments are passed in integer registers.
CCIfType<[i32], CCAssignToRegWithShadow<[ECX , EDX , R8D , R9D ],
[XMM0, XMM1, XMM2, XMM3]>>,
CCAssignToRegWithShadow<[XMM0, XMM1, XMM2, XMM3],
[RCX , RDX , R8 , R9 ]>>,
- // The first 4 MMX vector arguments are passed in GPRs.
- CCIfType<[v8i8, v4i16, v2i32, v1i64, v2f32],
- CCAssignToRegWithShadow<[RCX , RDX , R8 , R9 ],
- [XMM0, XMM1, XMM2, XMM3]>>,
-
// Integer/FP values get stored in stack slots that are 8 bytes in size and
- // 16-byte aligned if there are no more registers to hold them.
- CCIfType<[i32, i64, f32, f64], CCAssignToStack<8, 16>>,
+ // 8-byte aligned if there are no more registers to hold them.
+ CCIfType<[i32, i64, f32, f64], CCAssignToStack<8, 8>>,
// Long doubles get stack slots whose size and alignment depends on the
// subtarget.
CCIfType<[f80], CCAssignToStack<0, 0>>,
- // Vectors get 16-byte stack slots that are 16-byte aligned.
- CCIfType<[v16i8, v8i16, v4i32, v2i64, v4f32, v2f64], CCAssignToStack<16, 16>>,
-
- // __m64 vectors get 8-byte stack slots that are 16-byte aligned.
- CCIfType<[v8i8, v4i16, v2i32, v1i64], CCAssignToStack<8, 16>>
-]>;
-
-// Tail call convention (fast): One register is reserved for target address,
-// namely R9
-def CC_X86_64_TailCall : CallingConv<[
- // Handles byval parameters.
- CCIfByVal<CCPassByVal<8, 8>>,
-
- // Promote i8/i16 arguments to i32.
- CCIfType<[i8, i16], CCPromoteToType<i32>>,
-
- // The 'nest' parameter, if any, is passed in R10.
- CCIfNest<CCAssignToReg<[R10]>>,
-
- // The first 6 integer arguments are passed in integer registers.
- CCIfType<[i32], CCAssignToReg<[EDI, ESI, EDX, ECX, R8D]>>,
- CCIfType<[i64], CCAssignToReg<[RDI, RSI, RDX, RCX, R8]>>,
-
- // The first 8 FP/Vector arguments are passed in XMM registers.
- CCIfType<[f32, f64, v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
- CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>,
-
- // The first 8 MMX (except for v1i64) vector arguments are passed in XMM
- // registers on Darwin.
- CCIfType<[v8i8, v4i16, v2i32, v2f32],
- CCIfSubtarget<"isTargetDarwin()",
- CCAssignToReg<[XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7]>>>,
-
- // The first 8 v1i64 vector arguments are passed in GPRs on Darwin.
- CCIfType<[v1i64],
- CCIfSubtarget<"isTargetDarwin()",
- CCAssignToReg<[RDI, RSI, RDX, RCX, R8]>>>,
-
- // Integer/FP values get stored in stack slots that are 8 bytes in size and
- // 8-byte aligned if there are no more registers to hold them.
- CCIfType<[i32, i64, f32, f64], CCAssignToStack<8, 8>>,
-
- // Vectors get 16-byte stack slots that are 16-byte aligned.
- 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>>
]>;
-
//===----------------------------------------------------------------------===//
// X86 C Calling Convention
//===----------------------------------------------------------------------===//
def CC_X86_32_FastCC : CallingConv<[
// Handles byval parameters. Note that we can't rely on the delegation
// to CC_X86_32_Common for this because that happens after code that
- // handles i32 arguments.
+ // puts arguments in registers.
CCIfByVal<CCPassByVal<4, 4>>,
// Promote i8/i16 arguments to i32.
projects / oota-llvm.git / blobdiff